diff --git a/logs_oct11/LICENSE b/logs_oct11/LICENSE new file mode 100644 index 0000000000000000000000000000000000000000..ddb59dfd4c7db6c3fff68e9fc46196d57021e0bc --- /dev/null +++ b/logs_oct11/LICENSE @@ -0,0 +1,201 @@ + Apache License + Version 2.0, January 2004 + http://www.apache.org/licenses/ + + TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION + + 1. Definitions. + + "License" shall mean the terms and conditions for use, reproduction, + and distribution as defined by Sections 1 through 9 of this document. + + "Licensor" shall mean the copyright owner or entity authorized by + the copyright owner that is granting the License. + + "Legal Entity" shall mean the union of the acting entity and all + other entities that control, are controlled by, or are under common + control with that entity. For the purposes of this definition, + "control" means (i) the power, direct or indirect, to cause the + direction or management of such entity, whether by contract or + otherwise, or (ii) ownership of fifty percent (50%) or more of the + outstanding shares, or (iii) beneficial ownership of such entity. + + "You" (or "Your") shall mean an individual or Legal Entity + exercising permissions granted by this License. + + "Source" form shall mean the preferred form for making modifications, + including but not limited to software source code, documentation + source, and configuration files. + + "Object" form shall mean any form resulting from mechanical + transformation or translation of a Source form, including but + not limited to compiled object code, generated documentation, + and conversions to other media types. + + "Work" shall mean the work of authorship, whether in Source or + Object form, made available under the License, as indicated by a + copyright notice that is included in or attached to the work + (an example is provided in the Appendix below). + + "Derivative Works" shall mean any work, whether in Source or Object + form, that is based on (or derived from) the Work and for which the + editorial revisions, annotations, elaborations, or other modifications + represent, as a whole, an original work of authorship. For the purposes + of this License, Derivative Works shall not include works that remain + separable from, or merely link (or bind by name) to the interfaces of, + the Work and Derivative Works thereof. + + "Contribution" shall mean any work of authorship, including + the original version of the Work and any modifications or additions + to that Work or Derivative Works thereof, that is intentionally + submitted to Licensor for inclusion in the Work by the copyright owner + or by an individual or Legal Entity authorized to submit on behalf of + the copyright owner. For the purposes of this definition, "submitted" + means any form of electronic, verbal, or written communication sent + to the Licensor or its representatives, including but not limited to + communication on electronic mailing lists, source code control systems, + and issue tracking systems that are managed by, or on behalf of, the + Licensor for the purpose of discussing and improving the Work, but + excluding communication that is conspicuously marked or otherwise + designated in writing by the copyright owner as "Not a Contribution." + + "Contributor" shall mean Licensor and any individual or Legal Entity + on behalf of whom a Contribution has been received by Licensor and + subsequently incorporated within the Work. + + 2. Grant of Copyright License. Subject to the terms and conditions of + this License, each Contributor hereby grants to You a perpetual, + worldwide, non-exclusive, no-charge, royalty-free, irrevocable + copyright license to reproduce, prepare Derivative Works of, + publicly display, publicly perform, sublicense, and distribute the + Work and such Derivative Works in Source or Object form. + + 3. Grant of Patent License. Subject to the terms and conditions of + this License, each Contributor hereby grants to You a perpetual, + worldwide, non-exclusive, no-charge, royalty-free, irrevocable + (except as stated in this section) patent license to make, have made, + use, offer to sell, sell, import, and otherwise transfer the Work, + where such license applies only to those patent claims licensable + by such Contributor that are necessarily infringed by their + Contribution(s) alone or by combination of their Contribution(s) + with the Work to which such Contribution(s) was submitted. If You + institute patent litigation against any entity (including a + cross-claim or counterclaim in a lawsuit) alleging that the Work + or a Contribution incorporated within the Work constitutes direct + or contributory patent infringement, then any patent licenses + granted to You under this License for that Work shall terminate + as of the date such litigation is filed. + + 4. Redistribution. You may reproduce and distribute copies of the + Work or Derivative Works thereof in any medium, with or without + modifications, and in Source or Object form, provided that You + meet the following conditions: + + (a) You must give any other recipients of the Work or + Derivative Works a copy of this License; and + + (b) You must cause any modified files to carry prominent notices + stating that You changed the files; and + + (c) You must retain, in the Source form of any Derivative Works + that You distribute, all copyright, patent, trademark, and + attribution notices from the Source form of the Work, + excluding those notices that do not pertain to any part of + the Derivative Works; and + + (d) If the Work includes a "NOTICE" text file as part of its + distribution, then any Derivative Works that You distribute must + include a readable copy of the attribution notices contained + within such NOTICE file, excluding those notices that do not + pertain to any part of the Derivative Works, in at least one + of the following places: within a NOTICE text file distributed + as part of the Derivative Works; within the Source form or + documentation, if provided along with the Derivative Works; or, + within a display generated by the Derivative Works, if and + wherever such third-party notices normally appear. The contents + of the NOTICE file are for informational purposes only and + do not modify the License. You may add Your own attribution + notices within Derivative Works that You distribute, alongside + or as an addendum to the NOTICE text from the Work, provided + that such additional attribution notices cannot be construed + as modifying the License. + + You may add Your own copyright statement to Your modifications and + may provide additional or different license terms and conditions + for use, reproduction, or distribution of Your modifications, or + for any such Derivative Works as a whole, provided Your use, + reproduction, and distribution of the Work otherwise complies with + the conditions stated in this License. + + 5. Submission of Contributions. Unless You explicitly state otherwise, + any Contribution intentionally submitted for inclusion in the Work + by You to the Licensor shall be under the terms and conditions of + this License, without any additional terms or conditions. + Notwithstanding the above, nothing herein shall supersede or modify + the terms of any separate license agreement you may have executed + with Licensor regarding such Contributions. + + 6. Trademarks. This License does not grant permission to use the trade + names, trademarks, service marks, or product names of the Licensor, + except as required for reasonable and customary use in describing the + origin of the Work and reproducing the content of the NOTICE file. + + 7. Disclaimer of Warranty. Unless required by applicable law or + agreed to in writing, Licensor provides the Work (and each + Contributor provides its Contributions) on an "AS IS" BASIS, + WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or + implied, including, without limitation, any warranties or conditions + of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A + PARTICULAR PURPOSE. You are solely responsible for determining the + appropriateness of using or redistributing the Work and assume any + risks associated with Your exercise of permissions under this License. + + 8. Limitation of Liability. In no event and under no legal theory, + whether in tort (including negligence), contract, or otherwise, + unless required by applicable law (such as deliberate and grossly + negligent acts) or agreed to in writing, shall any Contributor be + liable to You for damages, including any direct, indirect, special, + incidental, or consequential damages of any character arising as a + result of this License or out of the use or inability to use the + Work (including but not limited to damages for loss of goodwill, + work stoppage, computer failure or malfunction, or any and all + other commercial damages or losses), even if such Contributor + has been advised of the possibility of such damages. + + 9. Accepting Warranty or Additional Liability. While redistributing + the Work or Derivative Works thereof, You may choose to offer, + and charge a fee for, acceptance of support, warranty, indemnity, + or other liability obligations and/or rights consistent with this + License. However, in accepting such obligations, You may act only + on Your own behalf and on Your sole responsibility, not on behalf + of any other Contributor, and only if You agree to indemnify, + defend, and hold each Contributor harmless for any liability + incurred by, or claims asserted against, such Contributor by reason + of your accepting any such warranty or additional liability. + + END OF TERMS AND CONDITIONS + + APPENDIX: How to apply the Apache License to your work. + + To apply the Apache License to your work, attach the following + boilerplate notice, with the fields enclosed by brackets "[]" + replaced with your own identifying information. (Don't include + the brackets!) The text should be enclosed in the appropriate + comment syntax for the file format. We also recommend that a + file or class name and description of purpose be included on the + same "printed page" as the copyright notice for easier + identification within third-party archives. + + Copyright [TinyLLaVA] + + Licensed under the Apache License, Version 2.0 (the "License"); + you may not use this file except in compliance with the License. + You may obtain a copy of the License at + + http://www.apache.org/licenses/LICENSE-2.0 + + Unless required by applicable law or agreed to in writing, software + distributed under the License is distributed on an "AS IS" BASIS, + WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + See the License for the specific language governing permissions and + limitations under the License. diff --git a/logs_oct11/eval_qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_1.9_2e-1_connector-1.0_1.9_2e-1_ablation_20251012_155204.log b/logs_oct11/eval_qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_1.9_2e-1_connector-1.0_1.9_2e-1_ablation_20251012_155204.log new file mode 100644 index 0000000000000000000000000000000000000000..94f4d4820659a030590fc1adcca11a9dfc3c692d --- /dev/null +++ b/logs_oct11/eval_qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_1.9_2e-1_connector-1.0_1.9_2e-1_ablation_20251012_155204.log @@ -0,0 +1,759 @@ +==== STARTING EXPERIMENT: eval_qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_1.9_2e-1_connector-1.0_1.9_2e-1_ablation ==== +Log File: eval_qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_1.9_2e-1_connector-1.0_1.9_2e-1_ablation_20251012_155204.log +Timestamp: 2025-10-12 15:52:04 +===================================== +Processing: /nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_1.9_2e-1_connector-1.0_1.9_2e-1_ablation +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 15:52:07,149] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +config_mask.torch_dtype: torch.bfloat16 +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +Load mask model from /nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_1.9_2e-1_connector-1.0_1.9_2e-1_ablation over. +TinyLlavaConfig { + "architectures": [ + "TinyLlavaForConditionalGeneration" + ], + "backward_type_connector": "normal", + "cache_dir": null, + "connector_type": "mlp2x_gelu", + "hidden_size": 896, + "ignore_index": -100, + "image_aspect_ratio": "square", + "image_token_index": -200, + "llm_model_name_or_path": "Qwen/Qwen2.5-0.5B", + "mask_model": [ + "llm", + "connector" + ], + "mask_type_connector": "soft", + "model_type": "tinyllava", + "num_queries": 128, + "num_resampler_layers": 3, + "pad_token": "<|endoftext|>", + "resampler_hidden_size": 768, + "sparsity_connector": null, + "subnet_type_connector": "global", + "temperature_connector": 1.9, + "text_config": { + "_name_or_path": "Qwen/Qwen2.5-0.5B", + "architectures": [ + "Qwen2ForCausalLM" + ], + "backward_type": "normal", + "bos_token_id": 151643, + "eos_token_id": 151643, + "hidden_size": 896, + "intermediate_size": 4864, + "mask_type": "soft", + "masked_layers": "all", + "max_position_embeddings": 32768, + "max_window_layers": 24, + "model_type": "qwen2", + "num_attention_heads": 14, + "num_hidden_layers": 24, + "num_key_value_heads": 2, + "rope_theta": 1000000.0, + "sliding_window": 32768, + "subnet_mode": "both", + "subnet_type": "None", + "temperature_attn": 1.9, + "temperature_mlp": 1.9, + "tie_word_embeddings": true, + "torch_dtype": "bfloat16", + "use_mrope": false, + "use_sliding_window": false, + "vocab_size": 151936 + }, + "threshold_connector": null, + "tokenizer_model_max_length": 2048, + "tokenizer_name_or_path": "Qwen/Qwen2.5-0.5B", + "tokenizer_padding_side": "right", + "tokenizer_use_fast": false, + "torch_dtype": "bfloat16", + "transformers_version": "4.40.1", + "tune_type_connector": "full", + "tune_type_llm": "full", + "tune_type_vision_tower": "frozen", + "tune_vision_tower_from_layer": 0, + "use_cache": true, + "vision_config": { + "hidden_act": "gelu_pytorch_tanh", + "hidden_size": 1152, + "image_size": 384, + "intermediate_size": 4304, + "layer_norm_eps": 1e-06, + "model_name_or_path": "google/siglip-so400m-patch14-384", + "model_name_or_path2": "", + "model_type": "siglip_vision_model", + "num_attention_heads": 16, + "num_hidden_layers": 27, + "patch_size": 14 + }, + "vision_feature_layer": -2, + "vision_feature_select_strategy": "patch", + "vision_hidden_size": 1152, + "vision_model_name_or_path": "google/siglip-so400m-patch14-384", + "vision_model_name_or_path2": "", + "vocab_size": 151936 +} + +TinyLlavaForConditionalGeneration( + (language_model): Qwen2ForCausalLM( + (model): Qwen2Model( + (embed_tokens): Embedding(151936, 896) + (layers): ModuleList( + (0-23): 24 x Qwen2DecoderLayer( + (self_attn): Qwen2Attention( + (q_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=True) + (k_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=128, bias=True) + (v_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=128, bias=True) + (o_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=False) + (rotary_emb): Qwen2RotaryEmbedding() + ) + (mlp): Qwen2MLP( + (gate_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=4864, bias=False) + (up_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=4864, bias=False) + (down_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=4864, out_features=896, bias=False) + (act_fn): SiLU() + ) + (input_layernorm): Qwen2RMSNorm() + (post_attention_layernorm): Qwen2RMSNorm() + ) + ) + (norm): Qwen2RMSNorm() + ) + (lm_head): Linear(in_features=896, out_features=151936, bias=False) + ) + (vision_tower): SIGLIPVisionTower( + (_vision_tower): SiglipVisionModel( + (vision_model): SiglipVisionTransformer( + (embeddings): SiglipVisionEmbeddings( + (patch_embedding): Conv2d(3, 1152, kernel_size=(14, 14), stride=(14, 14), padding=valid) + (position_embedding): Embedding(729, 1152) + ) + (encoder): SiglipEncoder( + (layers): ModuleList( + (0-26): 27 x SiglipEncoderLayer( + (self_attn): SiglipAttention( + (k_proj): Linear(in_features=1152, out_features=1152, bias=True) + (v_proj): Linear(in_features=1152, out_features=1152, bias=True) + (q_proj): Linear(in_features=1152, out_features=1152, bias=True) + (out_proj): Linear(in_features=1152, out_features=1152, bias=True) + ) + (layer_norm1): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + (layer_norm2): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + ) + ) + ) + (post_layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (head): SiglipMultiheadAttentionPoolingHead( + (attention): MultiheadAttention( + (out_proj): NonDynamicallyQuantizableLinear(in_features=1152, out_features=1152, bias=True) + ) + (layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + ) + ) + ) + ) + (connector): MLPConnector( + (_connector): Sequential( + (0): SupermaskLinearSparsity_SoftForward_Normal(in_features=1152, out_features=896, bias=True) + (1): GELU(approximate='none') + (2): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=True) + ) + ) +) +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Load base model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain over. +TinyLlavaConfig { + "cache_dir": null, + "connector_type": "mlp2x_gelu", + "hidden_size": 896, + "ignore_index": -100, + "image_aspect_ratio": "square", + "image_token_index": -200, + "llm_model_name_or_path": "Qwen/Qwen2.5-0.5B", + "model_type": "tinyllava", + "num_queries": 128, + "num_resampler_layers": 3, + "pad_token": "<|endoftext|>", + "pad_token_id": 151643, + "resampler_hidden_size": 768, + "text_config": { + "_name_or_path": "Qwen/Qwen2.5-0.5B", + "architectures": [ + "Qwen2ForCausalLM" + ], + "bos_token_id": 151643, + "eos_token_id": 151643, + "hidden_size": 896, + "intermediate_size": 4864, + "max_position_embeddings": 32768, + "max_window_layers": 24, + "model_type": "qwen2", + "num_attention_heads": 14, + "num_hidden_layers": 24, + "num_key_value_heads": 2, + "rope_theta": 1000000.0, + "sliding_window": 32768, + "tie_word_embeddings": true, + "use_mrope": false, + "use_sliding_window": false, + "vocab_size": 151936 + }, + "tokenizer_model_max_length": 2048, + "tokenizer_name_or_path": "Qwen/Qwen2.5-0.5B", + "tokenizer_padding_side": "right", + "tokenizer_use_fast": false, + "transformers_version": "4.40.1", + "tune_type_connector": "full", + "tune_type_llm": "frozen", + "tune_type_vision_tower": "frozen", + "tune_vision_tower_from_layer": 0, + "use_cache": true, + "vision_config": { + "hidden_act": "gelu_pytorch_tanh", + "hidden_size": 1152, + "image_size": 384, + "intermediate_size": 4304, + "layer_norm_eps": 1e-06, + "model_name_or_path": "google/siglip-so400m-patch14-384", + "model_name_or_path2": "", + "model_type": "siglip_vision_model", + "num_attention_heads": 16, + "num_hidden_layers": 27, + "patch_size": 14 + }, + "vision_feature_layer": -2, + "vision_feature_select_strategy": "patch", + "vision_hidden_size": 1152, + "vision_model_name_or_path": "google/siglip-so400m-patch14-384", + "vision_model_name_or_path2": "", + "vocab_size": 151936 +} + +TinyLlavaForConditionalGeneration( + (language_model): Qwen2ForCausalLM( + (model): Qwen2Model( + (embed_tokens): Embedding(151936, 896) + (layers): ModuleList( + (0-23): 24 x Qwen2DecoderLayer( + (self_attn): Qwen2Attention( + (q_proj): Linear(in_features=896, out_features=896, bias=True) + (k_proj): Linear(in_features=896, out_features=128, bias=True) + (v_proj): Linear(in_features=896, out_features=128, bias=True) + (o_proj): Linear(in_features=896, out_features=896, bias=False) + (rotary_emb): Qwen2RotaryEmbedding() + ) + (mlp): Qwen2MLP( + (gate_proj): Linear(in_features=896, out_features=4864, bias=False) + (up_proj): Linear(in_features=896, out_features=4864, bias=False) + (down_proj): Linear(in_features=4864, out_features=896, bias=False) + (act_fn): SiLU() + ) + (input_layernorm): Qwen2RMSNorm() + (post_attention_layernorm): Qwen2RMSNorm() + ) + ) + (norm): Qwen2RMSNorm() + ) + (lm_head): Linear(in_features=896, out_features=151936, bias=False) + ) + (vision_tower): SIGLIPVisionTower( + (_vision_tower): SiglipVisionModel( + (vision_model): SiglipVisionTransformer( + (embeddings): SiglipVisionEmbeddings( + (patch_embedding): Conv2d(3, 1152, kernel_size=(14, 14), stride=(14, 14), padding=valid) + (position_embedding): Embedding(729, 1152) + ) + (encoder): SiglipEncoder( + (layers): ModuleList( + (0-26): 27 x SiglipEncoderLayer( + (self_attn): SiglipAttention( + (k_proj): Linear(in_features=1152, out_features=1152, bias=True) + (v_proj): Linear(in_features=1152, out_features=1152, bias=True) + (q_proj): Linear(in_features=1152, out_features=1152, bias=True) + (out_proj): Linear(in_features=1152, out_features=1152, bias=True) + ) + (layer_norm1): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + (layer_norm2): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + ) + ) + ) + (post_layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (head): SiglipMultiheadAttentionPoolingHead( + (attention): MultiheadAttention( + (out_proj): NonDynamicallyQuantizableLinear(in_features=1152, out_features=1152, bias=True) + ) + (layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + ) + ) + ) + ) + (connector): MLPConnector( + (_connector): Sequential( + (0): Linear(in_features=1152, out_features=896, bias=True) + (1): GELU(approximate='none') + (2): Linear(in_features=896, out_features=896, bias=True) + ) + ) +) +Collect masks for language model over. +Collect masks for connector over. +Applying mask on model.layers.0.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.self_attn.q_proj. +Applying mask on model.layers.0.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.self_attn.k_proj. +Applying mask on model.layers.0.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.self_attn.v_proj. +Applying mask on model.layers.0.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.self_attn.o_proj. +Applying mask on model.layers.0.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.mlp.gate_proj. +Applying mask on model.layers.0.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.mlp.up_proj. +Applying mask on model.layers.0.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.mlp.down_proj. +Applying mask on model.layers.1.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.self_attn.q_proj. +Applying mask on model.layers.1.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.self_attn.k_proj. +Applying mask on model.layers.1.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.self_attn.v_proj. +Applying mask on model.layers.1.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.self_attn.o_proj. +Applying mask on model.layers.1.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.mlp.gate_proj. +Applying mask on model.layers.1.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.mlp.up_proj. +Applying mask on model.layers.1.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.mlp.down_proj. +Applying mask on model.layers.2.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.self_attn.q_proj. +Applying mask on model.layers.2.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.self_attn.k_proj. +Applying mask on model.layers.2.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.self_attn.v_proj. +Applying mask on model.layers.2.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.self_attn.o_proj. +Applying mask on model.layers.2.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.mlp.gate_proj. +Applying mask on model.layers.2.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.mlp.up_proj. +Applying mask on model.layers.2.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.mlp.down_proj. +Applying mask on model.layers.3.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.self_attn.q_proj. +Applying mask on model.layers.3.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.self_attn.k_proj. +Applying mask on model.layers.3.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.self_attn.v_proj. +Applying mask on model.layers.3.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.self_attn.o_proj. +Applying mask on model.layers.3.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.mlp.gate_proj. +Applying mask on model.layers.3.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.mlp.up_proj. +Applying mask on model.layers.3.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.mlp.down_proj. +Applying mask on model.layers.4.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.self_attn.q_proj. +Applying mask on model.layers.4.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.self_attn.k_proj. +Applying mask on model.layers.4.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.self_attn.v_proj. +Applying mask on model.layers.4.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.self_attn.o_proj. +Applying mask on model.layers.4.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.mlp.gate_proj. +Applying mask on model.layers.4.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.mlp.up_proj. +Applying mask on model.layers.4.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.mlp.down_proj. +Applying mask on model.layers.5.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.self_attn.q_proj. +Applying mask on model.layers.5.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.self_attn.k_proj. +Applying mask on model.layers.5.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.self_attn.v_proj. +Applying mask on model.layers.5.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.self_attn.o_proj. +Applying mask on model.layers.5.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.mlp.gate_proj. +Applying mask on model.layers.5.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.mlp.up_proj. +Applying mask on model.layers.5.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.mlp.down_proj. +Applying mask on model.layers.6.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.self_attn.q_proj. +Applying mask on model.layers.6.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.self_attn.k_proj. +Applying mask on model.layers.6.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.self_attn.v_proj. +Applying mask on model.layers.6.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.self_attn.o_proj. +Applying mask on model.layers.6.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.mlp.gate_proj. +Applying mask on model.layers.6.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.mlp.up_proj. +Applying mask on model.layers.6.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.mlp.down_proj. +Applying mask on model.layers.7.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.self_attn.q_proj. +Applying mask on model.layers.7.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.self_attn.k_proj. +Applying mask on model.layers.7.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.self_attn.v_proj. +Applying mask on model.layers.7.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.self_attn.o_proj. +Applying mask on model.layers.7.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.mlp.gate_proj. +Applying mask on model.layers.7.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.mlp.up_proj. +Applying mask on model.layers.7.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.mlp.down_proj. +Applying mask on model.layers.8.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.self_attn.q_proj. +Applying mask on model.layers.8.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.self_attn.k_proj. +Applying mask on model.layers.8.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.self_attn.v_proj. +Applying mask on model.layers.8.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.self_attn.o_proj. +Applying mask on model.layers.8.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.mlp.gate_proj. +Applying mask on model.layers.8.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.mlp.up_proj. +Applying mask on model.layers.8.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.mlp.down_proj. +Applying mask on model.layers.9.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.self_attn.q_proj. +Applying mask on model.layers.9.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.self_attn.k_proj. +Applying mask on model.layers.9.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.self_attn.v_proj. +Applying mask on model.layers.9.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.self_attn.o_proj. +Applying mask on model.layers.9.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.mlp.gate_proj. +Applying mask on model.layers.9.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.mlp.up_proj. +Applying mask on model.layers.9.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.mlp.down_proj. +Applying mask on model.layers.10.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.self_attn.q_proj. +Applying mask on model.layers.10.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.self_attn.k_proj. +Applying mask on model.layers.10.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.self_attn.v_proj. +Applying mask on model.layers.10.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.self_attn.o_proj. +Applying mask on model.layers.10.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.mlp.gate_proj. +Applying mask on model.layers.10.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.mlp.up_proj. +Applying mask on model.layers.10.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.mlp.down_proj. +Applying mask on model.layers.11.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.self_attn.q_proj. +Applying mask on model.layers.11.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.self_attn.k_proj. +Applying mask on model.layers.11.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.self_attn.v_proj. +Applying mask on model.layers.11.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.self_attn.o_proj. +Applying mask on model.layers.11.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.mlp.gate_proj. +Applying mask on model.layers.11.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.mlp.up_proj. +Applying mask on model.layers.11.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.mlp.down_proj. +Applying mask on model.layers.12.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.self_attn.q_proj. +Applying mask on model.layers.12.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.self_attn.k_proj. +Applying mask on model.layers.12.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.self_attn.v_proj. +Applying mask on model.layers.12.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.self_attn.o_proj. +Applying mask on model.layers.12.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.mlp.gate_proj. +Applying mask on model.layers.12.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.mlp.up_proj. +Applying mask on model.layers.12.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.mlp.down_proj. +Applying mask on model.layers.13.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.self_attn.q_proj. +Applying mask on model.layers.13.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.self_attn.k_proj. +Applying mask on model.layers.13.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.self_attn.v_proj. +Applying mask on model.layers.13.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.self_attn.o_proj. +Applying mask on model.layers.13.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.mlp.gate_proj. +Applying mask on model.layers.13.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.mlp.up_proj. +Applying mask on model.layers.13.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.mlp.down_proj. +Applying mask on model.layers.14.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.self_attn.q_proj. +Applying mask on model.layers.14.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.self_attn.k_proj. +Applying mask on model.layers.14.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.self_attn.v_proj. +Applying mask on model.layers.14.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.self_attn.o_proj. +Applying mask on model.layers.14.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.mlp.gate_proj. +Applying mask on model.layers.14.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.mlp.up_proj. +Applying mask on model.layers.14.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.mlp.down_proj. +Applying mask on model.layers.15.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.self_attn.q_proj. +Applying mask on model.layers.15.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.self_attn.k_proj. +Applying mask on model.layers.15.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.self_attn.v_proj. +Applying mask on model.layers.15.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.self_attn.o_proj. +Applying mask on model.layers.15.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.mlp.gate_proj. +Applying mask on model.layers.15.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.mlp.up_proj. +Applying mask on model.layers.15.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.mlp.down_proj. +Applying mask on model.layers.16.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.self_attn.q_proj. +Applying mask on model.layers.16.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.self_attn.k_proj. +Applying mask on model.layers.16.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.self_attn.v_proj. +Applying mask on model.layers.16.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.self_attn.o_proj. +Applying mask on model.layers.16.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.mlp.gate_proj. +Applying mask on model.layers.16.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.mlp.up_proj. +Applying mask on model.layers.16.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.mlp.down_proj. +Applying mask on model.layers.17.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.self_attn.q_proj. +Applying mask on model.layers.17.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.self_attn.k_proj. +Applying mask on model.layers.17.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.self_attn.v_proj. +Applying mask on model.layers.17.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.self_attn.o_proj. +Applying mask on model.layers.17.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.mlp.gate_proj. +Applying mask on model.layers.17.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.mlp.up_proj. +Applying mask on model.layers.17.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.mlp.down_proj. +Applying mask on model.layers.18.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.self_attn.q_proj. +Applying mask on model.layers.18.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.self_attn.k_proj. +Applying mask on model.layers.18.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.self_attn.v_proj. +Applying mask on model.layers.18.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.self_attn.o_proj. +Applying mask on model.layers.18.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.mlp.gate_proj. +Applying mask on model.layers.18.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.mlp.up_proj. +Applying mask on model.layers.18.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.mlp.down_proj. +Applying mask on model.layers.19.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.self_attn.q_proj. +Applying mask on model.layers.19.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.self_attn.k_proj. +Applying mask on model.layers.19.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.self_attn.v_proj. +Applying mask on model.layers.19.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.self_attn.o_proj. +Applying mask on model.layers.19.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.mlp.gate_proj. +Applying mask on model.layers.19.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.mlp.up_proj. +Applying mask on model.layers.19.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.mlp.down_proj. +Applying mask on model.layers.20.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.self_attn.q_proj. +Applying mask on model.layers.20.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.self_attn.k_proj. +Applying mask on model.layers.20.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.self_attn.v_proj. +Applying mask on model.layers.20.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.self_attn.o_proj. +Applying mask on model.layers.20.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.mlp.gate_proj. +Applying mask on model.layers.20.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.mlp.up_proj. +Applying mask on model.layers.20.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.mlp.down_proj. +Applying mask on model.layers.21.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.self_attn.q_proj. +Applying mask on model.layers.21.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.self_attn.k_proj. +Applying mask on model.layers.21.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.self_attn.v_proj. +Applying mask on model.layers.21.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.self_attn.o_proj. +Applying mask on model.layers.21.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.mlp.gate_proj. +Applying mask on model.layers.21.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.mlp.up_proj. +Applying mask on model.layers.21.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.mlp.down_proj. +Applying mask on model.layers.22.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.self_attn.q_proj. +Applying mask on model.layers.22.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.self_attn.k_proj. +Applying mask on model.layers.22.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.self_attn.v_proj. +Applying mask on model.layers.22.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.self_attn.o_proj. +Applying mask on model.layers.22.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.mlp.gate_proj. +Applying mask on model.layers.22.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.mlp.up_proj. +Applying mask on model.layers.22.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.mlp.down_proj. +Applying mask on model.layers.23.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.self_attn.q_proj. +Applying mask on model.layers.23.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.self_attn.k_proj. +Applying mask on model.layers.23.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.self_attn.v_proj. +Applying mask on model.layers.23.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.self_attn.o_proj. +Applying mask on model.layers.23.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.mlp.gate_proj. +Applying mask on model.layers.23.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.mlp.up_proj. +Applying mask on model.layers.23.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.mlp.down_proj. +Applying mask on _connector.0 with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on _connector.0. +Applying mask on _connector.2 with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on _connector.2. +Using cleaned config_mask (without mask parameters) for saving. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 15:52:49,500] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. + 0%| | 0/900 [00:00", + "resampler_hidden_size": 768, + "sparsity_connector": null, + "subnet_type_connector": "global", + "temperature_connector": 2.1, + "text_config": { + "_name_or_path": "Qwen/Qwen2.5-0.5B", + "architectures": [ + "Qwen2ForCausalLM" + ], + "backward_type": "normal", + "bos_token_id": 151643, + "eos_token_id": 151643, + "hidden_size": 896, + "intermediate_size": 4864, + "mask_type": "soft", + "masked_layers": "all", + "max_position_embeddings": 32768, + "max_window_layers": 24, + "model_type": "qwen2", + "num_attention_heads": 14, + "num_hidden_layers": 24, + "num_key_value_heads": 2, + "rope_theta": 1000000.0, + "sliding_window": 32768, + "subnet_mode": "both", + "subnet_type": "None", + "temperature_attn": 2.1, + "temperature_mlp": 2.1, + "tie_word_embeddings": true, + "torch_dtype": "bfloat16", + "use_mrope": false, + "use_sliding_window": false, + "vocab_size": 151936 + }, + "threshold_connector": null, + "tokenizer_model_max_length": 2048, + "tokenizer_name_or_path": "Qwen/Qwen2.5-0.5B", + "tokenizer_padding_side": "right", + "tokenizer_use_fast": false, + "torch_dtype": "bfloat16", + "transformers_version": "4.40.1", + "tune_type_connector": "full", + "tune_type_llm": "full", + "tune_type_vision_tower": "frozen", + "tune_vision_tower_from_layer": 0, + "use_cache": true, + "vision_config": { + "hidden_act": "gelu_pytorch_tanh", + "hidden_size": 1152, + "image_size": 384, + "intermediate_size": 4304, + "layer_norm_eps": 1e-06, + "model_name_or_path": "google/siglip-so400m-patch14-384", + "model_name_or_path2": "", + "model_type": "siglip_vision_model", + "num_attention_heads": 16, + "num_hidden_layers": 27, + "patch_size": 14 + }, + "vision_feature_layer": -2, + "vision_feature_select_strategy": "patch", + "vision_hidden_size": 1152, + "vision_model_name_or_path": "google/siglip-so400m-patch14-384", + "vision_model_name_or_path2": "", + "vocab_size": 151936 +} + +TinyLlavaForConditionalGeneration( + (language_model): Qwen2ForCausalLM( + (model): Qwen2Model( + (embed_tokens): Embedding(151936, 896) + (layers): ModuleList( + (0-23): 24 x Qwen2DecoderLayer( + (self_attn): Qwen2Attention( + (q_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=True) + (k_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=128, bias=True) + (v_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=128, bias=True) + (o_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=False) + (rotary_emb): Qwen2RotaryEmbedding() + ) + (mlp): Qwen2MLP( + (gate_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=4864, bias=False) + (up_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=4864, bias=False) + (down_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=4864, out_features=896, bias=False) + (act_fn): SiLU() + ) + (input_layernorm): Qwen2RMSNorm() + (post_attention_layernorm): Qwen2RMSNorm() + ) + ) + (norm): Qwen2RMSNorm() + ) + (lm_head): Linear(in_features=896, out_features=151936, bias=False) + ) + (vision_tower): SIGLIPVisionTower( + (_vision_tower): SiglipVisionModel( + (vision_model): SiglipVisionTransformer( + (embeddings): SiglipVisionEmbeddings( + (patch_embedding): Conv2d(3, 1152, kernel_size=(14, 14), stride=(14, 14), padding=valid) + (position_embedding): Embedding(729, 1152) + ) + (encoder): SiglipEncoder( + (layers): ModuleList( + (0-26): 27 x SiglipEncoderLayer( + (self_attn): SiglipAttention( + (k_proj): Linear(in_features=1152, out_features=1152, bias=True) + (v_proj): Linear(in_features=1152, out_features=1152, bias=True) + (q_proj): Linear(in_features=1152, out_features=1152, bias=True) + (out_proj): Linear(in_features=1152, out_features=1152, bias=True) + ) + (layer_norm1): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + (layer_norm2): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + ) + ) + ) + (post_layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (head): SiglipMultiheadAttentionPoolingHead( + (attention): MultiheadAttention( + (out_proj): NonDynamicallyQuantizableLinear(in_features=1152, out_features=1152, bias=True) + ) + (layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + ) + ) + ) + ) + (connector): MLPConnector( + (_connector): Sequential( + (0): SupermaskLinearSparsity_SoftForward_Normal(in_features=1152, out_features=896, bias=True) + (1): GELU(approximate='none') + (2): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=True) + ) + ) +) +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Load base model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain over. +TinyLlavaConfig { + "cache_dir": null, + "connector_type": "mlp2x_gelu", + "hidden_size": 896, + "ignore_index": -100, + "image_aspect_ratio": "square", + "image_token_index": -200, + "llm_model_name_or_path": "Qwen/Qwen2.5-0.5B", + "model_type": "tinyllava", + "num_queries": 128, + "num_resampler_layers": 3, + "pad_token": "<|endoftext|>", + "pad_token_id": 151643, + "resampler_hidden_size": 768, + "text_config": { + "_name_or_path": "Qwen/Qwen2.5-0.5B", + "architectures": [ + "Qwen2ForCausalLM" + ], + "bos_token_id": 151643, + "eos_token_id": 151643, + "hidden_size": 896, + "intermediate_size": 4864, + "max_position_embeddings": 32768, + "max_window_layers": 24, + "model_type": "qwen2", + "num_attention_heads": 14, + "num_hidden_layers": 24, + "num_key_value_heads": 2, + "rope_theta": 1000000.0, + "sliding_window": 32768, + "tie_word_embeddings": true, + "use_mrope": false, + "use_sliding_window": false, + "vocab_size": 151936 + }, + "tokenizer_model_max_length": 2048, + "tokenizer_name_or_path": "Qwen/Qwen2.5-0.5B", + "tokenizer_padding_side": "right", + "tokenizer_use_fast": false, + "transformers_version": "4.40.1", + "tune_type_connector": "full", + "tune_type_llm": "frozen", + "tune_type_vision_tower": "frozen", + "tune_vision_tower_from_layer": 0, + "use_cache": true, + "vision_config": { + "hidden_act": "gelu_pytorch_tanh", + "hidden_size": 1152, + "image_size": 384, + "intermediate_size": 4304, + "layer_norm_eps": 1e-06, + "model_name_or_path": "google/siglip-so400m-patch14-384", + "model_name_or_path2": "", + "model_type": "siglip_vision_model", + "num_attention_heads": 16, + "num_hidden_layers": 27, + "patch_size": 14 + }, + "vision_feature_layer": -2, + "vision_feature_select_strategy": "patch", + "vision_hidden_size": 1152, + "vision_model_name_or_path": "google/siglip-so400m-patch14-384", + "vision_model_name_or_path2": "", + "vocab_size": 151936 +} + +TinyLlavaForConditionalGeneration( + (language_model): Qwen2ForCausalLM( + (model): Qwen2Model( + (embed_tokens): Embedding(151936, 896) + (layers): ModuleList( + (0-23): 24 x Qwen2DecoderLayer( + (self_attn): Qwen2Attention( + (q_proj): Linear(in_features=896, out_features=896, bias=True) + (k_proj): Linear(in_features=896, out_features=128, bias=True) + (v_proj): Linear(in_features=896, out_features=128, bias=True) + (o_proj): Linear(in_features=896, out_features=896, bias=False) + (rotary_emb): Qwen2RotaryEmbedding() + ) + (mlp): Qwen2MLP( + (gate_proj): Linear(in_features=896, out_features=4864, bias=False) + (up_proj): Linear(in_features=896, out_features=4864, bias=False) + (down_proj): Linear(in_features=4864, out_features=896, bias=False) + (act_fn): SiLU() + ) + (input_layernorm): Qwen2RMSNorm() + (post_attention_layernorm): Qwen2RMSNorm() + ) + ) + (norm): Qwen2RMSNorm() + ) + (lm_head): Linear(in_features=896, out_features=151936, bias=False) + ) + (vision_tower): SIGLIPVisionTower( + (_vision_tower): SiglipVisionModel( + (vision_model): SiglipVisionTransformer( + (embeddings): SiglipVisionEmbeddings( + (patch_embedding): Conv2d(3, 1152, kernel_size=(14, 14), stride=(14, 14), padding=valid) + (position_embedding): Embedding(729, 1152) + ) + (encoder): SiglipEncoder( + (layers): ModuleList( + (0-26): 27 x SiglipEncoderLayer( + (self_attn): SiglipAttention( + (k_proj): Linear(in_features=1152, out_features=1152, bias=True) + (v_proj): Linear(in_features=1152, out_features=1152, bias=True) + (q_proj): Linear(in_features=1152, out_features=1152, bias=True) + (out_proj): Linear(in_features=1152, out_features=1152, bias=True) + ) + (layer_norm1): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + (layer_norm2): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + ) + ) + ) + (post_layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (head): SiglipMultiheadAttentionPoolingHead( + (attention): MultiheadAttention( + (out_proj): NonDynamicallyQuantizableLinear(in_features=1152, out_features=1152, bias=True) + ) + (layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + ) + ) + ) + ) + (connector): MLPConnector( + (_connector): Sequential( + (0): Linear(in_features=1152, out_features=896, bias=True) + (1): GELU(approximate='none') + (2): Linear(in_features=896, out_features=896, bias=True) + ) + ) +) +Collect masks for language model over. +Collect masks for connector over. +Applying mask on model.layers.0.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.self_attn.q_proj. +Applying mask on model.layers.0.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.self_attn.k_proj. +Applying mask on model.layers.0.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.self_attn.v_proj. +Applying mask on model.layers.0.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.self_attn.o_proj. +Applying mask on model.layers.0.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.mlp.gate_proj. +Applying mask on model.layers.0.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.mlp.up_proj. +Applying mask on model.layers.0.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.mlp.down_proj. +Applying mask on model.layers.1.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.self_attn.q_proj. +Applying mask on model.layers.1.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.self_attn.k_proj. +Applying mask on model.layers.1.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.self_attn.v_proj. +Applying mask on model.layers.1.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.self_attn.o_proj. +Applying mask on model.layers.1.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.mlp.gate_proj. +Applying mask on model.layers.1.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.mlp.up_proj. +Applying mask on model.layers.1.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.mlp.down_proj. +Applying mask on model.layers.2.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.self_attn.q_proj. +Applying mask on model.layers.2.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.self_attn.k_proj. +Applying mask on model.layers.2.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.self_attn.v_proj. +Applying mask on model.layers.2.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.self_attn.o_proj. +Applying mask on model.layers.2.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.mlp.gate_proj. +Applying mask on model.layers.2.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.mlp.up_proj. +Applying mask on model.layers.2.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.mlp.down_proj. +Applying mask on model.layers.3.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.self_attn.q_proj. +Applying mask on model.layers.3.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.self_attn.k_proj. +Applying mask on model.layers.3.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.self_attn.v_proj. +Applying mask on model.layers.3.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.self_attn.o_proj. +Applying mask on model.layers.3.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.mlp.gate_proj. +Applying mask on model.layers.3.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.mlp.up_proj. +Applying mask on model.layers.3.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.mlp.down_proj. +Applying mask on model.layers.4.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.self_attn.q_proj. +Applying mask on model.layers.4.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.self_attn.k_proj. +Applying mask on model.layers.4.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.self_attn.v_proj. +Applying mask on model.layers.4.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.self_attn.o_proj. +Applying mask on model.layers.4.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.mlp.gate_proj. +Applying mask on model.layers.4.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.mlp.up_proj. +Applying mask on model.layers.4.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.mlp.down_proj. +Applying mask on model.layers.5.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.self_attn.q_proj. +Applying mask on model.layers.5.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.self_attn.k_proj. +Applying mask on model.layers.5.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.self_attn.v_proj. +Applying mask on model.layers.5.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.self_attn.o_proj. +Applying mask on model.layers.5.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.mlp.gate_proj. +Applying mask on model.layers.5.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.mlp.up_proj. +Applying mask on model.layers.5.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.mlp.down_proj. +Applying mask on model.layers.6.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.self_attn.q_proj. +Applying mask on model.layers.6.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.self_attn.k_proj. +Applying mask on model.layers.6.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.self_attn.v_proj. +Applying mask on model.layers.6.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.self_attn.o_proj. +Applying mask on model.layers.6.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.mlp.gate_proj. +Applying mask on model.layers.6.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.mlp.up_proj. +Applying mask on model.layers.6.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.mlp.down_proj. +Applying mask on model.layers.7.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.self_attn.q_proj. +Applying mask on model.layers.7.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.self_attn.k_proj. +Applying mask on model.layers.7.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.self_attn.v_proj. +Applying mask on model.layers.7.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.self_attn.o_proj. +Applying mask on model.layers.7.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.mlp.gate_proj. +Applying mask on model.layers.7.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.mlp.up_proj. +Applying mask on model.layers.7.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.mlp.down_proj. +Applying mask on model.layers.8.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.self_attn.q_proj. +Applying mask on model.layers.8.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.self_attn.k_proj. +Applying mask on model.layers.8.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.self_attn.v_proj. +Applying mask on model.layers.8.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.self_attn.o_proj. +Applying mask on model.layers.8.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.mlp.gate_proj. +Applying mask on model.layers.8.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.mlp.up_proj. +Applying mask on model.layers.8.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.mlp.down_proj. +Applying mask on model.layers.9.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.self_attn.q_proj. +Applying mask on model.layers.9.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.self_attn.k_proj. +Applying mask on model.layers.9.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.self_attn.v_proj. +Applying mask on model.layers.9.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.self_attn.o_proj. +Applying mask on model.layers.9.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.mlp.gate_proj. +Applying mask on model.layers.9.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.mlp.up_proj. +Applying mask on model.layers.9.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.mlp.down_proj. +Applying mask on model.layers.10.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.self_attn.q_proj. +Applying mask on model.layers.10.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.self_attn.k_proj. +Applying mask on model.layers.10.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.self_attn.v_proj. +Applying mask on model.layers.10.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.self_attn.o_proj. +Applying mask on model.layers.10.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.mlp.gate_proj. +Applying mask on model.layers.10.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.mlp.up_proj. +Applying mask on model.layers.10.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.mlp.down_proj. +Applying mask on model.layers.11.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.self_attn.q_proj. +Applying mask on model.layers.11.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.self_attn.k_proj. +Applying mask on model.layers.11.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.self_attn.v_proj. +Applying mask on model.layers.11.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.self_attn.o_proj. +Applying mask on model.layers.11.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.mlp.gate_proj. +Applying mask on model.layers.11.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.mlp.up_proj. +Applying mask on model.layers.11.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.mlp.down_proj. +Applying mask on model.layers.12.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.self_attn.q_proj. +Applying mask on model.layers.12.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.self_attn.k_proj. +Applying mask on model.layers.12.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.self_attn.v_proj. +Applying mask on model.layers.12.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.self_attn.o_proj. +Applying mask on model.layers.12.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.mlp.gate_proj. +Applying mask on model.layers.12.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.mlp.up_proj. +Applying mask on model.layers.12.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.mlp.down_proj. +Applying mask on model.layers.13.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.self_attn.q_proj. +Applying mask on model.layers.13.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.self_attn.k_proj. +Applying mask on model.layers.13.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.self_attn.v_proj. +Applying mask on model.layers.13.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.self_attn.o_proj. +Applying mask on model.layers.13.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.mlp.gate_proj. +Applying mask on model.layers.13.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.mlp.up_proj. +Applying mask on model.layers.13.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.mlp.down_proj. +Applying mask on model.layers.14.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.self_attn.q_proj. +Applying mask on model.layers.14.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.self_attn.k_proj. +Applying mask on model.layers.14.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.self_attn.v_proj. +Applying mask on model.layers.14.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.self_attn.o_proj. +Applying mask on model.layers.14.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.mlp.gate_proj. +Applying mask on model.layers.14.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.mlp.up_proj. +Applying mask on model.layers.14.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.mlp.down_proj. +Applying mask on model.layers.15.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.self_attn.q_proj. +Applying mask on model.layers.15.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.self_attn.k_proj. +Applying mask on model.layers.15.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.self_attn.v_proj. +Applying mask on model.layers.15.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.self_attn.o_proj. +Applying mask on model.layers.15.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.mlp.gate_proj. +Applying mask on model.layers.15.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.mlp.up_proj. +Applying mask on model.layers.15.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.mlp.down_proj. +Applying mask on model.layers.16.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.self_attn.q_proj. +Applying mask on model.layers.16.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.self_attn.k_proj. +Applying mask on model.layers.16.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.self_attn.v_proj. +Applying mask on model.layers.16.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.self_attn.o_proj. +Applying mask on model.layers.16.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.mlp.gate_proj. +Applying mask on model.layers.16.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.mlp.up_proj. +Applying mask on model.layers.16.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.mlp.down_proj. +Applying mask on model.layers.17.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.self_attn.q_proj. +Applying mask on model.layers.17.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.self_attn.k_proj. +Applying mask on model.layers.17.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.self_attn.v_proj. +Applying mask on model.layers.17.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.self_attn.o_proj. +Applying mask on model.layers.17.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.mlp.gate_proj. +Applying mask on model.layers.17.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.mlp.up_proj. +Applying mask on model.layers.17.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.mlp.down_proj. +Applying mask on model.layers.18.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.self_attn.q_proj. +Applying mask on model.layers.18.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.self_attn.k_proj. +Applying mask on model.layers.18.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.self_attn.v_proj. +Applying mask on model.layers.18.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.self_attn.o_proj. +Applying mask on model.layers.18.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.mlp.gate_proj. +Applying mask on model.layers.18.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.mlp.up_proj. +Applying mask on model.layers.18.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.mlp.down_proj. +Applying mask on model.layers.19.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.self_attn.q_proj. +Applying mask on model.layers.19.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.self_attn.k_proj. +Applying mask on model.layers.19.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.self_attn.v_proj. +Applying mask on model.layers.19.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.self_attn.o_proj. +Applying mask on model.layers.19.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.mlp.gate_proj. +Applying mask on model.layers.19.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.mlp.up_proj. +Applying mask on model.layers.19.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.mlp.down_proj. +Applying mask on model.layers.20.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.self_attn.q_proj. +Applying mask on model.layers.20.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.self_attn.k_proj. +Applying mask on model.layers.20.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.self_attn.v_proj. +Applying mask on model.layers.20.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.self_attn.o_proj. +Applying mask on model.layers.20.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.mlp.gate_proj. +Applying mask on model.layers.20.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.mlp.up_proj. +Applying mask on model.layers.20.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.mlp.down_proj. +Applying mask on model.layers.21.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.self_attn.q_proj. +Applying mask on model.layers.21.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.self_attn.k_proj. +Applying mask on model.layers.21.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.self_attn.v_proj. +Applying mask on model.layers.21.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.self_attn.o_proj. +Applying mask on model.layers.21.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.mlp.gate_proj. +Applying mask on model.layers.21.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.mlp.up_proj. +Applying mask on model.layers.21.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.mlp.down_proj. +Applying mask on model.layers.22.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.self_attn.q_proj. +Applying mask on model.layers.22.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.self_attn.k_proj. +Applying mask on model.layers.22.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.self_attn.v_proj. +Applying mask on model.layers.22.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.self_attn.o_proj. +Applying mask on model.layers.22.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.mlp.gate_proj. +Applying mask on model.layers.22.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.mlp.up_proj. +Applying mask on model.layers.22.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.mlp.down_proj. +Applying mask on model.layers.23.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.self_attn.q_proj. +Applying mask on model.layers.23.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.self_attn.k_proj. +Applying mask on model.layers.23.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.self_attn.v_proj. +Applying mask on model.layers.23.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.self_attn.o_proj. +Applying mask on model.layers.23.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.mlp.gate_proj. +Applying mask on model.layers.23.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.mlp.up_proj. +Applying mask on model.layers.23.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.mlp.down_proj. +Applying mask on _connector.0 with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on _connector.0. +Applying mask on _connector.2 with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on _connector.2. +Using cleaned config_mask (without mask parameters) for saving. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 16:22:12,939] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. + 0%| | 0/900 [00:00 + main() + File "/nfs/ywang29/TinyLLaVA/scripts/apply_masks.py", line 123, in main + config_mask = TinyLlavaConfig.from_pretrained(model_args.mask_model_name_or_path) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/configuration_utils.py", line 602, in from_pretrained + config_dict, kwargs = cls.get_config_dict(pretrained_model_name_or_path, **kwargs) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/configuration_utils.py", line 631, in get_config_dict + config_dict, kwargs = cls._get_config_dict(pretrained_model_name_or_path, **kwargs) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/configuration_utils.py", line 686, in _get_config_dict + resolved_config_file = cached_file( + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/utils/hub.py", line 369, in cached_file + raise EnvironmentError( +OSError: /nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.3_2e-1_connector-1.0_2.3_2e-1_ablation does not appear to have a file named config.json. Checkout 'https://huggingface.co//nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.3_2e-1_connector-1.0_2.3_2e-1_ablation/tree/main' for available files. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 16:46:21,768] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +Traceback (most recent call last): + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/utils/hub.py", line 398, in cached_file + resolved_file = hf_hub_download( + File "/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/utils/_validators.py", line 106, in _inner_fn + validate_repo_id(arg_value) + File "/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/utils/_validators.py", line 154, in validate_repo_id + raise HFValidationError( +huggingface_hub.errors.HFValidationError: Repo id must be in the form 'repo_name' or 'namespace/repo_name': '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.3_2e-1_connector-1.0_2.3_2e-1_ablation/mask_applied'. Use `repo_type` argument if needed. + +The above exception was the direct cause of the following exception: + +Traceback (most recent call last): + File "/nfs/ywang29/TinyLLaVA/tinyllava/model/load_model.py", line 38, in load_pretrained_model + model = TinyLlavaForConditionalGeneration.from_pretrained(model_name_or_path,low_cpu_mem_usage=True) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/modeling_utils.py", line 3015, in from_pretrained + resolved_config_file = cached_file( + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/utils/hub.py", line 462, in cached_file + raise EnvironmentError( +OSError: Incorrect path_or_model_id: '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.3_2e-1_connector-1.0_2.3_2e-1_ablation/mask_applied'. Please provide either the path to a local folder or the repo_id of a model on the Hub. + +During handling of the above exception, another exception occurred: + +Traceback (most recent call last): + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/utils/hub.py", line 398, in cached_file + resolved_file = hf_hub_download( + File "/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/utils/_validators.py", line 106, in _inner_fn + validate_repo_id(arg_value) + File "/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/utils/_validators.py", line 154, in validate_repo_id + raise HFValidationError( +huggingface_hub.errors.HFValidationError: Repo id must be in the form 'repo_name' or 'namespace/repo_name': '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.3_2e-1_connector-1.0_2.3_2e-1_ablation/mask_applied'. Use `repo_type` argument if needed. + +The above exception was the direct cause of the following exception: + +Traceback (most recent call last): + File "/opt/conda/envs/tinyllava/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/opt/conda/envs/tinyllava/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/nfs/ywang29/TinyLLaVA/tinyllava/eval/model_vqa_mmmu.py", line 180, in + eval_model(args) + File "/nfs/ywang29/TinyLLaVA/tinyllava/eval/model_vqa_mmmu.py", line 88, in eval_model + model, tokenizer, image_processor, context_len = load_pretrained_model(model_path) + File "/nfs/ywang29/TinyLLaVA/tinyllava/model/load_model.py", line 40, in load_pretrained_model + model_config = TinyLlavaConfig.from_pretrained(model_name_or_path) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/configuration_utils.py", line 602, in from_pretrained + config_dict, kwargs = cls.get_config_dict(pretrained_model_name_or_path, **kwargs) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/configuration_utils.py", line 631, in get_config_dict + config_dict, kwargs = cls._get_config_dict(pretrained_model_name_or_path, **kwargs) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/configuration_utils.py", line 686, in _get_config_dict + resolved_config_file = cached_file( + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/utils/hub.py", line 462, in cached_file + raise EnvironmentError( +OSError: Incorrect path_or_model_id: '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.3_2e-1_connector-1.0_2.3_2e-1_ablation/mask_applied'. Please provide either the path to a local folder or the repo_id of a model on the Hub. +Traceback (most recent call last): + File "/nfs/ywang29/TinyLLaVA/scripts/convert_answer_to_mmmu.py", line 31, in + eval_model(args) + File "/nfs/ywang29/TinyLLaVA/scripts/convert_answer_to_mmmu.py", line 7, in eval_model + answers = [json.loads(q) for q in open(os.path.expanduser(args.answers_file), "r")] +FileNotFoundError: [Errno 2] No such file or directory: '/s3-code/ywang29/datasets/tinyllava/eval/MMMU/answers/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.3_2e-1_connector-1.0_2.3_2e-1_ablation-mask_applied.jsonl' +Traceback (most recent call last): + File "/s3-code/ywang29/datasets/tinyllava/eval/MMMU/eval/main_eval_only.py", line 19, in + output_dict = json.load(open(args.output_path)) +FileNotFoundError: [Errno 2] No such file or directory: '/s3-code/ywang29/datasets/tinyllava/eval/MMMU/answers/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.3_2e-1_connector-1.0_2.3_2e-1_ablation-mask_applied_output.json' +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 16:46:28,066] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +Traceback (most recent call last): + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/utils/hub.py", line 398, in cached_file + resolved_file = hf_hub_download( + File "/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/utils/_validators.py", line 106, in _inner_fn + validate_repo_id(arg_value) + File "/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/utils/_validators.py", line 154, in validate_repo_id + raise HFValidationError( +huggingface_hub.errors.HFValidationError: Repo id must be in the form 'repo_name' or 'namespace/repo_name': '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.3_2e-1_connector-1.0_2.3_2e-1_ablation/mask_applied'. Use `repo_type` argument if needed. + +The above exception was the direct cause of the following exception: + +Traceback (most recent call last): + File "/nfs/ywang29/TinyLLaVA/tinyllava/model/load_model.py", line 38, in load_pretrained_model + model = TinyLlavaForConditionalGeneration.from_pretrained(model_name_or_path,low_cpu_mem_usage=True) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/modeling_utils.py", line 3015, in from_pretrained + resolved_config_file = cached_file( + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/utils/hub.py", line 462, in cached_file + raise EnvironmentError( +OSError: Incorrect path_or_model_id: '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.3_2e-1_connector-1.0_2.3_2e-1_ablation/mask_applied'. Please provide either the path to a local folder or the repo_id of a model on the Hub. + +During handling of the above exception, another exception occurred: + +Traceback (most recent call last): + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/utils/hub.py", line 398, in cached_file + resolved_file = hf_hub_download( + File "/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/utils/_validators.py", line 106, in _inner_fn + validate_repo_id(arg_value) + File "/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/utils/_validators.py", line 154, in validate_repo_id + raise HFValidationError( +huggingface_hub.errors.HFValidationError: Repo id must be in the form 'repo_name' or 'namespace/repo_name': '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.3_2e-1_connector-1.0_2.3_2e-1_ablation/mask_applied'. Use `repo_type` argument if needed. + +The above exception was the direct cause of the following exception: + +Traceback (most recent call last): + File "/opt/conda/envs/tinyllava/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/opt/conda/envs/tinyllava/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/nfs/ywang29/TinyLLaVA/tinyllava/eval/model_vqa_loader.py", line 146, in + eval_model(args) + File "/nfs/ywang29/TinyLLaVA/tinyllava/eval/model_vqa_loader.py", line 79, in eval_model + model, tokenizer, image_processor, context_len = load_pretrained_model(model_path) + File "/nfs/ywang29/TinyLLaVA/tinyllava/model/load_model.py", line 40, in load_pretrained_model + model_config = TinyLlavaConfig.from_pretrained(model_name_or_path) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/configuration_utils.py", line 602, in from_pretrained + config_dict, kwargs = cls.get_config_dict(pretrained_model_name_or_path, **kwargs) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/configuration_utils.py", line 631, in get_config_dict + config_dict, kwargs = cls._get_config_dict(pretrained_model_name_or_path, **kwargs) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/configuration_utils.py", line 686, in _get_config_dict + resolved_config_file = cached_file( + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/utils/hub.py", line 462, in cached_file + raise EnvironmentError( +OSError: Incorrect path_or_model_id: '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.3_2e-1_connector-1.0_2.3_2e-1_ablation/mask_applied'. Please provide either the path to a local folder or the repo_id of a model on the Hub. +Traceback (most recent call last): + File "/s3-code/ywang29/datasets/tinyllava/eval/MME/convert_answer_to_mme.py", line 52, in + answers = [json.loads(line) for line in open(os.path.join('answers', f'{experiment}.jsonl'))] +FileNotFoundError: [Errno 2] No such file or directory: 'answers/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.3_2e-1_connector-1.0_2.3_2e-1_ablation-mask_applied.jsonl' +=========== Perception =========== +Traceback (most recent call last): + File "/s3-code/ywang29/datasets/tinyllava/eval/MME/eval_tool/calculation.py", line 164, in + cal.process_result(results_dir) + File "/s3-code/ywang29/datasets/tinyllava/eval/MME/eval_tool/calculation.py", line 98, in process_result + lines = open(task_txt, 'r').readlines() +FileNotFoundError: [Errno 2] No such file or directory: 'answers/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.3_2e-1_connector-1.0_2.3_2e-1_ablation-mask_applied/existence.txt' +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 16:47:34,409] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +Traceback (most recent call last): + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/utils/hub.py", line 398, in cached_file + resolved_file = hf_hub_download( + File "/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/utils/_validators.py", line 106, in _inner_fn + validate_repo_id(arg_value) + File "/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/utils/_validators.py", line 154, in validate_repo_id + raise HFValidationError( +huggingface_hub.errors.HFValidationError: Repo id must be in the form 'repo_name' or 'namespace/repo_name': '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.3_2e-1_connector-1.0_2.3_2e-1_ablation/mask_applied'. Use `repo_type` argument if needed. + +The above exception was the direct cause of the following exception: + +Traceback (most recent call last): + File "/nfs/ywang29/TinyLLaVA/tinyllava/model/load_model.py", line 38, in load_pretrained_model + model = TinyLlavaForConditionalGeneration.from_pretrained(model_name_or_path,low_cpu_mem_usage=True) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/modeling_utils.py", line 3015, in from_pretrained + resolved_config_file = cached_file( + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/utils/hub.py", line 462, in cached_file + raise EnvironmentError( +OSError: Incorrect path_or_model_id: '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.3_2e-1_connector-1.0_2.3_2e-1_ablation/mask_applied'. Please provide either the path to a local folder or the repo_id of a model on the Hub. + +During handling of the above exception, another exception occurred: + +Traceback (most recent call last): + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/utils/hub.py", line 398, in cached_file + resolved_file = hf_hub_download( + File "/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/utils/_validators.py", line 106, in _inner_fn + validate_repo_id(arg_value) + File "/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/utils/_validators.py", line 154, in validate_repo_id + raise HFValidationError( +huggingface_hub.errors.HFValidationError: Repo id must be in the form 'repo_name' or 'namespace/repo_name': '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.3_2e-1_connector-1.0_2.3_2e-1_ablation/mask_applied'. Use `repo_type` argument if needed. + +The above exception was the direct cause of the following exception: + +Traceback (most recent call last): + File "/opt/conda/envs/tinyllava/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/opt/conda/envs/tinyllava/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/nfs/ywang29/TinyLLaVA/tinyllava/eval/model_vqa_science.py", line 108, in + eval_model(args) + File "/nfs/ywang29/TinyLLaVA/tinyllava/eval/model_vqa_science.py", line 31, in eval_model + model, tokenizer, image_processor, context_len = load_pretrained_model(model_path) + File "/nfs/ywang29/TinyLLaVA/tinyllava/model/load_model.py", line 40, in load_pretrained_model + model_config = TinyLlavaConfig.from_pretrained(model_name_or_path) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/configuration_utils.py", line 602, in from_pretrained + config_dict, kwargs = cls.get_config_dict(pretrained_model_name_or_path, **kwargs) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/configuration_utils.py", line 631, in get_config_dict + config_dict, kwargs = cls._get_config_dict(pretrained_model_name_or_path, **kwargs) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/configuration_utils.py", line 686, in _get_config_dict + resolved_config_file = cached_file( + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/utils/hub.py", line 462, in cached_file + raise EnvironmentError( +OSError: Incorrect path_or_model_id: '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.3_2e-1_connector-1.0_2.3_2e-1_ablation/mask_applied'. Please provide either the path to a local folder or the repo_id of a model on the Hub. +Traceback (most recent call last): + File "/nfs/ywang29/TinyLLaVA/tinyllava/eval/eval_science_qa.py", line 45, in + predictions = [json.loads(line) for line in open(args.result_file)] +FileNotFoundError: [Errno 2] No such file or directory: '/s3-code/ywang29/datasets/tinyllava/eval/scienceqa/answers/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.3_2e-1_connector-1.0_2.3_2e-1_ablation-mask_applied.jsonl' +==== EXPERIMENT COMPLETED: eval_qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.3_2e-1_connector-1.0_2.3_2e-1_ablation ==== +Log File: eval_qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.3_2e-1_connector-1.0_2.3_2e-1_ablation_20251012_164611.log +Timestamp: 2025-10-12 16:47:37 +===================================== diff --git a/logs_oct11/eval_qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.5_2e-1_connector-1.0_2.5_2e-1_ablation_20251012_164737.log b/logs_oct11/eval_qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.5_2e-1_connector-1.0_2.5_2e-1_ablation_20251012_164737.log new file mode 100644 index 0000000000000000000000000000000000000000..4d3f88fcfa60afff3ca19080865e9f7b3ef7d4cd --- /dev/null +++ b/logs_oct11/eval_qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.5_2e-1_connector-1.0_2.5_2e-1_ablation_20251012_164737.log @@ -0,0 +1,759 @@ +==== STARTING EXPERIMENT: eval_qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.5_2e-1_connector-1.0_2.5_2e-1_ablation ==== +Log File: eval_qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.5_2e-1_connector-1.0_2.5_2e-1_ablation_20251012_164737.log +Timestamp: 2025-10-12 16:47:37 +===================================== +Processing: /nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.5_2e-1_connector-1.0_2.5_2e-1_ablation +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 16:47:40,587] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +config_mask.torch_dtype: torch.bfloat16 +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +Load mask model from /nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.5_2e-1_connector-1.0_2.5_2e-1_ablation over. +TinyLlavaConfig { + "architectures": [ + "TinyLlavaForConditionalGeneration" + ], + "backward_type_connector": "normal", + "cache_dir": null, + "connector_type": "mlp2x_gelu", + "hidden_size": 896, + "ignore_index": -100, + "image_aspect_ratio": "square", + "image_token_index": -200, + "llm_model_name_or_path": "Qwen/Qwen2.5-0.5B", + "mask_model": [ + "llm", + "connector" + ], + "mask_type_connector": "soft", + "model_type": "tinyllava", + "num_queries": 128, + "num_resampler_layers": 3, + "pad_token": "<|endoftext|>", + "resampler_hidden_size": 768, + "sparsity_connector": null, + "subnet_type_connector": "global", + "temperature_connector": 2.5, + "text_config": { + "_name_or_path": "Qwen/Qwen2.5-0.5B", + "architectures": [ + "Qwen2ForCausalLM" + ], + "backward_type": "normal", + "bos_token_id": 151643, + "eos_token_id": 151643, + "hidden_size": 896, + "intermediate_size": 4864, + "mask_type": "soft", + "masked_layers": "all", + "max_position_embeddings": 32768, + "max_window_layers": 24, + "model_type": "qwen2", + "num_attention_heads": 14, + "num_hidden_layers": 24, + "num_key_value_heads": 2, + "rope_theta": 1000000.0, + "sliding_window": 32768, + "subnet_mode": "both", + "subnet_type": "None", + "temperature_attn": 2.5, + "temperature_mlp": 2.5, + "tie_word_embeddings": true, + "torch_dtype": "bfloat16", + "use_mrope": false, + "use_sliding_window": false, + "vocab_size": 151936 + }, + "threshold_connector": null, + "tokenizer_model_max_length": 2048, + "tokenizer_name_or_path": "Qwen/Qwen2.5-0.5B", + "tokenizer_padding_side": "right", + "tokenizer_use_fast": false, + "torch_dtype": "bfloat16", + "transformers_version": "4.40.1", + "tune_type_connector": "full", + "tune_type_llm": "full", + "tune_type_vision_tower": "frozen", + "tune_vision_tower_from_layer": 0, + "use_cache": true, + "vision_config": { + "hidden_act": "gelu_pytorch_tanh", + "hidden_size": 1152, + "image_size": 384, + "intermediate_size": 4304, + "layer_norm_eps": 1e-06, + "model_name_or_path": "google/siglip-so400m-patch14-384", + "model_name_or_path2": "", + "model_type": "siglip_vision_model", + "num_attention_heads": 16, + "num_hidden_layers": 27, + "patch_size": 14 + }, + "vision_feature_layer": -2, + "vision_feature_select_strategy": "patch", + "vision_hidden_size": 1152, + "vision_model_name_or_path": "google/siglip-so400m-patch14-384", + "vision_model_name_or_path2": "", + "vocab_size": 151936 +} + +TinyLlavaForConditionalGeneration( + (language_model): Qwen2ForCausalLM( + (model): Qwen2Model( + (embed_tokens): Embedding(151936, 896) + (layers): ModuleList( + (0-23): 24 x Qwen2DecoderLayer( + (self_attn): Qwen2Attention( + (q_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=True) + (k_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=128, bias=True) + (v_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=128, bias=True) + (o_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=False) + (rotary_emb): Qwen2RotaryEmbedding() + ) + (mlp): Qwen2MLP( + (gate_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=4864, bias=False) + (up_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=4864, bias=False) + (down_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=4864, out_features=896, bias=False) + (act_fn): SiLU() + ) + (input_layernorm): Qwen2RMSNorm() + (post_attention_layernorm): Qwen2RMSNorm() + ) + ) + (norm): Qwen2RMSNorm() + ) + (lm_head): Linear(in_features=896, out_features=151936, bias=False) + ) + (vision_tower): SIGLIPVisionTower( + (_vision_tower): SiglipVisionModel( + (vision_model): SiglipVisionTransformer( + (embeddings): SiglipVisionEmbeddings( + (patch_embedding): Conv2d(3, 1152, kernel_size=(14, 14), stride=(14, 14), padding=valid) + (position_embedding): Embedding(729, 1152) + ) + (encoder): SiglipEncoder( + (layers): ModuleList( + (0-26): 27 x SiglipEncoderLayer( + (self_attn): SiglipAttention( + (k_proj): Linear(in_features=1152, out_features=1152, bias=True) + (v_proj): Linear(in_features=1152, out_features=1152, bias=True) + (q_proj): Linear(in_features=1152, out_features=1152, bias=True) + (out_proj): Linear(in_features=1152, out_features=1152, bias=True) + ) + (layer_norm1): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + (layer_norm2): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + ) + ) + ) + (post_layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (head): SiglipMultiheadAttentionPoolingHead( + (attention): MultiheadAttention( + (out_proj): NonDynamicallyQuantizableLinear(in_features=1152, out_features=1152, bias=True) + ) + (layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + ) + ) + ) + ) + (connector): MLPConnector( + (_connector): Sequential( + (0): SupermaskLinearSparsity_SoftForward_Normal(in_features=1152, out_features=896, bias=True) + (1): GELU(approximate='none') + (2): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=True) + ) + ) +) +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Load base model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain over. +TinyLlavaConfig { + "cache_dir": null, + "connector_type": "mlp2x_gelu", + "hidden_size": 896, + "ignore_index": -100, + "image_aspect_ratio": "square", + "image_token_index": -200, + "llm_model_name_or_path": "Qwen/Qwen2.5-0.5B", + "model_type": "tinyllava", + "num_queries": 128, + "num_resampler_layers": 3, + "pad_token": "<|endoftext|>", + "pad_token_id": 151643, + "resampler_hidden_size": 768, + "text_config": { + "_name_or_path": "Qwen/Qwen2.5-0.5B", + "architectures": [ + "Qwen2ForCausalLM" + ], + "bos_token_id": 151643, + "eos_token_id": 151643, + "hidden_size": 896, + "intermediate_size": 4864, + "max_position_embeddings": 32768, + "max_window_layers": 24, + "model_type": "qwen2", + "num_attention_heads": 14, + "num_hidden_layers": 24, + "num_key_value_heads": 2, + "rope_theta": 1000000.0, + "sliding_window": 32768, + "tie_word_embeddings": true, + "use_mrope": false, + "use_sliding_window": false, + "vocab_size": 151936 + }, + "tokenizer_model_max_length": 2048, + "tokenizer_name_or_path": "Qwen/Qwen2.5-0.5B", + "tokenizer_padding_side": "right", + "tokenizer_use_fast": false, + "transformers_version": "4.40.1", + "tune_type_connector": "full", + "tune_type_llm": "frozen", + "tune_type_vision_tower": "frozen", + "tune_vision_tower_from_layer": 0, + "use_cache": true, + "vision_config": { + "hidden_act": "gelu_pytorch_tanh", + "hidden_size": 1152, + "image_size": 384, + "intermediate_size": 4304, + "layer_norm_eps": 1e-06, + "model_name_or_path": "google/siglip-so400m-patch14-384", + "model_name_or_path2": "", + "model_type": "siglip_vision_model", + "num_attention_heads": 16, + "num_hidden_layers": 27, + "patch_size": 14 + }, + "vision_feature_layer": -2, + "vision_feature_select_strategy": "patch", + "vision_hidden_size": 1152, + "vision_model_name_or_path": "google/siglip-so400m-patch14-384", + "vision_model_name_or_path2": "", + "vocab_size": 151936 +} + +TinyLlavaForConditionalGeneration( + (language_model): Qwen2ForCausalLM( + (model): Qwen2Model( + (embed_tokens): Embedding(151936, 896) + (layers): ModuleList( + (0-23): 24 x Qwen2DecoderLayer( + (self_attn): Qwen2Attention( + (q_proj): Linear(in_features=896, out_features=896, bias=True) + (k_proj): Linear(in_features=896, out_features=128, bias=True) + (v_proj): Linear(in_features=896, out_features=128, bias=True) + (o_proj): Linear(in_features=896, out_features=896, bias=False) + (rotary_emb): Qwen2RotaryEmbedding() + ) + (mlp): Qwen2MLP( + (gate_proj): Linear(in_features=896, out_features=4864, bias=False) + (up_proj): Linear(in_features=896, out_features=4864, bias=False) + (down_proj): Linear(in_features=4864, out_features=896, bias=False) + (act_fn): SiLU() + ) + (input_layernorm): Qwen2RMSNorm() + (post_attention_layernorm): Qwen2RMSNorm() + ) + ) + (norm): Qwen2RMSNorm() + ) + (lm_head): Linear(in_features=896, out_features=151936, bias=False) + ) + (vision_tower): SIGLIPVisionTower( + (_vision_tower): SiglipVisionModel( + (vision_model): SiglipVisionTransformer( + (embeddings): SiglipVisionEmbeddings( + (patch_embedding): Conv2d(3, 1152, kernel_size=(14, 14), stride=(14, 14), padding=valid) + (position_embedding): Embedding(729, 1152) + ) + (encoder): SiglipEncoder( + (layers): ModuleList( + (0-26): 27 x SiglipEncoderLayer( + (self_attn): SiglipAttention( + (k_proj): Linear(in_features=1152, out_features=1152, bias=True) + (v_proj): Linear(in_features=1152, out_features=1152, bias=True) + (q_proj): Linear(in_features=1152, out_features=1152, bias=True) + (out_proj): Linear(in_features=1152, out_features=1152, bias=True) + ) + (layer_norm1): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + (layer_norm2): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + ) + ) + ) + (post_layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (head): SiglipMultiheadAttentionPoolingHead( + (attention): MultiheadAttention( + (out_proj): NonDynamicallyQuantizableLinear(in_features=1152, out_features=1152, bias=True) + ) + (layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + ) + ) + ) + ) + (connector): MLPConnector( + (_connector): Sequential( + (0): Linear(in_features=1152, out_features=896, bias=True) + (1): GELU(approximate='none') + (2): Linear(in_features=896, out_features=896, bias=True) + ) + ) +) +Collect masks for language model over. +Collect masks for connector over. +Applying mask on model.layers.0.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.self_attn.q_proj. +Applying mask on model.layers.0.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.self_attn.k_proj. +Applying mask on model.layers.0.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.self_attn.v_proj. +Applying mask on model.layers.0.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.self_attn.o_proj. +Applying mask on model.layers.0.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.mlp.gate_proj. +Applying mask on model.layers.0.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.mlp.up_proj. +Applying mask on model.layers.0.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.mlp.down_proj. +Applying mask on model.layers.1.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.self_attn.q_proj. +Applying mask on model.layers.1.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.self_attn.k_proj. +Applying mask on model.layers.1.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.self_attn.v_proj. +Applying mask on model.layers.1.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.self_attn.o_proj. +Applying mask on model.layers.1.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.mlp.gate_proj. +Applying mask on model.layers.1.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.mlp.up_proj. +Applying mask on model.layers.1.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.mlp.down_proj. +Applying mask on model.layers.2.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.self_attn.q_proj. +Applying mask on model.layers.2.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.self_attn.k_proj. +Applying mask on model.layers.2.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.self_attn.v_proj. +Applying mask on model.layers.2.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.self_attn.o_proj. +Applying mask on model.layers.2.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.mlp.gate_proj. +Applying mask on model.layers.2.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.mlp.up_proj. +Applying mask on model.layers.2.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.mlp.down_proj. +Applying mask on model.layers.3.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.self_attn.q_proj. +Applying mask on model.layers.3.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.self_attn.k_proj. +Applying mask on model.layers.3.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.self_attn.v_proj. +Applying mask on model.layers.3.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.self_attn.o_proj. +Applying mask on model.layers.3.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.mlp.gate_proj. +Applying mask on model.layers.3.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.mlp.up_proj. +Applying mask on model.layers.3.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.mlp.down_proj. +Applying mask on model.layers.4.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.self_attn.q_proj. +Applying mask on model.layers.4.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.self_attn.k_proj. +Applying mask on model.layers.4.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.self_attn.v_proj. +Applying mask on model.layers.4.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.self_attn.o_proj. +Applying mask on model.layers.4.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.mlp.gate_proj. +Applying mask on model.layers.4.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.mlp.up_proj. +Applying mask on model.layers.4.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.mlp.down_proj. +Applying mask on model.layers.5.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.self_attn.q_proj. +Applying mask on model.layers.5.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.self_attn.k_proj. +Applying mask on model.layers.5.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.self_attn.v_proj. +Applying mask on model.layers.5.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.self_attn.o_proj. +Applying mask on model.layers.5.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.mlp.gate_proj. +Applying mask on model.layers.5.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.mlp.up_proj. +Applying mask on model.layers.5.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.mlp.down_proj. +Applying mask on model.layers.6.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.self_attn.q_proj. +Applying mask on model.layers.6.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.self_attn.k_proj. +Applying mask on model.layers.6.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.self_attn.v_proj. +Applying mask on model.layers.6.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.self_attn.o_proj. +Applying mask on model.layers.6.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.mlp.gate_proj. +Applying mask on model.layers.6.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.mlp.up_proj. +Applying mask on model.layers.6.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.mlp.down_proj. +Applying mask on model.layers.7.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.self_attn.q_proj. +Applying mask on model.layers.7.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.self_attn.k_proj. +Applying mask on model.layers.7.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.self_attn.v_proj. +Applying mask on model.layers.7.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.self_attn.o_proj. +Applying mask on model.layers.7.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.mlp.gate_proj. +Applying mask on model.layers.7.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.mlp.up_proj. +Applying mask on model.layers.7.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.mlp.down_proj. +Applying mask on model.layers.8.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.self_attn.q_proj. +Applying mask on model.layers.8.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.self_attn.k_proj. +Applying mask on model.layers.8.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.self_attn.v_proj. +Applying mask on model.layers.8.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.self_attn.o_proj. +Applying mask on model.layers.8.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.mlp.gate_proj. +Applying mask on model.layers.8.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.mlp.up_proj. +Applying mask on model.layers.8.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.mlp.down_proj. +Applying mask on model.layers.9.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.self_attn.q_proj. +Applying mask on model.layers.9.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.self_attn.k_proj. +Applying mask on model.layers.9.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.self_attn.v_proj. +Applying mask on model.layers.9.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.self_attn.o_proj. +Applying mask on model.layers.9.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.mlp.gate_proj. +Applying mask on model.layers.9.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.mlp.up_proj. +Applying mask on model.layers.9.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.mlp.down_proj. +Applying mask on model.layers.10.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.self_attn.q_proj. +Applying mask on model.layers.10.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.self_attn.k_proj. +Applying mask on model.layers.10.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.self_attn.v_proj. +Applying mask on model.layers.10.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.self_attn.o_proj. +Applying mask on model.layers.10.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.mlp.gate_proj. +Applying mask on model.layers.10.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.mlp.up_proj. +Applying mask on model.layers.10.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.mlp.down_proj. +Applying mask on model.layers.11.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.self_attn.q_proj. +Applying mask on model.layers.11.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.self_attn.k_proj. +Applying mask on model.layers.11.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.self_attn.v_proj. +Applying mask on model.layers.11.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.self_attn.o_proj. +Applying mask on model.layers.11.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.mlp.gate_proj. +Applying mask on model.layers.11.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.mlp.up_proj. +Applying mask on model.layers.11.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.mlp.down_proj. +Applying mask on model.layers.12.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.self_attn.q_proj. +Applying mask on model.layers.12.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.self_attn.k_proj. +Applying mask on model.layers.12.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.self_attn.v_proj. +Applying mask on model.layers.12.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.self_attn.o_proj. +Applying mask on model.layers.12.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.mlp.gate_proj. +Applying mask on model.layers.12.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.mlp.up_proj. +Applying mask on model.layers.12.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.mlp.down_proj. +Applying mask on model.layers.13.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.self_attn.q_proj. +Applying mask on model.layers.13.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.self_attn.k_proj. +Applying mask on model.layers.13.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.self_attn.v_proj. +Applying mask on model.layers.13.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.self_attn.o_proj. +Applying mask on model.layers.13.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.mlp.gate_proj. +Applying mask on model.layers.13.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.mlp.up_proj. +Applying mask on model.layers.13.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.mlp.down_proj. +Applying mask on model.layers.14.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.self_attn.q_proj. +Applying mask on model.layers.14.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.self_attn.k_proj. +Applying mask on model.layers.14.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.self_attn.v_proj. +Applying mask on model.layers.14.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.self_attn.o_proj. +Applying mask on model.layers.14.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.mlp.gate_proj. +Applying mask on model.layers.14.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.mlp.up_proj. +Applying mask on model.layers.14.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.mlp.down_proj. +Applying mask on model.layers.15.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.self_attn.q_proj. +Applying mask on model.layers.15.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.self_attn.k_proj. +Applying mask on model.layers.15.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.self_attn.v_proj. +Applying mask on model.layers.15.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.self_attn.o_proj. +Applying mask on model.layers.15.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.mlp.gate_proj. +Applying mask on model.layers.15.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.mlp.up_proj. +Applying mask on model.layers.15.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.mlp.down_proj. +Applying mask on model.layers.16.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.self_attn.q_proj. +Applying mask on model.layers.16.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.self_attn.k_proj. +Applying mask on model.layers.16.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.self_attn.v_proj. +Applying mask on model.layers.16.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.self_attn.o_proj. +Applying mask on model.layers.16.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.mlp.gate_proj. +Applying mask on model.layers.16.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.mlp.up_proj. +Applying mask on model.layers.16.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.mlp.down_proj. +Applying mask on model.layers.17.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.self_attn.q_proj. +Applying mask on model.layers.17.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.self_attn.k_proj. +Applying mask on model.layers.17.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.self_attn.v_proj. +Applying mask on model.layers.17.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.self_attn.o_proj. +Applying mask on model.layers.17.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.mlp.gate_proj. +Applying mask on model.layers.17.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.mlp.up_proj. +Applying mask on model.layers.17.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.mlp.down_proj. +Applying mask on model.layers.18.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.self_attn.q_proj. +Applying mask on model.layers.18.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.self_attn.k_proj. +Applying mask on model.layers.18.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.self_attn.v_proj. +Applying mask on model.layers.18.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.self_attn.o_proj. +Applying mask on model.layers.18.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.mlp.gate_proj. +Applying mask on model.layers.18.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.mlp.up_proj. +Applying mask on model.layers.18.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.mlp.down_proj. +Applying mask on model.layers.19.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.self_attn.q_proj. +Applying mask on model.layers.19.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.self_attn.k_proj. +Applying mask on model.layers.19.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.self_attn.v_proj. +Applying mask on model.layers.19.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.self_attn.o_proj. +Applying mask on model.layers.19.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.mlp.gate_proj. +Applying mask on model.layers.19.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.mlp.up_proj. +Applying mask on model.layers.19.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.mlp.down_proj. +Applying mask on model.layers.20.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.self_attn.q_proj. +Applying mask on model.layers.20.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.self_attn.k_proj. +Applying mask on model.layers.20.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.self_attn.v_proj. +Applying mask on model.layers.20.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.self_attn.o_proj. +Applying mask on model.layers.20.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.mlp.gate_proj. +Applying mask on model.layers.20.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.mlp.up_proj. +Applying mask on model.layers.20.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.mlp.down_proj. +Applying mask on model.layers.21.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.self_attn.q_proj. +Applying mask on model.layers.21.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.self_attn.k_proj. +Applying mask on model.layers.21.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.self_attn.v_proj. +Applying mask on model.layers.21.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.self_attn.o_proj. +Applying mask on model.layers.21.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.mlp.gate_proj. +Applying mask on model.layers.21.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.mlp.up_proj. +Applying mask on model.layers.21.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.mlp.down_proj. +Applying mask on model.layers.22.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.self_attn.q_proj. +Applying mask on model.layers.22.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.self_attn.k_proj. +Applying mask on model.layers.22.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.self_attn.v_proj. +Applying mask on model.layers.22.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.self_attn.o_proj. +Applying mask on model.layers.22.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.mlp.gate_proj. +Applying mask on model.layers.22.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.mlp.up_proj. +Applying mask on model.layers.22.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.mlp.down_proj. +Applying mask on model.layers.23.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.self_attn.q_proj. +Applying mask on model.layers.23.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.self_attn.k_proj. +Applying mask on model.layers.23.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.self_attn.v_proj. +Applying mask on model.layers.23.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.self_attn.o_proj. +Applying mask on model.layers.23.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.mlp.gate_proj. +Applying mask on model.layers.23.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.mlp.up_proj. +Applying mask on model.layers.23.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.mlp.down_proj. +Applying mask on _connector.0 with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on _connector.0. +Applying mask on _connector.2 with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on _connector.2. +Using cleaned config_mask (without mask parameters) for saving. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 16:48:22,480] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. + 0%| | 0/900 [00:00", + "resampler_hidden_size": 768, + "sparsity_connector": null, + "subnet_type_connector": "global", + "temperature_connector": 2.7, + "text_config": { + "_name_or_path": "Qwen/Qwen2.5-0.5B", + "architectures": [ + "Qwen2ForCausalLM" + ], + "backward_type": "normal", + "bos_token_id": 151643, + "eos_token_id": 151643, + "hidden_size": 896, + "intermediate_size": 4864, + "mask_type": "soft", + "masked_layers": "all", + "max_position_embeddings": 32768, + "max_window_layers": 24, + "model_type": "qwen2", + "num_attention_heads": 14, + "num_hidden_layers": 24, + "num_key_value_heads": 2, + "rope_theta": 1000000.0, + "sliding_window": 32768, + "subnet_mode": "both", + "subnet_type": "None", + "temperature_attn": 2.7, + "temperature_mlp": 2.7, + "tie_word_embeddings": true, + "torch_dtype": "bfloat16", + "use_mrope": false, + "use_sliding_window": false, + "vocab_size": 151936 + }, + "threshold_connector": null, + "tokenizer_model_max_length": 2048, + "tokenizer_name_or_path": "Qwen/Qwen2.5-0.5B", + "tokenizer_padding_side": "right", + "tokenizer_use_fast": false, + "torch_dtype": "bfloat16", + "transformers_version": "4.40.1", + "tune_type_connector": "full", + "tune_type_llm": "full", + "tune_type_vision_tower": "frozen", + "tune_vision_tower_from_layer": 0, + "use_cache": true, + "vision_config": { + "hidden_act": "gelu_pytorch_tanh", + "hidden_size": 1152, + "image_size": 384, + "intermediate_size": 4304, + "layer_norm_eps": 1e-06, + "model_name_or_path": "google/siglip-so400m-patch14-384", + "model_name_or_path2": "", + "model_type": "siglip_vision_model", + "num_attention_heads": 16, + "num_hidden_layers": 27, + "patch_size": 14 + }, + "vision_feature_layer": -2, + "vision_feature_select_strategy": "patch", + "vision_hidden_size": 1152, + "vision_model_name_or_path": "google/siglip-so400m-patch14-384", + "vision_model_name_or_path2": "", + "vocab_size": 151936 +} + +TinyLlavaForConditionalGeneration( + (language_model): Qwen2ForCausalLM( + (model): Qwen2Model( + (embed_tokens): Embedding(151936, 896) + (layers): ModuleList( + (0-23): 24 x Qwen2DecoderLayer( + (self_attn): Qwen2Attention( + (q_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=True) + (k_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=128, bias=True) + (v_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=128, bias=True) + (o_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=False) + (rotary_emb): Qwen2RotaryEmbedding() + ) + (mlp): Qwen2MLP( + (gate_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=4864, bias=False) + (up_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=4864, bias=False) + (down_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=4864, out_features=896, bias=False) + (act_fn): SiLU() + ) + (input_layernorm): Qwen2RMSNorm() + (post_attention_layernorm): Qwen2RMSNorm() + ) + ) + (norm): Qwen2RMSNorm() + ) + (lm_head): Linear(in_features=896, out_features=151936, bias=False) + ) + (vision_tower): SIGLIPVisionTower( + (_vision_tower): SiglipVisionModel( + (vision_model): SiglipVisionTransformer( + (embeddings): SiglipVisionEmbeddings( + (patch_embedding): Conv2d(3, 1152, kernel_size=(14, 14), stride=(14, 14), padding=valid) + (position_embedding): Embedding(729, 1152) + ) + (encoder): SiglipEncoder( + (layers): ModuleList( + (0-26): 27 x SiglipEncoderLayer( + (self_attn): SiglipAttention( + (k_proj): Linear(in_features=1152, out_features=1152, bias=True) + (v_proj): Linear(in_features=1152, out_features=1152, bias=True) + (q_proj): Linear(in_features=1152, out_features=1152, bias=True) + (out_proj): Linear(in_features=1152, out_features=1152, bias=True) + ) + (layer_norm1): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + (layer_norm2): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + ) + ) + ) + (post_layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (head): SiglipMultiheadAttentionPoolingHead( + (attention): MultiheadAttention( + (out_proj): NonDynamicallyQuantizableLinear(in_features=1152, out_features=1152, bias=True) + ) + (layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + ) + ) + ) + ) + (connector): MLPConnector( + (_connector): Sequential( + (0): SupermaskLinearSparsity_SoftForward_Normal(in_features=1152, out_features=896, bias=True) + (1): GELU(approximate='none') + (2): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=True) + ) + ) +) +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Load base model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain over. +TinyLlavaConfig { + "cache_dir": null, + "connector_type": "mlp2x_gelu", + "hidden_size": 896, + "ignore_index": -100, + "image_aspect_ratio": "square", + "image_token_index": -200, + "llm_model_name_or_path": "Qwen/Qwen2.5-0.5B", + "model_type": "tinyllava", + "num_queries": 128, + "num_resampler_layers": 3, + "pad_token": "<|endoftext|>", + "pad_token_id": 151643, + "resampler_hidden_size": 768, + "text_config": { + "_name_or_path": "Qwen/Qwen2.5-0.5B", + "architectures": [ + "Qwen2ForCausalLM" + ], + "bos_token_id": 151643, + "eos_token_id": 151643, + "hidden_size": 896, + "intermediate_size": 4864, + "max_position_embeddings": 32768, + "max_window_layers": 24, + "model_type": "qwen2", + "num_attention_heads": 14, + "num_hidden_layers": 24, + "num_key_value_heads": 2, + "rope_theta": 1000000.0, + "sliding_window": 32768, + "tie_word_embeddings": true, + "use_mrope": false, + "use_sliding_window": false, + "vocab_size": 151936 + }, + "tokenizer_model_max_length": 2048, + "tokenizer_name_or_path": "Qwen/Qwen2.5-0.5B", + "tokenizer_padding_side": "right", + "tokenizer_use_fast": false, + "transformers_version": "4.40.1", + "tune_type_connector": "full", + "tune_type_llm": "frozen", + "tune_type_vision_tower": "frozen", + "tune_vision_tower_from_layer": 0, + "use_cache": true, + "vision_config": { + "hidden_act": "gelu_pytorch_tanh", + "hidden_size": 1152, + "image_size": 384, + "intermediate_size": 4304, + "layer_norm_eps": 1e-06, + "model_name_or_path": "google/siglip-so400m-patch14-384", + "model_name_or_path2": "", + "model_type": "siglip_vision_model", + "num_attention_heads": 16, + "num_hidden_layers": 27, + "patch_size": 14 + }, + "vision_feature_layer": -2, + "vision_feature_select_strategy": "patch", + "vision_hidden_size": 1152, + "vision_model_name_or_path": "google/siglip-so400m-patch14-384", + "vision_model_name_or_path2": "", + "vocab_size": 151936 +} + +TinyLlavaForConditionalGeneration( + (language_model): Qwen2ForCausalLM( + (model): Qwen2Model( + (embed_tokens): Embedding(151936, 896) + (layers): ModuleList( + (0-23): 24 x Qwen2DecoderLayer( + (self_attn): Qwen2Attention( + (q_proj): Linear(in_features=896, out_features=896, bias=True) + (k_proj): Linear(in_features=896, out_features=128, bias=True) + (v_proj): Linear(in_features=896, out_features=128, bias=True) + (o_proj): Linear(in_features=896, out_features=896, bias=False) + (rotary_emb): Qwen2RotaryEmbedding() + ) + (mlp): Qwen2MLP( + (gate_proj): Linear(in_features=896, out_features=4864, bias=False) + (up_proj): Linear(in_features=896, out_features=4864, bias=False) + (down_proj): Linear(in_features=4864, out_features=896, bias=False) + (act_fn): SiLU() + ) + (input_layernorm): Qwen2RMSNorm() + (post_attention_layernorm): Qwen2RMSNorm() + ) + ) + (norm): Qwen2RMSNorm() + ) + (lm_head): Linear(in_features=896, out_features=151936, bias=False) + ) + (vision_tower): SIGLIPVisionTower( + (_vision_tower): SiglipVisionModel( + (vision_model): SiglipVisionTransformer( + (embeddings): SiglipVisionEmbeddings( + (patch_embedding): Conv2d(3, 1152, kernel_size=(14, 14), stride=(14, 14), padding=valid) + (position_embedding): Embedding(729, 1152) + ) + (encoder): SiglipEncoder( + (layers): ModuleList( + (0-26): 27 x SiglipEncoderLayer( + (self_attn): SiglipAttention( + (k_proj): Linear(in_features=1152, out_features=1152, bias=True) + (v_proj): Linear(in_features=1152, out_features=1152, bias=True) + (q_proj): Linear(in_features=1152, out_features=1152, bias=True) + (out_proj): Linear(in_features=1152, out_features=1152, bias=True) + ) + (layer_norm1): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + (layer_norm2): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + ) + ) + ) + (post_layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (head): SiglipMultiheadAttentionPoolingHead( + (attention): MultiheadAttention( + (out_proj): NonDynamicallyQuantizableLinear(in_features=1152, out_features=1152, bias=True) + ) + (layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + ) + ) + ) + ) + (connector): MLPConnector( + (_connector): Sequential( + (0): Linear(in_features=1152, out_features=896, bias=True) + (1): GELU(approximate='none') + (2): Linear(in_features=896, out_features=896, bias=True) + ) + ) +) +Collect masks for language model over. +Collect masks for connector over. +Applying mask on model.layers.0.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.self_attn.q_proj. +Applying mask on model.layers.0.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.self_attn.k_proj. +Applying mask on model.layers.0.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.self_attn.v_proj. +Applying mask on model.layers.0.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.self_attn.o_proj. +Applying mask on model.layers.0.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.mlp.gate_proj. +Applying mask on model.layers.0.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.mlp.up_proj. +Applying mask on model.layers.0.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.mlp.down_proj. +Applying mask on model.layers.1.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.self_attn.q_proj. +Applying mask on model.layers.1.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.self_attn.k_proj. +Applying mask on model.layers.1.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.self_attn.v_proj. +Applying mask on model.layers.1.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.self_attn.o_proj. +Applying mask on model.layers.1.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.mlp.gate_proj. +Applying mask on model.layers.1.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.mlp.up_proj. +Applying mask on model.layers.1.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.mlp.down_proj. +Applying mask on model.layers.2.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.self_attn.q_proj. +Applying mask on model.layers.2.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.self_attn.k_proj. +Applying mask on model.layers.2.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.self_attn.v_proj. +Applying mask on model.layers.2.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.self_attn.o_proj. +Applying mask on model.layers.2.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.mlp.gate_proj. +Applying mask on model.layers.2.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.mlp.up_proj. +Applying mask on model.layers.2.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.mlp.down_proj. +Applying mask on model.layers.3.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.self_attn.q_proj. +Applying mask on model.layers.3.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.self_attn.k_proj. +Applying mask on model.layers.3.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.self_attn.v_proj. +Applying mask on model.layers.3.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.self_attn.o_proj. +Applying mask on model.layers.3.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.mlp.gate_proj. +Applying mask on model.layers.3.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.mlp.up_proj. +Applying mask on model.layers.3.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.mlp.down_proj. +Applying mask on model.layers.4.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.self_attn.q_proj. +Applying mask on model.layers.4.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.self_attn.k_proj. +Applying mask on model.layers.4.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.self_attn.v_proj. +Applying mask on model.layers.4.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.self_attn.o_proj. +Applying mask on model.layers.4.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.mlp.gate_proj. +Applying mask on model.layers.4.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.mlp.up_proj. +Applying mask on model.layers.4.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.mlp.down_proj. +Applying mask on model.layers.5.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.self_attn.q_proj. +Applying mask on model.layers.5.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.self_attn.k_proj. +Applying mask on model.layers.5.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.self_attn.v_proj. +Applying mask on model.layers.5.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.self_attn.o_proj. +Applying mask on model.layers.5.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.mlp.gate_proj. +Applying mask on model.layers.5.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.mlp.up_proj. +Applying mask on model.layers.5.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.mlp.down_proj. +Applying mask on model.layers.6.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.self_attn.q_proj. +Applying mask on model.layers.6.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.self_attn.k_proj. +Applying mask on model.layers.6.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.self_attn.v_proj. +Applying mask on model.layers.6.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.self_attn.o_proj. +Applying mask on model.layers.6.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.mlp.gate_proj. +Applying mask on model.layers.6.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.mlp.up_proj. +Applying mask on model.layers.6.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.mlp.down_proj. +Applying mask on model.layers.7.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.self_attn.q_proj. +Applying mask on model.layers.7.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.self_attn.k_proj. +Applying mask on model.layers.7.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.self_attn.v_proj. +Applying mask on model.layers.7.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.self_attn.o_proj. +Applying mask on model.layers.7.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.mlp.gate_proj. +Applying mask on model.layers.7.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.mlp.up_proj. +Applying mask on model.layers.7.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.mlp.down_proj. +Applying mask on model.layers.8.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.self_attn.q_proj. +Applying mask on model.layers.8.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.self_attn.k_proj. +Applying mask on model.layers.8.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.self_attn.v_proj. +Applying mask on model.layers.8.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.self_attn.o_proj. +Applying mask on model.layers.8.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.mlp.gate_proj. +Applying mask on model.layers.8.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.mlp.up_proj. +Applying mask on model.layers.8.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.mlp.down_proj. +Applying mask on model.layers.9.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.self_attn.q_proj. +Applying mask on model.layers.9.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.self_attn.k_proj. +Applying mask on model.layers.9.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.self_attn.v_proj. +Applying mask on model.layers.9.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.self_attn.o_proj. +Applying mask on model.layers.9.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.mlp.gate_proj. +Applying mask on model.layers.9.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.mlp.up_proj. +Applying mask on model.layers.9.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.mlp.down_proj. +Applying mask on model.layers.10.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.self_attn.q_proj. +Applying mask on model.layers.10.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.self_attn.k_proj. +Applying mask on model.layers.10.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.self_attn.v_proj. +Applying mask on model.layers.10.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.self_attn.o_proj. +Applying mask on model.layers.10.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.mlp.gate_proj. +Applying mask on model.layers.10.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.mlp.up_proj. +Applying mask on model.layers.10.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.mlp.down_proj. +Applying mask on model.layers.11.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.self_attn.q_proj. +Applying mask on model.layers.11.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.self_attn.k_proj. +Applying mask on model.layers.11.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.self_attn.v_proj. +Applying mask on model.layers.11.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.self_attn.o_proj. +Applying mask on model.layers.11.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.mlp.gate_proj. +Applying mask on model.layers.11.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.mlp.up_proj. +Applying mask on model.layers.11.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.mlp.down_proj. +Applying mask on model.layers.12.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.self_attn.q_proj. +Applying mask on model.layers.12.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.self_attn.k_proj. +Applying mask on model.layers.12.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.self_attn.v_proj. +Applying mask on model.layers.12.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.self_attn.o_proj. +Applying mask on model.layers.12.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.mlp.gate_proj. +Applying mask on model.layers.12.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.mlp.up_proj. +Applying mask on model.layers.12.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.mlp.down_proj. +Applying mask on model.layers.13.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.self_attn.q_proj. +Applying mask on model.layers.13.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.self_attn.k_proj. +Applying mask on model.layers.13.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.self_attn.v_proj. +Applying mask on model.layers.13.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.self_attn.o_proj. +Applying mask on model.layers.13.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.mlp.gate_proj. +Applying mask on model.layers.13.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.mlp.up_proj. +Applying mask on model.layers.13.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.mlp.down_proj. +Applying mask on model.layers.14.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.self_attn.q_proj. +Applying mask on model.layers.14.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.self_attn.k_proj. +Applying mask on model.layers.14.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.self_attn.v_proj. +Applying mask on model.layers.14.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.self_attn.o_proj. +Applying mask on model.layers.14.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.mlp.gate_proj. +Applying mask on model.layers.14.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.mlp.up_proj. +Applying mask on model.layers.14.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.mlp.down_proj. +Applying mask on model.layers.15.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.self_attn.q_proj. +Applying mask on model.layers.15.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.self_attn.k_proj. +Applying mask on model.layers.15.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.self_attn.v_proj. +Applying mask on model.layers.15.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.self_attn.o_proj. +Applying mask on model.layers.15.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.mlp.gate_proj. +Applying mask on model.layers.15.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.mlp.up_proj. +Applying mask on model.layers.15.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.mlp.down_proj. +Applying mask on model.layers.16.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.self_attn.q_proj. +Applying mask on model.layers.16.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.self_attn.k_proj. +Applying mask on model.layers.16.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.self_attn.v_proj. +Applying mask on model.layers.16.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.self_attn.o_proj. +Applying mask on model.layers.16.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.mlp.gate_proj. +Applying mask on model.layers.16.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.mlp.up_proj. +Applying mask on model.layers.16.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.mlp.down_proj. +Applying mask on model.layers.17.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.self_attn.q_proj. +Applying mask on model.layers.17.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.self_attn.k_proj. +Applying mask on model.layers.17.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.self_attn.v_proj. +Applying mask on model.layers.17.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.self_attn.o_proj. +Applying mask on model.layers.17.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.mlp.gate_proj. +Applying mask on model.layers.17.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.mlp.up_proj. +Applying mask on model.layers.17.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.mlp.down_proj. +Applying mask on model.layers.18.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.self_attn.q_proj. +Applying mask on model.layers.18.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.self_attn.k_proj. +Applying mask on model.layers.18.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.self_attn.v_proj. +Applying mask on model.layers.18.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.self_attn.o_proj. +Applying mask on model.layers.18.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.mlp.gate_proj. +Applying mask on model.layers.18.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.mlp.up_proj. +Applying mask on model.layers.18.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.mlp.down_proj. +Applying mask on model.layers.19.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.self_attn.q_proj. +Applying mask on model.layers.19.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.self_attn.k_proj. +Applying mask on model.layers.19.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.self_attn.v_proj. +Applying mask on model.layers.19.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.self_attn.o_proj. +Applying mask on model.layers.19.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.mlp.gate_proj. +Applying mask on model.layers.19.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.mlp.up_proj. +Applying mask on model.layers.19.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.mlp.down_proj. +Applying mask on model.layers.20.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.self_attn.q_proj. +Applying mask on model.layers.20.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.self_attn.k_proj. +Applying mask on model.layers.20.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.self_attn.v_proj. +Applying mask on model.layers.20.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.self_attn.o_proj. +Applying mask on model.layers.20.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.mlp.gate_proj. +Applying mask on model.layers.20.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.mlp.up_proj. +Applying mask on model.layers.20.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.mlp.down_proj. +Applying mask on model.layers.21.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.self_attn.q_proj. +Applying mask on model.layers.21.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.self_attn.k_proj. +Applying mask on model.layers.21.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.self_attn.v_proj. +Applying mask on model.layers.21.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.self_attn.o_proj. +Applying mask on model.layers.21.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.mlp.gate_proj. +Applying mask on model.layers.21.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.mlp.up_proj. +Applying mask on model.layers.21.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.mlp.down_proj. +Applying mask on model.layers.22.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.self_attn.q_proj. +Applying mask on model.layers.22.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.self_attn.k_proj. +Applying mask on model.layers.22.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.self_attn.v_proj. +Applying mask on model.layers.22.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.self_attn.o_proj. +Applying mask on model.layers.22.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.mlp.gate_proj. +Applying mask on model.layers.22.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.mlp.up_proj. +Applying mask on model.layers.22.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.mlp.down_proj. +Applying mask on model.layers.23.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.self_attn.q_proj. +Applying mask on model.layers.23.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.self_attn.k_proj. +Applying mask on model.layers.23.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.self_attn.v_proj. +Applying mask on model.layers.23.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.self_attn.o_proj. +Applying mask on model.layers.23.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.mlp.gate_proj. +Applying mask on model.layers.23.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.mlp.up_proj. +Applying mask on model.layers.23.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.mlp.down_proj. +Applying mask on _connector.0 with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on _connector.0. +Applying mask on _connector.2 with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on _connector.2. +Using cleaned config_mask (without mask parameters) for saving. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 17:13:23,519] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. + 0%| | 0/900 [00:00 + main() + File "/nfs/ywang29/TinyLLaVA/scripts/apply_masks.py", line 123, in main + config_mask = TinyLlavaConfig.from_pretrained(model_args.mask_model_name_or_path) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/configuration_utils.py", line 602, in from_pretrained + config_dict, kwargs = cls.get_config_dict(pretrained_model_name_or_path, **kwargs) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/configuration_utils.py", line 631, in get_config_dict + config_dict, kwargs = cls._get_config_dict(pretrained_model_name_or_path, **kwargs) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/configuration_utils.py", line 686, in _get_config_dict + resolved_config_file = cached_file( + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/utils/hub.py", line 369, in cached_file + raise EnvironmentError( +OSError: /nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.9_2e-1_connector-1.0_2.9_2e-1_ablation does not appear to have a file named config.json. Checkout 'https://huggingface.co//nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.9_2e-1_connector-1.0_2.9_2e-1_ablation/tree/main' for available files. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 17:40:08,208] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +Traceback (most recent call last): + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/utils/hub.py", line 398, in cached_file + resolved_file = hf_hub_download( + File "/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/utils/_validators.py", line 106, in _inner_fn + validate_repo_id(arg_value) + File "/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/utils/_validators.py", line 154, in validate_repo_id + raise HFValidationError( +huggingface_hub.errors.HFValidationError: Repo id must be in the form 'repo_name' or 'namespace/repo_name': '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.9_2e-1_connector-1.0_2.9_2e-1_ablation/mask_applied'. Use `repo_type` argument if needed. + +The above exception was the direct cause of the following exception: + +Traceback (most recent call last): + File "/nfs/ywang29/TinyLLaVA/tinyllava/model/load_model.py", line 38, in load_pretrained_model + model = TinyLlavaForConditionalGeneration.from_pretrained(model_name_or_path,low_cpu_mem_usage=True) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/modeling_utils.py", line 3015, in from_pretrained + resolved_config_file = cached_file( + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/utils/hub.py", line 462, in cached_file + raise EnvironmentError( +OSError: Incorrect path_or_model_id: '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.9_2e-1_connector-1.0_2.9_2e-1_ablation/mask_applied'. Please provide either the path to a local folder or the repo_id of a model on the Hub. + +During handling of the above exception, another exception occurred: + +Traceback (most recent call last): + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/utils/hub.py", line 398, in cached_file + resolved_file = hf_hub_download( + File "/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/utils/_validators.py", line 106, in _inner_fn + validate_repo_id(arg_value) + File "/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/utils/_validators.py", line 154, in validate_repo_id + raise HFValidationError( +huggingface_hub.errors.HFValidationError: Repo id must be in the form 'repo_name' or 'namespace/repo_name': '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.9_2e-1_connector-1.0_2.9_2e-1_ablation/mask_applied'. Use `repo_type` argument if needed. + +The above exception was the direct cause of the following exception: + +Traceback (most recent call last): + File "/opt/conda/envs/tinyllava/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/opt/conda/envs/tinyllava/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/nfs/ywang29/TinyLLaVA/tinyllava/eval/model_vqa_mmmu.py", line 180, in + eval_model(args) + File "/nfs/ywang29/TinyLLaVA/tinyllava/eval/model_vqa_mmmu.py", line 88, in eval_model + model, tokenizer, image_processor, context_len = load_pretrained_model(model_path) + File "/nfs/ywang29/TinyLLaVA/tinyllava/model/load_model.py", line 40, in load_pretrained_model + model_config = TinyLlavaConfig.from_pretrained(model_name_or_path) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/configuration_utils.py", line 602, in from_pretrained + config_dict, kwargs = cls.get_config_dict(pretrained_model_name_or_path, **kwargs) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/configuration_utils.py", line 631, in get_config_dict + config_dict, kwargs = cls._get_config_dict(pretrained_model_name_or_path, **kwargs) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/configuration_utils.py", line 686, in _get_config_dict + resolved_config_file = cached_file( + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/utils/hub.py", line 462, in cached_file + raise EnvironmentError( +OSError: Incorrect path_or_model_id: '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.9_2e-1_connector-1.0_2.9_2e-1_ablation/mask_applied'. Please provide either the path to a local folder or the repo_id of a model on the Hub. +Traceback (most recent call last): + File "/nfs/ywang29/TinyLLaVA/scripts/convert_answer_to_mmmu.py", line 31, in + eval_model(args) + File "/nfs/ywang29/TinyLLaVA/scripts/convert_answer_to_mmmu.py", line 7, in eval_model + answers = [json.loads(q) for q in open(os.path.expanduser(args.answers_file), "r")] +FileNotFoundError: [Errno 2] No such file or directory: '/s3-code/ywang29/datasets/tinyllava/eval/MMMU/answers/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.9_2e-1_connector-1.0_2.9_2e-1_ablation-mask_applied.jsonl' +Traceback (most recent call last): + File "/s3-code/ywang29/datasets/tinyllava/eval/MMMU/eval/main_eval_only.py", line 19, in + output_dict = json.load(open(args.output_path)) +FileNotFoundError: [Errno 2] No such file or directory: '/s3-code/ywang29/datasets/tinyllava/eval/MMMU/answers/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.9_2e-1_connector-1.0_2.9_2e-1_ablation-mask_applied_output.json' +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 17:40:14,451] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +Traceback (most recent call last): + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/utils/hub.py", line 398, in cached_file + resolved_file = hf_hub_download( + File "/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/utils/_validators.py", line 106, in _inner_fn + validate_repo_id(arg_value) + File "/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/utils/_validators.py", line 154, in validate_repo_id + raise HFValidationError( +huggingface_hub.errors.HFValidationError: Repo id must be in the form 'repo_name' or 'namespace/repo_name': '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.9_2e-1_connector-1.0_2.9_2e-1_ablation/mask_applied'. Use `repo_type` argument if needed. + +The above exception was the direct cause of the following exception: + +Traceback (most recent call last): + File "/nfs/ywang29/TinyLLaVA/tinyllava/model/load_model.py", line 38, in load_pretrained_model + model = TinyLlavaForConditionalGeneration.from_pretrained(model_name_or_path,low_cpu_mem_usage=True) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/modeling_utils.py", line 3015, in from_pretrained + resolved_config_file = cached_file( + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/utils/hub.py", line 462, in cached_file + raise EnvironmentError( +OSError: Incorrect path_or_model_id: '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.9_2e-1_connector-1.0_2.9_2e-1_ablation/mask_applied'. Please provide either the path to a local folder or the repo_id of a model on the Hub. + +During handling of the above exception, another exception occurred: + +Traceback (most recent call last): + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/utils/hub.py", line 398, in cached_file + resolved_file = hf_hub_download( + File "/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/utils/_validators.py", line 106, in _inner_fn + validate_repo_id(arg_value) + File "/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/utils/_validators.py", line 154, in validate_repo_id + raise HFValidationError( +huggingface_hub.errors.HFValidationError: Repo id must be in the form 'repo_name' or 'namespace/repo_name': '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.9_2e-1_connector-1.0_2.9_2e-1_ablation/mask_applied'. Use `repo_type` argument if needed. + +The above exception was the direct cause of the following exception: + +Traceback (most recent call last): + File "/opt/conda/envs/tinyllava/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/opt/conda/envs/tinyllava/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/nfs/ywang29/TinyLLaVA/tinyllava/eval/model_vqa_loader.py", line 146, in + eval_model(args) + File "/nfs/ywang29/TinyLLaVA/tinyllava/eval/model_vqa_loader.py", line 79, in eval_model + model, tokenizer, image_processor, context_len = load_pretrained_model(model_path) + File "/nfs/ywang29/TinyLLaVA/tinyllava/model/load_model.py", line 40, in load_pretrained_model + model_config = TinyLlavaConfig.from_pretrained(model_name_or_path) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/configuration_utils.py", line 602, in from_pretrained + config_dict, kwargs = cls.get_config_dict(pretrained_model_name_or_path, **kwargs) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/configuration_utils.py", line 631, in get_config_dict + config_dict, kwargs = cls._get_config_dict(pretrained_model_name_or_path, **kwargs) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/configuration_utils.py", line 686, in _get_config_dict + resolved_config_file = cached_file( + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/utils/hub.py", line 462, in cached_file + raise EnvironmentError( +OSError: Incorrect path_or_model_id: '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.9_2e-1_connector-1.0_2.9_2e-1_ablation/mask_applied'. Please provide either the path to a local folder or the repo_id of a model on the Hub. +Traceback (most recent call last): + File "/s3-code/ywang29/datasets/tinyllava/eval/MME/convert_answer_to_mme.py", line 52, in + answers = [json.loads(line) for line in open(os.path.join('answers', f'{experiment}.jsonl'))] +FileNotFoundError: [Errno 2] No such file or directory: 'answers/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.9_2e-1_connector-1.0_2.9_2e-1_ablation-mask_applied.jsonl' +=========== Perception =========== +Traceback (most recent call last): + File "/s3-code/ywang29/datasets/tinyllava/eval/MME/eval_tool/calculation.py", line 164, in + cal.process_result(results_dir) + File "/s3-code/ywang29/datasets/tinyllava/eval/MME/eval_tool/calculation.py", line 98, in process_result + lines = open(task_txt, 'r').readlines() +FileNotFoundError: [Errno 2] No such file or directory: 'answers/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.9_2e-1_connector-1.0_2.9_2e-1_ablation-mask_applied/existence.txt' +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 17:41:19,450] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +Traceback (most recent call last): + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/utils/hub.py", line 398, in cached_file + resolved_file = hf_hub_download( + File "/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/utils/_validators.py", line 106, in _inner_fn + validate_repo_id(arg_value) + File "/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/utils/_validators.py", line 154, in validate_repo_id + raise HFValidationError( +huggingface_hub.errors.HFValidationError: Repo id must be in the form 'repo_name' or 'namespace/repo_name': '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.9_2e-1_connector-1.0_2.9_2e-1_ablation/mask_applied'. Use `repo_type` argument if needed. + +The above exception was the direct cause of the following exception: + +Traceback (most recent call last): + File "/nfs/ywang29/TinyLLaVA/tinyllava/model/load_model.py", line 38, in load_pretrained_model + model = TinyLlavaForConditionalGeneration.from_pretrained(model_name_or_path,low_cpu_mem_usage=True) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/modeling_utils.py", line 3015, in from_pretrained + resolved_config_file = cached_file( + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/utils/hub.py", line 462, in cached_file + raise EnvironmentError( +OSError: Incorrect path_or_model_id: '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.9_2e-1_connector-1.0_2.9_2e-1_ablation/mask_applied'. Please provide either the path to a local folder or the repo_id of a model on the Hub. + +During handling of the above exception, another exception occurred: + +Traceback (most recent call last): + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/utils/hub.py", line 398, in cached_file + resolved_file = hf_hub_download( + File "/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/utils/_validators.py", line 106, in _inner_fn + validate_repo_id(arg_value) + File "/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/utils/_validators.py", line 154, in validate_repo_id + raise HFValidationError( +huggingface_hub.errors.HFValidationError: Repo id must be in the form 'repo_name' or 'namespace/repo_name': '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.9_2e-1_connector-1.0_2.9_2e-1_ablation/mask_applied'. Use `repo_type` argument if needed. + +The above exception was the direct cause of the following exception: + +Traceback (most recent call last): + File "/opt/conda/envs/tinyllava/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/opt/conda/envs/tinyllava/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/nfs/ywang29/TinyLLaVA/tinyllava/eval/model_vqa_science.py", line 108, in + eval_model(args) + File "/nfs/ywang29/TinyLLaVA/tinyllava/eval/model_vqa_science.py", line 31, in eval_model + model, tokenizer, image_processor, context_len = load_pretrained_model(model_path) + File "/nfs/ywang29/TinyLLaVA/tinyllava/model/load_model.py", line 40, in load_pretrained_model + model_config = TinyLlavaConfig.from_pretrained(model_name_or_path) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/configuration_utils.py", line 602, in from_pretrained + config_dict, kwargs = cls.get_config_dict(pretrained_model_name_or_path, **kwargs) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/configuration_utils.py", line 631, in get_config_dict + config_dict, kwargs = cls._get_config_dict(pretrained_model_name_or_path, **kwargs) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/configuration_utils.py", line 686, in _get_config_dict + resolved_config_file = cached_file( + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/utils/hub.py", line 462, in cached_file + raise EnvironmentError( +OSError: Incorrect path_or_model_id: '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.9_2e-1_connector-1.0_2.9_2e-1_ablation/mask_applied'. Please provide either the path to a local folder or the repo_id of a model on the Hub. +Traceback (most recent call last): + File "/nfs/ywang29/TinyLLaVA/tinyllava/eval/eval_science_qa.py", line 45, in + predictions = [json.loads(line) for line in open(args.result_file)] +FileNotFoundError: [Errno 2] No such file or directory: '/s3-code/ywang29/datasets/tinyllava/eval/scienceqa/answers/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.9_2e-1_connector-1.0_2.9_2e-1_ablation-mask_applied.jsonl' +==== EXPERIMENT COMPLETED: eval_qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.9_2e-1_connector-1.0_2.9_2e-1_ablation ==== +Log File: eval_qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.9_2e-1_connector-1.0_2.9_2e-1_ablation_20251012_173957.log +Timestamp: 2025-10-12 17:41:22 +===================================== diff --git a/logs_oct11/eval_qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_1e-2_connector-3.0_0.5_1e-2_ablation_20251012_174122.log b/logs_oct11/eval_qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_1e-2_connector-3.0_0.5_1e-2_ablation_20251012_174122.log new file mode 100644 index 0000000000000000000000000000000000000000..5908f1cb04fb9c56bced0570ababcd95c32d3959 --- /dev/null +++ b/logs_oct11/eval_qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_1e-2_connector-3.0_0.5_1e-2_ablation_20251012_174122.log @@ -0,0 +1,759 @@ +==== STARTING EXPERIMENT: eval_qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_1e-2_connector-3.0_0.5_1e-2_ablation ==== +Log File: eval_qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_1e-2_connector-3.0_0.5_1e-2_ablation_20251012_174122.log +Timestamp: 2025-10-12 17:41:22 +===================================== +Processing: /nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_1e-2_connector-3.0_0.5_1e-2_ablation +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 17:41:25,499] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +config_mask.torch_dtype: torch.bfloat16 +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +Load mask model from /nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_1e-2_connector-3.0_0.5_1e-2_ablation over. +TinyLlavaConfig { + "architectures": [ + "TinyLlavaForConditionalGeneration" + ], + "backward_type_connector": "normal", + "cache_dir": null, + "connector_type": "mlp2x_gelu", + "hidden_size": 896, + "ignore_index": -100, + "image_aspect_ratio": "square", + "image_token_index": -200, + "llm_model_name_or_path": "Qwen/Qwen2.5-0.5B", + "mask_model": [ + "llm", + "connector" + ], + "mask_type_connector": "soft", + "model_type": "tinyllava", + "num_queries": 128, + "num_resampler_layers": 3, + "pad_token": "<|endoftext|>", + "resampler_hidden_size": 768, + "sparsity_connector": null, + "subnet_type_connector": "global", + "temperature_connector": 0.5, + "text_config": { + "_name_or_path": "Qwen/Qwen2.5-0.5B", + "architectures": [ + "Qwen2ForCausalLM" + ], + "backward_type": "normal", + "bos_token_id": 151643, + "eos_token_id": 151643, + "hidden_size": 896, + "intermediate_size": 4864, + "mask_type": "soft", + "masked_layers": "all", + "max_position_embeddings": 32768, + "max_window_layers": 24, + "model_type": "qwen2", + "num_attention_heads": 14, + "num_hidden_layers": 24, + "num_key_value_heads": 2, + "rope_theta": 1000000.0, + "sliding_window": 32768, + "subnet_mode": "both", + "subnet_type": "None", + "temperature_attn": 0.5, + "temperature_mlp": 0.5, + "tie_word_embeddings": true, + "torch_dtype": "bfloat16", + "use_mrope": false, + "use_sliding_window": false, + "vocab_size": 151936 + }, + "threshold_connector": null, + "tokenizer_model_max_length": 2048, + "tokenizer_name_or_path": "Qwen/Qwen2.5-0.5B", + "tokenizer_padding_side": "right", + "tokenizer_use_fast": false, + "torch_dtype": "bfloat16", + "transformers_version": "4.40.1", + "tune_type_connector": "full", + "tune_type_llm": "full", + "tune_type_vision_tower": "frozen", + "tune_vision_tower_from_layer": 0, + "use_cache": true, + "vision_config": { + "hidden_act": "gelu_pytorch_tanh", + "hidden_size": 1152, + "image_size": 384, + "intermediate_size": 4304, + "layer_norm_eps": 1e-06, + "model_name_or_path": "google/siglip-so400m-patch14-384", + "model_name_or_path2": "", + "model_type": "siglip_vision_model", + "num_attention_heads": 16, + "num_hidden_layers": 27, + "patch_size": 14 + }, + "vision_feature_layer": -2, + "vision_feature_select_strategy": "patch", + "vision_hidden_size": 1152, + "vision_model_name_or_path": "google/siglip-so400m-patch14-384", + "vision_model_name_or_path2": "", + "vocab_size": 151936 +} + +TinyLlavaForConditionalGeneration( + (language_model): Qwen2ForCausalLM( + (model): Qwen2Model( + (embed_tokens): Embedding(151936, 896) + (layers): ModuleList( + (0-23): 24 x Qwen2DecoderLayer( + (self_attn): Qwen2Attention( + (q_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=True) + (k_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=128, bias=True) + (v_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=128, bias=True) + (o_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=False) + (rotary_emb): Qwen2RotaryEmbedding() + ) + (mlp): Qwen2MLP( + (gate_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=4864, bias=False) + (up_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=4864, bias=False) + (down_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=4864, out_features=896, bias=False) + (act_fn): SiLU() + ) + (input_layernorm): Qwen2RMSNorm() + (post_attention_layernorm): Qwen2RMSNorm() + ) + ) + (norm): Qwen2RMSNorm() + ) + (lm_head): Linear(in_features=896, out_features=151936, bias=False) + ) + (vision_tower): SIGLIPVisionTower( + (_vision_tower): SiglipVisionModel( + (vision_model): SiglipVisionTransformer( + (embeddings): SiglipVisionEmbeddings( + (patch_embedding): Conv2d(3, 1152, kernel_size=(14, 14), stride=(14, 14), padding=valid) + (position_embedding): Embedding(729, 1152) + ) + (encoder): SiglipEncoder( + (layers): ModuleList( + (0-26): 27 x SiglipEncoderLayer( + (self_attn): SiglipAttention( + (k_proj): Linear(in_features=1152, out_features=1152, bias=True) + (v_proj): Linear(in_features=1152, out_features=1152, bias=True) + (q_proj): Linear(in_features=1152, out_features=1152, bias=True) + (out_proj): Linear(in_features=1152, out_features=1152, bias=True) + ) + (layer_norm1): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + (layer_norm2): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + ) + ) + ) + (post_layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (head): SiglipMultiheadAttentionPoolingHead( + (attention): MultiheadAttention( + (out_proj): NonDynamicallyQuantizableLinear(in_features=1152, out_features=1152, bias=True) + ) + (layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + ) + ) + ) + ) + (connector): MLPConnector( + (_connector): Sequential( + (0): SupermaskLinearSparsity_SoftForward_Normal(in_features=1152, out_features=896, bias=True) + (1): GELU(approximate='none') + (2): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=True) + ) + ) +) +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Load base model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain over. +TinyLlavaConfig { + "cache_dir": null, + "connector_type": "mlp2x_gelu", + "hidden_size": 896, + "ignore_index": -100, + "image_aspect_ratio": "square", + "image_token_index": -200, + "llm_model_name_or_path": "Qwen/Qwen2.5-0.5B", + "model_type": "tinyllava", + "num_queries": 128, + "num_resampler_layers": 3, + "pad_token": "<|endoftext|>", + "pad_token_id": 151643, + "resampler_hidden_size": 768, + "text_config": { + "_name_or_path": "Qwen/Qwen2.5-0.5B", + "architectures": [ + "Qwen2ForCausalLM" + ], + "bos_token_id": 151643, + "eos_token_id": 151643, + "hidden_size": 896, + "intermediate_size": 4864, + "max_position_embeddings": 32768, + "max_window_layers": 24, + "model_type": "qwen2", + "num_attention_heads": 14, + "num_hidden_layers": 24, + "num_key_value_heads": 2, + "rope_theta": 1000000.0, + "sliding_window": 32768, + "tie_word_embeddings": true, + "use_mrope": false, + "use_sliding_window": false, + "vocab_size": 151936 + }, + "tokenizer_model_max_length": 2048, + "tokenizer_name_or_path": "Qwen/Qwen2.5-0.5B", + "tokenizer_padding_side": "right", + "tokenizer_use_fast": false, + "transformers_version": "4.40.1", + "tune_type_connector": "full", + "tune_type_llm": "frozen", + "tune_type_vision_tower": "frozen", + "tune_vision_tower_from_layer": 0, + "use_cache": true, + "vision_config": { + "hidden_act": "gelu_pytorch_tanh", + "hidden_size": 1152, + "image_size": 384, + "intermediate_size": 4304, + "layer_norm_eps": 1e-06, + "model_name_or_path": "google/siglip-so400m-patch14-384", + "model_name_or_path2": "", + "model_type": "siglip_vision_model", + "num_attention_heads": 16, + "num_hidden_layers": 27, + "patch_size": 14 + }, + "vision_feature_layer": -2, + "vision_feature_select_strategy": "patch", + "vision_hidden_size": 1152, + "vision_model_name_or_path": "google/siglip-so400m-patch14-384", + "vision_model_name_or_path2": "", + "vocab_size": 151936 +} + +TinyLlavaForConditionalGeneration( + (language_model): Qwen2ForCausalLM( + (model): Qwen2Model( + (embed_tokens): Embedding(151936, 896) + (layers): ModuleList( + (0-23): 24 x Qwen2DecoderLayer( + (self_attn): Qwen2Attention( + (q_proj): Linear(in_features=896, out_features=896, bias=True) + (k_proj): Linear(in_features=896, out_features=128, bias=True) + (v_proj): Linear(in_features=896, out_features=128, bias=True) + (o_proj): Linear(in_features=896, out_features=896, bias=False) + (rotary_emb): Qwen2RotaryEmbedding() + ) + (mlp): Qwen2MLP( + (gate_proj): Linear(in_features=896, out_features=4864, bias=False) + (up_proj): Linear(in_features=896, out_features=4864, bias=False) + (down_proj): Linear(in_features=4864, out_features=896, bias=False) + (act_fn): SiLU() + ) + (input_layernorm): Qwen2RMSNorm() + (post_attention_layernorm): Qwen2RMSNorm() + ) + ) + (norm): Qwen2RMSNorm() + ) + (lm_head): Linear(in_features=896, out_features=151936, bias=False) + ) + (vision_tower): SIGLIPVisionTower( + (_vision_tower): SiglipVisionModel( + (vision_model): SiglipVisionTransformer( + (embeddings): SiglipVisionEmbeddings( + (patch_embedding): Conv2d(3, 1152, kernel_size=(14, 14), stride=(14, 14), padding=valid) + (position_embedding): Embedding(729, 1152) + ) + (encoder): SiglipEncoder( + (layers): ModuleList( + (0-26): 27 x SiglipEncoderLayer( + (self_attn): SiglipAttention( + (k_proj): Linear(in_features=1152, out_features=1152, bias=True) + (v_proj): Linear(in_features=1152, out_features=1152, bias=True) + (q_proj): Linear(in_features=1152, out_features=1152, bias=True) + (out_proj): Linear(in_features=1152, out_features=1152, bias=True) + ) + (layer_norm1): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + (layer_norm2): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + ) + ) + ) + (post_layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (head): SiglipMultiheadAttentionPoolingHead( + (attention): MultiheadAttention( + (out_proj): NonDynamicallyQuantizableLinear(in_features=1152, out_features=1152, bias=True) + ) + (layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + ) + ) + ) + ) + (connector): MLPConnector( + (_connector): Sequential( + (0): Linear(in_features=1152, out_features=896, bias=True) + (1): GELU(approximate='none') + (2): Linear(in_features=896, out_features=896, bias=True) + ) + ) +) +Collect masks for language model over. +Collect masks for connector over. +Applying mask on model.layers.0.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.self_attn.q_proj. +Applying mask on model.layers.0.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.self_attn.k_proj. +Applying mask on model.layers.0.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.self_attn.v_proj. +Applying mask on model.layers.0.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.self_attn.o_proj. +Applying mask on model.layers.0.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.mlp.gate_proj. +Applying mask on model.layers.0.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.mlp.up_proj. +Applying mask on model.layers.0.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.mlp.down_proj. +Applying mask on model.layers.1.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.self_attn.q_proj. +Applying mask on model.layers.1.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.self_attn.k_proj. +Applying mask on model.layers.1.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.self_attn.v_proj. +Applying mask on model.layers.1.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.self_attn.o_proj. +Applying mask on model.layers.1.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.mlp.gate_proj. +Applying mask on model.layers.1.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.mlp.up_proj. +Applying mask on model.layers.1.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.mlp.down_proj. +Applying mask on model.layers.2.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.self_attn.q_proj. +Applying mask on model.layers.2.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.self_attn.k_proj. +Applying mask on model.layers.2.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.self_attn.v_proj. +Applying mask on model.layers.2.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.self_attn.o_proj. +Applying mask on model.layers.2.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.mlp.gate_proj. +Applying mask on model.layers.2.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.mlp.up_proj. +Applying mask on model.layers.2.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.mlp.down_proj. +Applying mask on model.layers.3.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.self_attn.q_proj. +Applying mask on model.layers.3.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.self_attn.k_proj. +Applying mask on model.layers.3.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.self_attn.v_proj. +Applying mask on model.layers.3.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.self_attn.o_proj. +Applying mask on model.layers.3.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.mlp.gate_proj. +Applying mask on model.layers.3.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.mlp.up_proj. +Applying mask on model.layers.3.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.mlp.down_proj. +Applying mask on model.layers.4.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.self_attn.q_proj. +Applying mask on model.layers.4.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.self_attn.k_proj. +Applying mask on model.layers.4.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.self_attn.v_proj. +Applying mask on model.layers.4.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.self_attn.o_proj. +Applying mask on model.layers.4.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.mlp.gate_proj. +Applying mask on model.layers.4.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.mlp.up_proj. +Applying mask on model.layers.4.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.mlp.down_proj. +Applying mask on model.layers.5.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.self_attn.q_proj. +Applying mask on model.layers.5.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.self_attn.k_proj. +Applying mask on model.layers.5.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.self_attn.v_proj. +Applying mask on model.layers.5.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.self_attn.o_proj. +Applying mask on model.layers.5.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.mlp.gate_proj. +Applying mask on model.layers.5.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.mlp.up_proj. +Applying mask on model.layers.5.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.mlp.down_proj. +Applying mask on model.layers.6.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.self_attn.q_proj. +Applying mask on model.layers.6.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.self_attn.k_proj. +Applying mask on model.layers.6.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.self_attn.v_proj. +Applying mask on model.layers.6.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.self_attn.o_proj. +Applying mask on model.layers.6.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.mlp.gate_proj. +Applying mask on model.layers.6.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.mlp.up_proj. +Applying mask on model.layers.6.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.mlp.down_proj. +Applying mask on model.layers.7.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.self_attn.q_proj. +Applying mask on model.layers.7.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.self_attn.k_proj. +Applying mask on model.layers.7.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.self_attn.v_proj. +Applying mask on model.layers.7.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.self_attn.o_proj. +Applying mask on model.layers.7.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.mlp.gate_proj. +Applying mask on model.layers.7.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.mlp.up_proj. +Applying mask on model.layers.7.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.mlp.down_proj. +Applying mask on model.layers.8.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.self_attn.q_proj. +Applying mask on model.layers.8.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.self_attn.k_proj. +Applying mask on model.layers.8.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.self_attn.v_proj. +Applying mask on model.layers.8.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.self_attn.o_proj. +Applying mask on model.layers.8.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.mlp.gate_proj. +Applying mask on model.layers.8.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.mlp.up_proj. +Applying mask on model.layers.8.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.mlp.down_proj. +Applying mask on model.layers.9.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.self_attn.q_proj. +Applying mask on model.layers.9.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.self_attn.k_proj. +Applying mask on model.layers.9.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.self_attn.v_proj. +Applying mask on model.layers.9.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.self_attn.o_proj. +Applying mask on model.layers.9.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.mlp.gate_proj. +Applying mask on model.layers.9.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.mlp.up_proj. +Applying mask on model.layers.9.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.mlp.down_proj. +Applying mask on model.layers.10.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.self_attn.q_proj. +Applying mask on model.layers.10.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.self_attn.k_proj. +Applying mask on model.layers.10.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.self_attn.v_proj. +Applying mask on model.layers.10.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.self_attn.o_proj. +Applying mask on model.layers.10.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.mlp.gate_proj. +Applying mask on model.layers.10.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.mlp.up_proj. +Applying mask on model.layers.10.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.mlp.down_proj. +Applying mask on model.layers.11.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.self_attn.q_proj. +Applying mask on model.layers.11.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.self_attn.k_proj. +Applying mask on model.layers.11.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.self_attn.v_proj. +Applying mask on model.layers.11.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.self_attn.o_proj. +Applying mask on model.layers.11.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.mlp.gate_proj. +Applying mask on model.layers.11.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.mlp.up_proj. +Applying mask on model.layers.11.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.mlp.down_proj. +Applying mask on model.layers.12.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.self_attn.q_proj. +Applying mask on model.layers.12.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.self_attn.k_proj. +Applying mask on model.layers.12.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.self_attn.v_proj. +Applying mask on model.layers.12.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.self_attn.o_proj. +Applying mask on model.layers.12.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.mlp.gate_proj. +Applying mask on model.layers.12.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.mlp.up_proj. +Applying mask on model.layers.12.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.mlp.down_proj. +Applying mask on model.layers.13.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.self_attn.q_proj. +Applying mask on model.layers.13.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.self_attn.k_proj. +Applying mask on model.layers.13.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.self_attn.v_proj. +Applying mask on model.layers.13.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.self_attn.o_proj. +Applying mask on model.layers.13.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.mlp.gate_proj. +Applying mask on model.layers.13.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.mlp.up_proj. +Applying mask on model.layers.13.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.mlp.down_proj. +Applying mask on model.layers.14.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.self_attn.q_proj. +Applying mask on model.layers.14.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.self_attn.k_proj. +Applying mask on model.layers.14.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.self_attn.v_proj. +Applying mask on model.layers.14.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.self_attn.o_proj. +Applying mask on model.layers.14.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.mlp.gate_proj. +Applying mask on model.layers.14.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.mlp.up_proj. +Applying mask on model.layers.14.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.mlp.down_proj. +Applying mask on model.layers.15.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.self_attn.q_proj. +Applying mask on model.layers.15.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.self_attn.k_proj. +Applying mask on model.layers.15.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.self_attn.v_proj. +Applying mask on model.layers.15.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.self_attn.o_proj. +Applying mask on model.layers.15.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.mlp.gate_proj. +Applying mask on model.layers.15.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.mlp.up_proj. +Applying mask on model.layers.15.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.mlp.down_proj. +Applying mask on model.layers.16.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.self_attn.q_proj. +Applying mask on model.layers.16.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.self_attn.k_proj. +Applying mask on model.layers.16.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.self_attn.v_proj. +Applying mask on model.layers.16.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.self_attn.o_proj. +Applying mask on model.layers.16.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.mlp.gate_proj. +Applying mask on model.layers.16.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.mlp.up_proj. +Applying mask on model.layers.16.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.mlp.down_proj. +Applying mask on model.layers.17.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.self_attn.q_proj. +Applying mask on model.layers.17.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.self_attn.k_proj. +Applying mask on model.layers.17.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.self_attn.v_proj. +Applying mask on model.layers.17.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.self_attn.o_proj. +Applying mask on model.layers.17.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.mlp.gate_proj. +Applying mask on model.layers.17.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.mlp.up_proj. +Applying mask on model.layers.17.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.mlp.down_proj. +Applying mask on model.layers.18.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.self_attn.q_proj. +Applying mask on model.layers.18.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.self_attn.k_proj. +Applying mask on model.layers.18.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.self_attn.v_proj. +Applying mask on model.layers.18.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.self_attn.o_proj. +Applying mask on model.layers.18.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.mlp.gate_proj. +Applying mask on model.layers.18.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.mlp.up_proj. +Applying mask on model.layers.18.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.mlp.down_proj. +Applying mask on model.layers.19.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.self_attn.q_proj. +Applying mask on model.layers.19.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.self_attn.k_proj. +Applying mask on model.layers.19.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.self_attn.v_proj. +Applying mask on model.layers.19.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.self_attn.o_proj. +Applying mask on model.layers.19.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.mlp.gate_proj. +Applying mask on model.layers.19.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.mlp.up_proj. +Applying mask on model.layers.19.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.mlp.down_proj. +Applying mask on model.layers.20.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.self_attn.q_proj. +Applying mask on model.layers.20.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.self_attn.k_proj. +Applying mask on model.layers.20.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.self_attn.v_proj. +Applying mask on model.layers.20.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.self_attn.o_proj. +Applying mask on model.layers.20.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.mlp.gate_proj. +Applying mask on model.layers.20.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.mlp.up_proj. +Applying mask on model.layers.20.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.mlp.down_proj. +Applying mask on model.layers.21.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.self_attn.q_proj. +Applying mask on model.layers.21.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.self_attn.k_proj. +Applying mask on model.layers.21.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.self_attn.v_proj. +Applying mask on model.layers.21.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.self_attn.o_proj. +Applying mask on model.layers.21.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.mlp.gate_proj. +Applying mask on model.layers.21.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.mlp.up_proj. +Applying mask on model.layers.21.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.mlp.down_proj. +Applying mask on model.layers.22.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.self_attn.q_proj. +Applying mask on model.layers.22.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.self_attn.k_proj. +Applying mask on model.layers.22.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.self_attn.v_proj. +Applying mask on model.layers.22.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.self_attn.o_proj. +Applying mask on model.layers.22.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.mlp.gate_proj. +Applying mask on model.layers.22.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.mlp.up_proj. +Applying mask on model.layers.22.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.mlp.down_proj. +Applying mask on model.layers.23.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.self_attn.q_proj. +Applying mask on model.layers.23.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.self_attn.k_proj. +Applying mask on model.layers.23.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.self_attn.v_proj. +Applying mask on model.layers.23.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.self_attn.o_proj. +Applying mask on model.layers.23.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.mlp.gate_proj. +Applying mask on model.layers.23.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.mlp.up_proj. +Applying mask on model.layers.23.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.mlp.down_proj. +Applying mask on _connector.0 with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on _connector.0. +Applying mask on _connector.2 with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on _connector.2. +Using cleaned config_mask (without mask parameters) for saving. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 17:42:07,532] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. + 0%| | 0/900 [00:00 + main() + File "/nfs/ywang29/TinyLLaVA/scripts/apply_masks.py", line 123, in main + config_mask = TinyLlavaConfig.from_pretrained(model_args.mask_model_name_or_path) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/configuration_utils.py", line 602, in from_pretrained + config_dict, kwargs = cls.get_config_dict(pretrained_model_name_or_path, **kwargs) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/configuration_utils.py", line 631, in get_config_dict + config_dict, kwargs = cls._get_config_dict(pretrained_model_name_or_path, **kwargs) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/configuration_utils.py", line 686, in _get_config_dict + resolved_config_file = cached_file( + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/utils/hub.py", line 369, in cached_file + raise EnvironmentError( +OSError: /nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_2e-1_connector-3.0_0.5_2e-1_ablation does not appear to have a file named config.json. Checkout 'https://huggingface.co//nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_2e-1_connector-3.0_0.5_2e-1_ablation/tree/main' for available files. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 18:06:39,994] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +Traceback (most recent call last): + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/utils/hub.py", line 398, in cached_file + resolved_file = hf_hub_download( + File "/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/utils/_validators.py", line 106, in _inner_fn + validate_repo_id(arg_value) + File "/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/utils/_validators.py", line 154, in validate_repo_id + raise HFValidationError( +huggingface_hub.errors.HFValidationError: Repo id must be in the form 'repo_name' or 'namespace/repo_name': '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_2e-1_connector-3.0_0.5_2e-1_ablation/mask_applied'. Use `repo_type` argument if needed. + +The above exception was the direct cause of the following exception: + +Traceback (most recent call last): + File "/nfs/ywang29/TinyLLaVA/tinyllava/model/load_model.py", line 38, in load_pretrained_model + model = TinyLlavaForConditionalGeneration.from_pretrained(model_name_or_path,low_cpu_mem_usage=True) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/modeling_utils.py", line 3015, in from_pretrained + resolved_config_file = cached_file( + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/utils/hub.py", line 462, in cached_file + raise EnvironmentError( +OSError: Incorrect path_or_model_id: '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_2e-1_connector-3.0_0.5_2e-1_ablation/mask_applied'. Please provide either the path to a local folder or the repo_id of a model on the Hub. + +During handling of the above exception, another exception occurred: + +Traceback (most recent call last): + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/utils/hub.py", line 398, in cached_file + resolved_file = hf_hub_download( + File "/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/utils/_validators.py", line 106, in _inner_fn + validate_repo_id(arg_value) + File "/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/utils/_validators.py", line 154, in validate_repo_id + raise HFValidationError( +huggingface_hub.errors.HFValidationError: Repo id must be in the form 'repo_name' or 'namespace/repo_name': '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_2e-1_connector-3.0_0.5_2e-1_ablation/mask_applied'. Use `repo_type` argument if needed. + +The above exception was the direct cause of the following exception: + +Traceback (most recent call last): + File "/opt/conda/envs/tinyllava/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/opt/conda/envs/tinyllava/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/nfs/ywang29/TinyLLaVA/tinyllava/eval/model_vqa_mmmu.py", line 180, in + eval_model(args) + File "/nfs/ywang29/TinyLLaVA/tinyllava/eval/model_vqa_mmmu.py", line 88, in eval_model + model, tokenizer, image_processor, context_len = load_pretrained_model(model_path) + File "/nfs/ywang29/TinyLLaVA/tinyllava/model/load_model.py", line 40, in load_pretrained_model + model_config = TinyLlavaConfig.from_pretrained(model_name_or_path) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/configuration_utils.py", line 602, in from_pretrained + config_dict, kwargs = cls.get_config_dict(pretrained_model_name_or_path, **kwargs) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/configuration_utils.py", line 631, in get_config_dict + config_dict, kwargs = cls._get_config_dict(pretrained_model_name_or_path, **kwargs) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/configuration_utils.py", line 686, in _get_config_dict + resolved_config_file = cached_file( + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/utils/hub.py", line 462, in cached_file + raise EnvironmentError( +OSError: Incorrect path_or_model_id: '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_2e-1_connector-3.0_0.5_2e-1_ablation/mask_applied'. Please provide either the path to a local folder or the repo_id of a model on the Hub. +Traceback (most recent call last): + File "/nfs/ywang29/TinyLLaVA/scripts/convert_answer_to_mmmu.py", line 31, in + eval_model(args) + File "/nfs/ywang29/TinyLLaVA/scripts/convert_answer_to_mmmu.py", line 7, in eval_model + answers = [json.loads(q) for q in open(os.path.expanduser(args.answers_file), "r")] +FileNotFoundError: [Errno 2] No such file or directory: '/s3-code/ywang29/datasets/tinyllava/eval/MMMU/answers/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_2e-1_connector-3.0_0.5_2e-1_ablation-mask_applied.jsonl' +Traceback (most recent call last): + File "/s3-code/ywang29/datasets/tinyllava/eval/MMMU/eval/main_eval_only.py", line 19, in + output_dict = json.load(open(args.output_path)) +FileNotFoundError: [Errno 2] No such file or directory: '/s3-code/ywang29/datasets/tinyllava/eval/MMMU/answers/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_2e-1_connector-3.0_0.5_2e-1_ablation-mask_applied_output.json' +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 18:06:46,191] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +Traceback (most recent call last): + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/utils/hub.py", line 398, in cached_file + resolved_file = hf_hub_download( + File "/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/utils/_validators.py", line 106, in _inner_fn + validate_repo_id(arg_value) + File "/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/utils/_validators.py", line 154, in validate_repo_id + raise HFValidationError( +huggingface_hub.errors.HFValidationError: Repo id must be in the form 'repo_name' or 'namespace/repo_name': '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_2e-1_connector-3.0_0.5_2e-1_ablation/mask_applied'. Use `repo_type` argument if needed. + +The above exception was the direct cause of the following exception: + +Traceback (most recent call last): + File "/nfs/ywang29/TinyLLaVA/tinyllava/model/load_model.py", line 38, in load_pretrained_model + model = TinyLlavaForConditionalGeneration.from_pretrained(model_name_or_path,low_cpu_mem_usage=True) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/modeling_utils.py", line 3015, in from_pretrained + resolved_config_file = cached_file( + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/utils/hub.py", line 462, in cached_file + raise EnvironmentError( +OSError: Incorrect path_or_model_id: '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_2e-1_connector-3.0_0.5_2e-1_ablation/mask_applied'. Please provide either the path to a local folder or the repo_id of a model on the Hub. + +During handling of the above exception, another exception occurred: + +Traceback (most recent call last): + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/utils/hub.py", line 398, in cached_file + resolved_file = hf_hub_download( + File "/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/utils/_validators.py", line 106, in _inner_fn + validate_repo_id(arg_value) + File "/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/utils/_validators.py", line 154, in validate_repo_id + raise HFValidationError( +huggingface_hub.errors.HFValidationError: Repo id must be in the form 'repo_name' or 'namespace/repo_name': '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_2e-1_connector-3.0_0.5_2e-1_ablation/mask_applied'. Use `repo_type` argument if needed. + +The above exception was the direct cause of the following exception: + +Traceback (most recent call last): + File "/opt/conda/envs/tinyllava/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/opt/conda/envs/tinyllava/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/nfs/ywang29/TinyLLaVA/tinyllava/eval/model_vqa_loader.py", line 146, in + eval_model(args) + File "/nfs/ywang29/TinyLLaVA/tinyllava/eval/model_vqa_loader.py", line 79, in eval_model + model, tokenizer, image_processor, context_len = load_pretrained_model(model_path) + File "/nfs/ywang29/TinyLLaVA/tinyllava/model/load_model.py", line 40, in load_pretrained_model + model_config = TinyLlavaConfig.from_pretrained(model_name_or_path) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/configuration_utils.py", line 602, in from_pretrained + config_dict, kwargs = cls.get_config_dict(pretrained_model_name_or_path, **kwargs) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/configuration_utils.py", line 631, in get_config_dict + config_dict, kwargs = cls._get_config_dict(pretrained_model_name_or_path, **kwargs) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/configuration_utils.py", line 686, in _get_config_dict + resolved_config_file = cached_file( + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/utils/hub.py", line 462, in cached_file + raise EnvironmentError( +OSError: Incorrect path_or_model_id: '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_2e-1_connector-3.0_0.5_2e-1_ablation/mask_applied'. Please provide either the path to a local folder or the repo_id of a model on the Hub. +Traceback (most recent call last): + File "/s3-code/ywang29/datasets/tinyllava/eval/MME/convert_answer_to_mme.py", line 52, in + answers = [json.loads(line) for line in open(os.path.join('answers', f'{experiment}.jsonl'))] +FileNotFoundError: [Errno 2] No such file or directory: 'answers/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_2e-1_connector-3.0_0.5_2e-1_ablation-mask_applied.jsonl' +=========== Perception =========== +Traceback (most recent call last): + File "/s3-code/ywang29/datasets/tinyllava/eval/MME/eval_tool/calculation.py", line 164, in + cal.process_result(results_dir) + File "/s3-code/ywang29/datasets/tinyllava/eval/MME/eval_tool/calculation.py", line 98, in process_result + lines = open(task_txt, 'r').readlines() +FileNotFoundError: [Errno 2] No such file or directory: 'answers/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_2e-1_connector-3.0_0.5_2e-1_ablation-mask_applied/existence.txt' +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 18:07:51,101] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +Traceback (most recent call last): + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/utils/hub.py", line 398, in cached_file + resolved_file = hf_hub_download( + File "/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/utils/_validators.py", line 106, in _inner_fn + validate_repo_id(arg_value) + File "/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/utils/_validators.py", line 154, in validate_repo_id + raise HFValidationError( +huggingface_hub.errors.HFValidationError: Repo id must be in the form 'repo_name' or 'namespace/repo_name': '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_2e-1_connector-3.0_0.5_2e-1_ablation/mask_applied'. Use `repo_type` argument if needed. + +The above exception was the direct cause of the following exception: + +Traceback (most recent call last): + File "/nfs/ywang29/TinyLLaVA/tinyllava/model/load_model.py", line 38, in load_pretrained_model + model = TinyLlavaForConditionalGeneration.from_pretrained(model_name_or_path,low_cpu_mem_usage=True) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/modeling_utils.py", line 3015, in from_pretrained + resolved_config_file = cached_file( + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/utils/hub.py", line 462, in cached_file + raise EnvironmentError( +OSError: Incorrect path_or_model_id: '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_2e-1_connector-3.0_0.5_2e-1_ablation/mask_applied'. Please provide either the path to a local folder or the repo_id of a model on the Hub. + +During handling of the above exception, another exception occurred: + +Traceback (most recent call last): + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/utils/hub.py", line 398, in cached_file + resolved_file = hf_hub_download( + File "/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/utils/_validators.py", line 106, in _inner_fn + validate_repo_id(arg_value) + File "/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/utils/_validators.py", line 154, in validate_repo_id + raise HFValidationError( +huggingface_hub.errors.HFValidationError: Repo id must be in the form 'repo_name' or 'namespace/repo_name': '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_2e-1_connector-3.0_0.5_2e-1_ablation/mask_applied'. Use `repo_type` argument if needed. + +The above exception was the direct cause of the following exception: + +Traceback (most recent call last): + File "/opt/conda/envs/tinyllava/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/opt/conda/envs/tinyllava/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/nfs/ywang29/TinyLLaVA/tinyllava/eval/model_vqa_science.py", line 108, in + eval_model(args) + File "/nfs/ywang29/TinyLLaVA/tinyllava/eval/model_vqa_science.py", line 31, in eval_model + model, tokenizer, image_processor, context_len = load_pretrained_model(model_path) + File "/nfs/ywang29/TinyLLaVA/tinyllava/model/load_model.py", line 40, in load_pretrained_model + model_config = TinyLlavaConfig.from_pretrained(model_name_or_path) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/configuration_utils.py", line 602, in from_pretrained + config_dict, kwargs = cls.get_config_dict(pretrained_model_name_or_path, **kwargs) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/configuration_utils.py", line 631, in get_config_dict + config_dict, kwargs = cls._get_config_dict(pretrained_model_name_or_path, **kwargs) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/configuration_utils.py", line 686, in _get_config_dict + resolved_config_file = cached_file( + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/utils/hub.py", line 462, in cached_file + raise EnvironmentError( +OSError: Incorrect path_or_model_id: '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_2e-1_connector-3.0_0.5_2e-1_ablation/mask_applied'. Please provide either the path to a local folder or the repo_id of a model on the Hub. +Traceback (most recent call last): + File "/nfs/ywang29/TinyLLaVA/tinyllava/eval/eval_science_qa.py", line 45, in + predictions = [json.loads(line) for line in open(args.result_file)] +FileNotFoundError: [Errno 2] No such file or directory: '/s3-code/ywang29/datasets/tinyllava/eval/scienceqa/answers/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_2e-1_connector-3.0_0.5_2e-1_ablation-mask_applied.jsonl' +==== EXPERIMENT COMPLETED: eval_qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_2e-1_connector-3.0_0.5_2e-1_ablation ==== +Log File: eval_qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_2e-1_connector-3.0_0.5_2e-1_ablation_20251012_180629.log +Timestamp: 2025-10-12 18:07:54 +===================================== diff --git a/logs_oct11/eval_qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_3e-2_connector-3.0_0.5_3e-2_ablation_20251012_180754.log b/logs_oct11/eval_qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_3e-2_connector-3.0_0.5_3e-2_ablation_20251012_180754.log new file mode 100644 index 0000000000000000000000000000000000000000..eaab3694728f1fda025650e7a617d637fd806182 --- /dev/null +++ b/logs_oct11/eval_qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_3e-2_connector-3.0_0.5_3e-2_ablation_20251012_180754.log @@ -0,0 +1,759 @@ +==== STARTING EXPERIMENT: eval_qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_3e-2_connector-3.0_0.5_3e-2_ablation ==== +Log File: eval_qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_3e-2_connector-3.0_0.5_3e-2_ablation_20251012_180754.log +Timestamp: 2025-10-12 18:07:54 +===================================== +Processing: /nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_3e-2_connector-3.0_0.5_3e-2_ablation +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 18:07:57,141] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +config_mask.torch_dtype: torch.bfloat16 +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +Load mask model from /nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_3e-2_connector-3.0_0.5_3e-2_ablation over. +TinyLlavaConfig { + "architectures": [ + "TinyLlavaForConditionalGeneration" + ], + "backward_type_connector": "normal", + "cache_dir": null, + "connector_type": "mlp2x_gelu", + "hidden_size": 896, + "ignore_index": -100, + "image_aspect_ratio": "square", + "image_token_index": -200, + "llm_model_name_or_path": "Qwen/Qwen2.5-0.5B", + "mask_model": [ + "llm", + "connector" + ], + "mask_type_connector": "soft", + "model_type": "tinyllava", + "num_queries": 128, + "num_resampler_layers": 3, + "pad_token": "<|endoftext|>", + "resampler_hidden_size": 768, + "sparsity_connector": null, + "subnet_type_connector": "global", + "temperature_connector": 0.5, + "text_config": { + "_name_or_path": "Qwen/Qwen2.5-0.5B", + "architectures": [ + "Qwen2ForCausalLM" + ], + "backward_type": "normal", + "bos_token_id": 151643, + "eos_token_id": 151643, + "hidden_size": 896, + "intermediate_size": 4864, + "mask_type": "soft", + "masked_layers": "all", + "max_position_embeddings": 32768, + "max_window_layers": 24, + "model_type": "qwen2", + "num_attention_heads": 14, + "num_hidden_layers": 24, + "num_key_value_heads": 2, + "rope_theta": 1000000.0, + "sliding_window": 32768, + "subnet_mode": "both", + "subnet_type": "None", + "temperature_attn": 0.5, + "temperature_mlp": 0.5, + "tie_word_embeddings": true, + "torch_dtype": "bfloat16", + "use_mrope": false, + "use_sliding_window": false, + "vocab_size": 151936 + }, + "threshold_connector": null, + "tokenizer_model_max_length": 2048, + "tokenizer_name_or_path": "Qwen/Qwen2.5-0.5B", + "tokenizer_padding_side": "right", + "tokenizer_use_fast": false, + "torch_dtype": "bfloat16", + "transformers_version": "4.40.1", + "tune_type_connector": "full", + "tune_type_llm": "full", + "tune_type_vision_tower": "frozen", + "tune_vision_tower_from_layer": 0, + "use_cache": true, + "vision_config": { + "hidden_act": "gelu_pytorch_tanh", + "hidden_size": 1152, + "image_size": 384, + "intermediate_size": 4304, + "layer_norm_eps": 1e-06, + "model_name_or_path": "google/siglip-so400m-patch14-384", + "model_name_or_path2": "", + "model_type": "siglip_vision_model", + "num_attention_heads": 16, + "num_hidden_layers": 27, + "patch_size": 14 + }, + "vision_feature_layer": -2, + "vision_feature_select_strategy": "patch", + "vision_hidden_size": 1152, + "vision_model_name_or_path": "google/siglip-so400m-patch14-384", + "vision_model_name_or_path2": "", + "vocab_size": 151936 +} + +TinyLlavaForConditionalGeneration( + (language_model): Qwen2ForCausalLM( + (model): Qwen2Model( + (embed_tokens): Embedding(151936, 896) + (layers): ModuleList( + (0-23): 24 x Qwen2DecoderLayer( + (self_attn): Qwen2Attention( + (q_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=True) + (k_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=128, bias=True) + (v_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=128, bias=True) + (o_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=False) + (rotary_emb): Qwen2RotaryEmbedding() + ) + (mlp): Qwen2MLP( + (gate_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=4864, bias=False) + (up_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=4864, bias=False) + (down_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=4864, out_features=896, bias=False) + (act_fn): SiLU() + ) + (input_layernorm): Qwen2RMSNorm() + (post_attention_layernorm): Qwen2RMSNorm() + ) + ) + (norm): Qwen2RMSNorm() + ) + (lm_head): Linear(in_features=896, out_features=151936, bias=False) + ) + (vision_tower): SIGLIPVisionTower( + (_vision_tower): SiglipVisionModel( + (vision_model): SiglipVisionTransformer( + (embeddings): SiglipVisionEmbeddings( + (patch_embedding): Conv2d(3, 1152, kernel_size=(14, 14), stride=(14, 14), padding=valid) + (position_embedding): Embedding(729, 1152) + ) + (encoder): SiglipEncoder( + (layers): ModuleList( + (0-26): 27 x SiglipEncoderLayer( + (self_attn): SiglipAttention( + (k_proj): Linear(in_features=1152, out_features=1152, bias=True) + (v_proj): Linear(in_features=1152, out_features=1152, bias=True) + (q_proj): Linear(in_features=1152, out_features=1152, bias=True) + (out_proj): Linear(in_features=1152, out_features=1152, bias=True) + ) + (layer_norm1): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + (layer_norm2): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + ) + ) + ) + (post_layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (head): SiglipMultiheadAttentionPoolingHead( + (attention): MultiheadAttention( + (out_proj): NonDynamicallyQuantizableLinear(in_features=1152, out_features=1152, bias=True) + ) + (layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + ) + ) + ) + ) + (connector): MLPConnector( + (_connector): Sequential( + (0): SupermaskLinearSparsity_SoftForward_Normal(in_features=1152, out_features=896, bias=True) + (1): GELU(approximate='none') + (2): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=True) + ) + ) +) +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Load base model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain over. +TinyLlavaConfig { + "cache_dir": null, + "connector_type": "mlp2x_gelu", + "hidden_size": 896, + "ignore_index": -100, + "image_aspect_ratio": "square", + "image_token_index": -200, + "llm_model_name_or_path": "Qwen/Qwen2.5-0.5B", + "model_type": "tinyllava", + "num_queries": 128, + "num_resampler_layers": 3, + "pad_token": "<|endoftext|>", + "pad_token_id": 151643, + "resampler_hidden_size": 768, + "text_config": { + "_name_or_path": "Qwen/Qwen2.5-0.5B", + "architectures": [ + "Qwen2ForCausalLM" + ], + "bos_token_id": 151643, + "eos_token_id": 151643, + "hidden_size": 896, + "intermediate_size": 4864, + "max_position_embeddings": 32768, + "max_window_layers": 24, + "model_type": "qwen2", + "num_attention_heads": 14, + "num_hidden_layers": 24, + "num_key_value_heads": 2, + "rope_theta": 1000000.0, + "sliding_window": 32768, + "tie_word_embeddings": true, + "use_mrope": false, + "use_sliding_window": false, + "vocab_size": 151936 + }, + "tokenizer_model_max_length": 2048, + "tokenizer_name_or_path": "Qwen/Qwen2.5-0.5B", + "tokenizer_padding_side": "right", + "tokenizer_use_fast": false, + "transformers_version": "4.40.1", + "tune_type_connector": "full", + "tune_type_llm": "frozen", + "tune_type_vision_tower": "frozen", + "tune_vision_tower_from_layer": 0, + "use_cache": true, + "vision_config": { + "hidden_act": "gelu_pytorch_tanh", + "hidden_size": 1152, + "image_size": 384, + "intermediate_size": 4304, + "layer_norm_eps": 1e-06, + "model_name_or_path": "google/siglip-so400m-patch14-384", + "model_name_or_path2": "", + "model_type": "siglip_vision_model", + "num_attention_heads": 16, + "num_hidden_layers": 27, + "patch_size": 14 + }, + "vision_feature_layer": -2, + "vision_feature_select_strategy": "patch", + "vision_hidden_size": 1152, + "vision_model_name_or_path": "google/siglip-so400m-patch14-384", + "vision_model_name_or_path2": "", + "vocab_size": 151936 +} + +TinyLlavaForConditionalGeneration( + (language_model): Qwen2ForCausalLM( + (model): Qwen2Model( + (embed_tokens): Embedding(151936, 896) + (layers): ModuleList( + (0-23): 24 x Qwen2DecoderLayer( + (self_attn): Qwen2Attention( + (q_proj): Linear(in_features=896, out_features=896, bias=True) + (k_proj): Linear(in_features=896, out_features=128, bias=True) + (v_proj): Linear(in_features=896, out_features=128, bias=True) + (o_proj): Linear(in_features=896, out_features=896, bias=False) + (rotary_emb): Qwen2RotaryEmbedding() + ) + (mlp): Qwen2MLP( + (gate_proj): Linear(in_features=896, out_features=4864, bias=False) + (up_proj): Linear(in_features=896, out_features=4864, bias=False) + (down_proj): Linear(in_features=4864, out_features=896, bias=False) + (act_fn): SiLU() + ) + (input_layernorm): Qwen2RMSNorm() + (post_attention_layernorm): Qwen2RMSNorm() + ) + ) + (norm): Qwen2RMSNorm() + ) + (lm_head): Linear(in_features=896, out_features=151936, bias=False) + ) + (vision_tower): SIGLIPVisionTower( + (_vision_tower): SiglipVisionModel( + (vision_model): SiglipVisionTransformer( + (embeddings): SiglipVisionEmbeddings( + (patch_embedding): Conv2d(3, 1152, kernel_size=(14, 14), stride=(14, 14), padding=valid) + (position_embedding): Embedding(729, 1152) + ) + (encoder): SiglipEncoder( + (layers): ModuleList( + (0-26): 27 x SiglipEncoderLayer( + (self_attn): SiglipAttention( + (k_proj): Linear(in_features=1152, out_features=1152, bias=True) + (v_proj): Linear(in_features=1152, out_features=1152, bias=True) + (q_proj): Linear(in_features=1152, out_features=1152, bias=True) + (out_proj): Linear(in_features=1152, out_features=1152, bias=True) + ) + (layer_norm1): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + (layer_norm2): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + ) + ) + ) + (post_layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (head): SiglipMultiheadAttentionPoolingHead( + (attention): MultiheadAttention( + (out_proj): NonDynamicallyQuantizableLinear(in_features=1152, out_features=1152, bias=True) + ) + (layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + ) + ) + ) + ) + (connector): MLPConnector( + (_connector): Sequential( + (0): Linear(in_features=1152, out_features=896, bias=True) + (1): GELU(approximate='none') + (2): Linear(in_features=896, out_features=896, bias=True) + ) + ) +) +Collect masks for language model over. +Collect masks for connector over. +Applying mask on model.layers.0.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.self_attn.q_proj. +Applying mask on model.layers.0.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.self_attn.k_proj. +Applying mask on model.layers.0.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.self_attn.v_proj. +Applying mask on model.layers.0.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.self_attn.o_proj. +Applying mask on model.layers.0.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.mlp.gate_proj. +Applying mask on model.layers.0.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.mlp.up_proj. +Applying mask on model.layers.0.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.mlp.down_proj. +Applying mask on model.layers.1.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.self_attn.q_proj. +Applying mask on model.layers.1.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.self_attn.k_proj. +Applying mask on model.layers.1.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.self_attn.v_proj. +Applying mask on model.layers.1.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.self_attn.o_proj. +Applying mask on model.layers.1.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.mlp.gate_proj. +Applying mask on model.layers.1.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.mlp.up_proj. +Applying mask on model.layers.1.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.mlp.down_proj. +Applying mask on model.layers.2.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.self_attn.q_proj. +Applying mask on model.layers.2.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.self_attn.k_proj. +Applying mask on model.layers.2.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.self_attn.v_proj. +Applying mask on model.layers.2.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.self_attn.o_proj. +Applying mask on model.layers.2.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.mlp.gate_proj. +Applying mask on model.layers.2.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.mlp.up_proj. +Applying mask on model.layers.2.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.mlp.down_proj. +Applying mask on model.layers.3.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.self_attn.q_proj. +Applying mask on model.layers.3.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.self_attn.k_proj. +Applying mask on model.layers.3.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.self_attn.v_proj. +Applying mask on model.layers.3.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.self_attn.o_proj. +Applying mask on model.layers.3.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.mlp.gate_proj. +Applying mask on model.layers.3.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.mlp.up_proj. +Applying mask on model.layers.3.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.mlp.down_proj. +Applying mask on model.layers.4.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.self_attn.q_proj. +Applying mask on model.layers.4.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.self_attn.k_proj. +Applying mask on model.layers.4.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.self_attn.v_proj. +Applying mask on model.layers.4.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.self_attn.o_proj. +Applying mask on model.layers.4.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.mlp.gate_proj. +Applying mask on model.layers.4.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.mlp.up_proj. +Applying mask on model.layers.4.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.mlp.down_proj. +Applying mask on model.layers.5.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.self_attn.q_proj. +Applying mask on model.layers.5.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.self_attn.k_proj. +Applying mask on model.layers.5.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.self_attn.v_proj. +Applying mask on model.layers.5.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.self_attn.o_proj. +Applying mask on model.layers.5.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.mlp.gate_proj. +Applying mask on model.layers.5.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.mlp.up_proj. +Applying mask on model.layers.5.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.mlp.down_proj. +Applying mask on model.layers.6.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.self_attn.q_proj. +Applying mask on model.layers.6.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.self_attn.k_proj. +Applying mask on model.layers.6.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.self_attn.v_proj. +Applying mask on model.layers.6.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.self_attn.o_proj. +Applying mask on model.layers.6.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.mlp.gate_proj. +Applying mask on model.layers.6.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.mlp.up_proj. +Applying mask on model.layers.6.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.mlp.down_proj. +Applying mask on model.layers.7.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.self_attn.q_proj. +Applying mask on model.layers.7.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.self_attn.k_proj. +Applying mask on model.layers.7.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.self_attn.v_proj. +Applying mask on model.layers.7.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.self_attn.o_proj. +Applying mask on model.layers.7.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.mlp.gate_proj. +Applying mask on model.layers.7.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.mlp.up_proj. +Applying mask on model.layers.7.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.mlp.down_proj. +Applying mask on model.layers.8.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.self_attn.q_proj. +Applying mask on model.layers.8.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.self_attn.k_proj. +Applying mask on model.layers.8.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.self_attn.v_proj. +Applying mask on model.layers.8.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.self_attn.o_proj. +Applying mask on model.layers.8.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.mlp.gate_proj. +Applying mask on model.layers.8.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.mlp.up_proj. +Applying mask on model.layers.8.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.mlp.down_proj. +Applying mask on model.layers.9.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.self_attn.q_proj. +Applying mask on model.layers.9.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.self_attn.k_proj. +Applying mask on model.layers.9.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.self_attn.v_proj. +Applying mask on model.layers.9.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.self_attn.o_proj. +Applying mask on model.layers.9.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.mlp.gate_proj. +Applying mask on model.layers.9.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.mlp.up_proj. +Applying mask on model.layers.9.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.mlp.down_proj. +Applying mask on model.layers.10.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.self_attn.q_proj. +Applying mask on model.layers.10.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.self_attn.k_proj. +Applying mask on model.layers.10.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.self_attn.v_proj. +Applying mask on model.layers.10.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.self_attn.o_proj. +Applying mask on model.layers.10.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.mlp.gate_proj. +Applying mask on model.layers.10.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.mlp.up_proj. +Applying mask on model.layers.10.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.mlp.down_proj. +Applying mask on model.layers.11.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.self_attn.q_proj. +Applying mask on model.layers.11.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.self_attn.k_proj. +Applying mask on model.layers.11.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.self_attn.v_proj. +Applying mask on model.layers.11.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.self_attn.o_proj. +Applying mask on model.layers.11.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.mlp.gate_proj. +Applying mask on model.layers.11.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.mlp.up_proj. +Applying mask on model.layers.11.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.mlp.down_proj. +Applying mask on model.layers.12.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.self_attn.q_proj. +Applying mask on model.layers.12.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.self_attn.k_proj. +Applying mask on model.layers.12.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.self_attn.v_proj. +Applying mask on model.layers.12.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.self_attn.o_proj. +Applying mask on model.layers.12.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.mlp.gate_proj. +Applying mask on model.layers.12.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.mlp.up_proj. +Applying mask on model.layers.12.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.mlp.down_proj. +Applying mask on model.layers.13.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.self_attn.q_proj. +Applying mask on model.layers.13.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.self_attn.k_proj. +Applying mask on model.layers.13.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.self_attn.v_proj. +Applying mask on model.layers.13.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.self_attn.o_proj. +Applying mask on model.layers.13.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.mlp.gate_proj. +Applying mask on model.layers.13.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.mlp.up_proj. +Applying mask on model.layers.13.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.mlp.down_proj. +Applying mask on model.layers.14.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.self_attn.q_proj. +Applying mask on model.layers.14.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.self_attn.k_proj. +Applying mask on model.layers.14.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.self_attn.v_proj. +Applying mask on model.layers.14.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.self_attn.o_proj. +Applying mask on model.layers.14.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.mlp.gate_proj. +Applying mask on model.layers.14.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.mlp.up_proj. +Applying mask on model.layers.14.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.mlp.down_proj. +Applying mask on model.layers.15.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.self_attn.q_proj. +Applying mask on model.layers.15.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.self_attn.k_proj. +Applying mask on model.layers.15.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.self_attn.v_proj. +Applying mask on model.layers.15.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.self_attn.o_proj. +Applying mask on model.layers.15.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.mlp.gate_proj. +Applying mask on model.layers.15.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.mlp.up_proj. +Applying mask on model.layers.15.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.mlp.down_proj. +Applying mask on model.layers.16.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.self_attn.q_proj. +Applying mask on model.layers.16.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.self_attn.k_proj. +Applying mask on model.layers.16.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.self_attn.v_proj. +Applying mask on model.layers.16.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.self_attn.o_proj. +Applying mask on model.layers.16.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.mlp.gate_proj. +Applying mask on model.layers.16.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.mlp.up_proj. +Applying mask on model.layers.16.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.mlp.down_proj. +Applying mask on model.layers.17.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.self_attn.q_proj. +Applying mask on model.layers.17.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.self_attn.k_proj. +Applying mask on model.layers.17.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.self_attn.v_proj. +Applying mask on model.layers.17.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.self_attn.o_proj. +Applying mask on model.layers.17.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.mlp.gate_proj. +Applying mask on model.layers.17.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.mlp.up_proj. +Applying mask on model.layers.17.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.mlp.down_proj. +Applying mask on model.layers.18.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.self_attn.q_proj. +Applying mask on model.layers.18.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.self_attn.k_proj. +Applying mask on model.layers.18.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.self_attn.v_proj. +Applying mask on model.layers.18.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.self_attn.o_proj. +Applying mask on model.layers.18.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.mlp.gate_proj. +Applying mask on model.layers.18.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.mlp.up_proj. +Applying mask on model.layers.18.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.mlp.down_proj. +Applying mask on model.layers.19.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.self_attn.q_proj. +Applying mask on model.layers.19.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.self_attn.k_proj. +Applying mask on model.layers.19.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.self_attn.v_proj. +Applying mask on model.layers.19.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.self_attn.o_proj. +Applying mask on model.layers.19.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.mlp.gate_proj. +Applying mask on model.layers.19.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.mlp.up_proj. +Applying mask on model.layers.19.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.mlp.down_proj. +Applying mask on model.layers.20.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.self_attn.q_proj. +Applying mask on model.layers.20.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.self_attn.k_proj. +Applying mask on model.layers.20.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.self_attn.v_proj. +Applying mask on model.layers.20.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.self_attn.o_proj. +Applying mask on model.layers.20.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.mlp.gate_proj. +Applying mask on model.layers.20.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.mlp.up_proj. +Applying mask on model.layers.20.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.mlp.down_proj. +Applying mask on model.layers.21.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.self_attn.q_proj. +Applying mask on model.layers.21.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.self_attn.k_proj. +Applying mask on model.layers.21.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.self_attn.v_proj. +Applying mask on model.layers.21.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.self_attn.o_proj. +Applying mask on model.layers.21.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.mlp.gate_proj. +Applying mask on model.layers.21.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.mlp.up_proj. +Applying mask on model.layers.21.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.mlp.down_proj. +Applying mask on model.layers.22.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.self_attn.q_proj. +Applying mask on model.layers.22.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.self_attn.k_proj. +Applying mask on model.layers.22.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.self_attn.v_proj. +Applying mask on model.layers.22.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.self_attn.o_proj. +Applying mask on model.layers.22.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.mlp.gate_proj. +Applying mask on model.layers.22.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.mlp.up_proj. +Applying mask on model.layers.22.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.mlp.down_proj. +Applying mask on model.layers.23.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.self_attn.q_proj. +Applying mask on model.layers.23.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.self_attn.k_proj. +Applying mask on model.layers.23.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.self_attn.v_proj. +Applying mask on model.layers.23.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.self_attn.o_proj. +Applying mask on model.layers.23.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.mlp.gate_proj. +Applying mask on model.layers.23.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.mlp.up_proj. +Applying mask on model.layers.23.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.mlp.down_proj. +Applying mask on _connector.0 with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on _connector.0. +Applying mask on _connector.2 with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on _connector.2. +Using cleaned config_mask (without mask parameters) for saving. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 18:08:38,618] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. + 0%| | 0/900 [00:00", + "resampler_hidden_size": 768, + "sparsity_connector": null, + "subnet_type_connector": "global", + "temperature_connector": 0.5, + "text_config": { + "_name_or_path": "Qwen/Qwen2.5-0.5B", + "architectures": [ + "Qwen2ForCausalLM" + ], + "backward_type": "normal", + "bos_token_id": 151643, + "eos_token_id": 151643, + "hidden_size": 896, + "intermediate_size": 4864, + "mask_type": "soft", + "masked_layers": "all", + "max_position_embeddings": 32768, + "max_window_layers": 24, + "model_type": "qwen2", + "num_attention_heads": 14, + "num_hidden_layers": 24, + "num_key_value_heads": 2, + "rope_theta": 1000000.0, + "sliding_window": 32768, + "subnet_mode": "both", + "subnet_type": "None", + "temperature_attn": 0.5, + "temperature_mlp": 0.5, + "tie_word_embeddings": true, + "torch_dtype": "bfloat16", + "use_mrope": false, + "use_sliding_window": false, + "vocab_size": 151936 + }, + "threshold_connector": null, + "tokenizer_model_max_length": 2048, + "tokenizer_name_or_path": "Qwen/Qwen2.5-0.5B", + "tokenizer_padding_side": "right", + "tokenizer_use_fast": false, + "torch_dtype": "bfloat16", + "transformers_version": "4.40.1", + "tune_type_connector": "full", + "tune_type_llm": "full", + "tune_type_vision_tower": "frozen", + "tune_vision_tower_from_layer": 0, + "use_cache": true, + "vision_config": { + "hidden_act": "gelu_pytorch_tanh", + "hidden_size": 1152, + "image_size": 384, + "intermediate_size": 4304, + "layer_norm_eps": 1e-06, + "model_name_or_path": "google/siglip-so400m-patch14-384", + "model_name_or_path2": "", + "model_type": "siglip_vision_model", + "num_attention_heads": 16, + "num_hidden_layers": 27, + "patch_size": 14 + }, + "vision_feature_layer": -2, + "vision_feature_select_strategy": "patch", + "vision_hidden_size": 1152, + "vision_model_name_or_path": "google/siglip-so400m-patch14-384", + "vision_model_name_or_path2": "", + "vocab_size": 151936 +} + +TinyLlavaForConditionalGeneration( + (language_model): Qwen2ForCausalLM( + (model): Qwen2Model( + (embed_tokens): Embedding(151936, 896) + (layers): ModuleList( + (0-23): 24 x Qwen2DecoderLayer( + (self_attn): Qwen2Attention( + (q_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=True) + (k_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=128, bias=True) + (v_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=128, bias=True) + (o_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=False) + (rotary_emb): Qwen2RotaryEmbedding() + ) + (mlp): Qwen2MLP( + (gate_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=4864, bias=False) + (up_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=4864, bias=False) + (down_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=4864, out_features=896, bias=False) + (act_fn): SiLU() + ) + (input_layernorm): Qwen2RMSNorm() + (post_attention_layernorm): Qwen2RMSNorm() + ) + ) + (norm): Qwen2RMSNorm() + ) + (lm_head): Linear(in_features=896, out_features=151936, bias=False) + ) + (vision_tower): SIGLIPVisionTower( + (_vision_tower): SiglipVisionModel( + (vision_model): SiglipVisionTransformer( + (embeddings): SiglipVisionEmbeddings( + (patch_embedding): Conv2d(3, 1152, kernel_size=(14, 14), stride=(14, 14), padding=valid) + (position_embedding): Embedding(729, 1152) + ) + (encoder): SiglipEncoder( + (layers): ModuleList( + (0-26): 27 x SiglipEncoderLayer( + (self_attn): SiglipAttention( + (k_proj): Linear(in_features=1152, out_features=1152, bias=True) + (v_proj): Linear(in_features=1152, out_features=1152, bias=True) + (q_proj): Linear(in_features=1152, out_features=1152, bias=True) + (out_proj): Linear(in_features=1152, out_features=1152, bias=True) + ) + (layer_norm1): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + (layer_norm2): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + ) + ) + ) + (post_layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (head): SiglipMultiheadAttentionPoolingHead( + (attention): MultiheadAttention( + (out_proj): NonDynamicallyQuantizableLinear(in_features=1152, out_features=1152, bias=True) + ) + (layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + ) + ) + ) + ) + (connector): MLPConnector( + (_connector): Sequential( + (0): SupermaskLinearSparsity_SoftForward_Normal(in_features=1152, out_features=896, bias=True) + (1): GELU(approximate='none') + (2): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=True) + ) + ) +) +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Load base model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain over. +TinyLlavaConfig { + "cache_dir": null, + "connector_type": "mlp2x_gelu", + "hidden_size": 896, + "ignore_index": -100, + "image_aspect_ratio": "square", + "image_token_index": -200, + "llm_model_name_or_path": "Qwen/Qwen2.5-0.5B", + "model_type": "tinyllava", + "num_queries": 128, + "num_resampler_layers": 3, + "pad_token": "<|endoftext|>", + "pad_token_id": 151643, + "resampler_hidden_size": 768, + "text_config": { + "_name_or_path": "Qwen/Qwen2.5-0.5B", + "architectures": [ + "Qwen2ForCausalLM" + ], + "bos_token_id": 151643, + "eos_token_id": 151643, + "hidden_size": 896, + "intermediate_size": 4864, + "max_position_embeddings": 32768, + "max_window_layers": 24, + "model_type": "qwen2", + "num_attention_heads": 14, + "num_hidden_layers": 24, + "num_key_value_heads": 2, + "rope_theta": 1000000.0, + "sliding_window": 32768, + "tie_word_embeddings": true, + "use_mrope": false, + "use_sliding_window": false, + "vocab_size": 151936 + }, + "tokenizer_model_max_length": 2048, + "tokenizer_name_or_path": "Qwen/Qwen2.5-0.5B", + "tokenizer_padding_side": "right", + "tokenizer_use_fast": false, + "transformers_version": "4.40.1", + "tune_type_connector": "full", + "tune_type_llm": "frozen", + "tune_type_vision_tower": "frozen", + "tune_vision_tower_from_layer": 0, + "use_cache": true, + "vision_config": { + "hidden_act": "gelu_pytorch_tanh", + "hidden_size": 1152, + "image_size": 384, + "intermediate_size": 4304, + "layer_norm_eps": 1e-06, + "model_name_or_path": "google/siglip-so400m-patch14-384", + "model_name_or_path2": "", + "model_type": "siglip_vision_model", + "num_attention_heads": 16, + "num_hidden_layers": 27, + "patch_size": 14 + }, + "vision_feature_layer": -2, + "vision_feature_select_strategy": "patch", + "vision_hidden_size": 1152, + "vision_model_name_or_path": "google/siglip-so400m-patch14-384", + "vision_model_name_or_path2": "", + "vocab_size": 151936 +} + +TinyLlavaForConditionalGeneration( + (language_model): Qwen2ForCausalLM( + (model): Qwen2Model( + (embed_tokens): Embedding(151936, 896) + (layers): ModuleList( + (0-23): 24 x Qwen2DecoderLayer( + (self_attn): Qwen2Attention( + (q_proj): Linear(in_features=896, out_features=896, bias=True) + (k_proj): Linear(in_features=896, out_features=128, bias=True) + (v_proj): Linear(in_features=896, out_features=128, bias=True) + (o_proj): Linear(in_features=896, out_features=896, bias=False) + (rotary_emb): Qwen2RotaryEmbedding() + ) + (mlp): Qwen2MLP( + (gate_proj): Linear(in_features=896, out_features=4864, bias=False) + (up_proj): Linear(in_features=896, out_features=4864, bias=False) + (down_proj): Linear(in_features=4864, out_features=896, bias=False) + (act_fn): SiLU() + ) + (input_layernorm): Qwen2RMSNorm() + (post_attention_layernorm): Qwen2RMSNorm() + ) + ) + (norm): Qwen2RMSNorm() + ) + (lm_head): Linear(in_features=896, out_features=151936, bias=False) + ) + (vision_tower): SIGLIPVisionTower( + (_vision_tower): SiglipVisionModel( + (vision_model): SiglipVisionTransformer( + (embeddings): SiglipVisionEmbeddings( + (patch_embedding): Conv2d(3, 1152, kernel_size=(14, 14), stride=(14, 14), padding=valid) + (position_embedding): Embedding(729, 1152) + ) + (encoder): SiglipEncoder( + (layers): ModuleList( + (0-26): 27 x SiglipEncoderLayer( + (self_attn): SiglipAttention( + (k_proj): Linear(in_features=1152, out_features=1152, bias=True) + (v_proj): Linear(in_features=1152, out_features=1152, bias=True) + (q_proj): Linear(in_features=1152, out_features=1152, bias=True) + (out_proj): Linear(in_features=1152, out_features=1152, bias=True) + ) + (layer_norm1): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + (layer_norm2): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + ) + ) + ) + (post_layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (head): SiglipMultiheadAttentionPoolingHead( + (attention): MultiheadAttention( + (out_proj): NonDynamicallyQuantizableLinear(in_features=1152, out_features=1152, bias=True) + ) + (layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + ) + ) + ) + ) + (connector): MLPConnector( + (_connector): Sequential( + (0): Linear(in_features=1152, out_features=896, bias=True) + (1): GELU(approximate='none') + (2): Linear(in_features=896, out_features=896, bias=True) + ) + ) +) +Collect masks for language model over. +Collect masks for connector over. +Applying mask on model.layers.0.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.self_attn.q_proj. +Applying mask on model.layers.0.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.self_attn.k_proj. +Applying mask on model.layers.0.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.self_attn.v_proj. +Applying mask on model.layers.0.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.self_attn.o_proj. +Applying mask on model.layers.0.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.mlp.gate_proj. +Applying mask on model.layers.0.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.mlp.up_proj. +Applying mask on model.layers.0.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.mlp.down_proj. +Applying mask on model.layers.1.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.self_attn.q_proj. +Applying mask on model.layers.1.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.self_attn.k_proj. +Applying mask on model.layers.1.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.self_attn.v_proj. +Applying mask on model.layers.1.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.self_attn.o_proj. +Applying mask on model.layers.1.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.mlp.gate_proj. +Applying mask on model.layers.1.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.mlp.up_proj. +Applying mask on model.layers.1.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.mlp.down_proj. +Applying mask on model.layers.2.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.self_attn.q_proj. +Applying mask on model.layers.2.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.self_attn.k_proj. +Applying mask on model.layers.2.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.self_attn.v_proj. +Applying mask on model.layers.2.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.self_attn.o_proj. +Applying mask on model.layers.2.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.mlp.gate_proj. +Applying mask on model.layers.2.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.mlp.up_proj. +Applying mask on model.layers.2.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.mlp.down_proj. +Applying mask on model.layers.3.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.self_attn.q_proj. +Applying mask on model.layers.3.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.self_attn.k_proj. +Applying mask on model.layers.3.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.self_attn.v_proj. +Applying mask on model.layers.3.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.self_attn.o_proj. +Applying mask on model.layers.3.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.mlp.gate_proj. +Applying mask on model.layers.3.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.mlp.up_proj. +Applying mask on model.layers.3.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.mlp.down_proj. +Applying mask on model.layers.4.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.self_attn.q_proj. +Applying mask on model.layers.4.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.self_attn.k_proj. +Applying mask on model.layers.4.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.self_attn.v_proj. +Applying mask on model.layers.4.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.self_attn.o_proj. +Applying mask on model.layers.4.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.mlp.gate_proj. +Applying mask on model.layers.4.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.mlp.up_proj. +Applying mask on model.layers.4.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.mlp.down_proj. +Applying mask on model.layers.5.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.self_attn.q_proj. +Applying mask on model.layers.5.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.self_attn.k_proj. +Applying mask on model.layers.5.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.self_attn.v_proj. +Applying mask on model.layers.5.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.self_attn.o_proj. +Applying mask on model.layers.5.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.mlp.gate_proj. +Applying mask on model.layers.5.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.mlp.up_proj. +Applying mask on model.layers.5.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.mlp.down_proj. +Applying mask on model.layers.6.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.self_attn.q_proj. +Applying mask on model.layers.6.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.self_attn.k_proj. +Applying mask on model.layers.6.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.self_attn.v_proj. +Applying mask on model.layers.6.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.self_attn.o_proj. +Applying mask on model.layers.6.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.mlp.gate_proj. +Applying mask on model.layers.6.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.mlp.up_proj. +Applying mask on model.layers.6.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.mlp.down_proj. +Applying mask on model.layers.7.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.self_attn.q_proj. +Applying mask on model.layers.7.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.self_attn.k_proj. +Applying mask on model.layers.7.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.self_attn.v_proj. +Applying mask on model.layers.7.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.self_attn.o_proj. +Applying mask on model.layers.7.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.mlp.gate_proj. +Applying mask on model.layers.7.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.mlp.up_proj. +Applying mask on model.layers.7.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.mlp.down_proj. +Applying mask on model.layers.8.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.self_attn.q_proj. +Applying mask on model.layers.8.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.self_attn.k_proj. +Applying mask on model.layers.8.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.self_attn.v_proj. +Applying mask on model.layers.8.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.self_attn.o_proj. +Applying mask on model.layers.8.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.mlp.gate_proj. +Applying mask on model.layers.8.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.mlp.up_proj. +Applying mask on model.layers.8.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.mlp.down_proj. +Applying mask on model.layers.9.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.self_attn.q_proj. +Applying mask on model.layers.9.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.self_attn.k_proj. +Applying mask on model.layers.9.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.self_attn.v_proj. +Applying mask on model.layers.9.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.self_attn.o_proj. +Applying mask on model.layers.9.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.mlp.gate_proj. +Applying mask on model.layers.9.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.mlp.up_proj. +Applying mask on model.layers.9.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.mlp.down_proj. +Applying mask on model.layers.10.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.self_attn.q_proj. +Applying mask on model.layers.10.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.self_attn.k_proj. +Applying mask on model.layers.10.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.self_attn.v_proj. +Applying mask on model.layers.10.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.self_attn.o_proj. +Applying mask on model.layers.10.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.mlp.gate_proj. +Applying mask on model.layers.10.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.mlp.up_proj. +Applying mask on model.layers.10.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.mlp.down_proj. +Applying mask on model.layers.11.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.self_attn.q_proj. +Applying mask on model.layers.11.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.self_attn.k_proj. +Applying mask on model.layers.11.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.self_attn.v_proj. +Applying mask on model.layers.11.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.self_attn.o_proj. +Applying mask on model.layers.11.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.mlp.gate_proj. +Applying mask on model.layers.11.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.mlp.up_proj. +Applying mask on model.layers.11.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.mlp.down_proj. +Applying mask on model.layers.12.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.self_attn.q_proj. +Applying mask on model.layers.12.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.self_attn.k_proj. +Applying mask on model.layers.12.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.self_attn.v_proj. +Applying mask on model.layers.12.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.self_attn.o_proj. +Applying mask on model.layers.12.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.mlp.gate_proj. +Applying mask on model.layers.12.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.mlp.up_proj. +Applying mask on model.layers.12.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.mlp.down_proj. +Applying mask on model.layers.13.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.self_attn.q_proj. +Applying mask on model.layers.13.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.self_attn.k_proj. +Applying mask on model.layers.13.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.self_attn.v_proj. +Applying mask on model.layers.13.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.self_attn.o_proj. +Applying mask on model.layers.13.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.mlp.gate_proj. +Applying mask on model.layers.13.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.mlp.up_proj. +Applying mask on model.layers.13.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.mlp.down_proj. +Applying mask on model.layers.14.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.self_attn.q_proj. +Applying mask on model.layers.14.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.self_attn.k_proj. +Applying mask on model.layers.14.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.self_attn.v_proj. +Applying mask on model.layers.14.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.self_attn.o_proj. +Applying mask on model.layers.14.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.mlp.gate_proj. +Applying mask on model.layers.14.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.mlp.up_proj. +Applying mask on model.layers.14.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.mlp.down_proj. +Applying mask on model.layers.15.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.self_attn.q_proj. +Applying mask on model.layers.15.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.self_attn.k_proj. +Applying mask on model.layers.15.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.self_attn.v_proj. +Applying mask on model.layers.15.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.self_attn.o_proj. +Applying mask on model.layers.15.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.mlp.gate_proj. +Applying mask on model.layers.15.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.mlp.up_proj. +Applying mask on model.layers.15.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.mlp.down_proj. +Applying mask on model.layers.16.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.self_attn.q_proj. +Applying mask on model.layers.16.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.self_attn.k_proj. +Applying mask on model.layers.16.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.self_attn.v_proj. +Applying mask on model.layers.16.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.self_attn.o_proj. +Applying mask on model.layers.16.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.mlp.gate_proj. +Applying mask on model.layers.16.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.mlp.up_proj. +Applying mask on model.layers.16.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.mlp.down_proj. +Applying mask on model.layers.17.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.self_attn.q_proj. +Applying mask on model.layers.17.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.self_attn.k_proj. +Applying mask on model.layers.17.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.self_attn.v_proj. +Applying mask on model.layers.17.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.self_attn.o_proj. +Applying mask on model.layers.17.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.mlp.gate_proj. +Applying mask on model.layers.17.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.mlp.up_proj. +Applying mask on model.layers.17.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.mlp.down_proj. +Applying mask on model.layers.18.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.self_attn.q_proj. +Applying mask on model.layers.18.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.self_attn.k_proj. +Applying mask on model.layers.18.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.self_attn.v_proj. +Applying mask on model.layers.18.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.self_attn.o_proj. +Applying mask on model.layers.18.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.mlp.gate_proj. +Applying mask on model.layers.18.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.mlp.up_proj. +Applying mask on model.layers.18.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.mlp.down_proj. +Applying mask on model.layers.19.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.self_attn.q_proj. +Applying mask on model.layers.19.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.self_attn.k_proj. +Applying mask on model.layers.19.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.self_attn.v_proj. +Applying mask on model.layers.19.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.self_attn.o_proj. +Applying mask on model.layers.19.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.mlp.gate_proj. +Applying mask on model.layers.19.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.mlp.up_proj. +Applying mask on model.layers.19.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.mlp.down_proj. +Applying mask on model.layers.20.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.self_attn.q_proj. +Applying mask on model.layers.20.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.self_attn.k_proj. +Applying mask on model.layers.20.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.self_attn.v_proj. +Applying mask on model.layers.20.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.self_attn.o_proj. +Applying mask on model.layers.20.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.mlp.gate_proj. +Applying mask on model.layers.20.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.mlp.up_proj. +Applying mask on model.layers.20.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.mlp.down_proj. +Applying mask on model.layers.21.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.self_attn.q_proj. +Applying mask on model.layers.21.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.self_attn.k_proj. +Applying mask on model.layers.21.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.self_attn.v_proj. +Applying mask on model.layers.21.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.self_attn.o_proj. +Applying mask on model.layers.21.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.mlp.gate_proj. +Applying mask on model.layers.21.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.mlp.up_proj. +Applying mask on model.layers.21.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.mlp.down_proj. +Applying mask on model.layers.22.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.self_attn.q_proj. +Applying mask on model.layers.22.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.self_attn.k_proj. +Applying mask on model.layers.22.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.self_attn.v_proj. +Applying mask on model.layers.22.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.self_attn.o_proj. +Applying mask on model.layers.22.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.mlp.gate_proj. +Applying mask on model.layers.22.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.mlp.up_proj. +Applying mask on model.layers.22.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.mlp.down_proj. +Applying mask on model.layers.23.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.self_attn.q_proj. +Applying mask on model.layers.23.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.self_attn.k_proj. +Applying mask on model.layers.23.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.self_attn.v_proj. +Applying mask on model.layers.23.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.self_attn.o_proj. +Applying mask on model.layers.23.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.mlp.gate_proj. +Applying mask on model.layers.23.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.mlp.up_proj. +Applying mask on model.layers.23.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.mlp.down_proj. +Applying mask on _connector.0 with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on _connector.0. +Applying mask on _connector.2 with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on _connector.2. +Using cleaned config_mask (without mask parameters) for saving. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 18:34:03,389] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. + 0%| | 0/900 [00:00 + main() + File "/nfs/ywang29/TinyLLaVA/scripts/apply_masks.py", line 123, in main + config_mask = TinyLlavaConfig.from_pretrained(model_args.mask_model_name_or_path) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/configuration_utils.py", line 602, in from_pretrained + config_dict, kwargs = cls.get_config_dict(pretrained_model_name_or_path, **kwargs) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/configuration_utils.py", line 631, in get_config_dict + config_dict, kwargs = cls._get_config_dict(pretrained_model_name_or_path, **kwargs) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/configuration_utils.py", line 686, in _get_config_dict + resolved_config_file = cached_file( + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/utils/hub.py", line 369, in cached_file + raise EnvironmentError( +OSError: /nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_5e-2_connector-3.0_0.5_5e-2_ablation does not appear to have a file named config.json. Checkout 'https://huggingface.co//nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_5e-2_connector-3.0_0.5_5e-2_ablation/tree/main' for available files. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 19:01:34,395] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +Traceback (most recent call last): + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/utils/hub.py", line 398, in cached_file + resolved_file = hf_hub_download( + File "/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/utils/_validators.py", line 106, in _inner_fn + validate_repo_id(arg_value) + File "/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/utils/_validators.py", line 154, in validate_repo_id + raise HFValidationError( +huggingface_hub.errors.HFValidationError: Repo id must be in the form 'repo_name' or 'namespace/repo_name': '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_5e-2_connector-3.0_0.5_5e-2_ablation/mask_applied'. Use `repo_type` argument if needed. + +The above exception was the direct cause of the following exception: + +Traceback (most recent call last): + File "/nfs/ywang29/TinyLLaVA/tinyllava/model/load_model.py", line 38, in load_pretrained_model + model = TinyLlavaForConditionalGeneration.from_pretrained(model_name_or_path,low_cpu_mem_usage=True) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/modeling_utils.py", line 3015, in from_pretrained + resolved_config_file = cached_file( + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/utils/hub.py", line 462, in cached_file + raise EnvironmentError( +OSError: Incorrect path_or_model_id: '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_5e-2_connector-3.0_0.5_5e-2_ablation/mask_applied'. Please provide either the path to a local folder or the repo_id of a model on the Hub. + +During handling of the above exception, another exception occurred: + +Traceback (most recent call last): + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/utils/hub.py", line 398, in cached_file + resolved_file = hf_hub_download( + File "/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/utils/_validators.py", line 106, in _inner_fn + validate_repo_id(arg_value) + File "/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/utils/_validators.py", line 154, in validate_repo_id + raise HFValidationError( +huggingface_hub.errors.HFValidationError: Repo id must be in the form 'repo_name' or 'namespace/repo_name': '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_5e-2_connector-3.0_0.5_5e-2_ablation/mask_applied'. Use `repo_type` argument if needed. + +The above exception was the direct cause of the following exception: + +Traceback (most recent call last): + File "/opt/conda/envs/tinyllava/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/opt/conda/envs/tinyllava/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/nfs/ywang29/TinyLLaVA/tinyllava/eval/model_vqa_mmmu.py", line 180, in + eval_model(args) + File "/nfs/ywang29/TinyLLaVA/tinyllava/eval/model_vqa_mmmu.py", line 88, in eval_model + model, tokenizer, image_processor, context_len = load_pretrained_model(model_path) + File "/nfs/ywang29/TinyLLaVA/tinyllava/model/load_model.py", line 40, in load_pretrained_model + model_config = TinyLlavaConfig.from_pretrained(model_name_or_path) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/configuration_utils.py", line 602, in from_pretrained + config_dict, kwargs = cls.get_config_dict(pretrained_model_name_or_path, **kwargs) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/configuration_utils.py", line 631, in get_config_dict + config_dict, kwargs = cls._get_config_dict(pretrained_model_name_or_path, **kwargs) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/configuration_utils.py", line 686, in _get_config_dict + resolved_config_file = cached_file( + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/utils/hub.py", line 462, in cached_file + raise EnvironmentError( +OSError: Incorrect path_or_model_id: '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_5e-2_connector-3.0_0.5_5e-2_ablation/mask_applied'. Please provide either the path to a local folder or the repo_id of a model on the Hub. +Traceback (most recent call last): + File "/nfs/ywang29/TinyLLaVA/scripts/convert_answer_to_mmmu.py", line 31, in + eval_model(args) + File "/nfs/ywang29/TinyLLaVA/scripts/convert_answer_to_mmmu.py", line 7, in eval_model + answers = [json.loads(q) for q in open(os.path.expanduser(args.answers_file), "r")] +FileNotFoundError: [Errno 2] No such file or directory: '/s3-code/ywang29/datasets/tinyllava/eval/MMMU/answers/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_5e-2_connector-3.0_0.5_5e-2_ablation-mask_applied.jsonl' +Traceback (most recent call last): + File "/s3-code/ywang29/datasets/tinyllava/eval/MMMU/eval/main_eval_only.py", line 19, in + output_dict = json.load(open(args.output_path)) +FileNotFoundError: [Errno 2] No such file or directory: '/s3-code/ywang29/datasets/tinyllava/eval/MMMU/answers/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_5e-2_connector-3.0_0.5_5e-2_ablation-mask_applied_output.json' +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 19:01:40,892] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +Traceback (most recent call last): + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/utils/hub.py", line 398, in cached_file + resolved_file = hf_hub_download( + File "/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/utils/_validators.py", line 106, in _inner_fn + validate_repo_id(arg_value) + File "/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/utils/_validators.py", line 154, in validate_repo_id + raise HFValidationError( +huggingface_hub.errors.HFValidationError: Repo id must be in the form 'repo_name' or 'namespace/repo_name': '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_5e-2_connector-3.0_0.5_5e-2_ablation/mask_applied'. Use `repo_type` argument if needed. + +The above exception was the direct cause of the following exception: + +Traceback (most recent call last): + File "/nfs/ywang29/TinyLLaVA/tinyllava/model/load_model.py", line 38, in load_pretrained_model + model = TinyLlavaForConditionalGeneration.from_pretrained(model_name_or_path,low_cpu_mem_usage=True) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/modeling_utils.py", line 3015, in from_pretrained + resolved_config_file = cached_file( + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/utils/hub.py", line 462, in cached_file + raise EnvironmentError( +OSError: Incorrect path_or_model_id: '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_5e-2_connector-3.0_0.5_5e-2_ablation/mask_applied'. Please provide either the path to a local folder or the repo_id of a model on the Hub. + +During handling of the above exception, another exception occurred: + +Traceback (most recent call last): + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/utils/hub.py", line 398, in cached_file + resolved_file = hf_hub_download( + File "/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/utils/_validators.py", line 106, in _inner_fn + validate_repo_id(arg_value) + File "/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/utils/_validators.py", line 154, in validate_repo_id + raise HFValidationError( +huggingface_hub.errors.HFValidationError: Repo id must be in the form 'repo_name' or 'namespace/repo_name': '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_5e-2_connector-3.0_0.5_5e-2_ablation/mask_applied'. Use `repo_type` argument if needed. + +The above exception was the direct cause of the following exception: + +Traceback (most recent call last): + File "/opt/conda/envs/tinyllava/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/opt/conda/envs/tinyllava/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/nfs/ywang29/TinyLLaVA/tinyllava/eval/model_vqa_loader.py", line 146, in + eval_model(args) + File "/nfs/ywang29/TinyLLaVA/tinyllava/eval/model_vqa_loader.py", line 79, in eval_model + model, tokenizer, image_processor, context_len = load_pretrained_model(model_path) + File "/nfs/ywang29/TinyLLaVA/tinyllava/model/load_model.py", line 40, in load_pretrained_model + model_config = TinyLlavaConfig.from_pretrained(model_name_or_path) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/configuration_utils.py", line 602, in from_pretrained + config_dict, kwargs = cls.get_config_dict(pretrained_model_name_or_path, **kwargs) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/configuration_utils.py", line 631, in get_config_dict + config_dict, kwargs = cls._get_config_dict(pretrained_model_name_or_path, **kwargs) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/configuration_utils.py", line 686, in _get_config_dict + resolved_config_file = cached_file( + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/utils/hub.py", line 462, in cached_file + raise EnvironmentError( +OSError: Incorrect path_or_model_id: '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_5e-2_connector-3.0_0.5_5e-2_ablation/mask_applied'. Please provide either the path to a local folder or the repo_id of a model on the Hub. +Traceback (most recent call last): + File "/s3-code/ywang29/datasets/tinyllava/eval/MME/convert_answer_to_mme.py", line 52, in + answers = [json.loads(line) for line in open(os.path.join('answers', f'{experiment}.jsonl'))] +FileNotFoundError: [Errno 2] No such file or directory: 'answers/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_5e-2_connector-3.0_0.5_5e-2_ablation-mask_applied.jsonl' +=========== Perception =========== +Traceback (most recent call last): + File "/s3-code/ywang29/datasets/tinyllava/eval/MME/eval_tool/calculation.py", line 164, in + cal.process_result(results_dir) + File "/s3-code/ywang29/datasets/tinyllava/eval/MME/eval_tool/calculation.py", line 98, in process_result + lines = open(task_txt, 'r').readlines() +FileNotFoundError: [Errno 2] No such file or directory: 'answers/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_5e-2_connector-3.0_0.5_5e-2_ablation-mask_applied/existence.txt' +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 19:02:46,283] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +Traceback (most recent call last): + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/utils/hub.py", line 398, in cached_file + resolved_file = hf_hub_download( + File "/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/utils/_validators.py", line 106, in _inner_fn + validate_repo_id(arg_value) + File "/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/utils/_validators.py", line 154, in validate_repo_id + raise HFValidationError( +huggingface_hub.errors.HFValidationError: Repo id must be in the form 'repo_name' or 'namespace/repo_name': '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_5e-2_connector-3.0_0.5_5e-2_ablation/mask_applied'. Use `repo_type` argument if needed. + +The above exception was the direct cause of the following exception: + +Traceback (most recent call last): + File "/nfs/ywang29/TinyLLaVA/tinyllava/model/load_model.py", line 38, in load_pretrained_model + model = TinyLlavaForConditionalGeneration.from_pretrained(model_name_or_path,low_cpu_mem_usage=True) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/modeling_utils.py", line 3015, in from_pretrained + resolved_config_file = cached_file( + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/utils/hub.py", line 462, in cached_file + raise EnvironmentError( +OSError: Incorrect path_or_model_id: '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_5e-2_connector-3.0_0.5_5e-2_ablation/mask_applied'. Please provide either the path to a local folder or the repo_id of a model on the Hub. + +During handling of the above exception, another exception occurred: + +Traceback (most recent call last): + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/utils/hub.py", line 398, in cached_file + resolved_file = hf_hub_download( + File "/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/utils/_validators.py", line 106, in _inner_fn + validate_repo_id(arg_value) + File "/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/utils/_validators.py", line 154, in validate_repo_id + raise HFValidationError( +huggingface_hub.errors.HFValidationError: Repo id must be in the form 'repo_name' or 'namespace/repo_name': '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_5e-2_connector-3.0_0.5_5e-2_ablation/mask_applied'. Use `repo_type` argument if needed. + +The above exception was the direct cause of the following exception: + +Traceback (most recent call last): + File "/opt/conda/envs/tinyllava/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/opt/conda/envs/tinyllava/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/nfs/ywang29/TinyLLaVA/tinyllava/eval/model_vqa_science.py", line 108, in + eval_model(args) + File "/nfs/ywang29/TinyLLaVA/tinyllava/eval/model_vqa_science.py", line 31, in eval_model + model, tokenizer, image_processor, context_len = load_pretrained_model(model_path) + File "/nfs/ywang29/TinyLLaVA/tinyllava/model/load_model.py", line 40, in load_pretrained_model + model_config = TinyLlavaConfig.from_pretrained(model_name_or_path) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/configuration_utils.py", line 602, in from_pretrained + config_dict, kwargs = cls.get_config_dict(pretrained_model_name_or_path, **kwargs) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/configuration_utils.py", line 631, in get_config_dict + config_dict, kwargs = cls._get_config_dict(pretrained_model_name_or_path, **kwargs) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/configuration_utils.py", line 686, in _get_config_dict + resolved_config_file = cached_file( + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/utils/hub.py", line 462, in cached_file + raise EnvironmentError( +OSError: Incorrect path_or_model_id: '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_5e-2_connector-3.0_0.5_5e-2_ablation/mask_applied'. Please provide either the path to a local folder or the repo_id of a model on the Hub. +Traceback (most recent call last): + File "/nfs/ywang29/TinyLLaVA/tinyllava/eval/eval_science_qa.py", line 45, in + predictions = [json.loads(line) for line in open(args.result_file)] +FileNotFoundError: [Errno 2] No such file or directory: '/s3-code/ywang29/datasets/tinyllava/eval/scienceqa/answers/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_5e-2_connector-3.0_0.5_5e-2_ablation-mask_applied.jsonl' +==== EXPERIMENT COMPLETED: eval_qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_5e-2_connector-3.0_0.5_5e-2_ablation ==== +Log File: eval_qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_5e-2_connector-3.0_0.5_5e-2_ablation_20251012_190124.log +Timestamp: 2025-10-12 19:02:49 +===================================== diff --git a/logs_oct11/eval_qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_7e-1_connector-3.0_0.5_7e-1_ablation_20251012_190249.log b/logs_oct11/eval_qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_7e-1_connector-3.0_0.5_7e-1_ablation_20251012_190249.log new file mode 100644 index 0000000000000000000000000000000000000000..d279e38f0e904eb8940746cab0e343be906bcdc1 --- /dev/null +++ b/logs_oct11/eval_qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_7e-1_connector-3.0_0.5_7e-1_ablation_20251012_190249.log @@ -0,0 +1,759 @@ +==== STARTING EXPERIMENT: eval_qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_7e-1_connector-3.0_0.5_7e-1_ablation ==== +Log File: eval_qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_7e-1_connector-3.0_0.5_7e-1_ablation_20251012_190249.log +Timestamp: 2025-10-12 19:02:49 +===================================== +Processing: /nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_7e-1_connector-3.0_0.5_7e-1_ablation +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 19:02:52,484] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +config_mask.torch_dtype: torch.bfloat16 +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +Load mask model from /nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_7e-1_connector-3.0_0.5_7e-1_ablation over. +TinyLlavaConfig { + "architectures": [ + "TinyLlavaForConditionalGeneration" + ], + "backward_type_connector": "normal", + "cache_dir": null, + "connector_type": "mlp2x_gelu", + "hidden_size": 896, + "ignore_index": -100, + "image_aspect_ratio": "square", + "image_token_index": -200, + "llm_model_name_or_path": "Qwen/Qwen2.5-0.5B", + "mask_model": [ + "llm", + "connector" + ], + "mask_type_connector": "soft", + "model_type": "tinyllava", + "num_queries": 128, + "num_resampler_layers": 3, + "pad_token": "<|endoftext|>", + "resampler_hidden_size": 768, + "sparsity_connector": null, + "subnet_type_connector": "global", + "temperature_connector": 0.5, + "text_config": { + "_name_or_path": "Qwen/Qwen2.5-0.5B", + "architectures": [ + "Qwen2ForCausalLM" + ], + "backward_type": "normal", + "bos_token_id": 151643, + "eos_token_id": 151643, + "hidden_size": 896, + "intermediate_size": 4864, + "mask_type": "soft", + "masked_layers": "all", + "max_position_embeddings": 32768, + "max_window_layers": 24, + "model_type": "qwen2", + "num_attention_heads": 14, + "num_hidden_layers": 24, + "num_key_value_heads": 2, + "rope_theta": 1000000.0, + "sliding_window": 32768, + "subnet_mode": "both", + "subnet_type": "None", + "temperature_attn": 0.5, + "temperature_mlp": 0.5, + "tie_word_embeddings": true, + "torch_dtype": "bfloat16", + "use_mrope": false, + "use_sliding_window": false, + "vocab_size": 151936 + }, + "threshold_connector": null, + "tokenizer_model_max_length": 2048, + "tokenizer_name_or_path": "Qwen/Qwen2.5-0.5B", + "tokenizer_padding_side": "right", + "tokenizer_use_fast": false, + "torch_dtype": "bfloat16", + "transformers_version": "4.40.1", + "tune_type_connector": "full", + "tune_type_llm": "full", + "tune_type_vision_tower": "frozen", + "tune_vision_tower_from_layer": 0, + "use_cache": true, + "vision_config": { + "hidden_act": "gelu_pytorch_tanh", + "hidden_size": 1152, + "image_size": 384, + "intermediate_size": 4304, + "layer_norm_eps": 1e-06, + "model_name_or_path": "google/siglip-so400m-patch14-384", + "model_name_or_path2": "", + "model_type": "siglip_vision_model", + "num_attention_heads": 16, + "num_hidden_layers": 27, + "patch_size": 14 + }, + "vision_feature_layer": -2, + "vision_feature_select_strategy": "patch", + "vision_hidden_size": 1152, + "vision_model_name_or_path": "google/siglip-so400m-patch14-384", + "vision_model_name_or_path2": "", + "vocab_size": 151936 +} + +TinyLlavaForConditionalGeneration( + (language_model): Qwen2ForCausalLM( + (model): Qwen2Model( + (embed_tokens): Embedding(151936, 896) + (layers): ModuleList( + (0-23): 24 x Qwen2DecoderLayer( + (self_attn): Qwen2Attention( + (q_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=True) + (k_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=128, bias=True) + (v_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=128, bias=True) + (o_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=False) + (rotary_emb): Qwen2RotaryEmbedding() + ) + (mlp): Qwen2MLP( + (gate_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=4864, bias=False) + (up_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=4864, bias=False) + (down_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=4864, out_features=896, bias=False) + (act_fn): SiLU() + ) + (input_layernorm): Qwen2RMSNorm() + (post_attention_layernorm): Qwen2RMSNorm() + ) + ) + (norm): Qwen2RMSNorm() + ) + (lm_head): Linear(in_features=896, out_features=151936, bias=False) + ) + (vision_tower): SIGLIPVisionTower( + (_vision_tower): SiglipVisionModel( + (vision_model): SiglipVisionTransformer( + (embeddings): SiglipVisionEmbeddings( + (patch_embedding): Conv2d(3, 1152, kernel_size=(14, 14), stride=(14, 14), padding=valid) + (position_embedding): Embedding(729, 1152) + ) + (encoder): SiglipEncoder( + (layers): ModuleList( + (0-26): 27 x SiglipEncoderLayer( + (self_attn): SiglipAttention( + (k_proj): Linear(in_features=1152, out_features=1152, bias=True) + (v_proj): Linear(in_features=1152, out_features=1152, bias=True) + (q_proj): Linear(in_features=1152, out_features=1152, bias=True) + (out_proj): Linear(in_features=1152, out_features=1152, bias=True) + ) + (layer_norm1): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + (layer_norm2): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + ) + ) + ) + (post_layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (head): SiglipMultiheadAttentionPoolingHead( + (attention): MultiheadAttention( + (out_proj): NonDynamicallyQuantizableLinear(in_features=1152, out_features=1152, bias=True) + ) + (layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + ) + ) + ) + ) + (connector): MLPConnector( + (_connector): Sequential( + (0): SupermaskLinearSparsity_SoftForward_Normal(in_features=1152, out_features=896, bias=True) + (1): GELU(approximate='none') + (2): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=True) + ) + ) +) +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Load base model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain over. +TinyLlavaConfig { + "cache_dir": null, + "connector_type": "mlp2x_gelu", + "hidden_size": 896, + "ignore_index": -100, + "image_aspect_ratio": "square", + "image_token_index": -200, + "llm_model_name_or_path": "Qwen/Qwen2.5-0.5B", + "model_type": "tinyllava", + "num_queries": 128, + "num_resampler_layers": 3, + "pad_token": "<|endoftext|>", + "pad_token_id": 151643, + "resampler_hidden_size": 768, + "text_config": { + "_name_or_path": "Qwen/Qwen2.5-0.5B", + "architectures": [ + "Qwen2ForCausalLM" + ], + "bos_token_id": 151643, + "eos_token_id": 151643, + "hidden_size": 896, + "intermediate_size": 4864, + "max_position_embeddings": 32768, + "max_window_layers": 24, + "model_type": "qwen2", + "num_attention_heads": 14, + "num_hidden_layers": 24, + "num_key_value_heads": 2, + "rope_theta": 1000000.0, + "sliding_window": 32768, + "tie_word_embeddings": true, + "use_mrope": false, + "use_sliding_window": false, + "vocab_size": 151936 + }, + "tokenizer_model_max_length": 2048, + "tokenizer_name_or_path": "Qwen/Qwen2.5-0.5B", + "tokenizer_padding_side": "right", + "tokenizer_use_fast": false, + "transformers_version": "4.40.1", + "tune_type_connector": "full", + "tune_type_llm": "frozen", + "tune_type_vision_tower": "frozen", + "tune_vision_tower_from_layer": 0, + "use_cache": true, + "vision_config": { + "hidden_act": "gelu_pytorch_tanh", + "hidden_size": 1152, + "image_size": 384, + "intermediate_size": 4304, + "layer_norm_eps": 1e-06, + "model_name_or_path": "google/siglip-so400m-patch14-384", + "model_name_or_path2": "", + "model_type": "siglip_vision_model", + "num_attention_heads": 16, + "num_hidden_layers": 27, + "patch_size": 14 + }, + "vision_feature_layer": -2, + "vision_feature_select_strategy": "patch", + "vision_hidden_size": 1152, + "vision_model_name_or_path": "google/siglip-so400m-patch14-384", + "vision_model_name_or_path2": "", + "vocab_size": 151936 +} + +TinyLlavaForConditionalGeneration( + (language_model): Qwen2ForCausalLM( + (model): Qwen2Model( + (embed_tokens): Embedding(151936, 896) + (layers): ModuleList( + (0-23): 24 x Qwen2DecoderLayer( + (self_attn): Qwen2Attention( + (q_proj): Linear(in_features=896, out_features=896, bias=True) + (k_proj): Linear(in_features=896, out_features=128, bias=True) + (v_proj): Linear(in_features=896, out_features=128, bias=True) + (o_proj): Linear(in_features=896, out_features=896, bias=False) + (rotary_emb): Qwen2RotaryEmbedding() + ) + (mlp): Qwen2MLP( + (gate_proj): Linear(in_features=896, out_features=4864, bias=False) + (up_proj): Linear(in_features=896, out_features=4864, bias=False) + (down_proj): Linear(in_features=4864, out_features=896, bias=False) + (act_fn): SiLU() + ) + (input_layernorm): Qwen2RMSNorm() + (post_attention_layernorm): Qwen2RMSNorm() + ) + ) + (norm): Qwen2RMSNorm() + ) + (lm_head): Linear(in_features=896, out_features=151936, bias=False) + ) + (vision_tower): SIGLIPVisionTower( + (_vision_tower): SiglipVisionModel( + (vision_model): SiglipVisionTransformer( + (embeddings): SiglipVisionEmbeddings( + (patch_embedding): Conv2d(3, 1152, kernel_size=(14, 14), stride=(14, 14), padding=valid) + (position_embedding): Embedding(729, 1152) + ) + (encoder): SiglipEncoder( + (layers): ModuleList( + (0-26): 27 x SiglipEncoderLayer( + (self_attn): SiglipAttention( + (k_proj): Linear(in_features=1152, out_features=1152, bias=True) + (v_proj): Linear(in_features=1152, out_features=1152, bias=True) + (q_proj): Linear(in_features=1152, out_features=1152, bias=True) + (out_proj): Linear(in_features=1152, out_features=1152, bias=True) + ) + (layer_norm1): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + (layer_norm2): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + ) + ) + ) + (post_layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (head): SiglipMultiheadAttentionPoolingHead( + (attention): MultiheadAttention( + (out_proj): NonDynamicallyQuantizableLinear(in_features=1152, out_features=1152, bias=True) + ) + (layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + ) + ) + ) + ) + (connector): MLPConnector( + (_connector): Sequential( + (0): Linear(in_features=1152, out_features=896, bias=True) + (1): GELU(approximate='none') + (2): Linear(in_features=896, out_features=896, bias=True) + ) + ) +) +Collect masks for language model over. +Collect masks for connector over. +Applying mask on model.layers.0.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.self_attn.q_proj. +Applying mask on model.layers.0.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.self_attn.k_proj. +Applying mask on model.layers.0.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.self_attn.v_proj. +Applying mask on model.layers.0.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.self_attn.o_proj. +Applying mask on model.layers.0.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.mlp.gate_proj. +Applying mask on model.layers.0.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.mlp.up_proj. +Applying mask on model.layers.0.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.mlp.down_proj. +Applying mask on model.layers.1.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.self_attn.q_proj. +Applying mask on model.layers.1.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.self_attn.k_proj. +Applying mask on model.layers.1.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.self_attn.v_proj. +Applying mask on model.layers.1.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.self_attn.o_proj. +Applying mask on model.layers.1.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.mlp.gate_proj. +Applying mask on model.layers.1.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.mlp.up_proj. +Applying mask on model.layers.1.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.mlp.down_proj. +Applying mask on model.layers.2.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.self_attn.q_proj. +Applying mask on model.layers.2.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.self_attn.k_proj. +Applying mask on model.layers.2.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.self_attn.v_proj. +Applying mask on model.layers.2.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.self_attn.o_proj. +Applying mask on model.layers.2.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.mlp.gate_proj. +Applying mask on model.layers.2.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.mlp.up_proj. +Applying mask on model.layers.2.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.mlp.down_proj. +Applying mask on model.layers.3.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.self_attn.q_proj. +Applying mask on model.layers.3.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.self_attn.k_proj. +Applying mask on model.layers.3.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.self_attn.v_proj. +Applying mask on model.layers.3.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.self_attn.o_proj. +Applying mask on model.layers.3.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.mlp.gate_proj. +Applying mask on model.layers.3.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.mlp.up_proj. +Applying mask on model.layers.3.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.mlp.down_proj. +Applying mask on model.layers.4.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.self_attn.q_proj. +Applying mask on model.layers.4.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.self_attn.k_proj. +Applying mask on model.layers.4.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.self_attn.v_proj. +Applying mask on model.layers.4.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.self_attn.o_proj. +Applying mask on model.layers.4.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.mlp.gate_proj. +Applying mask on model.layers.4.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.mlp.up_proj. +Applying mask on model.layers.4.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.mlp.down_proj. +Applying mask on model.layers.5.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.self_attn.q_proj. +Applying mask on model.layers.5.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.self_attn.k_proj. +Applying mask on model.layers.5.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.self_attn.v_proj. +Applying mask on model.layers.5.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.self_attn.o_proj. +Applying mask on model.layers.5.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.mlp.gate_proj. +Applying mask on model.layers.5.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.mlp.up_proj. +Applying mask on model.layers.5.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.mlp.down_proj. +Applying mask on model.layers.6.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.self_attn.q_proj. +Applying mask on model.layers.6.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.self_attn.k_proj. +Applying mask on model.layers.6.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.self_attn.v_proj. +Applying mask on model.layers.6.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.self_attn.o_proj. +Applying mask on model.layers.6.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.mlp.gate_proj. +Applying mask on model.layers.6.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.mlp.up_proj. +Applying mask on model.layers.6.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.mlp.down_proj. +Applying mask on model.layers.7.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.self_attn.q_proj. +Applying mask on model.layers.7.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.self_attn.k_proj. +Applying mask on model.layers.7.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.self_attn.v_proj. +Applying mask on model.layers.7.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.self_attn.o_proj. +Applying mask on model.layers.7.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.mlp.gate_proj. +Applying mask on model.layers.7.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.mlp.up_proj. +Applying mask on model.layers.7.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.mlp.down_proj. +Applying mask on model.layers.8.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.self_attn.q_proj. +Applying mask on model.layers.8.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.self_attn.k_proj. +Applying mask on model.layers.8.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.self_attn.v_proj. +Applying mask on model.layers.8.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.self_attn.o_proj. +Applying mask on model.layers.8.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.mlp.gate_proj. +Applying mask on model.layers.8.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.mlp.up_proj. +Applying mask on model.layers.8.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.mlp.down_proj. +Applying mask on model.layers.9.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.self_attn.q_proj. +Applying mask on model.layers.9.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.self_attn.k_proj. +Applying mask on model.layers.9.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.self_attn.v_proj. +Applying mask on model.layers.9.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.self_attn.o_proj. +Applying mask on model.layers.9.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.mlp.gate_proj. +Applying mask on model.layers.9.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.mlp.up_proj. +Applying mask on model.layers.9.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.mlp.down_proj. +Applying mask on model.layers.10.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.self_attn.q_proj. +Applying mask on model.layers.10.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.self_attn.k_proj. +Applying mask on model.layers.10.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.self_attn.v_proj. +Applying mask on model.layers.10.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.self_attn.o_proj. +Applying mask on model.layers.10.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.mlp.gate_proj. +Applying mask on model.layers.10.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.mlp.up_proj. +Applying mask on model.layers.10.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.mlp.down_proj. +Applying mask on model.layers.11.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.self_attn.q_proj. +Applying mask on model.layers.11.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.self_attn.k_proj. +Applying mask on model.layers.11.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.self_attn.v_proj. +Applying mask on model.layers.11.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.self_attn.o_proj. +Applying mask on model.layers.11.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.mlp.gate_proj. +Applying mask on model.layers.11.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.mlp.up_proj. +Applying mask on model.layers.11.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.mlp.down_proj. +Applying mask on model.layers.12.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.self_attn.q_proj. +Applying mask on model.layers.12.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.self_attn.k_proj. +Applying mask on model.layers.12.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.self_attn.v_proj. +Applying mask on model.layers.12.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.self_attn.o_proj. +Applying mask on model.layers.12.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.mlp.gate_proj. +Applying mask on model.layers.12.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.mlp.up_proj. +Applying mask on model.layers.12.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.mlp.down_proj. +Applying mask on model.layers.13.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.self_attn.q_proj. +Applying mask on model.layers.13.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.self_attn.k_proj. +Applying mask on model.layers.13.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.self_attn.v_proj. +Applying mask on model.layers.13.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.self_attn.o_proj. +Applying mask on model.layers.13.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.mlp.gate_proj. +Applying mask on model.layers.13.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.mlp.up_proj. +Applying mask on model.layers.13.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.mlp.down_proj. +Applying mask on model.layers.14.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.self_attn.q_proj. +Applying mask on model.layers.14.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.self_attn.k_proj. +Applying mask on model.layers.14.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.self_attn.v_proj. +Applying mask on model.layers.14.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.self_attn.o_proj. +Applying mask on model.layers.14.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.mlp.gate_proj. +Applying mask on model.layers.14.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.mlp.up_proj. +Applying mask on model.layers.14.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.mlp.down_proj. +Applying mask on model.layers.15.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.self_attn.q_proj. +Applying mask on model.layers.15.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.self_attn.k_proj. +Applying mask on model.layers.15.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.self_attn.v_proj. +Applying mask on model.layers.15.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.self_attn.o_proj. +Applying mask on model.layers.15.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.mlp.gate_proj. +Applying mask on model.layers.15.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.mlp.up_proj. +Applying mask on model.layers.15.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.mlp.down_proj. +Applying mask on model.layers.16.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.self_attn.q_proj. +Applying mask on model.layers.16.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.self_attn.k_proj. +Applying mask on model.layers.16.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.self_attn.v_proj. +Applying mask on model.layers.16.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.self_attn.o_proj. +Applying mask on model.layers.16.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.mlp.gate_proj. +Applying mask on model.layers.16.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.mlp.up_proj. +Applying mask on model.layers.16.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.mlp.down_proj. +Applying mask on model.layers.17.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.self_attn.q_proj. +Applying mask on model.layers.17.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.self_attn.k_proj. +Applying mask on model.layers.17.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.self_attn.v_proj. +Applying mask on model.layers.17.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.self_attn.o_proj. +Applying mask on model.layers.17.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.mlp.gate_proj. +Applying mask on model.layers.17.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.mlp.up_proj. +Applying mask on model.layers.17.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.mlp.down_proj. +Applying mask on model.layers.18.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.self_attn.q_proj. +Applying mask on model.layers.18.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.self_attn.k_proj. +Applying mask on model.layers.18.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.self_attn.v_proj. +Applying mask on model.layers.18.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.self_attn.o_proj. +Applying mask on model.layers.18.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.mlp.gate_proj. +Applying mask on model.layers.18.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.mlp.up_proj. +Applying mask on model.layers.18.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.mlp.down_proj. +Applying mask on model.layers.19.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.self_attn.q_proj. +Applying mask on model.layers.19.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.self_attn.k_proj. +Applying mask on model.layers.19.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.self_attn.v_proj. +Applying mask on model.layers.19.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.self_attn.o_proj. +Applying mask on model.layers.19.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.mlp.gate_proj. +Applying mask on model.layers.19.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.mlp.up_proj. +Applying mask on model.layers.19.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.mlp.down_proj. +Applying mask on model.layers.20.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.self_attn.q_proj. +Applying mask on model.layers.20.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.self_attn.k_proj. +Applying mask on model.layers.20.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.self_attn.v_proj. +Applying mask on model.layers.20.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.self_attn.o_proj. +Applying mask on model.layers.20.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.mlp.gate_proj. +Applying mask on model.layers.20.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.mlp.up_proj. +Applying mask on model.layers.20.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.mlp.down_proj. +Applying mask on model.layers.21.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.self_attn.q_proj. +Applying mask on model.layers.21.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.self_attn.k_proj. +Applying mask on model.layers.21.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.self_attn.v_proj. +Applying mask on model.layers.21.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.self_attn.o_proj. +Applying mask on model.layers.21.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.mlp.gate_proj. +Applying mask on model.layers.21.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.mlp.up_proj. +Applying mask on model.layers.21.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.mlp.down_proj. +Applying mask on model.layers.22.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.self_attn.q_proj. +Applying mask on model.layers.22.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.self_attn.k_proj. +Applying mask on model.layers.22.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.self_attn.v_proj. +Applying mask on model.layers.22.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.self_attn.o_proj. +Applying mask on model.layers.22.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.mlp.gate_proj. +Applying mask on model.layers.22.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.mlp.up_proj. +Applying mask on model.layers.22.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.mlp.down_proj. +Applying mask on model.layers.23.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.self_attn.q_proj. +Applying mask on model.layers.23.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.self_attn.k_proj. +Applying mask on model.layers.23.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.self_attn.v_proj. +Applying mask on model.layers.23.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.self_attn.o_proj. +Applying mask on model.layers.23.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.mlp.gate_proj. +Applying mask on model.layers.23.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.mlp.up_proj. +Applying mask on model.layers.23.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.mlp.down_proj. +Applying mask on _connector.0 with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on _connector.0. +Applying mask on _connector.2 with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on _connector.2. +Using cleaned config_mask (without mask parameters) for saving. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 19:03:35,692] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. + 0%| | 0/900 [00:00", + "resampler_hidden_size": 768, + "sparsity_connector": null, + "subnet_type_connector": "global", + "temperature_connector": 0.5, + "text_config": { + "_name_or_path": "Qwen/Qwen2.5-0.5B", + "architectures": [ + "Qwen2ForCausalLM" + ], + "backward_type": "normal", + "bos_token_id": 151643, + "eos_token_id": 151643, + "hidden_size": 896, + "intermediate_size": 4864, + "mask_type": "soft", + "masked_layers": "all", + "max_position_embeddings": 32768, + "max_window_layers": 24, + "model_type": "qwen2", + "num_attention_heads": 14, + "num_hidden_layers": 24, + "num_key_value_heads": 2, + "rope_theta": 1000000.0, + "sliding_window": 32768, + "subnet_mode": "both", + "subnet_type": "None", + "temperature_attn": 0.5, + "temperature_mlp": 0.5, + "tie_word_embeddings": true, + "torch_dtype": "bfloat16", + "use_mrope": false, + "use_sliding_window": false, + "vocab_size": 151936 + }, + "threshold_connector": null, + "tokenizer_model_max_length": 2048, + "tokenizer_name_or_path": "Qwen/Qwen2.5-0.5B", + "tokenizer_padding_side": "right", + "tokenizer_use_fast": false, + "torch_dtype": "bfloat16", + "transformers_version": "4.40.1", + "tune_type_connector": "full", + "tune_type_llm": "full", + "tune_type_vision_tower": "frozen", + "tune_vision_tower_from_layer": 0, + "use_cache": true, + "vision_config": { + "hidden_act": "gelu_pytorch_tanh", + "hidden_size": 1152, + "image_size": 384, + "intermediate_size": 4304, + "layer_norm_eps": 1e-06, + "model_name_or_path": "google/siglip-so400m-patch14-384", + "model_name_or_path2": "", + "model_type": "siglip_vision_model", + "num_attention_heads": 16, + "num_hidden_layers": 27, + "patch_size": 14 + }, + "vision_feature_layer": -2, + "vision_feature_select_strategy": "patch", + "vision_hidden_size": 1152, + "vision_model_name_or_path": "google/siglip-so400m-patch14-384", + "vision_model_name_or_path2": "", + "vocab_size": 151936 +} + +TinyLlavaForConditionalGeneration( + (language_model): Qwen2ForCausalLM( + (model): Qwen2Model( + (embed_tokens): Embedding(151936, 896) + (layers): ModuleList( + (0-23): 24 x Qwen2DecoderLayer( + (self_attn): Qwen2Attention( + (q_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=True) + (k_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=128, bias=True) + (v_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=128, bias=True) + (o_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=False) + (rotary_emb): Qwen2RotaryEmbedding() + ) + (mlp): Qwen2MLP( + (gate_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=4864, bias=False) + (up_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=4864, bias=False) + (down_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=4864, out_features=896, bias=False) + (act_fn): SiLU() + ) + (input_layernorm): Qwen2RMSNorm() + (post_attention_layernorm): Qwen2RMSNorm() + ) + ) + (norm): Qwen2RMSNorm() + ) + (lm_head): Linear(in_features=896, out_features=151936, bias=False) + ) + (vision_tower): SIGLIPVisionTower( + (_vision_tower): SiglipVisionModel( + (vision_model): SiglipVisionTransformer( + (embeddings): SiglipVisionEmbeddings( + (patch_embedding): Conv2d(3, 1152, kernel_size=(14, 14), stride=(14, 14), padding=valid) + (position_embedding): Embedding(729, 1152) + ) + (encoder): SiglipEncoder( + (layers): ModuleList( + (0-26): 27 x SiglipEncoderLayer( + (self_attn): SiglipAttention( + (k_proj): Linear(in_features=1152, out_features=1152, bias=True) + (v_proj): Linear(in_features=1152, out_features=1152, bias=True) + (q_proj): Linear(in_features=1152, out_features=1152, bias=True) + (out_proj): Linear(in_features=1152, out_features=1152, bias=True) + ) + (layer_norm1): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + (layer_norm2): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + ) + ) + ) + (post_layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (head): SiglipMultiheadAttentionPoolingHead( + (attention): MultiheadAttention( + (out_proj): NonDynamicallyQuantizableLinear(in_features=1152, out_features=1152, bias=True) + ) + (layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + ) + ) + ) + ) + (connector): MLPConnector( + (_connector): Sequential( + (0): SupermaskLinearSparsity_SoftForward_Normal(in_features=1152, out_features=896, bias=True) + (1): GELU(approximate='none') + (2): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=True) + ) + ) +) +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Load base model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain over. +TinyLlavaConfig { + "cache_dir": null, + "connector_type": "mlp2x_gelu", + "hidden_size": 896, + "ignore_index": -100, + "image_aspect_ratio": "square", + "image_token_index": -200, + "llm_model_name_or_path": "Qwen/Qwen2.5-0.5B", + "model_type": "tinyllava", + "num_queries": 128, + "num_resampler_layers": 3, + "pad_token": "<|endoftext|>", + "pad_token_id": 151643, + "resampler_hidden_size": 768, + "text_config": { + "_name_or_path": "Qwen/Qwen2.5-0.5B", + "architectures": [ + "Qwen2ForCausalLM" + ], + "bos_token_id": 151643, + "eos_token_id": 151643, + "hidden_size": 896, + "intermediate_size": 4864, + "max_position_embeddings": 32768, + "max_window_layers": 24, + "model_type": "qwen2", + "num_attention_heads": 14, + "num_hidden_layers": 24, + "num_key_value_heads": 2, + "rope_theta": 1000000.0, + "sliding_window": 32768, + "tie_word_embeddings": true, + "use_mrope": false, + "use_sliding_window": false, + "vocab_size": 151936 + }, + "tokenizer_model_max_length": 2048, + "tokenizer_name_or_path": "Qwen/Qwen2.5-0.5B", + "tokenizer_padding_side": "right", + "tokenizer_use_fast": false, + "transformers_version": "4.40.1", + "tune_type_connector": "full", + "tune_type_llm": "frozen", + "tune_type_vision_tower": "frozen", + "tune_vision_tower_from_layer": 0, + "use_cache": true, + "vision_config": { + "hidden_act": "gelu_pytorch_tanh", + "hidden_size": 1152, + "image_size": 384, + "intermediate_size": 4304, + "layer_norm_eps": 1e-06, + "model_name_or_path": "google/siglip-so400m-patch14-384", + "model_name_or_path2": "", + "model_type": "siglip_vision_model", + "num_attention_heads": 16, + "num_hidden_layers": 27, + "patch_size": 14 + }, + "vision_feature_layer": -2, + "vision_feature_select_strategy": "patch", + "vision_hidden_size": 1152, + "vision_model_name_or_path": "google/siglip-so400m-patch14-384", + "vision_model_name_or_path2": "", + "vocab_size": 151936 +} + +TinyLlavaForConditionalGeneration( + (language_model): Qwen2ForCausalLM( + (model): Qwen2Model( + (embed_tokens): Embedding(151936, 896) + (layers): ModuleList( + (0-23): 24 x Qwen2DecoderLayer( + (self_attn): Qwen2Attention( + (q_proj): Linear(in_features=896, out_features=896, bias=True) + (k_proj): Linear(in_features=896, out_features=128, bias=True) + (v_proj): Linear(in_features=896, out_features=128, bias=True) + (o_proj): Linear(in_features=896, out_features=896, bias=False) + (rotary_emb): Qwen2RotaryEmbedding() + ) + (mlp): Qwen2MLP( + (gate_proj): Linear(in_features=896, out_features=4864, bias=False) + (up_proj): Linear(in_features=896, out_features=4864, bias=False) + (down_proj): Linear(in_features=4864, out_features=896, bias=False) + (act_fn): SiLU() + ) + (input_layernorm): Qwen2RMSNorm() + (post_attention_layernorm): Qwen2RMSNorm() + ) + ) + (norm): Qwen2RMSNorm() + ) + (lm_head): Linear(in_features=896, out_features=151936, bias=False) + ) + (vision_tower): SIGLIPVisionTower( + (_vision_tower): SiglipVisionModel( + (vision_model): SiglipVisionTransformer( + (embeddings): SiglipVisionEmbeddings( + (patch_embedding): Conv2d(3, 1152, kernel_size=(14, 14), stride=(14, 14), padding=valid) + (position_embedding): Embedding(729, 1152) + ) + (encoder): SiglipEncoder( + (layers): ModuleList( + (0-26): 27 x SiglipEncoderLayer( + (self_attn): SiglipAttention( + (k_proj): Linear(in_features=1152, out_features=1152, bias=True) + (v_proj): Linear(in_features=1152, out_features=1152, bias=True) + (q_proj): Linear(in_features=1152, out_features=1152, bias=True) + (out_proj): Linear(in_features=1152, out_features=1152, bias=True) + ) + (layer_norm1): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + (layer_norm2): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + ) + ) + ) + (post_layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (head): SiglipMultiheadAttentionPoolingHead( + (attention): MultiheadAttention( + (out_proj): NonDynamicallyQuantizableLinear(in_features=1152, out_features=1152, bias=True) + ) + (layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + ) + ) + ) + ) + (connector): MLPConnector( + (_connector): Sequential( + (0): Linear(in_features=1152, out_features=896, bias=True) + (1): GELU(approximate='none') + (2): Linear(in_features=896, out_features=896, bias=True) + ) + ) +) +Collect masks for language model over. +Collect masks for connector over. +Applying mask on model.layers.0.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.self_attn.q_proj. +Applying mask on model.layers.0.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.self_attn.k_proj. +Applying mask on model.layers.0.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.self_attn.v_proj. +Applying mask on model.layers.0.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.self_attn.o_proj. +Applying mask on model.layers.0.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.mlp.gate_proj. +Applying mask on model.layers.0.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.mlp.up_proj. +Applying mask on model.layers.0.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.mlp.down_proj. +Applying mask on model.layers.1.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.self_attn.q_proj. +Applying mask on model.layers.1.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.self_attn.k_proj. +Applying mask on model.layers.1.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.self_attn.v_proj. +Applying mask on model.layers.1.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.self_attn.o_proj. +Applying mask on model.layers.1.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.mlp.gate_proj. +Applying mask on model.layers.1.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.mlp.up_proj. +Applying mask on model.layers.1.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.mlp.down_proj. +Applying mask on model.layers.2.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.self_attn.q_proj. +Applying mask on model.layers.2.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.self_attn.k_proj. +Applying mask on model.layers.2.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.self_attn.v_proj. +Applying mask on model.layers.2.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.self_attn.o_proj. +Applying mask on model.layers.2.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.mlp.gate_proj. +Applying mask on model.layers.2.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.mlp.up_proj. +Applying mask on model.layers.2.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.mlp.down_proj. +Applying mask on model.layers.3.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.self_attn.q_proj. +Applying mask on model.layers.3.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.self_attn.k_proj. +Applying mask on model.layers.3.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.self_attn.v_proj. +Applying mask on model.layers.3.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.self_attn.o_proj. +Applying mask on model.layers.3.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.mlp.gate_proj. +Applying mask on model.layers.3.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.mlp.up_proj. +Applying mask on model.layers.3.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.mlp.down_proj. +Applying mask on model.layers.4.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.self_attn.q_proj. +Applying mask on model.layers.4.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.self_attn.k_proj. +Applying mask on model.layers.4.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.self_attn.v_proj. +Applying mask on model.layers.4.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.self_attn.o_proj. +Applying mask on model.layers.4.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.mlp.gate_proj. +Applying mask on model.layers.4.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.mlp.up_proj. +Applying mask on model.layers.4.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.mlp.down_proj. +Applying mask on model.layers.5.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.self_attn.q_proj. +Applying mask on model.layers.5.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.self_attn.k_proj. +Applying mask on model.layers.5.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.self_attn.v_proj. +Applying mask on model.layers.5.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.self_attn.o_proj. +Applying mask on model.layers.5.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.mlp.gate_proj. +Applying mask on model.layers.5.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.mlp.up_proj. +Applying mask on model.layers.5.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.mlp.down_proj. +Applying mask on model.layers.6.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.self_attn.q_proj. +Applying mask on model.layers.6.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.self_attn.k_proj. +Applying mask on model.layers.6.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.self_attn.v_proj. +Applying mask on model.layers.6.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.self_attn.o_proj. +Applying mask on model.layers.6.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.mlp.gate_proj. +Applying mask on model.layers.6.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.mlp.up_proj. +Applying mask on model.layers.6.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.mlp.down_proj. +Applying mask on model.layers.7.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.self_attn.q_proj. +Applying mask on model.layers.7.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.self_attn.k_proj. +Applying mask on model.layers.7.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.self_attn.v_proj. +Applying mask on model.layers.7.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.self_attn.o_proj. +Applying mask on model.layers.7.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.mlp.gate_proj. +Applying mask on model.layers.7.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.mlp.up_proj. +Applying mask on model.layers.7.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.mlp.down_proj. +Applying mask on model.layers.8.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.self_attn.q_proj. +Applying mask on model.layers.8.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.self_attn.k_proj. +Applying mask on model.layers.8.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.self_attn.v_proj. +Applying mask on model.layers.8.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.self_attn.o_proj. +Applying mask on model.layers.8.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.mlp.gate_proj. +Applying mask on model.layers.8.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.mlp.up_proj. +Applying mask on model.layers.8.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.mlp.down_proj. +Applying mask on model.layers.9.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.self_attn.q_proj. +Applying mask on model.layers.9.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.self_attn.k_proj. +Applying mask on model.layers.9.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.self_attn.v_proj. +Applying mask on model.layers.9.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.self_attn.o_proj. +Applying mask on model.layers.9.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.mlp.gate_proj. +Applying mask on model.layers.9.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.mlp.up_proj. +Applying mask on model.layers.9.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.mlp.down_proj. +Applying mask on model.layers.10.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.self_attn.q_proj. +Applying mask on model.layers.10.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.self_attn.k_proj. +Applying mask on model.layers.10.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.self_attn.v_proj. +Applying mask on model.layers.10.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.self_attn.o_proj. +Applying mask on model.layers.10.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.mlp.gate_proj. +Applying mask on model.layers.10.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.mlp.up_proj. +Applying mask on model.layers.10.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.mlp.down_proj. +Applying mask on model.layers.11.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.self_attn.q_proj. +Applying mask on model.layers.11.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.self_attn.k_proj. +Applying mask on model.layers.11.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.self_attn.v_proj. +Applying mask on model.layers.11.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.self_attn.o_proj. +Applying mask on model.layers.11.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.mlp.gate_proj. +Applying mask on model.layers.11.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.mlp.up_proj. +Applying mask on model.layers.11.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.mlp.down_proj. +Applying mask on model.layers.12.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.self_attn.q_proj. +Applying mask on model.layers.12.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.self_attn.k_proj. +Applying mask on model.layers.12.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.self_attn.v_proj. +Applying mask on model.layers.12.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.self_attn.o_proj. +Applying mask on model.layers.12.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.mlp.gate_proj. +Applying mask on model.layers.12.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.mlp.up_proj. +Applying mask on model.layers.12.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.mlp.down_proj. +Applying mask on model.layers.13.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.self_attn.q_proj. +Applying mask on model.layers.13.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.self_attn.k_proj. +Applying mask on model.layers.13.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.self_attn.v_proj. +Applying mask on model.layers.13.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.self_attn.o_proj. +Applying mask on model.layers.13.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.mlp.gate_proj. +Applying mask on model.layers.13.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.mlp.up_proj. +Applying mask on model.layers.13.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.mlp.down_proj. +Applying mask on model.layers.14.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.self_attn.q_proj. +Applying mask on model.layers.14.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.self_attn.k_proj. +Applying mask on model.layers.14.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.self_attn.v_proj. +Applying mask on model.layers.14.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.self_attn.o_proj. +Applying mask on model.layers.14.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.mlp.gate_proj. +Applying mask on model.layers.14.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.mlp.up_proj. +Applying mask on model.layers.14.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.mlp.down_proj. +Applying mask on model.layers.15.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.self_attn.q_proj. +Applying mask on model.layers.15.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.self_attn.k_proj. +Applying mask on model.layers.15.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.self_attn.v_proj. +Applying mask on model.layers.15.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.self_attn.o_proj. +Applying mask on model.layers.15.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.mlp.gate_proj. +Applying mask on model.layers.15.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.mlp.up_proj. +Applying mask on model.layers.15.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.mlp.down_proj. +Applying mask on model.layers.16.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.self_attn.q_proj. +Applying mask on model.layers.16.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.self_attn.k_proj. +Applying mask on model.layers.16.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.self_attn.v_proj. +Applying mask on model.layers.16.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.self_attn.o_proj. +Applying mask on model.layers.16.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.mlp.gate_proj. +Applying mask on model.layers.16.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.mlp.up_proj. +Applying mask on model.layers.16.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.mlp.down_proj. +Applying mask on model.layers.17.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.self_attn.q_proj. +Applying mask on model.layers.17.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.self_attn.k_proj. +Applying mask on model.layers.17.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.self_attn.v_proj. +Applying mask on model.layers.17.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.self_attn.o_proj. +Applying mask on model.layers.17.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.mlp.gate_proj. +Applying mask on model.layers.17.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.mlp.up_proj. +Applying mask on model.layers.17.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.mlp.down_proj. +Applying mask on model.layers.18.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.self_attn.q_proj. +Applying mask on model.layers.18.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.self_attn.k_proj. +Applying mask on model.layers.18.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.self_attn.v_proj. +Applying mask on model.layers.18.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.self_attn.o_proj. +Applying mask on model.layers.18.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.mlp.gate_proj. +Applying mask on model.layers.18.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.mlp.up_proj. +Applying mask on model.layers.18.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.mlp.down_proj. +Applying mask on model.layers.19.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.self_attn.q_proj. +Applying mask on model.layers.19.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.self_attn.k_proj. +Applying mask on model.layers.19.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.self_attn.v_proj. +Applying mask on model.layers.19.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.self_attn.o_proj. +Applying mask on model.layers.19.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.mlp.gate_proj. +Applying mask on model.layers.19.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.mlp.up_proj. +Applying mask on model.layers.19.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.mlp.down_proj. +Applying mask on model.layers.20.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.self_attn.q_proj. +Applying mask on model.layers.20.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.self_attn.k_proj. +Applying mask on model.layers.20.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.self_attn.v_proj. +Applying mask on model.layers.20.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.self_attn.o_proj. +Applying mask on model.layers.20.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.mlp.gate_proj. +Applying mask on model.layers.20.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.mlp.up_proj. +Applying mask on model.layers.20.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.mlp.down_proj. +Applying mask on model.layers.21.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.self_attn.q_proj. +Applying mask on model.layers.21.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.self_attn.k_proj. +Applying mask on model.layers.21.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.self_attn.v_proj. +Applying mask on model.layers.21.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.self_attn.o_proj. +Applying mask on model.layers.21.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.mlp.gate_proj. +Applying mask on model.layers.21.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.mlp.up_proj. +Applying mask on model.layers.21.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.mlp.down_proj. +Applying mask on model.layers.22.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.self_attn.q_proj. +Applying mask on model.layers.22.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.self_attn.k_proj. +Applying mask on model.layers.22.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.self_attn.v_proj. +Applying mask on model.layers.22.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.self_attn.o_proj. +Applying mask on model.layers.22.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.mlp.gate_proj. +Applying mask on model.layers.22.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.mlp.up_proj. +Applying mask on model.layers.22.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.mlp.down_proj. +Applying mask on model.layers.23.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.self_attn.q_proj. +Applying mask on model.layers.23.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.self_attn.k_proj. +Applying mask on model.layers.23.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.self_attn.v_proj. +Applying mask on model.layers.23.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.self_attn.o_proj. +Applying mask on model.layers.23.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.mlp.gate_proj. +Applying mask on model.layers.23.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.mlp.up_proj. +Applying mask on model.layers.23.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.mlp.down_proj. +Applying mask on _connector.0 with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on _connector.0. +Applying mask on _connector.2 with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on _connector.2. +Using cleaned config_mask (without mask parameters) for saving. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 19:27:53,527] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. + 0%| | 0/900 [00:00", + "resampler_hidden_size": 768, + "sparsity_connector": null, + "subnet_type_connector": "global", + "temperature_connector": 0.5, + "text_config": { + "_name_or_path": "Qwen/Qwen2.5-0.5B", + "architectures": [ + "Qwen2ForCausalLM" + ], + "backward_type": "normal", + "bos_token_id": 151643, + "eos_token_id": 151643, + "hidden_size": 896, + "intermediate_size": 4864, + "mask_type": "soft", + "masked_layers": "all", + "max_position_embeddings": 32768, + "max_window_layers": 24, + "model_type": "qwen2", + "num_attention_heads": 14, + "num_hidden_layers": 24, + "num_key_value_heads": 2, + "rope_theta": 1000000.0, + "sliding_window": 32768, + "subnet_mode": "both", + "subnet_type": "None", + "temperature_attn": 0.5, + "temperature_mlp": 0.5, + "tie_word_embeddings": true, + "torch_dtype": "bfloat16", + "use_mrope": false, + "use_sliding_window": false, + "vocab_size": 151936 + }, + "threshold_connector": null, + "tokenizer_model_max_length": 2048, + "tokenizer_name_or_path": "Qwen/Qwen2.5-0.5B", + "tokenizer_padding_side": "right", + "tokenizer_use_fast": false, + "torch_dtype": "bfloat16", + "transformers_version": "4.40.1", + "tune_type_connector": "full", + "tune_type_llm": "full", + "tune_type_vision_tower": "frozen", + "tune_vision_tower_from_layer": 0, + "use_cache": true, + "vision_config": { + "hidden_act": "gelu_pytorch_tanh", + "hidden_size": 1152, + "image_size": 384, + "intermediate_size": 4304, + "layer_norm_eps": 1e-06, + "model_name_or_path": "google/siglip-so400m-patch14-384", + "model_name_or_path2": "", + "model_type": "siglip_vision_model", + "num_attention_heads": 16, + "num_hidden_layers": 27, + "patch_size": 14 + }, + "vision_feature_layer": -2, + "vision_feature_select_strategy": "patch", + "vision_hidden_size": 1152, + "vision_model_name_or_path": "google/siglip-so400m-patch14-384", + "vision_model_name_or_path2": "", + "vocab_size": 151936 +} + +TinyLlavaForConditionalGeneration( + (language_model): Qwen2ForCausalLM( + (model): Qwen2Model( + (embed_tokens): Embedding(151936, 896) + (layers): ModuleList( + (0-23): 24 x Qwen2DecoderLayer( + (self_attn): Qwen2Attention( + (q_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=True) + (k_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=128, bias=True) + (v_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=128, bias=True) + (o_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=False) + (rotary_emb): Qwen2RotaryEmbedding() + ) + (mlp): Qwen2MLP( + (gate_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=4864, bias=False) + (up_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=4864, bias=False) + (down_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=4864, out_features=896, bias=False) + (act_fn): SiLU() + ) + (input_layernorm): Qwen2RMSNorm() + (post_attention_layernorm): Qwen2RMSNorm() + ) + ) + (norm): Qwen2RMSNorm() + ) + (lm_head): Linear(in_features=896, out_features=151936, bias=False) + ) + (vision_tower): SIGLIPVisionTower( + (_vision_tower): SiglipVisionModel( + (vision_model): SiglipVisionTransformer( + (embeddings): SiglipVisionEmbeddings( + (patch_embedding): Conv2d(3, 1152, kernel_size=(14, 14), stride=(14, 14), padding=valid) + (position_embedding): Embedding(729, 1152) + ) + (encoder): SiglipEncoder( + (layers): ModuleList( + (0-26): 27 x SiglipEncoderLayer( + (self_attn): SiglipAttention( + (k_proj): Linear(in_features=1152, out_features=1152, bias=True) + (v_proj): Linear(in_features=1152, out_features=1152, bias=True) + (q_proj): Linear(in_features=1152, out_features=1152, bias=True) + (out_proj): Linear(in_features=1152, out_features=1152, bias=True) + ) + (layer_norm1): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + (layer_norm2): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + ) + ) + ) + (post_layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (head): SiglipMultiheadAttentionPoolingHead( + (attention): MultiheadAttention( + (out_proj): NonDynamicallyQuantizableLinear(in_features=1152, out_features=1152, bias=True) + ) + (layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + ) + ) + ) + ) + (connector): MLPConnector( + (_connector): Sequential( + (0): SupermaskLinearSparsity_SoftForward_Normal(in_features=1152, out_features=896, bias=True) + (1): GELU(approximate='none') + (2): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=True) + ) + ) +) +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Load base model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain over. +TinyLlavaConfig { + "cache_dir": null, + "connector_type": "mlp2x_gelu", + "hidden_size": 896, + "ignore_index": -100, + "image_aspect_ratio": "square", + "image_token_index": -200, + "llm_model_name_or_path": "Qwen/Qwen2.5-0.5B", + "model_type": "tinyllava", + "num_queries": 128, + "num_resampler_layers": 3, + "pad_token": "<|endoftext|>", + "pad_token_id": 151643, + "resampler_hidden_size": 768, + "text_config": { + "_name_or_path": "Qwen/Qwen2.5-0.5B", + "architectures": [ + "Qwen2ForCausalLM" + ], + "bos_token_id": 151643, + "eos_token_id": 151643, + "hidden_size": 896, + "intermediate_size": 4864, + "max_position_embeddings": 32768, + "max_window_layers": 24, + "model_type": "qwen2", + "num_attention_heads": 14, + "num_hidden_layers": 24, + "num_key_value_heads": 2, + "rope_theta": 1000000.0, + "sliding_window": 32768, + "tie_word_embeddings": true, + "use_mrope": false, + "use_sliding_window": false, + "vocab_size": 151936 + }, + "tokenizer_model_max_length": 2048, + "tokenizer_name_or_path": "Qwen/Qwen2.5-0.5B", + "tokenizer_padding_side": "right", + "tokenizer_use_fast": false, + "transformers_version": "4.40.1", + "tune_type_connector": "full", + "tune_type_llm": "frozen", + "tune_type_vision_tower": "frozen", + "tune_vision_tower_from_layer": 0, + "use_cache": true, + "vision_config": { + "hidden_act": "gelu_pytorch_tanh", + "hidden_size": 1152, + "image_size": 384, + "intermediate_size": 4304, + "layer_norm_eps": 1e-06, + "model_name_or_path": "google/siglip-so400m-patch14-384", + "model_name_or_path2": "", + "model_type": "siglip_vision_model", + "num_attention_heads": 16, + "num_hidden_layers": 27, + "patch_size": 14 + }, + "vision_feature_layer": -2, + "vision_feature_select_strategy": "patch", + "vision_hidden_size": 1152, + "vision_model_name_or_path": "google/siglip-so400m-patch14-384", + "vision_model_name_or_path2": "", + "vocab_size": 151936 +} + +TinyLlavaForConditionalGeneration( + (language_model): Qwen2ForCausalLM( + (model): Qwen2Model( + (embed_tokens): Embedding(151936, 896) + (layers): ModuleList( + (0-23): 24 x Qwen2DecoderLayer( + (self_attn): Qwen2Attention( + (q_proj): Linear(in_features=896, out_features=896, bias=True) + (k_proj): Linear(in_features=896, out_features=128, bias=True) + (v_proj): Linear(in_features=896, out_features=128, bias=True) + (o_proj): Linear(in_features=896, out_features=896, bias=False) + (rotary_emb): Qwen2RotaryEmbedding() + ) + (mlp): Qwen2MLP( + (gate_proj): Linear(in_features=896, out_features=4864, bias=False) + (up_proj): Linear(in_features=896, out_features=4864, bias=False) + (down_proj): Linear(in_features=4864, out_features=896, bias=False) + (act_fn): SiLU() + ) + (input_layernorm): Qwen2RMSNorm() + (post_attention_layernorm): Qwen2RMSNorm() + ) + ) + (norm): Qwen2RMSNorm() + ) + (lm_head): Linear(in_features=896, out_features=151936, bias=False) + ) + (vision_tower): SIGLIPVisionTower( + (_vision_tower): SiglipVisionModel( + (vision_model): SiglipVisionTransformer( + (embeddings): SiglipVisionEmbeddings( + (patch_embedding): Conv2d(3, 1152, kernel_size=(14, 14), stride=(14, 14), padding=valid) + (position_embedding): Embedding(729, 1152) + ) + (encoder): SiglipEncoder( + (layers): ModuleList( + (0-26): 27 x SiglipEncoderLayer( + (self_attn): SiglipAttention( + (k_proj): Linear(in_features=1152, out_features=1152, bias=True) + (v_proj): Linear(in_features=1152, out_features=1152, bias=True) + (q_proj): Linear(in_features=1152, out_features=1152, bias=True) + (out_proj): Linear(in_features=1152, out_features=1152, bias=True) + ) + (layer_norm1): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + (layer_norm2): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + ) + ) + ) + (post_layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (head): SiglipMultiheadAttentionPoolingHead( + (attention): MultiheadAttention( + (out_proj): NonDynamicallyQuantizableLinear(in_features=1152, out_features=1152, bias=True) + ) + (layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + ) + ) + ) + ) + (connector): MLPConnector( + (_connector): Sequential( + (0): Linear(in_features=1152, out_features=896, bias=True) + (1): GELU(approximate='none') + (2): Linear(in_features=896, out_features=896, bias=True) + ) + ) +) +Collect masks for language model over. +Collect masks for connector over. +Applying mask on model.layers.0.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.self_attn.q_proj. +Applying mask on model.layers.0.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.self_attn.k_proj. +Applying mask on model.layers.0.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.self_attn.v_proj. +Applying mask on model.layers.0.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.self_attn.o_proj. +Applying mask on model.layers.0.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.mlp.gate_proj. +Applying mask on model.layers.0.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.mlp.up_proj. +Applying mask on model.layers.0.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.mlp.down_proj. +Applying mask on model.layers.1.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.self_attn.q_proj. +Applying mask on model.layers.1.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.self_attn.k_proj. +Applying mask on model.layers.1.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.self_attn.v_proj. +Applying mask on model.layers.1.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.self_attn.o_proj. +Applying mask on model.layers.1.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.mlp.gate_proj. +Applying mask on model.layers.1.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.mlp.up_proj. +Applying mask on model.layers.1.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.mlp.down_proj. +Applying mask on model.layers.2.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.self_attn.q_proj. +Applying mask on model.layers.2.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.self_attn.k_proj. +Applying mask on model.layers.2.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.self_attn.v_proj. +Applying mask on model.layers.2.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.self_attn.o_proj. +Applying mask on model.layers.2.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.mlp.gate_proj. +Applying mask on model.layers.2.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.mlp.up_proj. +Applying mask on model.layers.2.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.mlp.down_proj. +Applying mask on model.layers.3.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.self_attn.q_proj. +Applying mask on model.layers.3.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.self_attn.k_proj. +Applying mask on model.layers.3.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.self_attn.v_proj. +Applying mask on model.layers.3.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.self_attn.o_proj. +Applying mask on model.layers.3.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.mlp.gate_proj. +Applying mask on model.layers.3.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.mlp.up_proj. +Applying mask on model.layers.3.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.mlp.down_proj. +Applying mask on model.layers.4.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.self_attn.q_proj. +Applying mask on model.layers.4.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.self_attn.k_proj. +Applying mask on model.layers.4.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.self_attn.v_proj. +Applying mask on model.layers.4.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.self_attn.o_proj. +Applying mask on model.layers.4.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.mlp.gate_proj. +Applying mask on model.layers.4.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.mlp.up_proj. +Applying mask on model.layers.4.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.mlp.down_proj. +Applying mask on model.layers.5.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.self_attn.q_proj. +Applying mask on model.layers.5.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.self_attn.k_proj. +Applying mask on model.layers.5.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.self_attn.v_proj. +Applying mask on model.layers.5.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.self_attn.o_proj. +Applying mask on model.layers.5.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.mlp.gate_proj. +Applying mask on model.layers.5.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.mlp.up_proj. +Applying mask on model.layers.5.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.mlp.down_proj. +Applying mask on model.layers.6.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.self_attn.q_proj. +Applying mask on model.layers.6.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.self_attn.k_proj. +Applying mask on model.layers.6.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.self_attn.v_proj. +Applying mask on model.layers.6.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.self_attn.o_proj. +Applying mask on model.layers.6.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.mlp.gate_proj. +Applying mask on model.layers.6.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.mlp.up_proj. +Applying mask on model.layers.6.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.mlp.down_proj. +Applying mask on model.layers.7.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.self_attn.q_proj. +Applying mask on model.layers.7.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.self_attn.k_proj. +Applying mask on model.layers.7.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.self_attn.v_proj. +Applying mask on model.layers.7.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.self_attn.o_proj. +Applying mask on model.layers.7.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.mlp.gate_proj. +Applying mask on model.layers.7.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.mlp.up_proj. +Applying mask on model.layers.7.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.mlp.down_proj. +Applying mask on model.layers.8.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.self_attn.q_proj. +Applying mask on model.layers.8.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.self_attn.k_proj. +Applying mask on model.layers.8.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.self_attn.v_proj. +Applying mask on model.layers.8.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.self_attn.o_proj. +Applying mask on model.layers.8.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.mlp.gate_proj. +Applying mask on model.layers.8.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.mlp.up_proj. +Applying mask on model.layers.8.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.mlp.down_proj. +Applying mask on model.layers.9.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.self_attn.q_proj. +Applying mask on model.layers.9.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.self_attn.k_proj. +Applying mask on model.layers.9.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.self_attn.v_proj. +Applying mask on model.layers.9.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.self_attn.o_proj. +Applying mask on model.layers.9.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.mlp.gate_proj. +Applying mask on model.layers.9.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.mlp.up_proj. +Applying mask on model.layers.9.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.mlp.down_proj. +Applying mask on model.layers.10.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.self_attn.q_proj. +Applying mask on model.layers.10.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.self_attn.k_proj. +Applying mask on model.layers.10.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.self_attn.v_proj. +Applying mask on model.layers.10.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.self_attn.o_proj. +Applying mask on model.layers.10.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.mlp.gate_proj. +Applying mask on model.layers.10.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.mlp.up_proj. +Applying mask on model.layers.10.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.mlp.down_proj. +Applying mask on model.layers.11.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.self_attn.q_proj. +Applying mask on model.layers.11.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.self_attn.k_proj. +Applying mask on model.layers.11.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.self_attn.v_proj. +Applying mask on model.layers.11.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.self_attn.o_proj. +Applying mask on model.layers.11.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.mlp.gate_proj. +Applying mask on model.layers.11.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.mlp.up_proj. +Applying mask on model.layers.11.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.mlp.down_proj. +Applying mask on model.layers.12.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.self_attn.q_proj. +Applying mask on model.layers.12.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.self_attn.k_proj. +Applying mask on model.layers.12.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.self_attn.v_proj. +Applying mask on model.layers.12.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.self_attn.o_proj. +Applying mask on model.layers.12.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.mlp.gate_proj. +Applying mask on model.layers.12.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.mlp.up_proj. +Applying mask on model.layers.12.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.mlp.down_proj. +Applying mask on model.layers.13.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.self_attn.q_proj. +Applying mask on model.layers.13.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.self_attn.k_proj. +Applying mask on model.layers.13.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.self_attn.v_proj. +Applying mask on model.layers.13.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.self_attn.o_proj. +Applying mask on model.layers.13.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.mlp.gate_proj. +Applying mask on model.layers.13.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.mlp.up_proj. +Applying mask on model.layers.13.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.mlp.down_proj. +Applying mask on model.layers.14.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.self_attn.q_proj. +Applying mask on model.layers.14.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.self_attn.k_proj. +Applying mask on model.layers.14.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.self_attn.v_proj. +Applying mask on model.layers.14.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.self_attn.o_proj. +Applying mask on model.layers.14.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.mlp.gate_proj. +Applying mask on model.layers.14.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.mlp.up_proj. +Applying mask on model.layers.14.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.mlp.down_proj. +Applying mask on model.layers.15.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.self_attn.q_proj. +Applying mask on model.layers.15.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.self_attn.k_proj. +Applying mask on model.layers.15.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.self_attn.v_proj. +Applying mask on model.layers.15.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.self_attn.o_proj. +Applying mask on model.layers.15.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.mlp.gate_proj. +Applying mask on model.layers.15.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.mlp.up_proj. +Applying mask on model.layers.15.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.mlp.down_proj. +Applying mask on model.layers.16.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.self_attn.q_proj. +Applying mask on model.layers.16.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.self_attn.k_proj. +Applying mask on model.layers.16.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.self_attn.v_proj. +Applying mask on model.layers.16.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.self_attn.o_proj. +Applying mask on model.layers.16.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.mlp.gate_proj. +Applying mask on model.layers.16.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.mlp.up_proj. +Applying mask on model.layers.16.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.mlp.down_proj. +Applying mask on model.layers.17.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.self_attn.q_proj. +Applying mask on model.layers.17.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.self_attn.k_proj. +Applying mask on model.layers.17.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.self_attn.v_proj. +Applying mask on model.layers.17.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.self_attn.o_proj. +Applying mask on model.layers.17.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.mlp.gate_proj. +Applying mask on model.layers.17.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.mlp.up_proj. +Applying mask on model.layers.17.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.mlp.down_proj. +Applying mask on model.layers.18.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.self_attn.q_proj. +Applying mask on model.layers.18.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.self_attn.k_proj. +Applying mask on model.layers.18.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.self_attn.v_proj. +Applying mask on model.layers.18.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.self_attn.o_proj. +Applying mask on model.layers.18.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.mlp.gate_proj. +Applying mask on model.layers.18.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.mlp.up_proj. +Applying mask on model.layers.18.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.mlp.down_proj. +Applying mask on model.layers.19.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.self_attn.q_proj. +Applying mask on model.layers.19.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.self_attn.k_proj. +Applying mask on model.layers.19.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.self_attn.v_proj. +Applying mask on model.layers.19.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.self_attn.o_proj. +Applying mask on model.layers.19.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.mlp.gate_proj. +Applying mask on model.layers.19.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.mlp.up_proj. +Applying mask on model.layers.19.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.mlp.down_proj. +Applying mask on model.layers.20.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.self_attn.q_proj. +Applying mask on model.layers.20.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.self_attn.k_proj. +Applying mask on model.layers.20.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.self_attn.v_proj. +Applying mask on model.layers.20.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.self_attn.o_proj. +Applying mask on model.layers.20.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.mlp.gate_proj. +Applying mask on model.layers.20.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.mlp.up_proj. +Applying mask on model.layers.20.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.mlp.down_proj. +Applying mask on model.layers.21.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.self_attn.q_proj. +Applying mask on model.layers.21.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.self_attn.k_proj. +Applying mask on model.layers.21.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.self_attn.v_proj. +Applying mask on model.layers.21.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.self_attn.o_proj. +Applying mask on model.layers.21.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.mlp.gate_proj. +Applying mask on model.layers.21.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.mlp.up_proj. +Applying mask on model.layers.21.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.mlp.down_proj. +Applying mask on model.layers.22.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.self_attn.q_proj. +Applying mask on model.layers.22.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.self_attn.k_proj. +Applying mask on model.layers.22.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.self_attn.v_proj. +Applying mask on model.layers.22.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.self_attn.o_proj. +Applying mask on model.layers.22.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.mlp.gate_proj. +Applying mask on model.layers.22.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.mlp.up_proj. +Applying mask on model.layers.22.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.mlp.down_proj. +Applying mask on model.layers.23.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.self_attn.q_proj. +Applying mask on model.layers.23.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.self_attn.k_proj. +Applying mask on model.layers.23.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.self_attn.v_proj. +Applying mask on model.layers.23.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.self_attn.o_proj. +Applying mask on model.layers.23.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.mlp.gate_proj. +Applying mask on model.layers.23.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.mlp.up_proj. +Applying mask on model.layers.23.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.mlp.down_proj. +Applying mask on _connector.0 with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on _connector.0. +Applying mask on _connector.2 with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on _connector.2. +Using cleaned config_mask (without mask parameters) for saving. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 19:52:32,414] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. + 0%| | 0/900 [00:00", + "resampler_hidden_size": 768, + "sparsity_connector": null, + "subnet_type_connector": "global", + "temperature_connector": 0.5, + "text_config": { + "_name_or_path": "Qwen/Qwen2.5-0.5B", + "architectures": [ + "Qwen2ForCausalLM" + ], + "backward_type": "normal", + "bos_token_id": 151643, + "eos_token_id": 151643, + "hidden_size": 896, + "intermediate_size": 4864, + "mask_type": "soft", + "masked_layers": "all", + "max_position_embeddings": 32768, + "max_window_layers": 24, + "model_type": "qwen2", + "num_attention_heads": 14, + "num_hidden_layers": 24, + "num_key_value_heads": 2, + "rope_theta": 1000000.0, + "sliding_window": 32768, + "subnet_mode": "both", + "subnet_type": "None", + "temperature_attn": 0.5, + "temperature_mlp": 0.5, + "tie_word_embeddings": true, + "torch_dtype": "bfloat16", + "use_mrope": false, + "use_sliding_window": false, + "vocab_size": 151936 + }, + "threshold_connector": null, + "tokenizer_model_max_length": 2048, + "tokenizer_name_or_path": "Qwen/Qwen2.5-0.5B", + "tokenizer_padding_side": "right", + "tokenizer_use_fast": false, + "torch_dtype": "bfloat16", + "transformers_version": "4.40.1", + "tune_type_connector": "full", + "tune_type_llm": "full", + "tune_type_vision_tower": "frozen", + "tune_vision_tower_from_layer": 0, + "use_cache": true, + "vision_config": { + "hidden_act": "gelu_pytorch_tanh", + "hidden_size": 1152, + "image_size": 384, + "intermediate_size": 4304, + "layer_norm_eps": 1e-06, + "model_name_or_path": "google/siglip-so400m-patch14-384", + "model_name_or_path2": "", + "model_type": "siglip_vision_model", + "num_attention_heads": 16, + "num_hidden_layers": 27, + "patch_size": 14 + }, + "vision_feature_layer": -2, + "vision_feature_select_strategy": "patch", + "vision_hidden_size": 1152, + "vision_model_name_or_path": "google/siglip-so400m-patch14-384", + "vision_model_name_or_path2": "", + "vocab_size": 151936 +} + +TinyLlavaForConditionalGeneration( + (language_model): Qwen2ForCausalLM( + (model): Qwen2Model( + (embed_tokens): Embedding(151936, 896) + (layers): ModuleList( + (0-23): 24 x Qwen2DecoderLayer( + (self_attn): Qwen2Attention( + (q_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=True) + (k_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=128, bias=True) + (v_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=128, bias=True) + (o_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=False) + (rotary_emb): Qwen2RotaryEmbedding() + ) + (mlp): Qwen2MLP( + (gate_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=4864, bias=False) + (up_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=4864, bias=False) + (down_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=4864, out_features=896, bias=False) + (act_fn): SiLU() + ) + (input_layernorm): Qwen2RMSNorm() + (post_attention_layernorm): Qwen2RMSNorm() + ) + ) + (norm): Qwen2RMSNorm() + ) + (lm_head): Linear(in_features=896, out_features=151936, bias=False) + ) + (vision_tower): SIGLIPVisionTower( + (_vision_tower): SiglipVisionModel( + (vision_model): SiglipVisionTransformer( + (embeddings): SiglipVisionEmbeddings( + (patch_embedding): Conv2d(3, 1152, kernel_size=(14, 14), stride=(14, 14), padding=valid) + (position_embedding): Embedding(729, 1152) + ) + (encoder): SiglipEncoder( + (layers): ModuleList( + (0-26): 27 x SiglipEncoderLayer( + (self_attn): SiglipAttention( + (k_proj): Linear(in_features=1152, out_features=1152, bias=True) + (v_proj): Linear(in_features=1152, out_features=1152, bias=True) + (q_proj): Linear(in_features=1152, out_features=1152, bias=True) + (out_proj): Linear(in_features=1152, out_features=1152, bias=True) + ) + (layer_norm1): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + (layer_norm2): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + ) + ) + ) + (post_layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (head): SiglipMultiheadAttentionPoolingHead( + (attention): MultiheadAttention( + (out_proj): NonDynamicallyQuantizableLinear(in_features=1152, out_features=1152, bias=True) + ) + (layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + ) + ) + ) + ) + (connector): MLPConnector( + (_connector): Sequential( + (0): SupermaskLinearSparsity_SoftForward_Normal(in_features=1152, out_features=896, bias=True) + (1): GELU(approximate='none') + (2): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=True) + ) + ) +) +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Load base model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain over. +TinyLlavaConfig { + "cache_dir": null, + "connector_type": "mlp2x_gelu", + "hidden_size": 896, + "ignore_index": -100, + "image_aspect_ratio": "square", + "image_token_index": -200, + "llm_model_name_or_path": "Qwen/Qwen2.5-0.5B", + "model_type": "tinyllava", + "num_queries": 128, + "num_resampler_layers": 3, + "pad_token": "<|endoftext|>", + "pad_token_id": 151643, + "resampler_hidden_size": 768, + "text_config": { + "_name_or_path": "Qwen/Qwen2.5-0.5B", + "architectures": [ + "Qwen2ForCausalLM" + ], + "bos_token_id": 151643, + "eos_token_id": 151643, + "hidden_size": 896, + "intermediate_size": 4864, + "max_position_embeddings": 32768, + "max_window_layers": 24, + "model_type": "qwen2", + "num_attention_heads": 14, + "num_hidden_layers": 24, + "num_key_value_heads": 2, + "rope_theta": 1000000.0, + "sliding_window": 32768, + "tie_word_embeddings": true, + "use_mrope": false, + "use_sliding_window": false, + "vocab_size": 151936 + }, + "tokenizer_model_max_length": 2048, + "tokenizer_name_or_path": "Qwen/Qwen2.5-0.5B", + "tokenizer_padding_side": "right", + "tokenizer_use_fast": false, + "transformers_version": "4.40.1", + "tune_type_connector": "full", + "tune_type_llm": "frozen", + "tune_type_vision_tower": "frozen", + "tune_vision_tower_from_layer": 0, + "use_cache": true, + "vision_config": { + "hidden_act": "gelu_pytorch_tanh", + "hidden_size": 1152, + "image_size": 384, + "intermediate_size": 4304, + "layer_norm_eps": 1e-06, + "model_name_or_path": "google/siglip-so400m-patch14-384", + "model_name_or_path2": "", + "model_type": "siglip_vision_model", + "num_attention_heads": 16, + "num_hidden_layers": 27, + "patch_size": 14 + }, + "vision_feature_layer": -2, + "vision_feature_select_strategy": "patch", + "vision_hidden_size": 1152, + "vision_model_name_or_path": "google/siglip-so400m-patch14-384", + "vision_model_name_or_path2": "", + "vocab_size": 151936 +} + +TinyLlavaForConditionalGeneration( + (language_model): Qwen2ForCausalLM( + (model): Qwen2Model( + (embed_tokens): Embedding(151936, 896) + (layers): ModuleList( + (0-23): 24 x Qwen2DecoderLayer( + (self_attn): Qwen2Attention( + (q_proj): Linear(in_features=896, out_features=896, bias=True) + (k_proj): Linear(in_features=896, out_features=128, bias=True) + (v_proj): Linear(in_features=896, out_features=128, bias=True) + (o_proj): Linear(in_features=896, out_features=896, bias=False) + (rotary_emb): Qwen2RotaryEmbedding() + ) + (mlp): Qwen2MLP( + (gate_proj): Linear(in_features=896, out_features=4864, bias=False) + (up_proj): Linear(in_features=896, out_features=4864, bias=False) + (down_proj): Linear(in_features=4864, out_features=896, bias=False) + (act_fn): SiLU() + ) + (input_layernorm): Qwen2RMSNorm() + (post_attention_layernorm): Qwen2RMSNorm() + ) + ) + (norm): Qwen2RMSNorm() + ) + (lm_head): Linear(in_features=896, out_features=151936, bias=False) + ) + (vision_tower): SIGLIPVisionTower( + (_vision_tower): SiglipVisionModel( + (vision_model): SiglipVisionTransformer( + (embeddings): SiglipVisionEmbeddings( + (patch_embedding): Conv2d(3, 1152, kernel_size=(14, 14), stride=(14, 14), padding=valid) + (position_embedding): Embedding(729, 1152) + ) + (encoder): SiglipEncoder( + (layers): ModuleList( + (0-26): 27 x SiglipEncoderLayer( + (self_attn): SiglipAttention( + (k_proj): Linear(in_features=1152, out_features=1152, bias=True) + (v_proj): Linear(in_features=1152, out_features=1152, bias=True) + (q_proj): Linear(in_features=1152, out_features=1152, bias=True) + (out_proj): Linear(in_features=1152, out_features=1152, bias=True) + ) + (layer_norm1): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + (layer_norm2): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + ) + ) + ) + (post_layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (head): SiglipMultiheadAttentionPoolingHead( + (attention): MultiheadAttention( + (out_proj): NonDynamicallyQuantizableLinear(in_features=1152, out_features=1152, bias=True) + ) + (layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + ) + ) + ) + ) + (connector): MLPConnector( + (_connector): Sequential( + (0): Linear(in_features=1152, out_features=896, bias=True) + (1): GELU(approximate='none') + (2): Linear(in_features=896, out_features=896, bias=True) + ) + ) +) +Collect masks for language model over. +Collect masks for connector over. +Applying mask on model.layers.0.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.self_attn.q_proj. +Applying mask on model.layers.0.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.self_attn.k_proj. +Applying mask on model.layers.0.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.self_attn.v_proj. +Applying mask on model.layers.0.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.self_attn.o_proj. +Applying mask on model.layers.0.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.mlp.gate_proj. +Applying mask on model.layers.0.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.mlp.up_proj. +Applying mask on model.layers.0.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.mlp.down_proj. +Applying mask on model.layers.1.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.self_attn.q_proj. +Applying mask on model.layers.1.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.self_attn.k_proj. +Applying mask on model.layers.1.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.self_attn.v_proj. +Applying mask on model.layers.1.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.self_attn.o_proj. +Applying mask on model.layers.1.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.mlp.gate_proj. +Applying mask on model.layers.1.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.mlp.up_proj. +Applying mask on model.layers.1.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.mlp.down_proj. +Applying mask on model.layers.2.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.self_attn.q_proj. +Applying mask on model.layers.2.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.self_attn.k_proj. +Applying mask on model.layers.2.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.self_attn.v_proj. +Applying mask on model.layers.2.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.self_attn.o_proj. +Applying mask on model.layers.2.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.mlp.gate_proj. +Applying mask on model.layers.2.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.mlp.up_proj. +Applying mask on model.layers.2.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.mlp.down_proj. +Applying mask on model.layers.3.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.self_attn.q_proj. +Applying mask on model.layers.3.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.self_attn.k_proj. +Applying mask on model.layers.3.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.self_attn.v_proj. +Applying mask on model.layers.3.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.self_attn.o_proj. +Applying mask on model.layers.3.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.mlp.gate_proj. +Applying mask on model.layers.3.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.mlp.up_proj. +Applying mask on model.layers.3.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.mlp.down_proj. +Applying mask on model.layers.4.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.self_attn.q_proj. +Applying mask on model.layers.4.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.self_attn.k_proj. +Applying mask on model.layers.4.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.self_attn.v_proj. +Applying mask on model.layers.4.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.self_attn.o_proj. +Applying mask on model.layers.4.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.mlp.gate_proj. +Applying mask on model.layers.4.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.mlp.up_proj. +Applying mask on model.layers.4.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.mlp.down_proj. +Applying mask on model.layers.5.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.self_attn.q_proj. +Applying mask on model.layers.5.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.self_attn.k_proj. +Applying mask on model.layers.5.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.self_attn.v_proj. +Applying mask on model.layers.5.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.self_attn.o_proj. +Applying mask on model.layers.5.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.mlp.gate_proj. +Applying mask on model.layers.5.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.mlp.up_proj. +Applying mask on model.layers.5.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.mlp.down_proj. +Applying mask on model.layers.6.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.self_attn.q_proj. +Applying mask on model.layers.6.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.self_attn.k_proj. +Applying mask on model.layers.6.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.self_attn.v_proj. +Applying mask on model.layers.6.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.self_attn.o_proj. +Applying mask on model.layers.6.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.mlp.gate_proj. +Applying mask on model.layers.6.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.mlp.up_proj. +Applying mask on model.layers.6.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.mlp.down_proj. +Applying mask on model.layers.7.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.self_attn.q_proj. +Applying mask on model.layers.7.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.self_attn.k_proj. +Applying mask on model.layers.7.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.self_attn.v_proj. +Applying mask on model.layers.7.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.self_attn.o_proj. +Applying mask on model.layers.7.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.mlp.gate_proj. +Applying mask on model.layers.7.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.mlp.up_proj. +Applying mask on model.layers.7.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.mlp.down_proj. +Applying mask on model.layers.8.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.self_attn.q_proj. +Applying mask on model.layers.8.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.self_attn.k_proj. +Applying mask on model.layers.8.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.self_attn.v_proj. +Applying mask on model.layers.8.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.self_attn.o_proj. +Applying mask on model.layers.8.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.mlp.gate_proj. +Applying mask on model.layers.8.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.mlp.up_proj. +Applying mask on model.layers.8.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.mlp.down_proj. +Applying mask on model.layers.9.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.self_attn.q_proj. +Applying mask on model.layers.9.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.self_attn.k_proj. +Applying mask on model.layers.9.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.self_attn.v_proj. +Applying mask on model.layers.9.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.self_attn.o_proj. +Applying mask on model.layers.9.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.mlp.gate_proj. +Applying mask on model.layers.9.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.mlp.up_proj. +Applying mask on model.layers.9.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.mlp.down_proj. +Applying mask on model.layers.10.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.self_attn.q_proj. +Applying mask on model.layers.10.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.self_attn.k_proj. +Applying mask on model.layers.10.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.self_attn.v_proj. +Applying mask on model.layers.10.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.self_attn.o_proj. +Applying mask on model.layers.10.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.mlp.gate_proj. +Applying mask on model.layers.10.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.mlp.up_proj. +Applying mask on model.layers.10.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.mlp.down_proj. +Applying mask on model.layers.11.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.self_attn.q_proj. +Applying mask on model.layers.11.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.self_attn.k_proj. +Applying mask on model.layers.11.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.self_attn.v_proj. +Applying mask on model.layers.11.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.self_attn.o_proj. +Applying mask on model.layers.11.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.mlp.gate_proj. +Applying mask on model.layers.11.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.mlp.up_proj. +Applying mask on model.layers.11.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.mlp.down_proj. +Applying mask on model.layers.12.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.self_attn.q_proj. +Applying mask on model.layers.12.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.self_attn.k_proj. +Applying mask on model.layers.12.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.self_attn.v_proj. +Applying mask on model.layers.12.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.self_attn.o_proj. +Applying mask on model.layers.12.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.mlp.gate_proj. +Applying mask on model.layers.12.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.mlp.up_proj. +Applying mask on model.layers.12.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.mlp.down_proj. +Applying mask on model.layers.13.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.self_attn.q_proj. +Applying mask on model.layers.13.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.self_attn.k_proj. +Applying mask on model.layers.13.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.self_attn.v_proj. +Applying mask on model.layers.13.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.self_attn.o_proj. +Applying mask on model.layers.13.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.mlp.gate_proj. +Applying mask on model.layers.13.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.mlp.up_proj. +Applying mask on model.layers.13.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.mlp.down_proj. +Applying mask on model.layers.14.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.self_attn.q_proj. +Applying mask on model.layers.14.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.self_attn.k_proj. +Applying mask on model.layers.14.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.self_attn.v_proj. +Applying mask on model.layers.14.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.self_attn.o_proj. +Applying mask on model.layers.14.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.mlp.gate_proj. +Applying mask on model.layers.14.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.mlp.up_proj. +Applying mask on model.layers.14.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.mlp.down_proj. +Applying mask on model.layers.15.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.self_attn.q_proj. +Applying mask on model.layers.15.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.self_attn.k_proj. +Applying mask on model.layers.15.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.self_attn.v_proj. +Applying mask on model.layers.15.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.self_attn.o_proj. +Applying mask on model.layers.15.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.mlp.gate_proj. +Applying mask on model.layers.15.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.mlp.up_proj. +Applying mask on model.layers.15.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.mlp.down_proj. +Applying mask on model.layers.16.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.self_attn.q_proj. +Applying mask on model.layers.16.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.self_attn.k_proj. +Applying mask on model.layers.16.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.self_attn.v_proj. +Applying mask on model.layers.16.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.self_attn.o_proj. +Applying mask on model.layers.16.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.mlp.gate_proj. +Applying mask on model.layers.16.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.mlp.up_proj. +Applying mask on model.layers.16.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.mlp.down_proj. +Applying mask on model.layers.17.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.self_attn.q_proj. +Applying mask on model.layers.17.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.self_attn.k_proj. +Applying mask on model.layers.17.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.self_attn.v_proj. +Applying mask on model.layers.17.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.self_attn.o_proj. +Applying mask on model.layers.17.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.mlp.gate_proj. +Applying mask on model.layers.17.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.mlp.up_proj. +Applying mask on model.layers.17.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.mlp.down_proj. +Applying mask on model.layers.18.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.self_attn.q_proj. +Applying mask on model.layers.18.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.self_attn.k_proj. +Applying mask on model.layers.18.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.self_attn.v_proj. +Applying mask on model.layers.18.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.self_attn.o_proj. +Applying mask on model.layers.18.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.mlp.gate_proj. +Applying mask on model.layers.18.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.mlp.up_proj. +Applying mask on model.layers.18.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.mlp.down_proj. +Applying mask on model.layers.19.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.self_attn.q_proj. +Applying mask on model.layers.19.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.self_attn.k_proj. +Applying mask on model.layers.19.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.self_attn.v_proj. +Applying mask on model.layers.19.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.self_attn.o_proj. +Applying mask on model.layers.19.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.mlp.gate_proj. +Applying mask on model.layers.19.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.mlp.up_proj. +Applying mask on model.layers.19.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.mlp.down_proj. +Applying mask on model.layers.20.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.self_attn.q_proj. +Applying mask on model.layers.20.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.self_attn.k_proj. +Applying mask on model.layers.20.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.self_attn.v_proj. +Applying mask on model.layers.20.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.self_attn.o_proj. +Applying mask on model.layers.20.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.mlp.gate_proj. +Applying mask on model.layers.20.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.mlp.up_proj. +Applying mask on model.layers.20.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.mlp.down_proj. +Applying mask on model.layers.21.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.self_attn.q_proj. +Applying mask on model.layers.21.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.self_attn.k_proj. +Applying mask on model.layers.21.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.self_attn.v_proj. +Applying mask on model.layers.21.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.self_attn.o_proj. +Applying mask on model.layers.21.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.mlp.gate_proj. +Applying mask on model.layers.21.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.mlp.up_proj. +Applying mask on model.layers.21.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.mlp.down_proj. +Applying mask on model.layers.22.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.self_attn.q_proj. +Applying mask on model.layers.22.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.self_attn.k_proj. +Applying mask on model.layers.22.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.self_attn.v_proj. +Applying mask on model.layers.22.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.self_attn.o_proj. +Applying mask on model.layers.22.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.mlp.gate_proj. +Applying mask on model.layers.22.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.mlp.up_proj. +Applying mask on model.layers.22.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.mlp.down_proj. +Applying mask on model.layers.23.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.self_attn.q_proj. +Applying mask on model.layers.23.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.self_attn.k_proj. +Applying mask on model.layers.23.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.self_attn.v_proj. +Applying mask on model.layers.23.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.self_attn.o_proj. +Applying mask on model.layers.23.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.mlp.gate_proj. +Applying mask on model.layers.23.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.mlp.up_proj. +Applying mask on model.layers.23.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.mlp.down_proj. +Applying mask on _connector.0 with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on _connector.0. +Applying mask on _connector.2 with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on _connector.2. +Using cleaned config_mask (without mask parameters) for saving. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 20:22:44,279] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. + 0%| | 0/900 [00:00", + "resampler_hidden_size": 768, + "sparsity_connector": null, + "subnet_type_connector": "global", + "temperature_connector": 0.7, + "text_config": { + "_name_or_path": "Qwen/Qwen2.5-0.5B", + "architectures": [ + "Qwen2ForCausalLM" + ], + "backward_type": "normal", + "bos_token_id": 151643, + "eos_token_id": 151643, + "hidden_size": 896, + "intermediate_size": 4864, + "mask_type": "soft", + "masked_layers": "all", + "max_position_embeddings": 32768, + "max_window_layers": 24, + "model_type": "qwen2", + "num_attention_heads": 14, + "num_hidden_layers": 24, + "num_key_value_heads": 2, + "rope_theta": 1000000.0, + "sliding_window": 32768, + "subnet_mode": "both", + "subnet_type": "None", + "temperature_attn": 0.7, + "temperature_mlp": 0.7, + "tie_word_embeddings": true, + "torch_dtype": "bfloat16", + "use_mrope": false, + "use_sliding_window": false, + "vocab_size": 151936 + }, + "threshold_connector": null, + "tokenizer_model_max_length": 2048, + "tokenizer_name_or_path": "Qwen/Qwen2.5-0.5B", + "tokenizer_padding_side": "right", + "tokenizer_use_fast": false, + "torch_dtype": "bfloat16", + "transformers_version": "4.40.1", + "tune_type_connector": "full", + "tune_type_llm": "full", + "tune_type_vision_tower": "frozen", + "tune_vision_tower_from_layer": 0, + "use_cache": true, + "vision_config": { + "hidden_act": "gelu_pytorch_tanh", + "hidden_size": 1152, + "image_size": 384, + "intermediate_size": 4304, + "layer_norm_eps": 1e-06, + "model_name_or_path": "google/siglip-so400m-patch14-384", + "model_name_or_path2": "", + "model_type": "siglip_vision_model", + "num_attention_heads": 16, + "num_hidden_layers": 27, + "patch_size": 14 + }, + "vision_feature_layer": -2, + "vision_feature_select_strategy": "patch", + "vision_hidden_size": 1152, + "vision_model_name_or_path": "google/siglip-so400m-patch14-384", + "vision_model_name_or_path2": "", + "vocab_size": 151936 +} + +TinyLlavaForConditionalGeneration( + (language_model): Qwen2ForCausalLM( + (model): Qwen2Model( + (embed_tokens): Embedding(151936, 896) + (layers): ModuleList( + (0-23): 24 x Qwen2DecoderLayer( + (self_attn): Qwen2Attention( + (q_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=True) + (k_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=128, bias=True) + (v_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=128, bias=True) + (o_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=False) + (rotary_emb): Qwen2RotaryEmbedding() + ) + (mlp): Qwen2MLP( + (gate_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=4864, bias=False) + (up_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=4864, bias=False) + (down_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=4864, out_features=896, bias=False) + (act_fn): SiLU() + ) + (input_layernorm): Qwen2RMSNorm() + (post_attention_layernorm): Qwen2RMSNorm() + ) + ) + (norm): Qwen2RMSNorm() + ) + (lm_head): Linear(in_features=896, out_features=151936, bias=False) + ) + (vision_tower): SIGLIPVisionTower( + (_vision_tower): SiglipVisionModel( + (vision_model): SiglipVisionTransformer( + (embeddings): SiglipVisionEmbeddings( + (patch_embedding): Conv2d(3, 1152, kernel_size=(14, 14), stride=(14, 14), padding=valid) + (position_embedding): Embedding(729, 1152) + ) + (encoder): SiglipEncoder( + (layers): ModuleList( + (0-26): 27 x SiglipEncoderLayer( + (self_attn): SiglipAttention( + (k_proj): Linear(in_features=1152, out_features=1152, bias=True) + (v_proj): Linear(in_features=1152, out_features=1152, bias=True) + (q_proj): Linear(in_features=1152, out_features=1152, bias=True) + (out_proj): Linear(in_features=1152, out_features=1152, bias=True) + ) + (layer_norm1): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + (layer_norm2): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + ) + ) + ) + (post_layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (head): SiglipMultiheadAttentionPoolingHead( + (attention): MultiheadAttention( + (out_proj): NonDynamicallyQuantizableLinear(in_features=1152, out_features=1152, bias=True) + ) + (layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + ) + ) + ) + ) + (connector): MLPConnector( + (_connector): Sequential( + (0): SupermaskLinearSparsity_SoftForward_Normal(in_features=1152, out_features=896, bias=True) + (1): GELU(approximate='none') + (2): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=True) + ) + ) +) +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Load base model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain over. +TinyLlavaConfig { + "cache_dir": null, + "connector_type": "mlp2x_gelu", + "hidden_size": 896, + "ignore_index": -100, + "image_aspect_ratio": "square", + "image_token_index": -200, + "llm_model_name_or_path": "Qwen/Qwen2.5-0.5B", + "model_type": "tinyllava", + "num_queries": 128, + "num_resampler_layers": 3, + "pad_token": "<|endoftext|>", + "pad_token_id": 151643, + "resampler_hidden_size": 768, + "text_config": { + "_name_or_path": "Qwen/Qwen2.5-0.5B", + "architectures": [ + "Qwen2ForCausalLM" + ], + "bos_token_id": 151643, + "eos_token_id": 151643, + "hidden_size": 896, + "intermediate_size": 4864, + "max_position_embeddings": 32768, + "max_window_layers": 24, + "model_type": "qwen2", + "num_attention_heads": 14, + "num_hidden_layers": 24, + "num_key_value_heads": 2, + "rope_theta": 1000000.0, + "sliding_window": 32768, + "tie_word_embeddings": true, + "use_mrope": false, + "use_sliding_window": false, + "vocab_size": 151936 + }, + "tokenizer_model_max_length": 2048, + "tokenizer_name_or_path": "Qwen/Qwen2.5-0.5B", + "tokenizer_padding_side": "right", + "tokenizer_use_fast": false, + "transformers_version": "4.40.1", + "tune_type_connector": "full", + "tune_type_llm": "frozen", + "tune_type_vision_tower": "frozen", + "tune_vision_tower_from_layer": 0, + "use_cache": true, + "vision_config": { + "hidden_act": "gelu_pytorch_tanh", + "hidden_size": 1152, + "image_size": 384, + "intermediate_size": 4304, + "layer_norm_eps": 1e-06, + "model_name_or_path": "google/siglip-so400m-patch14-384", + "model_name_or_path2": "", + "model_type": "siglip_vision_model", + "num_attention_heads": 16, + "num_hidden_layers": 27, + "patch_size": 14 + }, + "vision_feature_layer": -2, + "vision_feature_select_strategy": "patch", + "vision_hidden_size": 1152, + "vision_model_name_or_path": "google/siglip-so400m-patch14-384", + "vision_model_name_or_path2": "", + "vocab_size": 151936 +} + +TinyLlavaForConditionalGeneration( + (language_model): Qwen2ForCausalLM( + (model): Qwen2Model( + (embed_tokens): Embedding(151936, 896) + (layers): ModuleList( + (0-23): 24 x Qwen2DecoderLayer( + (self_attn): Qwen2Attention( + (q_proj): Linear(in_features=896, out_features=896, bias=True) + (k_proj): Linear(in_features=896, out_features=128, bias=True) + (v_proj): Linear(in_features=896, out_features=128, bias=True) + (o_proj): Linear(in_features=896, out_features=896, bias=False) + (rotary_emb): Qwen2RotaryEmbedding() + ) + (mlp): Qwen2MLP( + (gate_proj): Linear(in_features=896, out_features=4864, bias=False) + (up_proj): Linear(in_features=896, out_features=4864, bias=False) + (down_proj): Linear(in_features=4864, out_features=896, bias=False) + (act_fn): SiLU() + ) + (input_layernorm): Qwen2RMSNorm() + (post_attention_layernorm): Qwen2RMSNorm() + ) + ) + (norm): Qwen2RMSNorm() + ) + (lm_head): Linear(in_features=896, out_features=151936, bias=False) + ) + (vision_tower): SIGLIPVisionTower( + (_vision_tower): SiglipVisionModel( + (vision_model): SiglipVisionTransformer( + (embeddings): SiglipVisionEmbeddings( + (patch_embedding): Conv2d(3, 1152, kernel_size=(14, 14), stride=(14, 14), padding=valid) + (position_embedding): Embedding(729, 1152) + ) + (encoder): SiglipEncoder( + (layers): ModuleList( + (0-26): 27 x SiglipEncoderLayer( + (self_attn): SiglipAttention( + (k_proj): Linear(in_features=1152, out_features=1152, bias=True) + (v_proj): Linear(in_features=1152, out_features=1152, bias=True) + (q_proj): Linear(in_features=1152, out_features=1152, bias=True) + (out_proj): Linear(in_features=1152, out_features=1152, bias=True) + ) + (layer_norm1): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + (layer_norm2): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + ) + ) + ) + (post_layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (head): SiglipMultiheadAttentionPoolingHead( + (attention): MultiheadAttention( + (out_proj): NonDynamicallyQuantizableLinear(in_features=1152, out_features=1152, bias=True) + ) + (layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + ) + ) + ) + ) + (connector): MLPConnector( + (_connector): Sequential( + (0): Linear(in_features=1152, out_features=896, bias=True) + (1): GELU(approximate='none') + (2): Linear(in_features=896, out_features=896, bias=True) + ) + ) +) +Collect masks for language model over. +Collect masks for connector over. +Applying mask on model.layers.0.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.self_attn.q_proj. +Applying mask on model.layers.0.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.self_attn.k_proj. +Applying mask on model.layers.0.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.self_attn.v_proj. +Applying mask on model.layers.0.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.self_attn.o_proj. +Applying mask on model.layers.0.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.mlp.gate_proj. +Applying mask on model.layers.0.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.mlp.up_proj. +Applying mask on model.layers.0.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.mlp.down_proj. +Applying mask on model.layers.1.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.self_attn.q_proj. +Applying mask on model.layers.1.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.self_attn.k_proj. +Applying mask on model.layers.1.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.self_attn.v_proj. +Applying mask on model.layers.1.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.self_attn.o_proj. +Applying mask on model.layers.1.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.mlp.gate_proj. +Applying mask on model.layers.1.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.mlp.up_proj. +Applying mask on model.layers.1.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.mlp.down_proj. +Applying mask on model.layers.2.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.self_attn.q_proj. +Applying mask on model.layers.2.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.self_attn.k_proj. +Applying mask on model.layers.2.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.self_attn.v_proj. +Applying mask on model.layers.2.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.self_attn.o_proj. +Applying mask on model.layers.2.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.mlp.gate_proj. +Applying mask on model.layers.2.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.mlp.up_proj. +Applying mask on model.layers.2.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.mlp.down_proj. +Applying mask on model.layers.3.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.self_attn.q_proj. +Applying mask on model.layers.3.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.self_attn.k_proj. +Applying mask on model.layers.3.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.self_attn.v_proj. +Applying mask on model.layers.3.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.self_attn.o_proj. +Applying mask on model.layers.3.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.mlp.gate_proj. +Applying mask on model.layers.3.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.mlp.up_proj. +Applying mask on model.layers.3.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.mlp.down_proj. +Applying mask on model.layers.4.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.self_attn.q_proj. +Applying mask on model.layers.4.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.self_attn.k_proj. +Applying mask on model.layers.4.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.self_attn.v_proj. +Applying mask on model.layers.4.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.self_attn.o_proj. +Applying mask on model.layers.4.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.mlp.gate_proj. +Applying mask on model.layers.4.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.mlp.up_proj. +Applying mask on model.layers.4.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.mlp.down_proj. +Applying mask on model.layers.5.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.self_attn.q_proj. +Applying mask on model.layers.5.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.self_attn.k_proj. +Applying mask on model.layers.5.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.self_attn.v_proj. +Applying mask on model.layers.5.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.self_attn.o_proj. +Applying mask on model.layers.5.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.mlp.gate_proj. +Applying mask on model.layers.5.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.mlp.up_proj. +Applying mask on model.layers.5.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.mlp.down_proj. +Applying mask on model.layers.6.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.self_attn.q_proj. +Applying mask on model.layers.6.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.self_attn.k_proj. +Applying mask on model.layers.6.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.self_attn.v_proj. +Applying mask on model.layers.6.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.self_attn.o_proj. +Applying mask on model.layers.6.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.mlp.gate_proj. +Applying mask on model.layers.6.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.mlp.up_proj. +Applying mask on model.layers.6.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.mlp.down_proj. +Applying mask on model.layers.7.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.self_attn.q_proj. +Applying mask on model.layers.7.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.self_attn.k_proj. +Applying mask on model.layers.7.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.self_attn.v_proj. +Applying mask on model.layers.7.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.self_attn.o_proj. +Applying mask on model.layers.7.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.mlp.gate_proj. +Applying mask on model.layers.7.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.mlp.up_proj. +Applying mask on model.layers.7.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.mlp.down_proj. +Applying mask on model.layers.8.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.self_attn.q_proj. +Applying mask on model.layers.8.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.self_attn.k_proj. +Applying mask on model.layers.8.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.self_attn.v_proj. +Applying mask on model.layers.8.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.self_attn.o_proj. +Applying mask on model.layers.8.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.mlp.gate_proj. +Applying mask on model.layers.8.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.mlp.up_proj. +Applying mask on model.layers.8.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.mlp.down_proj. +Applying mask on model.layers.9.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.self_attn.q_proj. +Applying mask on model.layers.9.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.self_attn.k_proj. +Applying mask on model.layers.9.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.self_attn.v_proj. +Applying mask on model.layers.9.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.self_attn.o_proj. +Applying mask on model.layers.9.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.mlp.gate_proj. +Applying mask on model.layers.9.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.mlp.up_proj. +Applying mask on model.layers.9.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.mlp.down_proj. +Applying mask on model.layers.10.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.self_attn.q_proj. +Applying mask on model.layers.10.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.self_attn.k_proj. +Applying mask on model.layers.10.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.self_attn.v_proj. +Applying mask on model.layers.10.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.self_attn.o_proj. +Applying mask on model.layers.10.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.mlp.gate_proj. +Applying mask on model.layers.10.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.mlp.up_proj. +Applying mask on model.layers.10.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.mlp.down_proj. +Applying mask on model.layers.11.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.self_attn.q_proj. +Applying mask on model.layers.11.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.self_attn.k_proj. +Applying mask on model.layers.11.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.self_attn.v_proj. +Applying mask on model.layers.11.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.self_attn.o_proj. +Applying mask on model.layers.11.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.mlp.gate_proj. +Applying mask on model.layers.11.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.mlp.up_proj. +Applying mask on model.layers.11.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.mlp.down_proj. +Applying mask on model.layers.12.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.self_attn.q_proj. +Applying mask on model.layers.12.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.self_attn.k_proj. +Applying mask on model.layers.12.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.self_attn.v_proj. +Applying mask on model.layers.12.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.self_attn.o_proj. +Applying mask on model.layers.12.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.mlp.gate_proj. +Applying mask on model.layers.12.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.mlp.up_proj. +Applying mask on model.layers.12.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.mlp.down_proj. +Applying mask on model.layers.13.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.self_attn.q_proj. +Applying mask on model.layers.13.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.self_attn.k_proj. +Applying mask on model.layers.13.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.self_attn.v_proj. +Applying mask on model.layers.13.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.self_attn.o_proj. +Applying mask on model.layers.13.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.mlp.gate_proj. +Applying mask on model.layers.13.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.mlp.up_proj. +Applying mask on model.layers.13.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.mlp.down_proj. +Applying mask on model.layers.14.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.self_attn.q_proj. +Applying mask on model.layers.14.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.self_attn.k_proj. +Applying mask on model.layers.14.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.self_attn.v_proj. +Applying mask on model.layers.14.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.self_attn.o_proj. +Applying mask on model.layers.14.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.mlp.gate_proj. +Applying mask on model.layers.14.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.mlp.up_proj. +Applying mask on model.layers.14.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.mlp.down_proj. +Applying mask on model.layers.15.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.self_attn.q_proj. +Applying mask on model.layers.15.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.self_attn.k_proj. +Applying mask on model.layers.15.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.self_attn.v_proj. +Applying mask on model.layers.15.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.self_attn.o_proj. +Applying mask on model.layers.15.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.mlp.gate_proj. +Applying mask on model.layers.15.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.mlp.up_proj. +Applying mask on model.layers.15.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.mlp.down_proj. +Applying mask on model.layers.16.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.self_attn.q_proj. +Applying mask on model.layers.16.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.self_attn.k_proj. +Applying mask on model.layers.16.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.self_attn.v_proj. +Applying mask on model.layers.16.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.self_attn.o_proj. +Applying mask on model.layers.16.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.mlp.gate_proj. +Applying mask on model.layers.16.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.mlp.up_proj. +Applying mask on model.layers.16.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.mlp.down_proj. +Applying mask on model.layers.17.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.self_attn.q_proj. +Applying mask on model.layers.17.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.self_attn.k_proj. +Applying mask on model.layers.17.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.self_attn.v_proj. +Applying mask on model.layers.17.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.self_attn.o_proj. +Applying mask on model.layers.17.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.mlp.gate_proj. +Applying mask on model.layers.17.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.mlp.up_proj. +Applying mask on model.layers.17.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.mlp.down_proj. +Applying mask on model.layers.18.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.self_attn.q_proj. +Applying mask on model.layers.18.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.self_attn.k_proj. +Applying mask on model.layers.18.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.self_attn.v_proj. +Applying mask on model.layers.18.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.self_attn.o_proj. +Applying mask on model.layers.18.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.mlp.gate_proj. +Applying mask on model.layers.18.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.mlp.up_proj. +Applying mask on model.layers.18.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.mlp.down_proj. +Applying mask on model.layers.19.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.self_attn.q_proj. +Applying mask on model.layers.19.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.self_attn.k_proj. +Applying mask on model.layers.19.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.self_attn.v_proj. +Applying mask on model.layers.19.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.self_attn.o_proj. +Applying mask on model.layers.19.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.mlp.gate_proj. +Applying mask on model.layers.19.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.mlp.up_proj. +Applying mask on model.layers.19.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.mlp.down_proj. +Applying mask on model.layers.20.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.self_attn.q_proj. +Applying mask on model.layers.20.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.self_attn.k_proj. +Applying mask on model.layers.20.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.self_attn.v_proj. +Applying mask on model.layers.20.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.self_attn.o_proj. +Applying mask on model.layers.20.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.mlp.gate_proj. +Applying mask on model.layers.20.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.mlp.up_proj. +Applying mask on model.layers.20.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.mlp.down_proj. +Applying mask on model.layers.21.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.self_attn.q_proj. +Applying mask on model.layers.21.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.self_attn.k_proj. +Applying mask on model.layers.21.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.self_attn.v_proj. +Applying mask on model.layers.21.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.self_attn.o_proj. +Applying mask on model.layers.21.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.mlp.gate_proj. +Applying mask on model.layers.21.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.mlp.up_proj. +Applying mask on model.layers.21.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.mlp.down_proj. +Applying mask on model.layers.22.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.self_attn.q_proj. +Applying mask on model.layers.22.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.self_attn.k_proj. +Applying mask on model.layers.22.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.self_attn.v_proj. +Applying mask on model.layers.22.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.self_attn.o_proj. +Applying mask on model.layers.22.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.mlp.gate_proj. +Applying mask on model.layers.22.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.mlp.up_proj. +Applying mask on model.layers.22.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.mlp.down_proj. +Applying mask on model.layers.23.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.self_attn.q_proj. +Applying mask on model.layers.23.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.self_attn.k_proj. +Applying mask on model.layers.23.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.self_attn.v_proj. +Applying mask on model.layers.23.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.self_attn.o_proj. +Applying mask on model.layers.23.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.mlp.gate_proj. +Applying mask on model.layers.23.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.mlp.up_proj. +Applying mask on model.layers.23.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.mlp.down_proj. +Applying mask on _connector.0 with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on _connector.0. +Applying mask on _connector.2 with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on _connector.2. +Using cleaned config_mask (without mask parameters) for saving. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 20:47:34,322] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. + 0%| | 0/900 [00:00", + "resampler_hidden_size": 768, + "sparsity_connector": null, + "subnet_type_connector": "global", + "temperature_connector": 0.9, + "text_config": { + "_name_or_path": "Qwen/Qwen2.5-0.5B", + "architectures": [ + "Qwen2ForCausalLM" + ], + "backward_type": "normal", + "bos_token_id": 151643, + "eos_token_id": 151643, + "hidden_size": 896, + "intermediate_size": 4864, + "mask_type": "soft", + "masked_layers": "all", + "max_position_embeddings": 32768, + "max_window_layers": 24, + "model_type": "qwen2", + "num_attention_heads": 14, + "num_hidden_layers": 24, + "num_key_value_heads": 2, + "rope_theta": 1000000.0, + "sliding_window": 32768, + "subnet_mode": "both", + "subnet_type": "None", + "temperature_attn": 0.9, + "temperature_mlp": 0.9, + "tie_word_embeddings": true, + "torch_dtype": "bfloat16", + "use_mrope": false, + "use_sliding_window": false, + "vocab_size": 151936 + }, + "threshold_connector": null, + "tokenizer_model_max_length": 2048, + "tokenizer_name_or_path": "Qwen/Qwen2.5-0.5B", + "tokenizer_padding_side": "right", + "tokenizer_use_fast": false, + "torch_dtype": "bfloat16", + "transformers_version": "4.40.1", + "tune_type_connector": "full", + "tune_type_llm": "full", + "tune_type_vision_tower": "frozen", + "tune_vision_tower_from_layer": 0, + "use_cache": true, + "vision_config": { + "hidden_act": "gelu_pytorch_tanh", + "hidden_size": 1152, + "image_size": 384, + "intermediate_size": 4304, + "layer_norm_eps": 1e-06, + "model_name_or_path": "google/siglip-so400m-patch14-384", + "model_name_or_path2": "", + "model_type": "siglip_vision_model", + "num_attention_heads": 16, + "num_hidden_layers": 27, + "patch_size": 14 + }, + "vision_feature_layer": -2, + "vision_feature_select_strategy": "patch", + "vision_hidden_size": 1152, + "vision_model_name_or_path": "google/siglip-so400m-patch14-384", + "vision_model_name_or_path2": "", + "vocab_size": 151936 +} + +TinyLlavaForConditionalGeneration( + (language_model): Qwen2ForCausalLM( + (model): Qwen2Model( + (embed_tokens): Embedding(151936, 896) + (layers): ModuleList( + (0-23): 24 x Qwen2DecoderLayer( + (self_attn): Qwen2Attention( + (q_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=True) + (k_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=128, bias=True) + (v_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=128, bias=True) + (o_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=False) + (rotary_emb): Qwen2RotaryEmbedding() + ) + (mlp): Qwen2MLP( + (gate_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=4864, bias=False) + (up_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=4864, bias=False) + (down_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=4864, out_features=896, bias=False) + (act_fn): SiLU() + ) + (input_layernorm): Qwen2RMSNorm() + (post_attention_layernorm): Qwen2RMSNorm() + ) + ) + (norm): Qwen2RMSNorm() + ) + (lm_head): Linear(in_features=896, out_features=151936, bias=False) + ) + (vision_tower): SIGLIPVisionTower( + (_vision_tower): SiglipVisionModel( + (vision_model): SiglipVisionTransformer( + (embeddings): SiglipVisionEmbeddings( + (patch_embedding): Conv2d(3, 1152, kernel_size=(14, 14), stride=(14, 14), padding=valid) + (position_embedding): Embedding(729, 1152) + ) + (encoder): SiglipEncoder( + (layers): ModuleList( + (0-26): 27 x SiglipEncoderLayer( + (self_attn): SiglipAttention( + (k_proj): Linear(in_features=1152, out_features=1152, bias=True) + (v_proj): Linear(in_features=1152, out_features=1152, bias=True) + (q_proj): Linear(in_features=1152, out_features=1152, bias=True) + (out_proj): Linear(in_features=1152, out_features=1152, bias=True) + ) + (layer_norm1): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + (layer_norm2): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + ) + ) + ) + (post_layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (head): SiglipMultiheadAttentionPoolingHead( + (attention): MultiheadAttention( + (out_proj): NonDynamicallyQuantizableLinear(in_features=1152, out_features=1152, bias=True) + ) + (layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + ) + ) + ) + ) + (connector): MLPConnector( + (_connector): Sequential( + (0): SupermaskLinearSparsity_SoftForward_Normal(in_features=1152, out_features=896, bias=True) + (1): GELU(approximate='none') + (2): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=True) + ) + ) +) +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Load base model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain over. +TinyLlavaConfig { + "cache_dir": null, + "connector_type": "mlp2x_gelu", + "hidden_size": 896, + "ignore_index": -100, + "image_aspect_ratio": "square", + "image_token_index": -200, + "llm_model_name_or_path": "Qwen/Qwen2.5-0.5B", + "model_type": "tinyllava", + "num_queries": 128, + "num_resampler_layers": 3, + "pad_token": "<|endoftext|>", + "pad_token_id": 151643, + "resampler_hidden_size": 768, + "text_config": { + "_name_or_path": "Qwen/Qwen2.5-0.5B", + "architectures": [ + "Qwen2ForCausalLM" + ], + "bos_token_id": 151643, + "eos_token_id": 151643, + "hidden_size": 896, + "intermediate_size": 4864, + "max_position_embeddings": 32768, + "max_window_layers": 24, + "model_type": "qwen2", + "num_attention_heads": 14, + "num_hidden_layers": 24, + "num_key_value_heads": 2, + "rope_theta": 1000000.0, + "sliding_window": 32768, + "tie_word_embeddings": true, + "use_mrope": false, + "use_sliding_window": false, + "vocab_size": 151936 + }, + "tokenizer_model_max_length": 2048, + "tokenizer_name_or_path": "Qwen/Qwen2.5-0.5B", + "tokenizer_padding_side": "right", + "tokenizer_use_fast": false, + "transformers_version": "4.40.1", + "tune_type_connector": "full", + "tune_type_llm": "frozen", + "tune_type_vision_tower": "frozen", + "tune_vision_tower_from_layer": 0, + "use_cache": true, + "vision_config": { + "hidden_act": "gelu_pytorch_tanh", + "hidden_size": 1152, + "image_size": 384, + "intermediate_size": 4304, + "layer_norm_eps": 1e-06, + "model_name_or_path": "google/siglip-so400m-patch14-384", + "model_name_or_path2": "", + "model_type": "siglip_vision_model", + "num_attention_heads": 16, + "num_hidden_layers": 27, + "patch_size": 14 + }, + "vision_feature_layer": -2, + "vision_feature_select_strategy": "patch", + "vision_hidden_size": 1152, + "vision_model_name_or_path": "google/siglip-so400m-patch14-384", + "vision_model_name_or_path2": "", + "vocab_size": 151936 +} + +TinyLlavaForConditionalGeneration( + (language_model): Qwen2ForCausalLM( + (model): Qwen2Model( + (embed_tokens): Embedding(151936, 896) + (layers): ModuleList( + (0-23): 24 x Qwen2DecoderLayer( + (self_attn): Qwen2Attention( + (q_proj): Linear(in_features=896, out_features=896, bias=True) + (k_proj): Linear(in_features=896, out_features=128, bias=True) + (v_proj): Linear(in_features=896, out_features=128, bias=True) + (o_proj): Linear(in_features=896, out_features=896, bias=False) + (rotary_emb): Qwen2RotaryEmbedding() + ) + (mlp): Qwen2MLP( + (gate_proj): Linear(in_features=896, out_features=4864, bias=False) + (up_proj): Linear(in_features=896, out_features=4864, bias=False) + (down_proj): Linear(in_features=4864, out_features=896, bias=False) + (act_fn): SiLU() + ) + (input_layernorm): Qwen2RMSNorm() + (post_attention_layernorm): Qwen2RMSNorm() + ) + ) + (norm): Qwen2RMSNorm() + ) + (lm_head): Linear(in_features=896, out_features=151936, bias=False) + ) + (vision_tower): SIGLIPVisionTower( + (_vision_tower): SiglipVisionModel( + (vision_model): SiglipVisionTransformer( + (embeddings): SiglipVisionEmbeddings( + (patch_embedding): Conv2d(3, 1152, kernel_size=(14, 14), stride=(14, 14), padding=valid) + (position_embedding): Embedding(729, 1152) + ) + (encoder): SiglipEncoder( + (layers): ModuleList( + (0-26): 27 x SiglipEncoderLayer( + (self_attn): SiglipAttention( + (k_proj): Linear(in_features=1152, out_features=1152, bias=True) + (v_proj): Linear(in_features=1152, out_features=1152, bias=True) + (q_proj): Linear(in_features=1152, out_features=1152, bias=True) + (out_proj): Linear(in_features=1152, out_features=1152, bias=True) + ) + (layer_norm1): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + (layer_norm2): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + ) + ) + ) + (post_layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (head): SiglipMultiheadAttentionPoolingHead( + (attention): MultiheadAttention( + (out_proj): NonDynamicallyQuantizableLinear(in_features=1152, out_features=1152, bias=True) + ) + (layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + ) + ) + ) + ) + (connector): MLPConnector( + (_connector): Sequential( + (0): Linear(in_features=1152, out_features=896, bias=True) + (1): GELU(approximate='none') + (2): Linear(in_features=896, out_features=896, bias=True) + ) + ) +) +Collect masks for language model over. +Collect masks for connector over. +Applying mask on model.layers.0.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.self_attn.q_proj. +Applying mask on model.layers.0.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.self_attn.k_proj. +Applying mask on model.layers.0.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.self_attn.v_proj. +Applying mask on model.layers.0.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.self_attn.o_proj. +Applying mask on model.layers.0.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.mlp.gate_proj. +Applying mask on model.layers.0.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.mlp.up_proj. +Applying mask on model.layers.0.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.mlp.down_proj. +Applying mask on model.layers.1.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.self_attn.q_proj. +Applying mask on model.layers.1.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.self_attn.k_proj. +Applying mask on model.layers.1.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.self_attn.v_proj. +Applying mask on model.layers.1.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.self_attn.o_proj. +Applying mask on model.layers.1.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.mlp.gate_proj. +Applying mask on model.layers.1.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.mlp.up_proj. +Applying mask on model.layers.1.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.mlp.down_proj. +Applying mask on model.layers.2.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.self_attn.q_proj. +Applying mask on model.layers.2.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.self_attn.k_proj. +Applying mask on model.layers.2.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.self_attn.v_proj. +Applying mask on model.layers.2.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.self_attn.o_proj. +Applying mask on model.layers.2.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.mlp.gate_proj. +Applying mask on model.layers.2.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.mlp.up_proj. +Applying mask on model.layers.2.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.mlp.down_proj. +Applying mask on model.layers.3.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.self_attn.q_proj. +Applying mask on model.layers.3.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.self_attn.k_proj. +Applying mask on model.layers.3.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.self_attn.v_proj. +Applying mask on model.layers.3.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.self_attn.o_proj. +Applying mask on model.layers.3.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.mlp.gate_proj. +Applying mask on model.layers.3.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.mlp.up_proj. +Applying mask on model.layers.3.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.mlp.down_proj. +Applying mask on model.layers.4.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.self_attn.q_proj. +Applying mask on model.layers.4.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.self_attn.k_proj. +Applying mask on model.layers.4.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.self_attn.v_proj. +Applying mask on model.layers.4.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.self_attn.o_proj. +Applying mask on model.layers.4.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.mlp.gate_proj. +Applying mask on model.layers.4.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.mlp.up_proj. +Applying mask on model.layers.4.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.mlp.down_proj. +Applying mask on model.layers.5.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.self_attn.q_proj. +Applying mask on model.layers.5.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.self_attn.k_proj. +Applying mask on model.layers.5.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.self_attn.v_proj. +Applying mask on model.layers.5.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.self_attn.o_proj. +Applying mask on model.layers.5.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.mlp.gate_proj. +Applying mask on model.layers.5.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.mlp.up_proj. +Applying mask on model.layers.5.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.mlp.down_proj. +Applying mask on model.layers.6.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.self_attn.q_proj. +Applying mask on model.layers.6.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.self_attn.k_proj. +Applying mask on model.layers.6.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.self_attn.v_proj. +Applying mask on model.layers.6.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.self_attn.o_proj. +Applying mask on model.layers.6.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.mlp.gate_proj. +Applying mask on model.layers.6.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.mlp.up_proj. +Applying mask on model.layers.6.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.mlp.down_proj. +Applying mask on model.layers.7.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.self_attn.q_proj. +Applying mask on model.layers.7.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.self_attn.k_proj. +Applying mask on model.layers.7.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.self_attn.v_proj. +Applying mask on model.layers.7.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.self_attn.o_proj. +Applying mask on model.layers.7.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.mlp.gate_proj. +Applying mask on model.layers.7.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.mlp.up_proj. +Applying mask on model.layers.7.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.mlp.down_proj. +Applying mask on model.layers.8.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.self_attn.q_proj. +Applying mask on model.layers.8.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.self_attn.k_proj. +Applying mask on model.layers.8.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.self_attn.v_proj. +Applying mask on model.layers.8.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.self_attn.o_proj. +Applying mask on model.layers.8.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.mlp.gate_proj. +Applying mask on model.layers.8.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.mlp.up_proj. +Applying mask on model.layers.8.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.mlp.down_proj. +Applying mask on model.layers.9.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.self_attn.q_proj. +Applying mask on model.layers.9.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.self_attn.k_proj. +Applying mask on model.layers.9.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.self_attn.v_proj. +Applying mask on model.layers.9.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.self_attn.o_proj. +Applying mask on model.layers.9.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.mlp.gate_proj. +Applying mask on model.layers.9.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.mlp.up_proj. +Applying mask on model.layers.9.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.mlp.down_proj. +Applying mask on model.layers.10.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.self_attn.q_proj. +Applying mask on model.layers.10.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.self_attn.k_proj. +Applying mask on model.layers.10.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.self_attn.v_proj. +Applying mask on model.layers.10.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.self_attn.o_proj. +Applying mask on model.layers.10.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.mlp.gate_proj. +Applying mask on model.layers.10.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.mlp.up_proj. +Applying mask on model.layers.10.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.mlp.down_proj. +Applying mask on model.layers.11.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.self_attn.q_proj. +Applying mask on model.layers.11.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.self_attn.k_proj. +Applying mask on model.layers.11.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.self_attn.v_proj. +Applying mask on model.layers.11.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.self_attn.o_proj. +Applying mask on model.layers.11.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.mlp.gate_proj. +Applying mask on model.layers.11.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.mlp.up_proj. +Applying mask on model.layers.11.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.mlp.down_proj. +Applying mask on model.layers.12.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.self_attn.q_proj. +Applying mask on model.layers.12.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.self_attn.k_proj. +Applying mask on model.layers.12.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.self_attn.v_proj. +Applying mask on model.layers.12.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.self_attn.o_proj. +Applying mask on model.layers.12.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.mlp.gate_proj. +Applying mask on model.layers.12.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.mlp.up_proj. +Applying mask on model.layers.12.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.mlp.down_proj. +Applying mask on model.layers.13.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.self_attn.q_proj. +Applying mask on model.layers.13.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.self_attn.k_proj. +Applying mask on model.layers.13.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.self_attn.v_proj. +Applying mask on model.layers.13.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.self_attn.o_proj. +Applying mask on model.layers.13.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.mlp.gate_proj. +Applying mask on model.layers.13.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.mlp.up_proj. +Applying mask on model.layers.13.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.mlp.down_proj. +Applying mask on model.layers.14.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.self_attn.q_proj. +Applying mask on model.layers.14.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.self_attn.k_proj. +Applying mask on model.layers.14.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.self_attn.v_proj. +Applying mask on model.layers.14.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.self_attn.o_proj. +Applying mask on model.layers.14.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.mlp.gate_proj. +Applying mask on model.layers.14.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.mlp.up_proj. +Applying mask on model.layers.14.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.mlp.down_proj. +Applying mask on model.layers.15.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.self_attn.q_proj. +Applying mask on model.layers.15.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.self_attn.k_proj. +Applying mask on model.layers.15.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.self_attn.v_proj. +Applying mask on model.layers.15.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.self_attn.o_proj. +Applying mask on model.layers.15.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.mlp.gate_proj. +Applying mask on model.layers.15.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.mlp.up_proj. +Applying mask on model.layers.15.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.mlp.down_proj. +Applying mask on model.layers.16.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.self_attn.q_proj. +Applying mask on model.layers.16.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.self_attn.k_proj. +Applying mask on model.layers.16.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.self_attn.v_proj. +Applying mask on model.layers.16.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.self_attn.o_proj. +Applying mask on model.layers.16.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.mlp.gate_proj. +Applying mask on model.layers.16.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.mlp.up_proj. +Applying mask on model.layers.16.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.mlp.down_proj. +Applying mask on model.layers.17.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.self_attn.q_proj. +Applying mask on model.layers.17.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.self_attn.k_proj. +Applying mask on model.layers.17.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.self_attn.v_proj. +Applying mask on model.layers.17.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.self_attn.o_proj. +Applying mask on model.layers.17.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.mlp.gate_proj. +Applying mask on model.layers.17.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.mlp.up_proj. +Applying mask on model.layers.17.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.mlp.down_proj. +Applying mask on model.layers.18.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.self_attn.q_proj. +Applying mask on model.layers.18.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.self_attn.k_proj. +Applying mask on model.layers.18.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.self_attn.v_proj. +Applying mask on model.layers.18.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.self_attn.o_proj. +Applying mask on model.layers.18.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.mlp.gate_proj. +Applying mask on model.layers.18.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.mlp.up_proj. +Applying mask on model.layers.18.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.mlp.down_proj. +Applying mask on model.layers.19.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.self_attn.q_proj. +Applying mask on model.layers.19.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.self_attn.k_proj. +Applying mask on model.layers.19.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.self_attn.v_proj. +Applying mask on model.layers.19.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.self_attn.o_proj. +Applying mask on model.layers.19.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.mlp.gate_proj. +Applying mask on model.layers.19.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.mlp.up_proj. +Applying mask on model.layers.19.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.mlp.down_proj. +Applying mask on model.layers.20.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.self_attn.q_proj. +Applying mask on model.layers.20.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.self_attn.k_proj. +Applying mask on model.layers.20.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.self_attn.v_proj. +Applying mask on model.layers.20.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.self_attn.o_proj. +Applying mask on model.layers.20.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.mlp.gate_proj. +Applying mask on model.layers.20.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.mlp.up_proj. +Applying mask on model.layers.20.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.mlp.down_proj. +Applying mask on model.layers.21.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.self_attn.q_proj. +Applying mask on model.layers.21.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.self_attn.k_proj. +Applying mask on model.layers.21.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.self_attn.v_proj. +Applying mask on model.layers.21.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.self_attn.o_proj. +Applying mask on model.layers.21.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.mlp.gate_proj. +Applying mask on model.layers.21.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.mlp.up_proj. +Applying mask on model.layers.21.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.mlp.down_proj. +Applying mask on model.layers.22.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.self_attn.q_proj. +Applying mask on model.layers.22.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.self_attn.k_proj. +Applying mask on model.layers.22.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.self_attn.v_proj. +Applying mask on model.layers.22.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.self_attn.o_proj. +Applying mask on model.layers.22.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.mlp.gate_proj. +Applying mask on model.layers.22.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.mlp.up_proj. +Applying mask on model.layers.22.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.mlp.down_proj. +Applying mask on model.layers.23.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.self_attn.q_proj. +Applying mask on model.layers.23.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.self_attn.k_proj. +Applying mask on model.layers.23.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.self_attn.v_proj. +Applying mask on model.layers.23.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.self_attn.o_proj. +Applying mask on model.layers.23.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.mlp.gate_proj. +Applying mask on model.layers.23.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.mlp.up_proj. +Applying mask on model.layers.23.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.mlp.down_proj. +Applying mask on _connector.0 with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on _connector.0. +Applying mask on _connector.2 with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on _connector.2. +Using cleaned config_mask (without mask parameters) for saving. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 21:11:57,686] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. + 0%| | 0/900 [00:00", + "resampler_hidden_size": 768, + "sparsity_connector": null, + "subnet_type_connector": "global", + "temperature_connector": 1.1, + "text_config": { + "_name_or_path": "Qwen/Qwen2.5-0.5B", + "architectures": [ + "Qwen2ForCausalLM" + ], + "backward_type": "normal", + "bos_token_id": 151643, + "eos_token_id": 151643, + "hidden_size": 896, + "intermediate_size": 4864, + "mask_type": "soft", + "masked_layers": "all", + "max_position_embeddings": 32768, + "max_window_layers": 24, + "model_type": "qwen2", + "num_attention_heads": 14, + "num_hidden_layers": 24, + "num_key_value_heads": 2, + "rope_theta": 1000000.0, + "sliding_window": 32768, + "subnet_mode": "both", + "subnet_type": "None", + "temperature_attn": 1.1, + "temperature_mlp": 1.1, + "tie_word_embeddings": true, + "torch_dtype": "bfloat16", + "use_mrope": false, + "use_sliding_window": false, + "vocab_size": 151936 + }, + "threshold_connector": null, + "tokenizer_model_max_length": 2048, + "tokenizer_name_or_path": "Qwen/Qwen2.5-0.5B", + "tokenizer_padding_side": "right", + "tokenizer_use_fast": false, + "torch_dtype": "bfloat16", + "transformers_version": "4.40.1", + "tune_type_connector": "full", + "tune_type_llm": "full", + "tune_type_vision_tower": "frozen", + "tune_vision_tower_from_layer": 0, + "use_cache": true, + "vision_config": { + "hidden_act": "gelu_pytorch_tanh", + "hidden_size": 1152, + "image_size": 384, + "intermediate_size": 4304, + "layer_norm_eps": 1e-06, + "model_name_or_path": "google/siglip-so400m-patch14-384", + "model_name_or_path2": "", + "model_type": "siglip_vision_model", + "num_attention_heads": 16, + "num_hidden_layers": 27, + "patch_size": 14 + }, + "vision_feature_layer": -2, + "vision_feature_select_strategy": "patch", + "vision_hidden_size": 1152, + "vision_model_name_or_path": "google/siglip-so400m-patch14-384", + "vision_model_name_or_path2": "", + "vocab_size": 151936 +} + +TinyLlavaForConditionalGeneration( + (language_model): Qwen2ForCausalLM( + (model): Qwen2Model( + (embed_tokens): Embedding(151936, 896) + (layers): ModuleList( + (0-23): 24 x Qwen2DecoderLayer( + (self_attn): Qwen2Attention( + (q_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=True) + (k_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=128, bias=True) + (v_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=128, bias=True) + (o_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=False) + (rotary_emb): Qwen2RotaryEmbedding() + ) + (mlp): Qwen2MLP( + (gate_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=4864, bias=False) + (up_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=4864, bias=False) + (down_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=4864, out_features=896, bias=False) + (act_fn): SiLU() + ) + (input_layernorm): Qwen2RMSNorm() + (post_attention_layernorm): Qwen2RMSNorm() + ) + ) + (norm): Qwen2RMSNorm() + ) + (lm_head): Linear(in_features=896, out_features=151936, bias=False) + ) + (vision_tower): SIGLIPVisionTower( + (_vision_tower): SiglipVisionModel( + (vision_model): SiglipVisionTransformer( + (embeddings): SiglipVisionEmbeddings( + (patch_embedding): Conv2d(3, 1152, kernel_size=(14, 14), stride=(14, 14), padding=valid) + (position_embedding): Embedding(729, 1152) + ) + (encoder): SiglipEncoder( + (layers): ModuleList( + (0-26): 27 x SiglipEncoderLayer( + (self_attn): SiglipAttention( + (k_proj): Linear(in_features=1152, out_features=1152, bias=True) + (v_proj): Linear(in_features=1152, out_features=1152, bias=True) + (q_proj): Linear(in_features=1152, out_features=1152, bias=True) + (out_proj): Linear(in_features=1152, out_features=1152, bias=True) + ) + (layer_norm1): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + (layer_norm2): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + ) + ) + ) + (post_layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (head): SiglipMultiheadAttentionPoolingHead( + (attention): MultiheadAttention( + (out_proj): NonDynamicallyQuantizableLinear(in_features=1152, out_features=1152, bias=True) + ) + (layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + ) + ) + ) + ) + (connector): MLPConnector( + (_connector): Sequential( + (0): SupermaskLinearSparsity_SoftForward_Normal(in_features=1152, out_features=896, bias=True) + (1): GELU(approximate='none') + (2): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=True) + ) + ) +) +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Load base model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain over. +TinyLlavaConfig { + "cache_dir": null, + "connector_type": "mlp2x_gelu", + "hidden_size": 896, + "ignore_index": -100, + "image_aspect_ratio": "square", + "image_token_index": -200, + "llm_model_name_or_path": "Qwen/Qwen2.5-0.5B", + "model_type": "tinyllava", + "num_queries": 128, + "num_resampler_layers": 3, + "pad_token": "<|endoftext|>", + "pad_token_id": 151643, + "resampler_hidden_size": 768, + "text_config": { + "_name_or_path": "Qwen/Qwen2.5-0.5B", + "architectures": [ + "Qwen2ForCausalLM" + ], + "bos_token_id": 151643, + "eos_token_id": 151643, + "hidden_size": 896, + "intermediate_size": 4864, + "max_position_embeddings": 32768, + "max_window_layers": 24, + "model_type": "qwen2", + "num_attention_heads": 14, + "num_hidden_layers": 24, + "num_key_value_heads": 2, + "rope_theta": 1000000.0, + "sliding_window": 32768, + "tie_word_embeddings": true, + "use_mrope": false, + "use_sliding_window": false, + "vocab_size": 151936 + }, + "tokenizer_model_max_length": 2048, + "tokenizer_name_or_path": "Qwen/Qwen2.5-0.5B", + "tokenizer_padding_side": "right", + "tokenizer_use_fast": false, + "transformers_version": "4.40.1", + "tune_type_connector": "full", + "tune_type_llm": "frozen", + "tune_type_vision_tower": "frozen", + "tune_vision_tower_from_layer": 0, + "use_cache": true, + "vision_config": { + "hidden_act": "gelu_pytorch_tanh", + "hidden_size": 1152, + "image_size": 384, + "intermediate_size": 4304, + "layer_norm_eps": 1e-06, + "model_name_or_path": "google/siglip-so400m-patch14-384", + "model_name_or_path2": "", + "model_type": "siglip_vision_model", + "num_attention_heads": 16, + "num_hidden_layers": 27, + "patch_size": 14 + }, + "vision_feature_layer": -2, + "vision_feature_select_strategy": "patch", + "vision_hidden_size": 1152, + "vision_model_name_or_path": "google/siglip-so400m-patch14-384", + "vision_model_name_or_path2": "", + "vocab_size": 151936 +} + +TinyLlavaForConditionalGeneration( + (language_model): Qwen2ForCausalLM( + (model): Qwen2Model( + (embed_tokens): Embedding(151936, 896) + (layers): ModuleList( + (0-23): 24 x Qwen2DecoderLayer( + (self_attn): Qwen2Attention( + (q_proj): Linear(in_features=896, out_features=896, bias=True) + (k_proj): Linear(in_features=896, out_features=128, bias=True) + (v_proj): Linear(in_features=896, out_features=128, bias=True) + (o_proj): Linear(in_features=896, out_features=896, bias=False) + (rotary_emb): Qwen2RotaryEmbedding() + ) + (mlp): Qwen2MLP( + (gate_proj): Linear(in_features=896, out_features=4864, bias=False) + (up_proj): Linear(in_features=896, out_features=4864, bias=False) + (down_proj): Linear(in_features=4864, out_features=896, bias=False) + (act_fn): SiLU() + ) + (input_layernorm): Qwen2RMSNorm() + (post_attention_layernorm): Qwen2RMSNorm() + ) + ) + (norm): Qwen2RMSNorm() + ) + (lm_head): Linear(in_features=896, out_features=151936, bias=False) + ) + (vision_tower): SIGLIPVisionTower( + (_vision_tower): SiglipVisionModel( + (vision_model): SiglipVisionTransformer( + (embeddings): SiglipVisionEmbeddings( + (patch_embedding): Conv2d(3, 1152, kernel_size=(14, 14), stride=(14, 14), padding=valid) + (position_embedding): Embedding(729, 1152) + ) + (encoder): SiglipEncoder( + (layers): ModuleList( + (0-26): 27 x SiglipEncoderLayer( + (self_attn): SiglipAttention( + (k_proj): Linear(in_features=1152, out_features=1152, bias=True) + (v_proj): Linear(in_features=1152, out_features=1152, bias=True) + (q_proj): Linear(in_features=1152, out_features=1152, bias=True) + (out_proj): Linear(in_features=1152, out_features=1152, bias=True) + ) + (layer_norm1): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + (layer_norm2): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + ) + ) + ) + (post_layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (head): SiglipMultiheadAttentionPoolingHead( + (attention): MultiheadAttention( + (out_proj): NonDynamicallyQuantizableLinear(in_features=1152, out_features=1152, bias=True) + ) + (layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + ) + ) + ) + ) + (connector): MLPConnector( + (_connector): Sequential( + (0): Linear(in_features=1152, out_features=896, bias=True) + (1): GELU(approximate='none') + (2): Linear(in_features=896, out_features=896, bias=True) + ) + ) +) +Collect masks for language model over. +Collect masks for connector over. +Applying mask on model.layers.0.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.self_attn.q_proj. +Applying mask on model.layers.0.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.self_attn.k_proj. +Applying mask on model.layers.0.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.self_attn.v_proj. +Applying mask on model.layers.0.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.self_attn.o_proj. +Applying mask on model.layers.0.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.mlp.gate_proj. +Applying mask on model.layers.0.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.mlp.up_proj. +Applying mask on model.layers.0.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.0.mlp.down_proj. +Applying mask on model.layers.1.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.self_attn.q_proj. +Applying mask on model.layers.1.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.self_attn.k_proj. +Applying mask on model.layers.1.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.self_attn.v_proj. +Applying mask on model.layers.1.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.self_attn.o_proj. +Applying mask on model.layers.1.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.mlp.gate_proj. +Applying mask on model.layers.1.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.mlp.up_proj. +Applying mask on model.layers.1.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.1.mlp.down_proj. +Applying mask on model.layers.2.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.self_attn.q_proj. +Applying mask on model.layers.2.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.self_attn.k_proj. +Applying mask on model.layers.2.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.self_attn.v_proj. +Applying mask on model.layers.2.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.self_attn.o_proj. +Applying mask on model.layers.2.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.mlp.gate_proj. +Applying mask on model.layers.2.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.mlp.up_proj. +Applying mask on model.layers.2.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.2.mlp.down_proj. +Applying mask on model.layers.3.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.self_attn.q_proj. +Applying mask on model.layers.3.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.self_attn.k_proj. +Applying mask on model.layers.3.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.self_attn.v_proj. +Applying mask on model.layers.3.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.self_attn.o_proj. +Applying mask on model.layers.3.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.mlp.gate_proj. +Applying mask on model.layers.3.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.mlp.up_proj. +Applying mask on model.layers.3.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.3.mlp.down_proj. +Applying mask on model.layers.4.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.self_attn.q_proj. +Applying mask on model.layers.4.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.self_attn.k_proj. +Applying mask on model.layers.4.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.self_attn.v_proj. +Applying mask on model.layers.4.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.self_attn.o_proj. +Applying mask on model.layers.4.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.mlp.gate_proj. +Applying mask on model.layers.4.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.mlp.up_proj. +Applying mask on model.layers.4.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.4.mlp.down_proj. +Applying mask on model.layers.5.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.self_attn.q_proj. +Applying mask on model.layers.5.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.self_attn.k_proj. +Applying mask on model.layers.5.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.self_attn.v_proj. +Applying mask on model.layers.5.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.self_attn.o_proj. +Applying mask on model.layers.5.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.mlp.gate_proj. +Applying mask on model.layers.5.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.mlp.up_proj. +Applying mask on model.layers.5.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.5.mlp.down_proj. +Applying mask on model.layers.6.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.self_attn.q_proj. +Applying mask on model.layers.6.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.self_attn.k_proj. +Applying mask on model.layers.6.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.self_attn.v_proj. +Applying mask on model.layers.6.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.self_attn.o_proj. +Applying mask on model.layers.6.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.mlp.gate_proj. +Applying mask on model.layers.6.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.mlp.up_proj. +Applying mask on model.layers.6.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.6.mlp.down_proj. +Applying mask on model.layers.7.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.self_attn.q_proj. +Applying mask on model.layers.7.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.self_attn.k_proj. +Applying mask on model.layers.7.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.self_attn.v_proj. +Applying mask on model.layers.7.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.self_attn.o_proj. +Applying mask on model.layers.7.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.mlp.gate_proj. +Applying mask on model.layers.7.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.mlp.up_proj. +Applying mask on model.layers.7.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.7.mlp.down_proj. +Applying mask on model.layers.8.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.self_attn.q_proj. +Applying mask on model.layers.8.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.self_attn.k_proj. +Applying mask on model.layers.8.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.self_attn.v_proj. +Applying mask on model.layers.8.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.self_attn.o_proj. +Applying mask on model.layers.8.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.mlp.gate_proj. +Applying mask on model.layers.8.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.mlp.up_proj. +Applying mask on model.layers.8.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.8.mlp.down_proj. +Applying mask on model.layers.9.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.self_attn.q_proj. +Applying mask on model.layers.9.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.self_attn.k_proj. +Applying mask on model.layers.9.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.self_attn.v_proj. +Applying mask on model.layers.9.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.self_attn.o_proj. +Applying mask on model.layers.9.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.mlp.gate_proj. +Applying mask on model.layers.9.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.mlp.up_proj. +Applying mask on model.layers.9.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.9.mlp.down_proj. +Applying mask on model.layers.10.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.self_attn.q_proj. +Applying mask on model.layers.10.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.self_attn.k_proj. +Applying mask on model.layers.10.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.self_attn.v_proj. +Applying mask on model.layers.10.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.self_attn.o_proj. +Applying mask on model.layers.10.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.mlp.gate_proj. +Applying mask on model.layers.10.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.mlp.up_proj. +Applying mask on model.layers.10.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.10.mlp.down_proj. +Applying mask on model.layers.11.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.self_attn.q_proj. +Applying mask on model.layers.11.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.self_attn.k_proj. +Applying mask on model.layers.11.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.self_attn.v_proj. +Applying mask on model.layers.11.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.self_attn.o_proj. +Applying mask on model.layers.11.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.mlp.gate_proj. +Applying mask on model.layers.11.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.mlp.up_proj. +Applying mask on model.layers.11.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.11.mlp.down_proj. +Applying mask on model.layers.12.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.self_attn.q_proj. +Applying mask on model.layers.12.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.self_attn.k_proj. +Applying mask on model.layers.12.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.self_attn.v_proj. +Applying mask on model.layers.12.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.self_attn.o_proj. +Applying mask on model.layers.12.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.mlp.gate_proj. +Applying mask on model.layers.12.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.mlp.up_proj. +Applying mask on model.layers.12.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.12.mlp.down_proj. +Applying mask on model.layers.13.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.self_attn.q_proj. +Applying mask on model.layers.13.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.self_attn.k_proj. +Applying mask on model.layers.13.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.self_attn.v_proj. +Applying mask on model.layers.13.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.self_attn.o_proj. +Applying mask on model.layers.13.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.mlp.gate_proj. +Applying mask on model.layers.13.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.mlp.up_proj. +Applying mask on model.layers.13.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.13.mlp.down_proj. +Applying mask on model.layers.14.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.self_attn.q_proj. +Applying mask on model.layers.14.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.self_attn.k_proj. +Applying mask on model.layers.14.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.self_attn.v_proj. +Applying mask on model.layers.14.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.self_attn.o_proj. +Applying mask on model.layers.14.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.mlp.gate_proj. +Applying mask on model.layers.14.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.mlp.up_proj. +Applying mask on model.layers.14.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.14.mlp.down_proj. +Applying mask on model.layers.15.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.self_attn.q_proj. +Applying mask on model.layers.15.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.self_attn.k_proj. +Applying mask on model.layers.15.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.self_attn.v_proj. +Applying mask on model.layers.15.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.self_attn.o_proj. +Applying mask on model.layers.15.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.mlp.gate_proj. +Applying mask on model.layers.15.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.mlp.up_proj. +Applying mask on model.layers.15.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.15.mlp.down_proj. +Applying mask on model.layers.16.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.self_attn.q_proj. +Applying mask on model.layers.16.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.self_attn.k_proj. +Applying mask on model.layers.16.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.self_attn.v_proj. +Applying mask on model.layers.16.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.self_attn.o_proj. +Applying mask on model.layers.16.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.mlp.gate_proj. +Applying mask on model.layers.16.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.mlp.up_proj. +Applying mask on model.layers.16.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.16.mlp.down_proj. +Applying mask on model.layers.17.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.self_attn.q_proj. +Applying mask on model.layers.17.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.self_attn.k_proj. +Applying mask on model.layers.17.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.self_attn.v_proj. +Applying mask on model.layers.17.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.self_attn.o_proj. +Applying mask on model.layers.17.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.mlp.gate_proj. +Applying mask on model.layers.17.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.mlp.up_proj. +Applying mask on model.layers.17.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.17.mlp.down_proj. +Applying mask on model.layers.18.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.self_attn.q_proj. +Applying mask on model.layers.18.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.self_attn.k_proj. +Applying mask on model.layers.18.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.self_attn.v_proj. +Applying mask on model.layers.18.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.self_attn.o_proj. +Applying mask on model.layers.18.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.mlp.gate_proj. +Applying mask on model.layers.18.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.mlp.up_proj. +Applying mask on model.layers.18.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.18.mlp.down_proj. +Applying mask on model.layers.19.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.self_attn.q_proj. +Applying mask on model.layers.19.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.self_attn.k_proj. +Applying mask on model.layers.19.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.self_attn.v_proj. +Applying mask on model.layers.19.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.self_attn.o_proj. +Applying mask on model.layers.19.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.mlp.gate_proj. +Applying mask on model.layers.19.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.mlp.up_proj. +Applying mask on model.layers.19.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.19.mlp.down_proj. +Applying mask on model.layers.20.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.self_attn.q_proj. +Applying mask on model.layers.20.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.self_attn.k_proj. +Applying mask on model.layers.20.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.self_attn.v_proj. +Applying mask on model.layers.20.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.self_attn.o_proj. +Applying mask on model.layers.20.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.mlp.gate_proj. +Applying mask on model.layers.20.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.mlp.up_proj. +Applying mask on model.layers.20.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.20.mlp.down_proj. +Applying mask on model.layers.21.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.self_attn.q_proj. +Applying mask on model.layers.21.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.self_attn.k_proj. +Applying mask on model.layers.21.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.self_attn.v_proj. +Applying mask on model.layers.21.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.self_attn.o_proj. +Applying mask on model.layers.21.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.mlp.gate_proj. +Applying mask on model.layers.21.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.mlp.up_proj. +Applying mask on model.layers.21.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.21.mlp.down_proj. +Applying mask on model.layers.22.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.self_attn.q_proj. +Applying mask on model.layers.22.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.self_attn.k_proj. +Applying mask on model.layers.22.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.self_attn.v_proj. +Applying mask on model.layers.22.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.self_attn.o_proj. +Applying mask on model.layers.22.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.mlp.gate_proj. +Applying mask on model.layers.22.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.mlp.up_proj. +Applying mask on model.layers.22.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.22.mlp.down_proj. +Applying mask on model.layers.23.self_attn.q_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.self_attn.q_proj. +Applying mask on model.layers.23.self_attn.k_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.self_attn.k_proj. +Applying mask on model.layers.23.self_attn.v_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.self_attn.v_proj. +Applying mask on model.layers.23.self_attn.o_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.self_attn.o_proj. +Applying mask on model.layers.23.mlp.gate_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.mlp.gate_proj. +Applying mask on model.layers.23.mlp.up_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.mlp.up_proj. +Applying mask on model.layers.23.mlp.down_proj with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on model.layers.23.mlp.down_proj. +Applying mask on _connector.0 with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on _connector.0. +Applying mask on _connector.2 with dtype, mask_dtype=torch.bfloat16, module_dtype=torch.bfloat16 +Applied soft mask on _connector.2. +Using cleaned config_mask (without mask parameters) for saving. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 21:36:51,008] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. + 0%| | 0/900 [00:00 + main() + File "/nfs/ywang29/TinyLLaVA/scripts/apply_masks.py", line 123, in main + config_mask = TinyLlavaConfig.from_pretrained(model_args.mask_model_name_or_path) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/configuration_utils.py", line 602, in from_pretrained + config_dict, kwargs = cls.get_config_dict(pretrained_model_name_or_path, **kwargs) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/configuration_utils.py", line 631, in get_config_dict + config_dict, kwargs = cls._get_config_dict(pretrained_model_name_or_path, **kwargs) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/configuration_utils.py", line 686, in _get_config_dict + resolved_config_file = cached_file( + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/utils/hub.py", line 369, in cached_file + raise EnvironmentError( +OSError: /nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.3_2e-1_connector-3.0_1.3_2e-1_ablation does not appear to have a file named config.json. Checkout 'https://huggingface.co//nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.3_2e-1_connector-3.0_1.3_2e-1_ablation/tree/main' for available files. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 22:01:25,210] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +Traceback (most recent call last): + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/utils/hub.py", line 398, in cached_file + resolved_file = hf_hub_download( + File "/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/utils/_validators.py", line 106, in _inner_fn + validate_repo_id(arg_value) + File "/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/utils/_validators.py", line 154, in validate_repo_id + raise HFValidationError( +huggingface_hub.errors.HFValidationError: Repo id must be in the form 'repo_name' or 'namespace/repo_name': '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.3_2e-1_connector-3.0_1.3_2e-1_ablation/mask_applied'. Use `repo_type` argument if needed. + +The above exception was the direct cause of the following exception: + +Traceback (most recent call last): + File "/nfs/ywang29/TinyLLaVA/tinyllava/model/load_model.py", line 38, in load_pretrained_model + model = TinyLlavaForConditionalGeneration.from_pretrained(model_name_or_path,low_cpu_mem_usage=True) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/modeling_utils.py", line 3015, in from_pretrained + resolved_config_file = cached_file( + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/utils/hub.py", line 462, in cached_file + raise EnvironmentError( +OSError: Incorrect path_or_model_id: '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.3_2e-1_connector-3.0_1.3_2e-1_ablation/mask_applied'. Please provide either the path to a local folder or the repo_id of a model on the Hub. + +During handling of the above exception, another exception occurred: + +Traceback (most recent call last): + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/utils/hub.py", line 398, in cached_file + resolved_file = hf_hub_download( + File "/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/utils/_validators.py", line 106, in _inner_fn + validate_repo_id(arg_value) + File "/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/utils/_validators.py", line 154, in validate_repo_id + raise HFValidationError( +huggingface_hub.errors.HFValidationError: Repo id must be in the form 'repo_name' or 'namespace/repo_name': '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.3_2e-1_connector-3.0_1.3_2e-1_ablation/mask_applied'. Use `repo_type` argument if needed. + +The above exception was the direct cause of the following exception: + +Traceback (most recent call last): + File "/opt/conda/envs/tinyllava/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/opt/conda/envs/tinyllava/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/nfs/ywang29/TinyLLaVA/tinyllava/eval/model_vqa_mmmu.py", line 180, in + eval_model(args) + File "/nfs/ywang29/TinyLLaVA/tinyllava/eval/model_vqa_mmmu.py", line 88, in eval_model + model, tokenizer, image_processor, context_len = load_pretrained_model(model_path) + File "/nfs/ywang29/TinyLLaVA/tinyllava/model/load_model.py", line 40, in load_pretrained_model + model_config = TinyLlavaConfig.from_pretrained(model_name_or_path) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/configuration_utils.py", line 602, in from_pretrained + config_dict, kwargs = cls.get_config_dict(pretrained_model_name_or_path, **kwargs) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/configuration_utils.py", line 631, in get_config_dict + config_dict, kwargs = cls._get_config_dict(pretrained_model_name_or_path, **kwargs) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/configuration_utils.py", line 686, in _get_config_dict + resolved_config_file = cached_file( + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/utils/hub.py", line 462, in cached_file + raise EnvironmentError( +OSError: Incorrect path_or_model_id: '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.3_2e-1_connector-3.0_1.3_2e-1_ablation/mask_applied'. Please provide either the path to a local folder or the repo_id of a model on the Hub. +Traceback (most recent call last): + File "/nfs/ywang29/TinyLLaVA/scripts/convert_answer_to_mmmu.py", line 31, in + eval_model(args) + File "/nfs/ywang29/TinyLLaVA/scripts/convert_answer_to_mmmu.py", line 7, in eval_model + answers = [json.loads(q) for q in open(os.path.expanduser(args.answers_file), "r")] +FileNotFoundError: [Errno 2] No such file or directory: '/s3-code/ywang29/datasets/tinyllava/eval/MMMU/answers/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.3_2e-1_connector-3.0_1.3_2e-1_ablation-mask_applied.jsonl' +Traceback (most recent call last): + File "/s3-code/ywang29/datasets/tinyllava/eval/MMMU/eval/main_eval_only.py", line 19, in + output_dict = json.load(open(args.output_path)) +FileNotFoundError: [Errno 2] No such file or directory: '/s3-code/ywang29/datasets/tinyllava/eval/MMMU/answers/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.3_2e-1_connector-3.0_1.3_2e-1_ablation-mask_applied_output.json' +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 22:01:31,630] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +Traceback (most recent call last): + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/utils/hub.py", line 398, in cached_file + resolved_file = hf_hub_download( + File "/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/utils/_validators.py", line 106, in _inner_fn + validate_repo_id(arg_value) + File "/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/utils/_validators.py", line 154, in validate_repo_id + raise HFValidationError( +huggingface_hub.errors.HFValidationError: Repo id must be in the form 'repo_name' or 'namespace/repo_name': '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.3_2e-1_connector-3.0_1.3_2e-1_ablation/mask_applied'. Use `repo_type` argument if needed. + +The above exception was the direct cause of the following exception: + +Traceback (most recent call last): + File "/nfs/ywang29/TinyLLaVA/tinyllava/model/load_model.py", line 38, in load_pretrained_model + model = TinyLlavaForConditionalGeneration.from_pretrained(model_name_or_path,low_cpu_mem_usage=True) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/modeling_utils.py", line 3015, in from_pretrained + resolved_config_file = cached_file( + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/utils/hub.py", line 462, in cached_file + raise EnvironmentError( +OSError: Incorrect path_or_model_id: '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.3_2e-1_connector-3.0_1.3_2e-1_ablation/mask_applied'. Please provide either the path to a local folder or the repo_id of a model on the Hub. + +During handling of the above exception, another exception occurred: + +Traceback (most recent call last): + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/utils/hub.py", line 398, in cached_file + resolved_file = hf_hub_download( + File "/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/utils/_validators.py", line 106, in _inner_fn + validate_repo_id(arg_value) + File "/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/utils/_validators.py", line 154, in validate_repo_id + raise HFValidationError( +huggingface_hub.errors.HFValidationError: Repo id must be in the form 'repo_name' or 'namespace/repo_name': '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.3_2e-1_connector-3.0_1.3_2e-1_ablation/mask_applied'. Use `repo_type` argument if needed. + +The above exception was the direct cause of the following exception: + +Traceback (most recent call last): + File "/opt/conda/envs/tinyllava/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/opt/conda/envs/tinyllava/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/nfs/ywang29/TinyLLaVA/tinyllava/eval/model_vqa_loader.py", line 146, in + eval_model(args) + File "/nfs/ywang29/TinyLLaVA/tinyllava/eval/model_vqa_loader.py", line 79, in eval_model + model, tokenizer, image_processor, context_len = load_pretrained_model(model_path) + File "/nfs/ywang29/TinyLLaVA/tinyllava/model/load_model.py", line 40, in load_pretrained_model + model_config = TinyLlavaConfig.from_pretrained(model_name_or_path) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/configuration_utils.py", line 602, in from_pretrained + config_dict, kwargs = cls.get_config_dict(pretrained_model_name_or_path, **kwargs) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/configuration_utils.py", line 631, in get_config_dict + config_dict, kwargs = cls._get_config_dict(pretrained_model_name_or_path, **kwargs) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/configuration_utils.py", line 686, in _get_config_dict + resolved_config_file = cached_file( + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/utils/hub.py", line 462, in cached_file + raise EnvironmentError( +OSError: Incorrect path_or_model_id: '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.3_2e-1_connector-3.0_1.3_2e-1_ablation/mask_applied'. Please provide either the path to a local folder or the repo_id of a model on the Hub. +Traceback (most recent call last): + File "/s3-code/ywang29/datasets/tinyllava/eval/MME/convert_answer_to_mme.py", line 52, in + answers = [json.loads(line) for line in open(os.path.join('answers', f'{experiment}.jsonl'))] +FileNotFoundError: [Errno 2] No such file or directory: 'answers/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.3_2e-1_connector-3.0_1.3_2e-1_ablation-mask_applied.jsonl' +=========== Perception =========== +Traceback (most recent call last): + File "/s3-code/ywang29/datasets/tinyllava/eval/MME/eval_tool/calculation.py", line 164, in + cal.process_result(results_dir) + File "/s3-code/ywang29/datasets/tinyllava/eval/MME/eval_tool/calculation.py", line 98, in process_result + lines = open(task_txt, 'r').readlines() +FileNotFoundError: [Errno 2] No such file or directory: 'answers/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.3_2e-1_connector-3.0_1.3_2e-1_ablation-mask_applied/existence.txt' +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 22:02:36,588] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +Traceback (most recent call last): + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/utils/hub.py", line 398, in cached_file + resolved_file = hf_hub_download( + File "/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/utils/_validators.py", line 106, in _inner_fn + validate_repo_id(arg_value) + File "/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/utils/_validators.py", line 154, in validate_repo_id + raise HFValidationError( +huggingface_hub.errors.HFValidationError: Repo id must be in the form 'repo_name' or 'namespace/repo_name': '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.3_2e-1_connector-3.0_1.3_2e-1_ablation/mask_applied'. Use `repo_type` argument if needed. + +The above exception was the direct cause of the following exception: + +Traceback (most recent call last): + File "/nfs/ywang29/TinyLLaVA/tinyllava/model/load_model.py", line 38, in load_pretrained_model + model = TinyLlavaForConditionalGeneration.from_pretrained(model_name_or_path,low_cpu_mem_usage=True) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/modeling_utils.py", line 3015, in from_pretrained + resolved_config_file = cached_file( + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/utils/hub.py", line 462, in cached_file + raise EnvironmentError( +OSError: Incorrect path_or_model_id: '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.3_2e-1_connector-3.0_1.3_2e-1_ablation/mask_applied'. Please provide either the path to a local folder or the repo_id of a model on the Hub. + +During handling of the above exception, another exception occurred: + +Traceback (most recent call last): + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/utils/hub.py", line 398, in cached_file + resolved_file = hf_hub_download( + File "/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/utils/_validators.py", line 106, in _inner_fn + validate_repo_id(arg_value) + File "/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/utils/_validators.py", line 154, in validate_repo_id + raise HFValidationError( +huggingface_hub.errors.HFValidationError: Repo id must be in the form 'repo_name' or 'namespace/repo_name': '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.3_2e-1_connector-3.0_1.3_2e-1_ablation/mask_applied'. Use `repo_type` argument if needed. + +The above exception was the direct cause of the following exception: + +Traceback (most recent call last): + File "/opt/conda/envs/tinyllava/lib/python3.10/runpy.py", line 196, in _run_module_as_main + return _run_code(code, main_globals, None, + File "/opt/conda/envs/tinyllava/lib/python3.10/runpy.py", line 86, in _run_code + exec(code, run_globals) + File "/nfs/ywang29/TinyLLaVA/tinyllava/eval/model_vqa_science.py", line 108, in + eval_model(args) + File "/nfs/ywang29/TinyLLaVA/tinyllava/eval/model_vqa_science.py", line 31, in eval_model + model, tokenizer, image_processor, context_len = load_pretrained_model(model_path) + File "/nfs/ywang29/TinyLLaVA/tinyllava/model/load_model.py", line 40, in load_pretrained_model + model_config = TinyLlavaConfig.from_pretrained(model_name_or_path) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/configuration_utils.py", line 602, in from_pretrained + config_dict, kwargs = cls.get_config_dict(pretrained_model_name_or_path, **kwargs) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/configuration_utils.py", line 631, in get_config_dict + config_dict, kwargs = cls._get_config_dict(pretrained_model_name_or_path, **kwargs) + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/configuration_utils.py", line 686, in _get_config_dict + resolved_config_file = cached_file( + File "/nfs/ywang29/TinyLLaVA/transformers/src/transformers/utils/hub.py", line 462, in cached_file + raise EnvironmentError( +OSError: Incorrect path_or_model_id: '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.3_2e-1_connector-3.0_1.3_2e-1_ablation/mask_applied'. Please provide either the path to a local folder or the repo_id of a model on the Hub. +Traceback (most recent call last): + File "/nfs/ywang29/TinyLLaVA/tinyllava/eval/eval_science_qa.py", line 45, in + predictions = [json.loads(line) for line in open(args.result_file)] +FileNotFoundError: [Errno 2] No such file or directory: '/s3-code/ywang29/datasets/tinyllava/eval/scienceqa/answers/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.3_2e-1_connector-3.0_1.3_2e-1_ablation-mask_applied.jsonl' +==== EXPERIMENT COMPLETED: eval_qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.3_2e-1_connector-3.0_1.3_2e-1_ablation ==== +Log File: eval_qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.3_2e-1_connector-3.0_1.3_2e-1_ablation_20251012_220114.log +Timestamp: 2025-10-12 22:02:39 +===================================== diff --git a/logs_oct11/logs_oct10.tar b/logs_oct11/logs_oct10.tar new file mode 100644 index 0000000000000000000000000000000000000000..84c827a750c94227e79a09bc16e12998fbbfae4a --- /dev/null +++ b/logs_oct11/logs_oct10.tar @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:49b2a4f137f8b5451c7553bd7aff99979ef50535267bd54c9681bcbc8c342cde +size 917335 diff --git a/logs_oct11/logs_oct9.tar b/logs_oct11/logs_oct9.tar new file mode 100644 index 0000000000000000000000000000000000000000..95c7cb70562c367df14fbabf5381e75b197bb7e7 --- /dev/null +++ b/logs_oct11/logs_oct9.tar @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:88c8911009bc76194a78b1b78dc3bebabc8069034e88401141577ac3aac0e6e4 +size 471303 diff --git a/logs_oct11/plot_ablation.py b/logs_oct11/plot_ablation.py new file mode 100644 index 0000000000000000000000000000000000000000..6c38e3b2b1c760ce893e77456ef91ae49747f90a --- /dev/null +++ b/logs_oct11/plot_ablation.py @@ -0,0 +1,120 @@ +import numpy as np +import matplotlib +import matplotlib.pyplot as plt +from mpl_toolkits.mplot3d import Axes3D + +# 使用Agg后端（无GUI） +matplotlib.use("Agg") + +# 定义x轴和y轴的数据 +x_values = np.array([1, 3, 5, 7, 9]) # init value +y_values = np.array([0.1, 0.3, 0.5, 0.7, 0.9, 1.1, 1.3, 1.5, 1.7, 1.9, 2.1, 2.3, 2.5, 2.7, 2.9]) # temperature + +# 创建网格 +X, Y = np.meshgrid(x_values, y_values) + +# Z轴数据（performance）- 这里使用示例数据 +# 请将这里替换为你的实际数据，应该是一个15x5的数组 +# 示例：使用一个简单的函数生成数据 +# Z = np.sin(np.sqrt(X**2 + Y**2)) * 100 + np.random.randn(15, 5) * 5 + +# 如果你有实际的performance数据，请按照以下格式替换： +Z = np.array([ + [11.9, 14.8, 13.7, 13.0, 10.8], # temperature=0.1时，对应5个init value的performance + [17.5, 10.8, 10.3, 14.3, 13.0], # temperature=0.3时，对应5个init value的performance + [29.2, 31.7, 13.5, 10.1, 14.8], # temperature=0.5时，对应5个init value的performance + [20.0, 30.8, 23.3, 14.2, 11.1], # temperature=0.7时，对应5个init value的performance + [10.9, 23.3, 25.0, 24.2, 10.9], # temperature=0.9时，对应5个init value的performance + [23.3, 24.2, 23.3, 29.2, 21.7], # temperature=1.1时，对应5个init value的performance + [21.7, 27.5, 22.5, 25.8, 25.0], # temperature=1.3时，对应5个init value的performance + [11.5, 23.3, 17.5, 26.7, 27.5], # temperature=1.5时，对应5个init value的performance + [24.2, 24.2, 16.7, 23.3, 29.2], # temperature=1.7时，对应5个init value的performance + [18.3, 24.2, 27.5, 20.0, 25.8], # temperature=1.9时，对应5个init value的performance + [25.8, 25.8, 26.7, 21.7, 28.3], # temperature=2.1时，对应5个init value的performance + [28.3, 26.2, 23.3, 20.8, 26.7], # temperature=2.3时，对应5个init value的performance + [20.0, 24.2, 20.0, 20.0, 28.3], # temperature=2.5时，对应5个init value的performance + [23.3, 26.2, 21.7, 24.2, 24.2], # temperature=2.7时，对应5个init value的performance + [15.8, 22.5, 20.8, 25.8, 23.3], # temperature=2.9时，对应5个init value的performance +]) + +# 创建3D图形 +fig = plt.figure(figsize=(12, 8)) +ax = fig.add_subplot(111, projection='3d') + +# 绘制3D表面图（不显示散点） +surf = ax.plot_surface(X, Y, Z, cmap='RdYlBu_r', alpha=0.8, edgecolor='none') + +# 设置轴标签 +ax.set_xlabel('Init Value', fontsize=12, labelpad=10) +ax.set_ylabel('Temperature', fontsize=12, labelpad=10) +ax.set_zlabel('Performance', fontsize=12, labelpad=10) + +# 设置标题（字体更大，间隙更小） +ax.set_title('Performance vs Init Value and Temperature', fontsize=18, pad=10) + +# 添加颜色条（更窄，离图更近） +colorbar = fig.colorbar(surf, ax=ax, shrink=0.5, aspect=10, pad=0.05) +colorbar.set_label('Performance', rotation=270, labelpad=15) + +# 设置视角（可以调整以获得最佳视图） +ax.view_init(elev=30, azim=45) + +# 明确设置x轴和y轴的刻度 +ax.set_xticks(x_values) +ax.set_yticks(y_values) +# 如果需要，也可以设置z轴的刻度 +# ax.set_zticks(np.linspace(Z.min(), Z.max(), 5)) + +# 移除网格线 +ax.grid(False) + +# 设置透明背景 +ax.xaxis.pane.fill = False +ax.yaxis.pane.fill = False +ax.zaxis.pane.fill = False +ax.xaxis.pane.set_edgecolor('none') +ax.yaxis.pane.set_edgecolor('none') +ax.zaxis.pane.set_edgecolor('none') + +# 保存3D图（透明背景） +plt.tight_layout() +plt.savefig('3d_surface_plot.png', dpi=300, bbox_inches='tight', transparent=True) +plt.close() + +print("3D表面图已保存为: 3d_surface_plot.png") + +# 创建等高线图 +plt.figure(figsize=(7, 5)) + +num_levels = 15 +# 使用 'RdYlBu_r' 颜色映射 +contourf = plt.contourf(X, Y, Z, levels=num_levels, cmap='RdYlBu_r', alpha=0.7) + +# 加粗的黑色等高线，不贴数值标签 +contour_lines = plt.contour(X, Y, Z, levels=num_levels, colors='black', alpha=0.5, linewidths=1.5) + +# 添加颜色条（更窄，离图更近） +cbar = plt.colorbar(contourf, shrink=0.8, aspect=15, pad=0.02) +cbar.set_label("Performance", rotation=270, labelpad=15) + +# 设置轴标签 +plt.xlabel('Init Value', fontsize=12) +plt.ylabel('Temperature', fontsize=12) + +# 设置标题（字体更大，间隙更小） +plt.title('Performance Contour Plot', fontsize=18, pad=10) + +# 移除网格 +plt.grid(False) + +# 显示x和y轴的刻度 +plt.xticks(x_values) +plt.yticks(y_values) + +# 保存等高线图（透明背景） +plt.tight_layout() +plt.savefig('contour_plot.png', dpi=300, bbox_inches='tight', transparent=True) +plt.close() + +print("等高线图已保存为: contour_plot.png") +print("所有图片已成功保存！") \ No newline at end of file diff --git a/logs_oct11/pretrain_qwen.zip b/logs_oct11/pretrain_qwen.zip new file mode 100644 index 0000000000000000000000000000000000000000..4bc25f9e59768712bae0bc9f89a375b3a045dab3 --- /dev/null +++ b/logs_oct11/pretrain_qwen.zip @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:420378c2943468d547bf10619389107e3e52783ba88768e51e07099df45eadc6 +size 1473310868 diff --git a/logs_oct11/pyproject.toml b/logs_oct11/pyproject.toml new file mode 100644 index 0000000000000000000000000000000000000000..ec1acfa8256328cc87d93162e7a184c8c8b7d77b --- /dev/null +++ b/logs_oct11/pyproject.toml @@ -0,0 +1,38 @@ +[build-system] +requires = ["setuptools>=61.0"] +build-backend = "setuptools.build_meta" + +[project] +name = "tinyllava" +version = "1.0.0" +description = "A Framework of Small-scale Large Multimodal Models." +readme = "README.md" +requires-python = ">=3.9" +classifiers = [ + "Programming Language :: Python :: 3", + "License :: OSI Approved :: Apache Software License", +] +dependencies = [ + "torch==2.1.0", "torchvision==0.16.0", "tiktoken", "openpyxl", "tensorboardX", + "transformers==4.40.1", "tokenizers==0.19.0", "sentencepiece==0.1.99", "shortuuid", + "accelerate==0.27.2", "bitsandbytes==0.41.0", "peft==0.10.0", + "pydantic<2,>=1", "markdown2[all]", "numpy==1.26.4", "scikit-learn==1.2.2", + "gradio==3.35.2", "gradio_client==0.2.9", + "requests", "httpx==0.24.0", "uvicorn", "fastapi", + "einops==0.6.1", "einops-exts==0.0.4", "timm==0.6.13", + "deepspeed==0.14.0", "ninja", "wandb", +] + +[project.optional-dependencies] +train = ["deepspeed==0.14.0", "ninja", "wandb"] + +[project.urls] +"Homepage" = "https://github.com/DLCV-BUAA/TinyLLaVABench" +"Bug Tracker" = "https://github.com/DLCV-BUAA/TinyLLaVABench/issues" + +[tool.setuptools.packages.find] +exclude = ["assets*", "benchmark*", "docs", "dist*", "playground*", "scripts*", "tests*"] + +[tool.wheel] +exclude = ["assets*", "benchmark*", "docs", "dist*", "playground*", "scripts*", "tests*"] + diff --git a/logs_oct11/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_1.9_2e-1_connector-1.0_1.9_2e-1_ablation_20251012_052820.log b/logs_oct11/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_1.9_2e-1_connector-1.0_1.9_2e-1_ablation_20251012_052820.log new file mode 100644 index 0000000000000000000000000000000000000000..80ad22ba843f32dca2e0b5d22db80424aca12a91 --- /dev/null +++ b/logs_oct11/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_1.9_2e-1_connector-1.0_1.9_2e-1_ablation_20251012_052820.log @@ -0,0 +1,2314 @@ +==== STARTING EXPERIMENT: qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_1.9_2e-1_connector-1.0_1.9_2e-1_ablation ==== +Log File: qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_1.9_2e-1_connector-1.0_1.9_2e-1_ablation_20251012_052820.log +Timestamp: 2025-10-12 05:28:20 +===================================== +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +The cache for model files in Transformers v4.22.0 has been updated. Migrating your old cache. This is a one-time only operation. You can interrupt this and resume the migration later on by calling `transformers.utils.move_cache()`. + 0it [00:00, ?it/s] 0it [00:00, ?it/s] +[2025-10-12 05:28:22,895] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 05:28:25,963] [WARNING] [runner.py:202:fetch_hostfile] Unable to find hostfile, will proceed with training with local resources only. +[2025-10-12 05:28:25,964] [INFO] [runner.py:568:main] cmd = /opt/conda/envs/tinyllava/bin/python3.10 -u -m deepspeed.launcher.launch --world_info=eyJsb2NhbGhvc3QiOiBbMCwgMSwgMiwgMywgNCwgNSwgNiwgN119 --master_addr=127.0.0.1 --master_port=29501 --enable_each_rank_log=None tinyllava/train/train.py --deepspeed ./scripts/zero3.json --data_path /s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json --image_folder /s3-code/ywang29/datasets/tinyllava --is_multimodal True --conv_version qwen2_base --model_name_or_path Qwen/Qwen2.5-0.5B --vision_tower google/siglip-so400m-patch14-384 --vision_tower2 --connector_type mlp2x_gelu --mm_vision_select_layer -2 --image_aspect_ratio square --attn_implementation flash_attention_2 --bf16 True --training_recipe common --tune_type_llm full --tune_type_vision_tower frozen --tune_vision_tower_from_layer 0 --tune_type_connector full --group_by_modality_length True --pretrained_model_path /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain --output_dir /nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_1.9_2e-1_connector-1.0_1.9_2e-1_ablation --num_train_epochs 1 --per_device_train_batch_size 4 --per_device_eval_batch_size 4 --gradient_accumulation_steps 4 --evaluation_strategy no --learning_rate 2e-1 --weight_decay 0. --warmup_ratio 0.03 --lr_scheduler_type cosine --logging_steps 1 --tf32 False --model_max_length 2048 --gradient_checkpointing True --dataloader_num_workers 8 --lazy_preprocess True --report_to tensorboard --tokenizer_use_fast False --run_name tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune --subnet_mode_text both --subnet_type_text None --mask_type_text soft --init_mean_text 1.0 --temperature_attn_text 1.9 --temperature_mlp_text 1.9 --backward_type_text normal --masked_layers_text all --subnet_mode_vision both --subnet_type_vision None --mask_type_vision soft --init_mean_vision 1.0 --temperature_attn_vision 1.9 --temperature_mlp_vision 1.9 --backward_type_vision normal --masked_layers_vision all --subnet_type_connector global --mask_type_connector soft --init_mean_connector 1.0 --temperature_connector 1.9 --backward_type_connector normal --mm_projector_lr 2e-1 --seed 42 --mask_model llm-connector --save_strategy steps --save_steps 50000 --save_total_limit 1 --train_data_ratio 0.1 +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 05:28:28,489] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 05:28:29,541] [INFO] [launch.py:138:main] 0 NCCL_VERSION=2.21.5 +[2025-10-12 05:28:29,541] [INFO] [launch.py:138:main] 0 NCCL_SOCKET_IFNAME=eth +[2025-10-12 05:28:29,541] [INFO] [launch.py:145:main] WORLD INFO DICT: {'localhost': [0, 1, 2, 3, 4, 5, 6, 7]} +[2025-10-12 05:28:29,541] [INFO] [launch.py:151:main] nnodes=1, num_local_procs=8, node_rank=0 +[2025-10-12 05:28:29,541] [INFO] [launch.py:162:main] global_rank_mapping=defaultdict(, {'localhost': [0, 1, 2, 3, 4, 5, 6, 7]}) +[2025-10-12 05:28:29,541] [INFO] [launch.py:163:main] dist_world_size=8 +[2025-10-12 05:28:29,541] [INFO] [launch.py:165:main] Setting CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 +[2025-10-12 05:28:29,544] [INFO] [launch.py:253:main] process 148696 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=0', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_1.9_2e-1_connector-1.0_1.9_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '1.0', '--temperature_attn_text', '1.9', '--temperature_mlp_text', '1.9', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '1.0', '--temperature_attn_vision', '1.9', '--temperature_mlp_vision', '1.9', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '1.0', '--temperature_connector', '1.9', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 05:28:29,546] [INFO] [launch.py:253:main] process 148697 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=1', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_1.9_2e-1_connector-1.0_1.9_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '1.0', '--temperature_attn_text', '1.9', '--temperature_mlp_text', '1.9', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '1.0', '--temperature_attn_vision', '1.9', '--temperature_mlp_vision', '1.9', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '1.0', '--temperature_connector', '1.9', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 05:28:29,548] [INFO] [launch.py:253:main] process 148698 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=2', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_1.9_2e-1_connector-1.0_1.9_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '1.0', '--temperature_attn_text', '1.9', '--temperature_mlp_text', '1.9', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '1.0', '--temperature_attn_vision', '1.9', '--temperature_mlp_vision', '1.9', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '1.0', '--temperature_connector', '1.9', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 05:28:29,550] [INFO] [launch.py:253:main] process 148699 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=3', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_1.9_2e-1_connector-1.0_1.9_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '1.0', '--temperature_attn_text', '1.9', '--temperature_mlp_text', '1.9', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '1.0', '--temperature_attn_vision', '1.9', '--temperature_mlp_vision', '1.9', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '1.0', '--temperature_connector', '1.9', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 05:28:29,552] [INFO] [launch.py:253:main] process 148700 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=4', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_1.9_2e-1_connector-1.0_1.9_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '1.0', '--temperature_attn_text', '1.9', '--temperature_mlp_text', '1.9', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '1.0', '--temperature_attn_vision', '1.9', '--temperature_mlp_vision', '1.9', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '1.0', '--temperature_connector', '1.9', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 05:28:29,555] [INFO] [launch.py:253:main] process 148701 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=5', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_1.9_2e-1_connector-1.0_1.9_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '1.0', '--temperature_attn_text', '1.9', '--temperature_mlp_text', '1.9', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '1.0', '--temperature_attn_vision', '1.9', '--temperature_mlp_vision', '1.9', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '1.0', '--temperature_connector', '1.9', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 05:28:29,557] [INFO] [launch.py:253:main] process 148702 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=6', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_1.9_2e-1_connector-1.0_1.9_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '1.0', '--temperature_attn_text', '1.9', '--temperature_mlp_text', '1.9', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '1.0', '--temperature_attn_vision', '1.9', '--temperature_mlp_vision', '1.9', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '1.0', '--temperature_connector', '1.9', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 05:28:29,559] [INFO] [launch.py:253:main] process 148703 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=7', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_1.9_2e-1_connector-1.0_1.9_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '1.0', '--temperature_attn_text', '1.9', '--temperature_mlp_text', '1.9', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '1.0', '--temperature_attn_vision', '1.9', '--temperature_mlp_vision', '1.9', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '1.0', '--temperature_connector', '1.9', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 05:28:36,305] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 05:28:36,326] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 05:28:36,330] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 05:28:36,343] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 05:28:36,360] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 05:28:36,389] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 05:28:36,411] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 05:28:36,412] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 05:28:36,795] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 05:28:36,795] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 05:28:36,795] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 05:28:36,795] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 05:28:36,799] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 05:28:36,820] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 05:28:36,827] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 05:28:36,827] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 05:28:36,827] [INFO] [comm.py:668:init_distributed] Initializing TorchBackend in DeepSpeed with backend nccl +Apply masks for the following modules: ['llm', 'connector'] +{'llm': {'model_name_or_path': 'Qwen/Qwen2.5-0.5B', 'cache_dir': None, 'attn_implementation': 'flash_attention_2', 'subnet_mode': 'both', 'subnet_type': 'None', 'sparsity_attn': None, 'sparsity_mlp': None, 'threshold_attn': None, 'threshold_mlp': None, 'temperature_attn': 1.9, 'temperature_mlp': 1.9, 'masked_layers': 'all', 'mask_type': 'soft', 'backward_type': 'normal'}, 'vision_tower': {'model_name_or_path': 'google/siglip-so400m-patch14-384'}, 'connector': {'connector_type': 'mlp2x_gelu', 'subnet_type': 'global', 'threshold': None, 'sparsity': None, 'temperature': 1.9, 'mask_type': 'soft', 'backward_type': 'normal'}} +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +TinyLlavaConfig { + "backward_type_connector": "normal", + "cache_dir": null, + "connector_type": "mlp2x_gelu", + "hidden_size": 896, + "ignore_index": -100, + "image_aspect_ratio": "square", + "image_token_index": -200, + "llm_model_name_or_path": "Qwen/Qwen2.5-0.5B", + "mask_model": [ + "llm", + "connector" + ], + "mask_type_connector": "soft", + "model_type": "tinyllava", + "num_queries": 128, + "num_resampler_layers": 3, + "pad_token": null, + "resampler_hidden_size": 768, + "sparsity_connector": null, + "subnet_type_connector": "global", + "temperature_connector": 1.9, + "text_config": { + "_name_or_path": "Qwen/Qwen2.5-0.5B", + "architectures": [ + "Qwen2ForCausalLM" + ], + "backward_type": "normal", + "bos_token_id": 151643, + "eos_token_id": 151643, + "hidden_size": 896, + "intermediate_size": 4864, + "mask_type": "soft", + "masked_layers": "all", + "max_position_embeddings": 32768, + "max_window_layers": 24, + "model_type": "qwen2", + "num_attention_heads": 14, + "num_hidden_layers": 24, + "num_key_value_heads": 2, + "rope_theta": 1000000.0, + "sliding_window": 32768, + "subnet_mode": "both", + "subnet_type": "None", + "temperature_attn": 1.9, + "temperature_mlp": 1.9, + "tie_word_embeddings": true, + "torch_dtype": "bfloat16", + "use_mrope": false, + "use_sliding_window": false, + "vocab_size": 151936 + }, + "threshold_connector": null, + "tokenizer_model_max_length": 2048, + "tokenizer_name_or_path": "Qwen/Qwen2.5-0.5B", + "tokenizer_padding_side": "right", + "tokenizer_use_fast": false, + "transformers_version": "4.40.1", + "tune_type_connector": "frozen", + "tune_type_llm": "frozen", + "tune_type_vision_tower": "frozen", + "tune_vision_tower_from_layer": -1, + "use_cache": false, + "vision_config": { + "hidden_act": "gelu_pytorch_tanh", + "hidden_size": 1152, + "image_size": 384, + "intermediate_size": 4304, + "layer_norm_eps": 1e-06, + "model_name_or_path": "google/siglip-so400m-patch14-384", + "model_name_or_path2": "", + "model_type": "siglip_vision_model", + "num_attention_heads": 16, + "num_hidden_layers": 27, + "patch_size": 14 + }, + "vision_feature_layer": -2, + "vision_feature_select_strategy": "patch", + "vision_hidden_size": 1152, + "vision_model_name_or_path": "google/siglip-so400m-patch14-384", + "vision_model_name_or_path2": "", + "vocab_size": 151936 +} + +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test1-worker-0:148696:148696 [0] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:148696:148696 [0] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:148696:148696 [0] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:148696:148696 [0] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:148696:148696 [0] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test1-worker-0:148696:148696 [0] NCCL INFO cudaDriverVersion 12040 +NCCL version 2.21.5+cuda12.1 +ywang29-vrdb-test1-worker-0:148696:150314 [0] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:148696:150314 [0] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:148696:150314 [0] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:148696:150314 [0] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:148696:150314 [0] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:148696:150314 [0] NCCL INFO Using network Socket +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test1-worker-0:148698:148698 [2] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test1-worker-0:148698:148698 [2] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:148698:148698 [2] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:148698:148698 [2] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:148698:148698 [2] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:148698:148698 [2] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test1-worker-0:148699:148699 [3] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test1-worker-0:148699:148699 [3] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:148699:148699 [3] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:148699:148699 [3] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:148699:148699 [3] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:148699:148699 [3] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test1-worker-0:148701:148701 [5] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test1-worker-0:148701:148701 [5] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:148701:148701 [5] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:148701:148701 [5] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:148701:148701 [5] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:148701:148701 [5] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test1-worker-0:148703:148703 [7] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test1-worker-0:148703:148703 [7] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:148703:148703 [7] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:148703:148703 [7] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:148703:148703 [7] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:148703:148703 [7] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test1-worker-0:148702:148702 [6] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test1-worker-0:148702:148702 [6] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:148702:148702 [6] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:148702:148702 [6] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:148702:148702 [6] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:148702:148702 [6] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test1-worker-0:148697:148697 [1] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test1-worker-0:148697:148697 [1] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:148697:148697 [1] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:148700:148700 [4] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test1-worker-0:148700:148700 [4] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:148700:148700 [4] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:148697:148697 [1] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:148697:148697 [1] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:148697:148697 [1] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test1-worker-0:148700:148700 [4] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:148700:148700 [4] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:148700:148700 [4] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test1-worker-0:148698:150331 [2] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:148698:150331 [2] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:148698:150331 [2] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:148698:150331 [2] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:148699:150332 [3] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:148698:150331 [2] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:148698:150331 [2] NCCL INFO Using network Socket +ywang29-vrdb-test1-worker-0:148699:150332 [3] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:148699:150332 [3] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:148699:150332 [3] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:148699:150332 [3] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:148699:150332 [3] NCCL INFO Using network Socket +ywang29-vrdb-test1-worker-0:148701:150333 [5] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:148701:150333 [5] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:148701:150333 [5] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:148701:150333 [5] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:148701:150333 [5] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:148701:150333 [5] NCCL INFO Using network Socket +ywang29-vrdb-test1-worker-0:148697:150336 [1] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:148697:150336 [1] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:148697:150336 [1] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:148702:150335 [6] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:148697:150336 [1] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:148697:150336 [1] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:148697:150336 [1] NCCL INFO Using network Socket +ywang29-vrdb-test1-worker-0:148702:150335 [6] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:148702:150335 [6] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:148702:150335 [6] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:148702:150335 [6] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:148702:150335 [6] NCCL INFO Using network Socket +ywang29-vrdb-test1-worker-0:148703:150334 [7] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:148703:150334 [7] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:148703:150334 [7] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:148703:150334 [7] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:148703:150334 [7] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:148703:150334 [7] NCCL INFO Using network Socket +ywang29-vrdb-test1-worker-0:148700:150337 [4] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:148700:150337 [4] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:148700:150337 [4] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:148700:150337 [4] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:148700:150337 [4] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:148700:150337 [4] NCCL INFO Using network Socket +ywang29-vrdb-test1-worker-0:148701:150333 [5] NCCL INFO ncclCommInitRank comm 0x5593a5bfdea0 rank 5 nranks 8 cudaDev 5 nvmlDev 5 busId 901d0 commId 0xdfdf20cd4d1d0e2e - Init START +ywang29-vrdb-test1-worker-0:148699:150332 [3] NCCL INFO ncclCommInitRank comm 0x55b197d3e510 rank 3 nranks 8 cudaDev 3 nvmlDev 3 busId 201d0 commId 0xdfdf20cd4d1d0e2e - Init START +ywang29-vrdb-test1-worker-0:148703:150334 [7] NCCL INFO ncclCommInitRank comm 0x55c5a694f7c0 rank 7 nranks 8 cudaDev 7 nvmlDev 7 busId a01d0 commId 0xdfdf20cd4d1d0e2e - Init START +ywang29-vrdb-test1-worker-0:148697:150336 [1] NCCL INFO ncclCommInitRank comm 0x55f59ae48a80 rank 1 nranks 8 cudaDev 1 nvmlDev 1 busId 101d0 commId 0xdfdf20cd4d1d0e2e - Init START +ywang29-vrdb-test1-worker-0:148696:150314 [0] NCCL INFO ncclCommInitRank comm 0x5569753dfc40 rank 0 nranks 8 cudaDev 0 nvmlDev 0 busId 101c0 commId 0xdfdf20cd4d1d0e2e - Init START +ywang29-vrdb-test1-worker-0:148700:150337 [4] NCCL INFO ncclCommInitRank comm 0x5585ce2758d0 rank 4 nranks 8 cudaDev 4 nvmlDev 4 busId 901c0 commId 0xdfdf20cd4d1d0e2e - Init START +ywang29-vrdb-test1-worker-0:148702:150335 [6] NCCL INFO ncclCommInitRank comm 0x562343388560 rank 6 nranks 8 cudaDev 6 nvmlDev 6 busId a01c0 commId 0xdfdf20cd4d1d0e2e - Init START +ywang29-vrdb-test1-worker-0:148698:150331 [2] NCCL INFO ncclCommInitRank comm 0x55da401e8df0 rank 2 nranks 8 cudaDev 2 nvmlDev 2 busId 201c0 commId 0xdfdf20cd4d1d0e2e - Init START +ywang29-vrdb-test1-worker-0:148696:150314 [0] NCCL INFO Setting affinity for GPU 0 to ff,ffff0000,00ffffff +ywang29-vrdb-test1-worker-0:148696:150314 [0] NCCL INFO NVLS multicast support is not available on dev 0 +ywang29-vrdb-test1-worker-0:148700:150337 [4] NCCL INFO Setting affinity for GPU 4 to ffffff00,0000ffff,ff000000 +ywang29-vrdb-test1-worker-0:148700:150337 [4] NCCL INFO NVLS multicast support is not available on dev 4 +ywang29-vrdb-test1-worker-0:148697:150336 [1] NCCL INFO Setting affinity for GPU 1 to ff,ffff0000,00ffffff +ywang29-vrdb-test1-worker-0:148697:150336 [1] NCCL INFO NVLS multicast support is not available on dev 1 +ywang29-vrdb-test1-worker-0:148699:150332 [3] NCCL INFO Setting affinity for GPU 3 to ff,ffff0000,00ffffff +ywang29-vrdb-test1-worker-0:148699:150332 [3] NCCL INFO NVLS multicast support is not available on dev 3 +ywang29-vrdb-test1-worker-0:148701:150333 [5] NCCL INFO Setting affinity for GPU 5 to ffffff00,0000ffff,ff000000 +ywang29-vrdb-test1-worker-0:148701:150333 [5] NCCL INFO NVLS multicast support is not available on dev 5 +ywang29-vrdb-test1-worker-0:148703:150334 [7] NCCL INFO Setting affinity for GPU 7 to ffffff00,0000ffff,ff000000 +ywang29-vrdb-test1-worker-0:148698:150331 [2] NCCL INFO Setting affinity for GPU 2 to ff,ffff0000,00ffffff +ywang29-vrdb-test1-worker-0:148698:150331 [2] NCCL INFO NVLS multicast support is not available on dev 2 +ywang29-vrdb-test1-worker-0:148702:150335 [6] NCCL INFO Setting affinity for GPU 6 to ffffff00,0000ffff,ff000000 +ywang29-vrdb-test1-worker-0:148703:150334 [7] NCCL INFO NVLS multicast support is not available on dev 7 +ywang29-vrdb-test1-worker-0:148702:150335 [6] NCCL INFO NVLS multicast support is not available on dev 6 +ywang29-vrdb-test1-worker-0:148702:150335 [6] NCCL INFO comm 0x562343388560 rank 6 nRanks 8 nNodes 1 localRanks 8 localRank 6 MNNVL 0 +ywang29-vrdb-test1-worker-0:148701:150333 [5] NCCL INFO comm 0x5593a5bfdea0 rank 5 nRanks 8 nNodes 1 localRanks 8 localRank 5 MNNVL 0 +ywang29-vrdb-test1-worker-0:148699:150332 [3] NCCL INFO comm 0x55b197d3e510 rank 3 nRanks 8 nNodes 1 localRanks 8 localRank 3 MNNVL 0 +ywang29-vrdb-test1-worker-0:148698:150331 [2] NCCL INFO comm 0x55da401e8df0 rank 2 nRanks 8 nNodes 1 localRanks 8 localRank 2 MNNVL 0 +ywang29-vrdb-test1-worker-0:148703:150334 [7] NCCL INFO comm 0x55c5a694f7c0 rank 7 nRanks 8 nNodes 1 localRanks 8 localRank 7 MNNVL 0 +ywang29-vrdb-test1-worker-0:148697:150336 [1] NCCL INFO comm 0x55f59ae48a80 rank 1 nRanks 8 nNodes 1 localRanks 8 localRank 1 MNNVL 0 +ywang29-vrdb-test1-worker-0:148700:150337 [4] NCCL INFO comm 0x5585ce2758d0 rank 4 nRanks 8 nNodes 1 localRanks 8 localRank 4 MNNVL 0 +ywang29-vrdb-test1-worker-0:148696:150314 [0] NCCL INFO comm 0x5569753dfc40 rank 0 nRanks 8 nNodes 1 localRanks 8 localRank 0 MNNVL 0 +ywang29-vrdb-test1-worker-0:148696:150314 [0] NCCL INFO Channel 00/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:148702:150335 [6] NCCL INFO Trees [0] 7/-1/-1->6->5 [1] 7/-1/-1->6->5 [2] 7/-1/-1->6->5 [3] 7/-1/-1->6->5 [4] 7/-1/-1->6->5 [5] 7/-1/-1->6->5 [6] 7/-1/-1->6->5 [7] 7/-1/-1->6->5 [8] 7/-1/-1->6->5 [9] 7/-1/-1->6->5 [10] 7/-1/-1->6->5 [11] 7/-1/-1->6->5 [12] 7/-1/-1->6->5 [13] 7/-1/-1->6->5 [14] 7/-1/-1->6->5 [15] 7/-1/-1->6->5 [16] 7/-1/-1->6->5 [17] 7/-1/-1->6->5 [18] 7/-1/-1->6->5 [19] 7/-1/-1->6->5 [20] 7/-1/-1->6->5 [21] 7/-1/-1->6->5 [22] 7/-1/-1->6->5 [23] 7/-1/-1->6->5 +ywang29-vrdb-test1-worker-0:148696:150314 [0] NCCL INFO Channel 01/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:148696:150314 [0] NCCL INFO Channel 02/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:148702:150335 [6] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:148701:150333 [5] NCCL INFO Trees [0] 6/-1/-1->5->4 [1] 6/-1/-1->5->4 [2] 6/-1/-1->5->4 [3] 6/-1/-1->5->4 [4] 6/-1/-1->5->4 [5] 6/-1/-1->5->4 [6] 6/-1/-1->5->4 [7] 6/-1/-1->5->4 [8] 6/-1/-1->5->4 [9] 6/-1/-1->5->4 [10] 6/-1/-1->5->4 [11] 6/-1/-1->5->4 [12] 6/-1/-1->5->4 [13] 6/-1/-1->5->4 [14] 6/-1/-1->5->4 [15] 6/-1/-1->5->4 [16] 6/-1/-1->5->4 [17] 6/-1/-1->5->4 [18] 6/-1/-1->5->4 [19] 6/-1/-1->5->4 [20] 6/-1/-1->5->4 [21] 6/-1/-1->5->4 [22] 6/-1/-1->5->4 [23] 6/-1/-1->5->4 +ywang29-vrdb-test1-worker-0:148703:150334 [7] NCCL INFO Trees [0] -1/-1/-1->7->6 [1] -1/-1/-1->7->6 [2] -1/-1/-1->7->6 [3] -1/-1/-1->7->6 [4] -1/-1/-1->7->6 [5] -1/-1/-1->7->6 [6] -1/-1/-1->7->6 [7] -1/-1/-1->7->6 [8] -1/-1/-1->7->6 [9] -1/-1/-1->7->6 [10] -1/-1/-1->7->6 [11] -1/-1/-1->7->6 [12] -1/-1/-1->7->6 [13] -1/-1/-1->7->6 [14] -1/-1/-1->7->6 [15] -1/-1/-1->7->6 [16] -1/-1/-1->7->6 [17] -1/-1/-1->7->6 [18] -1/-1/-1->7->6 [19] -1/-1/-1->7->6 [20] -1/-1/-1->7->6 [21] -1/-1/-1->7->6 [22] -1/-1/-1->7->6 [23] -1/-1/-1->7->6 +ywang29-vrdb-test1-worker-0:148699:150332 [3] NCCL INFO Trees [0] 4/-1/-1->3->2 [1] 4/-1/-1->3->2 [2] 4/-1/-1->3->2 [3] 4/-1/-1->3->2 [4] 4/-1/-1->3->2 [5] 4/-1/-1->3->2 [6] 4/-1/-1->3->2 [7] 4/-1/-1->3->2 [8] 4/-1/-1->3->2 [9] 4/-1/-1->3->2 [10] 4/-1/-1->3->2 [11] 4/-1/-1->3->2 [12] 4/-1/-1->3->2 [13] 4/-1/-1->3->2 [14] 4/-1/-1->3->2 [15] 4/-1/-1->3->2 [16] 4/-1/-1->3->2 [17] 4/-1/-1->3->2 [18] 4/-1/-1->3->2 [19] 4/-1/-1->3->2 [20] 4/-1/-1->3->2 [21] 4/-1/-1->3->2 [22] 4/-1/-1->3->2 [23] 4/-1/-1->3->2 +ywang29-vrdb-test1-worker-0:148696:150314 [0] NCCL INFO Channel 03/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:148698:150331 [2] NCCL INFO Trees [0] 3/-1/-1->2->1 [1] 3/-1/-1->2->1 [2] 3/-1/-1->2->1 [3] 3/-1/-1->2->1 [4] 3/-1/-1->2->1 [5] 3/-1/-1->2->1 [6] 3/-1/-1->2->1 [7] 3/-1/-1->2->1 [8] 3/-1/-1->2->1 [9] 3/-1/-1->2->1 [10] 3/-1/-1->2->1 [11] 3/-1/-1->2->1 [12] 3/-1/-1->2->1 [13] 3/-1/-1->2->1 [14] 3/-1/-1->2->1 [15] 3/-1/-1->2->1 [16] 3/-1/-1->2->1 [17] 3/-1/-1->2->1 [18] 3/-1/-1->2->1 [19] 3/-1/-1->2->1 [20] 3/-1/-1->2->1 [21] 3/-1/-1->2->1 [22] 3/-1/-1->2->1 [23] 3/-1/-1->2->1 +ywang29-vrdb-test1-worker-0:148696:150314 [0] NCCL INFO Channel 04/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:148700:150337 [4] NCCL INFO Trees [0] 5/-1/-1->4->3 [1] 5/-1/-1->4->3 [2] 5/-1/-1->4->3 [3] 5/-1/-1->4->3 [4] 5/-1/-1->4->3 [5] 5/-1/-1->4->3 [6] 5/-1/-1->4->3 [7] 5/-1/-1->4->3 [8] 5/-1/-1->4->3 [9] 5/-1/-1->4->3 [10] 5/-1/-1->4->3 [11] 5/-1/-1->4->3 [12] 5/-1/-1->4->3 [13] 5/-1/-1->4->3 [14] 5/-1/-1->4->3 [15] 5/-1/-1->4->3 [16] 5/-1/-1->4->3 [17] 5/-1/-1->4->3 [18] 5/-1/-1->4->3 [19] 5/-1/-1->4->3 [20] 5/-1/-1->4->3 [21] 5/-1/-1->4->3 [22] 5/-1/-1->4->3 [23] 5/-1/-1->4->3 +ywang29-vrdb-test1-worker-0:148701:150333 [5] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:148703:150334 [7] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:148696:150314 [0] NCCL INFO Channel 05/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:148699:150332 [3] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:148697:150336 [1] NCCL INFO Trees [0] 2/-1/-1->1->0 [1] 2/-1/-1->1->0 [2] 2/-1/-1->1->0 [3] 2/-1/-1->1->0 [4] 2/-1/-1->1->0 [5] 2/-1/-1->1->0 [6] 2/-1/-1->1->0 [7] 2/-1/-1->1->0 [8] 2/-1/-1->1->0 [9] 2/-1/-1->1->0 [10] 2/-1/-1->1->0 [11] 2/-1/-1->1->0 [12] 2/-1/-1->1->0 [13] 2/-1/-1->1->0 [14] 2/-1/-1->1->0 [15] 2/-1/-1->1->0 [16] 2/-1/-1->1->0 [17] 2/-1/-1->1->0 [18] 2/-1/-1->1->0 [19] 2/-1/-1->1->0 [20] 2/-1/-1->1->0 [21] 2/-1/-1->1->0 [22] 2/-1/-1->1->0 [23] 2/-1/-1->1->0 +ywang29-vrdb-test1-worker-0:148698:150331 [2] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:148696:150314 [0] NCCL INFO Channel 06/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:148700:150337 [4] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:148696:150314 [0] NCCL INFO Channel 07/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:148697:150336 [1] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:148696:150314 [0] NCCL INFO Channel 08/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:148696:150314 [0] NCCL INFO Channel 09/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:148696:150314 [0] NCCL INFO Channel 10/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:148696:150314 [0] NCCL INFO Channel 11/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:148696:150314 [0] NCCL INFO Channel 12/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:148696:150314 [0] NCCL INFO Channel 13/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:148696:150314 [0] NCCL INFO Channel 14/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:148696:150314 [0] NCCL INFO Channel 15/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:148696:150314 [0] NCCL INFO Channel 16/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:148696:150314 [0] NCCL INFO Channel 17/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:148696:150314 [0] NCCL INFO Channel 18/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:148696:150314 [0] NCCL INFO Channel 19/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:148696:150314 [0] NCCL INFO Channel 20/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:148696:150314 [0] NCCL INFO Channel 21/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:148696:150314 [0] NCCL INFO Channel 22/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:148696:150314 [0] NCCL INFO Channel 23/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:148696:150314 [0] NCCL INFO Trees [0] 1/-1/-1->0->-1 [1] 1/-1/-1->0->-1 [2] 1/-1/-1->0->-1 [3] 1/-1/-1->0->-1 [4] 1/-1/-1->0->-1 [5] 1/-1/-1->0->-1 [6] 1/-1/-1->0->-1 [7] 1/-1/-1->0->-1 [8] 1/-1/-1->0->-1 [9] 1/-1/-1->0->-1 [10] 1/-1/-1->0->-1 [11] 1/-1/-1->0->-1 [12] 1/-1/-1->0->-1 [13] 1/-1/-1->0->-1 [14] 1/-1/-1->0->-1 [15] 1/-1/-1->0->-1 [16] 1/-1/-1->0->-1 [17] 1/-1/-1->0->-1 [18] 1/-1/-1->0->-1 [19] 1/-1/-1->0->-1 [20] 1/-1/-1->0->-1 [21] 1/-1/-1->0->-1 [22] 1/-1/-1->0->-1 [23] 1/-1/-1->0->-1 +ywang29-vrdb-test1-worker-0:148696:150314 [0] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:148699:150332 [3] NCCL INFO Channel 00/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:150335 [6] NCCL INFO Channel 00/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:150334 [7] NCCL INFO Channel 00/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:150333 [5] NCCL INFO Channel 00/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:150336 [1] NCCL INFO Channel 00/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148696:150314 [0] NCCL INFO Channel 00/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:150332 [3] NCCL INFO Channel 01/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:150337 [4] NCCL INFO Channel 00/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:150335 [6] NCCL INFO Channel 01/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:150334 [7] NCCL INFO Channel 01/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:150333 [5] NCCL INFO Channel 01/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:150336 [1] NCCL INFO Channel 01/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148696:150314 [0] NCCL INFO Channel 01/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:150332 [3] NCCL INFO Channel 02/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:150337 [4] NCCL INFO Channel 01/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:150335 [6] NCCL INFO Channel 02/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:150334 [7] NCCL INFO Channel 02/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:150333 [5] NCCL INFO Channel 02/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:150336 [1] NCCL INFO Channel 02/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148696:150314 [0] NCCL INFO Channel 02/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:150332 [3] NCCL INFO Channel 03/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:150337 [4] NCCL INFO Channel 02/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:150335 [6] NCCL INFO Channel 03/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:150334 [7] NCCL INFO Channel 03/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:150333 [5] NCCL INFO Channel 03/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:150336 [1] NCCL INFO Channel 03/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148696:150314 [0] NCCL INFO Channel 03/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:150332 [3] NCCL INFO Channel 04/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:150337 [4] NCCL INFO Channel 03/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:150335 [6] NCCL INFO Channel 04/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:150334 [7] NCCL INFO Channel 04/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:150333 [5] NCCL INFO Channel 04/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:150336 [1] NCCL INFO Channel 04/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148696:150314 [0] NCCL INFO Channel 04/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:150332 [3] NCCL INFO Channel 05/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:150337 [4] NCCL INFO Channel 04/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:150335 [6] NCCL INFO Channel 05/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:150334 [7] NCCL INFO Channel 05/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:150333 [5] NCCL INFO Channel 05/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:150336 [1] NCCL INFO Channel 05/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148696:150314 [0] NCCL INFO Channel 05/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:150332 [3] NCCL INFO Channel 06/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:150337 [4] NCCL INFO Channel 05/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:150335 [6] NCCL INFO Channel 06/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:150334 [7] NCCL INFO Channel 06/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:150333 [5] NCCL INFO Channel 06/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:150336 [1] NCCL INFO Channel 06/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148696:150314 [0] NCCL INFO Channel 06/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:150332 [3] NCCL INFO Channel 07/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:150337 [4] NCCL INFO Channel 06/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:150335 [6] NCCL INFO Channel 07/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:150334 [7] NCCL INFO Channel 07/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:150333 [5] NCCL INFO Channel 07/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:150336 [1] NCCL INFO Channel 07/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148696:150314 [0] NCCL INFO Channel 07/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:150332 [3] NCCL INFO Channel 08/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:150337 [4] NCCL INFO Channel 07/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:150335 [6] NCCL INFO Channel 08/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:150334 [7] NCCL INFO Channel 08/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:150333 [5] NCCL INFO Channel 08/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:150336 [1] NCCL INFO Channel 08/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148696:150314 [0] NCCL INFO Channel 08/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:150332 [3] NCCL INFO Channel 09/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:150337 [4] NCCL INFO Channel 08/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:150335 [6] NCCL INFO Channel 09/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:150334 [7] NCCL INFO Channel 09/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:150333 [5] NCCL INFO Channel 09/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:150336 [1] NCCL INFO Channel 09/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148696:150314 [0] NCCL INFO Channel 09/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:150332 [3] NCCL INFO Channel 10/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:150337 [4] NCCL INFO Channel 09/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:150335 [6] NCCL INFO Channel 10/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:150334 [7] NCCL INFO Channel 10/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:150333 [5] NCCL INFO Channel 10/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:150336 [1] NCCL INFO Channel 10/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148696:150314 [0] NCCL INFO Channel 10/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:150332 [3] NCCL INFO Channel 11/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:150337 [4] NCCL INFO Channel 10/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:150335 [6] NCCL INFO Channel 11/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:150334 [7] NCCL INFO Channel 11/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:150333 [5] NCCL INFO Channel 11/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:150336 [1] NCCL INFO Channel 11/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148696:150314 [0] NCCL INFO Channel 11/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:150332 [3] NCCL INFO Channel 12/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:150337 [4] NCCL INFO Channel 11/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:150335 [6] NCCL INFO Channel 12/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:150334 [7] NCCL INFO Channel 12/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:150333 [5] NCCL INFO Channel 12/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:150336 [1] NCCL INFO Channel 12/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148696:150314 [0] NCCL INFO Channel 12/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:150332 [3] NCCL INFO Channel 13/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:150337 [4] NCCL INFO Channel 12/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:150335 [6] NCCL INFO Channel 13/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:150334 [7] NCCL INFO Channel 13/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:150333 [5] NCCL INFO Channel 13/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:150336 [1] NCCL INFO Channel 13/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148696:150314 [0] NCCL INFO Channel 13/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:150332 [3] NCCL INFO Channel 14/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:150337 [4] NCCL INFO Channel 13/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:150331 [2] NCCL INFO Channel 00/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:150335 [6] NCCL INFO Channel 14/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:150334 [7] NCCL INFO Channel 14/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:150333 [5] NCCL INFO Channel 14/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:150336 [1] NCCL INFO Channel 14/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148696:150314 [0] NCCL INFO Channel 14/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:150332 [3] NCCL INFO Channel 15/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:150337 [4] NCCL INFO Channel 14/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:150331 [2] NCCL INFO Channel 01/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:150335 [6] NCCL INFO Channel 15/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:150334 [7] NCCL INFO Channel 15/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:150336 [1] NCCL INFO Channel 15/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:150333 [5] NCCL INFO Channel 15/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148696:150314 [0] NCCL INFO Channel 15/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:150332 [3] NCCL INFO Channel 16/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:150337 [4] NCCL INFO Channel 15/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:150331 [2] NCCL INFO Channel 02/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:150335 [6] NCCL INFO Channel 16/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:150334 [7] NCCL INFO Channel 16/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:150336 [1] NCCL INFO Channel 16/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:150333 [5] NCCL INFO Channel 16/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148696:150314 [0] NCCL INFO Channel 16/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:150332 [3] NCCL INFO Channel 17/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:150337 [4] NCCL INFO Channel 16/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:150331 [2] NCCL INFO Channel 03/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:150335 [6] NCCL INFO Channel 17/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:150334 [7] NCCL INFO Channel 17/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:150336 [1] NCCL INFO Channel 17/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:150333 [5] NCCL INFO Channel 17/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148696:150314 [0] NCCL INFO Channel 17/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:150332 [3] NCCL INFO Channel 18/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:150337 [4] NCCL INFO Channel 17/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:150331 [2] NCCL INFO Channel 04/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:150335 [6] NCCL INFO Channel 18/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:150334 [7] NCCL INFO Channel 18/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:150336 [1] NCCL INFO Channel 18/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:150333 [5] NCCL INFO Channel 18/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148696:150314 [0] NCCL INFO Channel 18/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:150332 [3] NCCL INFO Channel 19/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:150337 [4] NCCL INFO Channel 18/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:150335 [6] NCCL INFO Channel 19/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:150331 [2] NCCL INFO Channel 05/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:150334 [7] NCCL INFO Channel 19/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:150336 [1] NCCL INFO Channel 19/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:150333 [5] NCCL INFO Channel 19/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148696:150314 [0] NCCL INFO Channel 19/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:150332 [3] NCCL INFO Channel 20/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:150337 [4] NCCL INFO Channel 19/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:150335 [6] NCCL INFO Channel 20/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:150331 [2] NCCL INFO Channel 06/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:150334 [7] NCCL INFO Channel 20/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:150336 [1] NCCL INFO Channel 20/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:150333 [5] NCCL INFO Channel 20/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148696:150314 [0] NCCL INFO Channel 20/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:150332 [3] NCCL INFO Channel 21/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:150337 [4] NCCL INFO Channel 20/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:150335 [6] NCCL INFO Channel 21/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:150331 [2] NCCL INFO Channel 07/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:150334 [7] NCCL INFO Channel 21/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:150336 [1] NCCL INFO Channel 21/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:150333 [5] NCCL INFO Channel 21/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148696:150314 [0] NCCL INFO Channel 21/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:150332 [3] NCCL INFO Channel 22/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:150337 [4] NCCL INFO Channel 21/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:150335 [6] NCCL INFO Channel 22/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:150331 [2] NCCL INFO Channel 08/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:150334 [7] NCCL INFO Channel 22/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:150336 [1] NCCL INFO Channel 22/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:150333 [5] NCCL INFO Channel 22/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148696:150314 [0] NCCL INFO Channel 22/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:150332 [3] NCCL INFO Channel 23/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:150337 [4] NCCL INFO Channel 22/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:150335 [6] NCCL INFO Channel 23/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:150331 [2] NCCL INFO Channel 09/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:150334 [7] NCCL INFO Channel 23/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:150336 [1] NCCL INFO Channel 23/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:150333 [5] NCCL INFO Channel 23/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148696:150314 [0] NCCL INFO Channel 23/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:150337 [4] NCCL INFO Channel 23/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:150331 [2] NCCL INFO Channel 10/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:150331 [2] NCCL INFO Channel 11/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:150331 [2] NCCL INFO Channel 12/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:150331 [2] NCCL INFO Channel 13/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:150331 [2] NCCL INFO Channel 14/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:150331 [2] NCCL INFO Channel 15/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:150331 [2] NCCL INFO Channel 16/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:150331 [2] NCCL INFO Channel 17/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:150331 [2] NCCL INFO Channel 18/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:150331 [2] NCCL INFO Channel 19/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:150331 [2] NCCL INFO Channel 20/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:150331 [2] NCCL INFO Channel 21/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:150331 [2] NCCL INFO Channel 22/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:150331 [2] NCCL INFO Channel 23/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:150336 [1] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:148696:150314 [0] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:148698:150331 [2] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:148699:150332 [3] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:148700:150337 [4] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:148703:150334 [7] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:148703:150334 [7] NCCL INFO Channel 00/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:150333 [5] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:148702:150335 [6] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:148703:150334 [7] NCCL INFO Channel 01/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:150334 [7] NCCL INFO Channel 02/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:150334 [7] NCCL INFO Channel 03/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:150334 [7] NCCL INFO Channel 04/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:150334 [7] NCCL INFO Channel 05/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:150334 [7] NCCL INFO Channel 06/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:150334 [7] NCCL INFO Channel 07/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:150334 [7] NCCL INFO Channel 08/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:150334 [7] NCCL INFO Channel 09/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:150332 [3] NCCL INFO Channel 00/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:150332 [3] NCCL INFO Channel 01/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:150332 [3] NCCL INFO Channel 02/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:150332 [3] NCCL INFO Channel 03/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:150332 [3] NCCL INFO Channel 04/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:150332 [3] NCCL INFO Channel 05/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:150336 [1] NCCL INFO Channel 00/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:150332 [3] NCCL INFO Channel 06/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:150336 [1] NCCL INFO Channel 01/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:150332 [3] NCCL INFO Channel 07/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:150337 [4] NCCL INFO Channel 00/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:150336 [1] NCCL INFO Channel 02/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:150332 [3] NCCL INFO Channel 08/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:150337 [4] NCCL INFO Channel 01/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:150334 [7] NCCL INFO Channel 10/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:150336 [1] NCCL INFO Channel 03/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:150337 [4] NCCL INFO Channel 02/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:150332 [3] NCCL INFO Channel 09/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:150334 [7] NCCL INFO Channel 11/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:150336 [1] NCCL INFO Channel 04/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:150335 [6] NCCL INFO Channel 00/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:150337 [4] NCCL INFO Channel 03/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:150332 [3] NCCL INFO Channel 10/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:150333 [5] NCCL INFO Channel 00/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:150334 [7] NCCL INFO Channel 12/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:150336 [1] NCCL INFO Channel 05/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:150335 [6] NCCL INFO Channel 01/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:150337 [4] NCCL INFO Channel 04/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:150332 [3] NCCL INFO Channel 11/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:150333 [5] NCCL INFO Channel 01/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:150334 [7] NCCL INFO Channel 13/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:150336 [1] NCCL INFO Channel 06/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:150335 [6] NCCL INFO Channel 02/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:150337 [4] NCCL INFO Channel 05/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:150332 [3] NCCL INFO Channel 12/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:150333 [5] NCCL INFO Channel 02/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:150334 [7] NCCL INFO Channel 14/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:150336 [1] NCCL INFO Channel 07/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:150335 [6] NCCL INFO Channel 03/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:150332 [3] NCCL INFO Channel 13/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:150333 [5] NCCL INFO Channel 03/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:150336 [1] NCCL INFO Channel 08/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:150335 [6] NCCL INFO Channel 04/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:150337 [4] NCCL INFO Channel 06/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:150332 [3] NCCL INFO Channel 14/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:150333 [5] NCCL INFO Channel 04/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:150336 [1] NCCL INFO Channel 09/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:150335 [6] NCCL INFO Channel 05/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:150337 [4] NCCL INFO Channel 07/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:150332 [3] NCCL INFO Channel 15/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:150333 [5] NCCL INFO Channel 05/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:150336 [1] NCCL INFO Channel 10/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:150335 [6] NCCL INFO Channel 06/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:150337 [4] NCCL INFO Channel 08/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:150332 [3] NCCL INFO Channel 16/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:150333 [5] NCCL INFO Channel 06/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:150336 [1] NCCL INFO Channel 11/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:150335 [6] NCCL INFO Channel 07/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:150337 [4] NCCL INFO Channel 09/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:150332 [3] NCCL INFO Channel 17/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:150333 [5] NCCL INFO Channel 07/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:150336 [1] NCCL INFO Channel 12/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:150335 [6] NCCL INFO Channel 08/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:150337 [4] NCCL INFO Channel 10/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:150332 [3] NCCL INFO Channel 18/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:150333 [5] NCCL INFO Channel 08/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:150336 [1] NCCL INFO Channel 13/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:150335 [6] NCCL INFO Channel 09/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:150337 [4] NCCL INFO Channel 11/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:150332 [3] NCCL INFO Channel 19/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:150333 [5] NCCL INFO Channel 09/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:150335 [6] NCCL INFO Channel 10/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:150336 [1] NCCL INFO Channel 14/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:150337 [4] NCCL INFO Channel 12/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:150332 [3] NCCL INFO Channel 20/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:150333 [5] NCCL INFO Channel 10/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:150335 [6] NCCL INFO Channel 11/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:150336 [1] NCCL INFO Channel 15/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:150337 [4] NCCL INFO Channel 13/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:150332 [3] NCCL INFO Channel 21/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:150334 [7] NCCL INFO Channel 15/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:150333 [5] NCCL INFO Channel 11/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:150335 [6] NCCL INFO Channel 12/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:150336 [1] NCCL INFO Channel 16/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:150337 [4] NCCL INFO Channel 14/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:150332 [3] NCCL INFO Channel 22/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:150334 [7] NCCL INFO Channel 16/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:150333 [5] NCCL INFO Channel 12/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:150335 [6] NCCL INFO Channel 13/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:150336 [1] NCCL INFO Channel 17/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:150331 [2] NCCL INFO Channel 00/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:150337 [4] NCCL INFO Channel 15/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:150332 [3] NCCL INFO Channel 23/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:150334 [7] NCCL INFO Channel 17/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:150333 [5] NCCL INFO Channel 13/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:150335 [6] NCCL INFO Channel 14/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:150336 [1] NCCL INFO Channel 18/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:150331 [2] NCCL INFO Channel 01/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:150337 [4] NCCL INFO Channel 16/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:150334 [7] NCCL INFO Channel 18/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:150335 [6] NCCL INFO Channel 15/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:150333 [5] NCCL INFO Channel 14/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:150336 [1] NCCL INFO Channel 19/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:150331 [2] NCCL INFO Channel 02/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:150337 [4] NCCL INFO Channel 17/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:150334 [7] NCCL INFO Channel 19/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:150335 [6] NCCL INFO Channel 16/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:150333 [5] NCCL INFO Channel 15/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:150336 [1] NCCL INFO Channel 20/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:150331 [2] NCCL INFO Channel 03/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:150337 [4] NCCL INFO Channel 18/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:150334 [7] NCCL INFO Channel 20/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:150335 [6] NCCL INFO Channel 17/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:150333 [5] NCCL INFO Channel 16/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:150336 [1] NCCL INFO Channel 21/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:150331 [2] NCCL INFO Channel 04/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:150337 [4] NCCL INFO Channel 19/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:150334 [7] NCCL INFO Channel 21/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:150335 [6] NCCL INFO Channel 18/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:150333 [5] NCCL INFO Channel 17/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:150336 [1] NCCL INFO Channel 22/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:150331 [2] NCCL INFO Channel 05/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:150337 [4] NCCL INFO Channel 20/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:150334 [7] NCCL INFO Channel 22/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:150335 [6] NCCL INFO Channel 19/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:150336 [1] NCCL INFO Channel 23/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:150331 [2] NCCL INFO Channel 06/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:150337 [4] NCCL INFO Channel 21/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:150334 [7] NCCL INFO Channel 23/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:150335 [6] NCCL INFO Channel 20/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:150331 [2] NCCL INFO Channel 07/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:150337 [4] NCCL INFO Channel 22/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:150335 [6] NCCL INFO Channel 21/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:150337 [4] NCCL INFO Channel 23/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:150331 [2] NCCL INFO Channel 08/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:150335 [6] NCCL INFO Channel 22/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:150333 [5] NCCL INFO Channel 18/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:150335 [6] NCCL INFO Channel 23/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:150331 [2] NCCL INFO Channel 09/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:150333 [5] NCCL INFO Channel 19/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:150331 [2] NCCL INFO Channel 10/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:150333 [5] NCCL INFO Channel 20/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:150331 [2] NCCL INFO Channel 11/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:150333 [5] NCCL INFO Channel 21/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:150331 [2] NCCL INFO Channel 12/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:150333 [5] NCCL INFO Channel 22/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:150331 [2] NCCL INFO Channel 13/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:150333 [5] NCCL INFO Channel 23/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:150331 [2] NCCL INFO Channel 14/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:150331 [2] NCCL INFO Channel 15/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:150331 [2] NCCL INFO Channel 16/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:150331 [2] NCCL INFO Channel 17/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:150331 [2] NCCL INFO Channel 18/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:150331 [2] NCCL INFO Channel 19/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:150331 [2] NCCL INFO Channel 20/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:150331 [2] NCCL INFO Channel 21/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:150331 [2] NCCL INFO Channel 22/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:150331 [2] NCCL INFO Channel 23/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148696:150314 [0] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:148696:150314 [0] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:148696:150314 [0] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:148697:150336 [1] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:148697:150336 [1] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:148697:150336 [1] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:148698:150331 [2] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:148698:150331 [2] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:148698:150331 [2] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:148699:150332 [3] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:148699:150332 [3] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:148699:150332 [3] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:148703:150334 [7] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:148703:150334 [7] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:148703:150334 [7] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:148700:150337 [4] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:148700:150337 [4] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:148700:150337 [4] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:148702:150335 [6] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:148701:150333 [5] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:148702:150335 [6] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:148702:150335 [6] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:148701:150333 [5] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:148701:150333 [5] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:148696:150314 [0] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:148699:150332 [3] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:148699:150332 [3] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:148700:150337 [4] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:148699:150332 [3] NCCL INFO ncclCommInitRank comm 0x55b197d3e510 rank 3 nranks 8 cudaDev 3 nvmlDev 3 busId 201d0 commId 0xdfdf20cd4d1d0e2e - Init COMPLETE +ywang29-vrdb-test1-worker-0:148696:150314 [0] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:148700:150337 [4] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:148701:150333 [5] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:148696:150314 [0] NCCL INFO ncclCommInitRank comm 0x5569753dfc40 rank 0 nranks 8 cudaDev 0 nvmlDev 0 busId 101c0 commId 0xdfdf20cd4d1d0e2e - Init COMPLETE +ywang29-vrdb-test1-worker-0:148700:150337 [4] NCCL INFO ncclCommInitRank comm 0x5585ce2758d0 rank 4 nranks 8 cudaDev 4 nvmlDev 4 busId 901c0 commId 0xdfdf20cd4d1d0e2e - Init COMPLETE +ywang29-vrdb-test1-worker-0:148697:150336 [1] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:148701:150333 [5] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:148697:150336 [1] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:148701:150333 [5] NCCL INFO ncclCommInitRank comm 0x5593a5bfdea0 rank 5 nranks 8 cudaDev 5 nvmlDev 5 busId 901d0 commId 0xdfdf20cd4d1d0e2e - Init COMPLETE +ywang29-vrdb-test1-worker-0:148697:150336 [1] NCCL INFO ncclCommInitRank comm 0x55f59ae48a80 rank 1 nranks 8 cudaDev 1 nvmlDev 1 busId 101d0 commId 0xdfdf20cd4d1d0e2e - Init COMPLETE +ywang29-vrdb-test1-worker-0:148703:150334 [7] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:148703:150334 [7] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:148703:150334 [7] NCCL INFO ncclCommInitRank comm 0x55c5a694f7c0 rank 7 nranks 8 cudaDev 7 nvmlDev 7 busId a01d0 commId 0xdfdf20cd4d1d0e2e - Init COMPLETE +ywang29-vrdb-test1-worker-0:148702:150335 [6] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:148698:150331 [2] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:148698:150331 [2] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:148702:150335 [6] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:148698:150331 [2] NCCL INFO ncclCommInitRank comm 0x55da401e8df0 rank 2 nranks 8 cudaDev 2 nvmlDev 2 busId 201c0 commId 0xdfdf20cd4d1d0e2e - Init COMPLETE +ywang29-vrdb-test1-worker-0:148702:150335 [6] NCCL INFO ncclCommInitRank comm 0x562343388560 rank 6 nranks 8 cudaDev 6 nvmlDev 6 busId a01c0 commId 0xdfdf20cd4d1d0e2e - Init COMPLETE +[2025-10-12 05:29:26,647] [INFO] [partition_parameters.py:343:__exit__] finished initializing model - num_params = 459, num_elems = 0.99B +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +[2025-10-12 05:29:29,286] [INFO] [partition_parameters.py:343:__exit__] finished initializing model - num_params = 907, num_elems = 1.42B +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +TinyLlavaForConditionalGeneration( + (language_model): Qwen2ForCausalLM( + (model): Qwen2Model( + (embed_tokens): Embedding(151936, 896) + (layers): ModuleList( + (0-23): 24 x Qwen2DecoderLayer( + (self_attn): Qwen2FlashAttention2( + (q_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=True) + (k_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=128, bias=True) + (v_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=128, bias=True) + (o_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=False) + (rotary_emb): Qwen2RotaryEmbedding() + ) + (mlp): Qwen2MLP( + (gate_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=4864, bias=False) + (up_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=4864, bias=False) + (down_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=4864, out_features=896, bias=False) + (act_fn): SiLU() + ) + (input_layernorm): Qwen2RMSNorm() + (post_attention_layernorm): Qwen2RMSNorm() + ) + ) + (norm): Qwen2RMSNorm() + ) + (lm_head): Linear(in_features=896, out_features=151936, bias=False) + ) + (vision_tower): SIGLIPVisionTower( + (_vision_tower): SiglipVisionModel( + (vision_model): SiglipVisionTransformer( + (embeddings): SiglipVisionEmbeddings( + (patch_embedding): Conv2d(3, 1152, kernel_size=(14, 14), stride=(14, 14), padding=valid) + (position_embedding): Embedding(729, 1152) + ) + (encoder): SiglipEncoder( + (layers): ModuleList( + (0-26): 27 x SiglipEncoderLayer( + (self_attn): SiglipAttention( + (k_proj): Linear(in_features=1152, out_features=1152, bias=True) + (v_proj): Linear(in_features=1152, out_features=1152, bias=True) + (q_proj): Linear(in_features=1152, out_features=1152, bias=True) + (out_proj): Linear(in_features=1152, out_features=1152, bias=True) + ) + (layer_norm1): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + (layer_norm2): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + ) + ) + ) + (post_layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (head): SiglipMultiheadAttentionPoolingHead( + (attention): MultiheadAttention( + (out_proj): NonDynamicallyQuantizableLinear(in_features=1152, out_features=1152, bias=True) + ) + (layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + ) + ) + ) + ) + (connector): MLPConnector( + (_connector): Sequential( + (0): SupermaskLinearSparsity_SoftForward_Normal(in_features=1152, out_features=896, bias=True) + (1): GELU(approximate='none') + (2): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=True) + ) + ) +) +Pre-training init language_model.model.layers.0.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.0.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.0.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.0.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.0.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.0.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.0.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.1.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.1.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.1.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.1.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.1.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.1.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.1.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.2.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.2.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.2.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.2.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.2.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.2.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.2.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.3.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.3.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.3.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.3.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.3.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.3.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.3.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.4.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.4.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.4.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.4.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.4.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.4.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.4.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.5.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.5.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.5.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.5.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.5.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.5.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.5.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.6.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.6.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.6.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.6.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.6.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.6.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.6.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.7.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.7.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.7.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.7.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.7.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.7.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.7.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.8.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.8.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.8.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.8.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.8.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.8.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.8.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.9.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.9.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.9.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.9.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.9.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.9.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.9.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.10.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.10.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.10.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.10.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.10.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.10.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.10.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.11.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.11.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.11.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.11.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.11.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.11.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.11.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.12.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.12.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.12.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.12.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.12.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.12.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.12.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.13.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.13.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.13.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.13.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.13.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.13.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.13.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.14.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.14.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.14.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.14.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.14.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.14.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.14.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.15.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.15.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.15.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.15.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.15.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.15.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.15.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.16.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.16.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.16.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.16.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.16.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.16.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.16.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.17.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.17.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.17.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.17.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.17.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.17.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.17.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.18.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.18.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.18.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.18.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.18.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.18.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.18.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.19.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.19.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.19.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.19.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.19.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.19.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.19.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.20.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.20.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.20.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.20.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.20.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.20.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.20.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.21.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.21.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.21.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.21.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.21.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.21.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.21.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.22.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.22.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.22.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.22.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.22.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.22.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.22.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.23.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.23.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.23.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.23.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.23.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.23.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.23.mlp.down_proj.scores: Mean=1.000000 +Pre-training init connector._connector.0.scores: Mean=1.000005 +Pre-training init connector._connector.2.scores: Mean=0.999970 +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +2025-10-12 05:29:46,756 | INFO: Total Parameters: 1283756736, Total Trainable Parameters: 359661568 +2025-10-12 05:29:46,763 | INFO: Trainable Parameters: +language_model.model.layers.0.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.0.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.0.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.0.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.0.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.0.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.0.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.1.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.1.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.1.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.1.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.1.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.1.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.1.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.2.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.2.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.2.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.2.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.2.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.2.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.2.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.3.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.3.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.3.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.3.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.3.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.3.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.3.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.4.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.4.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.4.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.4.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.4.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.4.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.4.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.5.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.5.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.5.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.5.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.5.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.5.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.5.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.6.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.6.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.6.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.6.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.6.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.6.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.6.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.7.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.7.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.7.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.7.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.7.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.7.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.7.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.8.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.8.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.8.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.8.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.8.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.8.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.8.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.9.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.9.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.9.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.9.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.9.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.9.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.9.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.10.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.10.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.10.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.10.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.10.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.10.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.10.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.11.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.11.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.11.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.11.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.11.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.11.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.11.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.12.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.12.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.12.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.12.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.12.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.12.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.12.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.13.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.13.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.13.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.13.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.13.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.13.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.13.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.14.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.14.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.14.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.14.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.14.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.14.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.14.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.15.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.15.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.15.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.15.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.15.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.15.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.15.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.16.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.16.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.16.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.16.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.16.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.16.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.16.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.17.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.17.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.17.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.17.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.17.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.17.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.17.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.18.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.18.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.18.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.18.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.18.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.18.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.18.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.19.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.19.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.19.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.19.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.19.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.19.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.19.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.20.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.20.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.20.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.20.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.20.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.20.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.20.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.21.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.21.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.21.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.21.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.21.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.21.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.21.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.22.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.22.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.22.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.22.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.22.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.22.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.22.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.23.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.23.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.23.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.23.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.23.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.23.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.23.mlp.down_proj.scores: 4358144 parameters +connector._connector.0.scores: 1032192 parameters +connector._connector.2.scores: 802816 parameters +Parameter Offload: Total persistent parameters: 486464 in 403 params + 0%| | 0/520 [00:006->5 [1] 7/-1/-1->6->5 [2] 7/-1/-1->6->5 [3] 7/-1/-1->6->5 [4] 7/-1/-1->6->5 [5] 7/-1/-1->6->5 [6] 7/-1/-1->6->5 [7] 7/-1/-1->6->5 [8] 7/-1/-1->6->5 [9] 7/-1/-1->6->5 [10] 7/-1/-1->6->5 [11] 7/-1/-1->6->5 [12] 7/-1/-1->6->5 [13] 7/-1/-1->6->5 [14] 7/-1/-1->6->5 [15] 7/-1/-1->6->5 [16] 7/-1/-1->6->5 [17] 7/-1/-1->6->5 [18] 7/-1/-1->6->5 [19] 7/-1/-1->6->5 [20] 7/-1/-1->6->5 [21] 7/-1/-1->6->5 [22] 7/-1/-1->6->5 [23] 7/-1/-1->6->5 +ywang29-vrdb-test1-worker-0:148702:155356 [6] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:148696:155354 [0] NCCL INFO Channel 00/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:148696:155354 [0] NCCL INFO Channel 01/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:148700:155355 [4] NCCL INFO Trees [0] 5/-1/-1->4->3 [1] 5/-1/-1->4->3 [2] 5/-1/-1->4->3 [3] 5/-1/-1->4->3 [4] 5/-1/-1->4->3 [5] 5/-1/-1->4->3 [6] 5/-1/-1->4->3 [7] 5/-1/-1->4->3 [8] 5/-1/-1->4->3 [9] 5/-1/-1->4->3 [10] 5/-1/-1->4->3 [11] 5/-1/-1->4->3 [12] 5/-1/-1->4->3 [13] 5/-1/-1->4->3 [14] 5/-1/-1->4->3 [15] 5/-1/-1->4->3 [16] 5/-1/-1->4->3 [17] 5/-1/-1->4->3 [18] 5/-1/-1->4->3 [19] 5/-1/-1->4->3 [20] 5/-1/-1->4->3 [21] 5/-1/-1->4->3 [22] 5/-1/-1->4->3 [23] 5/-1/-1->4->3 +ywang29-vrdb-test1-worker-0:148696:155354 [0] NCCL INFO Channel 02/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:148700:155355 [4] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:148699:155360 [3] NCCL INFO Trees [0] 4/-1/-1->3->2 [1] 4/-1/-1->3->2 [2] 4/-1/-1->3->2 [3] 4/-1/-1->3->2 [4] 4/-1/-1->3->2 [5] 4/-1/-1->3->2 [6] 4/-1/-1->3->2 [7] 4/-1/-1->3->2 [8] 4/-1/-1->3->2 [9] 4/-1/-1->3->2 [10] 4/-1/-1->3->2 [11] 4/-1/-1->3->2 [12] 4/-1/-1->3->2 [13] 4/-1/-1->3->2 [14] 4/-1/-1->3->2 [15] 4/-1/-1->3->2 [16] 4/-1/-1->3->2 [17] 4/-1/-1->3->2 [18] 4/-1/-1->3->2 [19] 4/-1/-1->3->2 [20] 4/-1/-1->3->2 [21] 4/-1/-1->3->2 [22] 4/-1/-1->3->2 [23] 4/-1/-1->3->2 +ywang29-vrdb-test1-worker-0:148696:155354 [0] NCCL INFO Channel 03/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:148699:155360 [3] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:148696:155354 [0] NCCL INFO Channel 04/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:148696:155354 [0] NCCL INFO Channel 05/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:148701:155358 [5] NCCL INFO Trees [0] 6/-1/-1->5->4 [1] 6/-1/-1->5->4 [2] 6/-1/-1->5->4 [3] 6/-1/-1->5->4 [4] 6/-1/-1->5->4 [5] 6/-1/-1->5->4 [6] 6/-1/-1->5->4 [7] 6/-1/-1->5->4 [8] 6/-1/-1->5->4 [9] 6/-1/-1->5->4 [10] 6/-1/-1->5->4 [11] 6/-1/-1->5->4 [12] 6/-1/-1->5->4 [13] 6/-1/-1->5->4 [14] 6/-1/-1->5->4 [15] 6/-1/-1->5->4 [16] 6/-1/-1->5->4 [17] 6/-1/-1->5->4 [18] 6/-1/-1->5->4 [19] 6/-1/-1->5->4 [20] 6/-1/-1->5->4 [21] 6/-1/-1->5->4 [22] 6/-1/-1->5->4 [23] 6/-1/-1->5->4 +ywang29-vrdb-test1-worker-0:148696:155354 [0] NCCL INFO Channel 06/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:148701:155358 [5] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:148696:155354 [0] NCCL INFO Channel 07/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:148696:155354 [0] NCCL INFO Channel 08/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:148696:155354 [0] NCCL INFO Channel 09/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:148696:155354 [0] NCCL INFO Channel 10/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:148696:155354 [0] NCCL INFO Channel 11/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:148696:155354 [0] NCCL INFO Channel 12/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:148703:155359 [7] NCCL INFO Trees [0] -1/-1/-1->7->6 [1] -1/-1/-1->7->6 [2] -1/-1/-1->7->6 [3] -1/-1/-1->7->6 [4] -1/-1/-1->7->6 [5] -1/-1/-1->7->6 [6] -1/-1/-1->7->6 [7] -1/-1/-1->7->6 [8] -1/-1/-1->7->6 [9] -1/-1/-1->7->6 [10] -1/-1/-1->7->6 [11] -1/-1/-1->7->6 [12] -1/-1/-1->7->6 [13] -1/-1/-1->7->6 [14] -1/-1/-1->7->6 [15] -1/-1/-1->7->6 [16] -1/-1/-1->7->6 [17] -1/-1/-1->7->6 [18] -1/-1/-1->7->6 [19] -1/-1/-1->7->6 [20] -1/-1/-1->7->6 [21] -1/-1/-1->7->6 [22] -1/-1/-1->7->6 [23] -1/-1/-1->7->6 +ywang29-vrdb-test1-worker-0:148696:155354 [0] NCCL INFO Channel 13/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:148703:155359 [7] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:148696:155354 [0] NCCL INFO Channel 14/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:148697:155361 [1] NCCL INFO Trees [0] 2/-1/-1->1->0 [1] 2/-1/-1->1->0 [2] 2/-1/-1->1->0 [3] 2/-1/-1->1->0 [4] 2/-1/-1->1->0 [5] 2/-1/-1->1->0 [6] 2/-1/-1->1->0 [7] 2/-1/-1->1->0 [8] 2/-1/-1->1->0 [9] 2/-1/-1->1->0 [10] 2/-1/-1->1->0 [11] 2/-1/-1->1->0 [12] 2/-1/-1->1->0 [13] 2/-1/-1->1->0 [14] 2/-1/-1->1->0 [15] 2/-1/-1->1->0 [16] 2/-1/-1->1->0 [17] 2/-1/-1->1->0 [18] 2/-1/-1->1->0 [19] 2/-1/-1->1->0 [20] 2/-1/-1->1->0 [21] 2/-1/-1->1->0 [22] 2/-1/-1->1->0 [23] 2/-1/-1->1->0 +ywang29-vrdb-test1-worker-0:148698:155357 [2] NCCL INFO Trees [0] 3/-1/-1->2->1 [1] 3/-1/-1->2->1 [2] 3/-1/-1->2->1 [3] 3/-1/-1->2->1 [4] 3/-1/-1->2->1 [5] 3/-1/-1->2->1 [6] 3/-1/-1->2->1 [7] 3/-1/-1->2->1 [8] 3/-1/-1->2->1 [9] 3/-1/-1->2->1 [10] 3/-1/-1->2->1 [11] 3/-1/-1->2->1 [12] 3/-1/-1->2->1 [13] 3/-1/-1->2->1 [14] 3/-1/-1->2->1 [15] 3/-1/-1->2->1 [16] 3/-1/-1->2->1 [17] 3/-1/-1->2->1 [18] 3/-1/-1->2->1 [19] 3/-1/-1->2->1 [20] 3/-1/-1->2->1 [21] 3/-1/-1->2->1 [22] 3/-1/-1->2->1 [23] 3/-1/-1->2->1 +ywang29-vrdb-test1-worker-0:148696:155354 [0] NCCL INFO Channel 15/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:148696:155354 [0] NCCL INFO Channel 16/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:148697:155361 [1] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:148698:155357 [2] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:148696:155354 [0] NCCL INFO Channel 17/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:148696:155354 [0] NCCL INFO Channel 18/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:148696:155354 [0] NCCL INFO Channel 19/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:148696:155354 [0] NCCL INFO Channel 20/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:148696:155354 [0] NCCL INFO Channel 21/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:148696:155354 [0] NCCL INFO Channel 22/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:148696:155354 [0] NCCL INFO Channel 23/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:148696:155354 [0] NCCL INFO Trees [0] 1/-1/-1->0->-1 [1] 1/-1/-1->0->-1 [2] 1/-1/-1->0->-1 [3] 1/-1/-1->0->-1 [4] 1/-1/-1->0->-1 [5] 1/-1/-1->0->-1 [6] 1/-1/-1->0->-1 [7] 1/-1/-1->0->-1 [8] 1/-1/-1->0->-1 [9] 1/-1/-1->0->-1 [10] 1/-1/-1->0->-1 [11] 1/-1/-1->0->-1 [12] 1/-1/-1->0->-1 [13] 1/-1/-1->0->-1 [14] 1/-1/-1->0->-1 [15] 1/-1/-1->0->-1 [16] 1/-1/-1->0->-1 [17] 1/-1/-1->0->-1 [18] 1/-1/-1->0->-1 [19] 1/-1/-1->0->-1 [20] 1/-1/-1->0->-1 [21] 1/-1/-1->0->-1 [22] 1/-1/-1->0->-1 [23] 1/-1/-1->0->-1 +ywang29-vrdb-test1-worker-0:148696:155354 [0] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:148697:155361 [1] NCCL INFO Channel 00/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:155361 [1] NCCL INFO Channel 01/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:155361 [1] NCCL INFO Channel 02/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:155361 [1] NCCL INFO Channel 03/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:155360 [3] NCCL INFO Channel 00/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148696:155354 [0] NCCL INFO Channel 00/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:155361 [1] NCCL INFO Channel 04/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:155360 [3] NCCL INFO Channel 01/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148696:155354 [0] NCCL INFO Channel 01/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:155358 [5] NCCL INFO Channel 00/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:155361 [1] NCCL INFO Channel 05/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:155360 [3] NCCL INFO Channel 02/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148696:155354 [0] NCCL INFO Channel 02/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:155359 [7] NCCL INFO Channel 00/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:155358 [5] NCCL INFO Channel 01/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148696:155354 [0] NCCL INFO Channel 03/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:155360 [3] NCCL INFO Channel 03/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:155361 [1] NCCL INFO Channel 06/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:155359 [7] NCCL INFO Channel 01/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:155358 [5] NCCL INFO Channel 02/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148696:155354 [0] NCCL INFO Channel 04/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:155360 [3] NCCL INFO Channel 04/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:155361 [1] NCCL INFO Channel 07/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:155359 [7] NCCL INFO Channel 02/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:155358 [5] NCCL INFO Channel 03/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148696:155354 [0] NCCL INFO Channel 05/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:155360 [3] NCCL INFO Channel 05/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:155361 [1] NCCL INFO Channel 08/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:155359 [7] NCCL INFO Channel 03/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:155358 [5] NCCL INFO Channel 04/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148696:155354 [0] NCCL INFO Channel 06/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:155360 [3] NCCL INFO Channel 06/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:155361 [1] NCCL INFO Channel 09/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:155359 [7] NCCL INFO Channel 04/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:155358 [5] NCCL INFO Channel 05/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148696:155354 [0] NCCL INFO Channel 07/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:155360 [3] NCCL INFO Channel 07/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:155361 [1] NCCL INFO Channel 10/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:155359 [7] NCCL INFO Channel 05/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:155358 [5] NCCL INFO Channel 06/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148696:155354 [0] NCCL INFO Channel 08/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:155360 [3] NCCL INFO Channel 08/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:155361 [1] NCCL INFO Channel 11/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:155359 [7] NCCL INFO Channel 06/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:155358 [5] NCCL INFO Channel 07/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148696:155354 [0] NCCL INFO Channel 09/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:155360 [3] NCCL INFO Channel 09/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:155361 [1] NCCL INFO Channel 12/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:155359 [7] NCCL INFO Channel 07/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:155358 [5] NCCL INFO Channel 08/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148696:155354 [0] NCCL INFO Channel 10/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:155360 [3] NCCL INFO Channel 10/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:155361 [1] NCCL INFO Channel 13/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:155359 [7] NCCL INFO Channel 08/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:155358 [5] NCCL INFO Channel 09/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:155356 [6] NCCL INFO Channel 00/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148696:155354 [0] NCCL INFO Channel 11/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:155360 [3] NCCL INFO Channel 11/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:155361 [1] NCCL INFO Channel 14/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:155359 [7] NCCL INFO Channel 09/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:155358 [5] NCCL INFO Channel 10/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:155356 [6] NCCL INFO Channel 01/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:155355 [4] NCCL INFO Channel 00/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148696:155354 [0] NCCL INFO Channel 12/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:155360 [3] NCCL INFO Channel 12/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:155361 [1] NCCL INFO Channel 15/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:155359 [7] NCCL INFO Channel 10/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:155358 [5] NCCL INFO Channel 11/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:155356 [6] NCCL INFO Channel 02/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:155355 [4] NCCL INFO Channel 01/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148696:155354 [0] NCCL INFO Channel 13/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:155360 [3] NCCL INFO Channel 13/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:155359 [7] NCCL INFO Channel 11/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:155361 [1] NCCL INFO Channel 16/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:155358 [5] NCCL INFO Channel 12/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:155356 [6] NCCL INFO Channel 03/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:155355 [4] NCCL INFO Channel 02/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148696:155354 [0] NCCL INFO Channel 14/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:155360 [3] NCCL INFO Channel 14/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:155359 [7] NCCL INFO Channel 12/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:155361 [1] NCCL INFO Channel 17/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:155357 [2] NCCL INFO Channel 00/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:155358 [5] NCCL INFO Channel 13/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:155356 [6] NCCL INFO Channel 04/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:155355 [4] NCCL INFO Channel 03/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148696:155354 [0] NCCL INFO Channel 15/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:155360 [3] NCCL INFO Channel 15/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:155359 [7] NCCL INFO Channel 13/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:155361 [1] NCCL INFO Channel 18/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:155357 [2] NCCL INFO Channel 01/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:155358 [5] NCCL INFO Channel 14/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:155356 [6] NCCL INFO Channel 05/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:155355 [4] NCCL INFO Channel 04/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148696:155354 [0] NCCL INFO Channel 16/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:155360 [3] NCCL INFO Channel 16/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:155359 [7] NCCL INFO Channel 14/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:155361 [1] NCCL INFO Channel 19/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:155357 [2] NCCL INFO Channel 02/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:155358 [5] NCCL INFO Channel 15/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:155356 [6] NCCL INFO Channel 06/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148696:155354 [0] NCCL INFO Channel 17/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:155355 [4] NCCL INFO Channel 05/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:155360 [3] NCCL INFO Channel 17/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:155359 [7] NCCL INFO Channel 15/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:155361 [1] NCCL INFO Channel 20/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:155357 [2] NCCL INFO Channel 03/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:155358 [5] NCCL INFO Channel 16/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:155356 [6] NCCL INFO Channel 07/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148696:155354 [0] NCCL INFO Channel 18/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:155355 [4] NCCL INFO Channel 06/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:155360 [3] NCCL INFO Channel 18/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:155359 [7] NCCL INFO Channel 16/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:155361 [1] NCCL INFO Channel 21/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:155357 [2] NCCL INFO Channel 04/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:155358 [5] NCCL INFO Channel 17/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:155356 [6] NCCL INFO Channel 08/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148696:155354 [0] NCCL INFO Channel 19/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:155355 [4] NCCL INFO Channel 07/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:155360 [3] NCCL INFO Channel 19/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:155361 [1] NCCL INFO Channel 22/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:155359 [7] NCCL INFO Channel 17/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:155357 [2] NCCL INFO Channel 05/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:155358 [5] NCCL INFO Channel 18/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:155356 [6] NCCL INFO Channel 09/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148696:155354 [0] NCCL INFO Channel 20/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:155355 [4] NCCL INFO Channel 08/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:155360 [3] NCCL INFO Channel 20/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:155361 [1] NCCL INFO Channel 23/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:155359 [7] NCCL INFO Channel 18/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:155357 [2] NCCL INFO Channel 06/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:155358 [5] NCCL INFO Channel 19/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148696:155354 [0] NCCL INFO Channel 21/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:155356 [6] NCCL INFO Channel 10/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:155355 [4] NCCL INFO Channel 09/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:155360 [3] NCCL INFO Channel 21/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:155359 [7] NCCL INFO Channel 19/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:155357 [2] NCCL INFO Channel 07/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:155358 [5] NCCL INFO Channel 20/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148696:155354 [0] NCCL INFO Channel 22/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:155356 [6] NCCL INFO Channel 11/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:155360 [3] NCCL INFO Channel 22/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:155355 [4] NCCL INFO Channel 10/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:155359 [7] NCCL INFO Channel 20/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:155357 [2] NCCL INFO Channel 08/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:155358 [5] NCCL INFO Channel 21/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148696:155354 [0] NCCL INFO Channel 23/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:155356 [6] NCCL INFO Channel 12/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:155360 [3] NCCL INFO Channel 23/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:155355 [4] NCCL INFO Channel 11/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:155359 [7] NCCL INFO Channel 21/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:155357 [2] NCCL INFO Channel 09/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:155358 [5] NCCL INFO Channel 22/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:155356 [6] NCCL INFO Channel 13/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:155355 [4] NCCL INFO Channel 12/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:155359 [7] NCCL INFO Channel 22/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:155357 [2] NCCL INFO Channel 10/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:155358 [5] NCCL INFO Channel 23/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:155356 [6] NCCL INFO Channel 14/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:155355 [4] NCCL INFO Channel 13/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:155359 [7] NCCL INFO Channel 23/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:155357 [2] NCCL INFO Channel 11/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:155356 [6] NCCL INFO Channel 15/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:155355 [4] NCCL INFO Channel 14/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:155356 [6] NCCL INFO Channel 16/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:155355 [4] NCCL INFO Channel 15/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:155357 [2] NCCL INFO Channel 12/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:155356 [6] NCCL INFO Channel 17/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:155355 [4] NCCL INFO Channel 16/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:155357 [2] NCCL INFO Channel 13/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:155356 [6] NCCL INFO Channel 18/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:155355 [4] NCCL INFO Channel 17/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:155357 [2] NCCL INFO Channel 14/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:155356 [6] NCCL INFO Channel 19/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:155355 [4] NCCL INFO Channel 18/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:155357 [2] NCCL INFO Channel 15/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:155355 [4] NCCL INFO Channel 19/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:155357 [2] NCCL INFO Channel 16/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:155356 [6] NCCL INFO Channel 20/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:155355 [4] NCCL INFO Channel 20/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:155357 [2] NCCL INFO Channel 17/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:155356 [6] NCCL INFO Channel 21/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:155355 [4] NCCL INFO Channel 21/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:155357 [2] NCCL INFO Channel 18/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:155356 [6] NCCL INFO Channel 22/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:155355 [4] NCCL INFO Channel 22/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:155357 [2] NCCL INFO Channel 19/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:155356 [6] NCCL INFO Channel 23/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:155355 [4] NCCL INFO Channel 23/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:155357 [2] NCCL INFO Channel 20/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:155357 [2] NCCL INFO Channel 21/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:155357 [2] NCCL INFO Channel 22/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:155357 [2] NCCL INFO Channel 23/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:155357 [2] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:148697:155361 [1] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:148696:155354 [0] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:148703:155359 [7] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:148703:155359 [7] NCCL INFO Channel 00/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:155359 [7] NCCL INFO Channel 01/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:155359 [7] NCCL INFO Channel 02/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:155360 [3] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:148703:155359 [7] NCCL INFO Channel 03/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:155359 [7] NCCL INFO Channel 04/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:155357 [2] NCCL INFO Channel 00/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:155359 [7] NCCL INFO Channel 05/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:155357 [2] NCCL INFO Channel 01/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:155357 [2] NCCL INFO Channel 02/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:155359 [7] NCCL INFO Channel 06/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:155357 [2] NCCL INFO Channel 03/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:155359 [7] NCCL INFO Channel 07/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:155359 [7] NCCL INFO Channel 08/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:155357 [2] NCCL INFO Channel 04/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:155359 [7] NCCL INFO Channel 09/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:155359 [7] NCCL INFO Channel 10/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:155357 [2] NCCL INFO Channel 05/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:155357 [2] NCCL INFO Channel 06/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:155357 [2] NCCL INFO Channel 07/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:155359 [7] NCCL INFO Channel 11/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:155357 [2] NCCL INFO Channel 08/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:155359 [7] NCCL INFO Channel 12/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:155357 [2] NCCL INFO Channel 09/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:155359 [7] NCCL INFO Channel 13/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:155357 [2] NCCL INFO Channel 10/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:155359 [7] NCCL INFO Channel 14/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:155357 [2] NCCL INFO Channel 11/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:155359 [7] NCCL INFO Channel 15/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:155361 [1] NCCL INFO Channel 00/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:155357 [2] NCCL INFO Channel 12/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:155359 [7] NCCL INFO Channel 16/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:155358 [5] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:148697:155361 [1] NCCL INFO Channel 01/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:155357 [2] NCCL INFO Channel 13/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:155359 [7] NCCL INFO Channel 17/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:155361 [1] NCCL INFO Channel 02/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:155357 [2] NCCL INFO Channel 14/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:155359 [7] NCCL INFO Channel 18/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:155361 [1] NCCL INFO Channel 03/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:155357 [2] NCCL INFO Channel 15/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:155359 [7] NCCL INFO Channel 19/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:155361 [1] NCCL INFO Channel 04/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:155355 [4] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:148698:155357 [2] NCCL INFO Channel 16/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:155359 [7] NCCL INFO Channel 20/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:155356 [6] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:148697:155361 [1] NCCL INFO Channel 05/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:155357 [2] NCCL INFO Channel 17/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:155359 [7] NCCL INFO Channel 21/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:155361 [1] NCCL INFO Channel 06/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:155357 [2] NCCL INFO Channel 18/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:155359 [7] NCCL INFO Channel 22/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:155361 [1] NCCL INFO Channel 07/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:155357 [2] NCCL INFO Channel 19/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148703:155359 [7] NCCL INFO Channel 23/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:155361 [1] NCCL INFO Channel 08/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:155357 [2] NCCL INFO Channel 20/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:155361 [1] NCCL INFO Channel 09/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:155357 [2] NCCL INFO Channel 21/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:155361 [1] NCCL INFO Channel 10/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:155357 [2] NCCL INFO Channel 22/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:155361 [1] NCCL INFO Channel 11/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148698:155357 [2] NCCL INFO Channel 23/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:155361 [1] NCCL INFO Channel 12/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:155361 [1] NCCL INFO Channel 13/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:155361 [1] NCCL INFO Channel 14/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:155361 [1] NCCL INFO Channel 15/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:155361 [1] NCCL INFO Channel 16/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:155361 [1] NCCL INFO Channel 17/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:155361 [1] NCCL INFO Channel 18/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:155361 [1] NCCL INFO Channel 19/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:155361 [1] NCCL INFO Channel 20/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:155361 [1] NCCL INFO Channel 21/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:155360 [3] NCCL INFO Channel 00/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:155361 [1] NCCL INFO Channel 22/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:155360 [3] NCCL INFO Channel 01/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148697:155361 [1] NCCL INFO Channel 23/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:155360 [3] NCCL INFO Channel 02/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:155358 [5] NCCL INFO Channel 00/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:155360 [3] NCCL INFO Channel 03/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:155358 [5] NCCL INFO Channel 01/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:155358 [5] NCCL INFO Channel 02/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:155360 [3] NCCL INFO Channel 04/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:155360 [3] NCCL INFO Channel 05/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:155358 [5] NCCL INFO Channel 03/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:155360 [3] NCCL INFO Channel 06/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:155358 [5] NCCL INFO Channel 04/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:155360 [3] NCCL INFO Channel 07/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:155358 [5] NCCL INFO Channel 05/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:155360 [3] NCCL INFO Channel 08/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:155358 [5] NCCL INFO Channel 06/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:155358 [5] NCCL INFO Channel 07/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:155358 [5] NCCL INFO Channel 08/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:155356 [6] NCCL INFO Channel 00/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:155358 [5] NCCL INFO Channel 09/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:155360 [3] NCCL INFO Channel 09/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:155356 [6] NCCL INFO Channel 01/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:155355 [4] NCCL INFO Channel 00/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:155358 [5] NCCL INFO Channel 10/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:155360 [3] NCCL INFO Channel 10/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:155355 [4] NCCL INFO Channel 01/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:155360 [3] NCCL INFO Channel 11/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:155355 [4] NCCL INFO Channel 02/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:155360 [3] NCCL INFO Channel 12/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:155356 [6] NCCL INFO Channel 02/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:155355 [4] NCCL INFO Channel 03/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:155356 [6] NCCL INFO Channel 03/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:155360 [3] NCCL INFO Channel 13/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:155355 [4] NCCL INFO Channel 04/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:155356 [6] NCCL INFO Channel 04/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:155360 [3] NCCL INFO Channel 14/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:155355 [4] NCCL INFO Channel 05/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:155356 [6] NCCL INFO Channel 05/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:155360 [3] NCCL INFO Channel 15/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:155358 [5] NCCL INFO Channel 11/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:155355 [4] NCCL INFO Channel 06/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:155360 [3] NCCL INFO Channel 16/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:155358 [5] NCCL INFO Channel 12/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:155355 [4] NCCL INFO Channel 07/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:155360 [3] NCCL INFO Channel 17/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:155358 [5] NCCL INFO Channel 13/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:155355 [4] NCCL INFO Channel 08/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:155360 [3] NCCL INFO Channel 18/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:155358 [5] NCCL INFO Channel 14/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:155360 [3] NCCL INFO Channel 19/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:155358 [5] NCCL INFO Channel 15/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:155355 [4] NCCL INFO Channel 09/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:155360 [3] NCCL INFO Channel 20/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:155358 [5] NCCL INFO Channel 16/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:155355 [4] NCCL INFO Channel 10/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:155356 [6] NCCL INFO Channel 06/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:155360 [3] NCCL INFO Channel 21/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:155358 [5] NCCL INFO Channel 17/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:155355 [4] NCCL INFO Channel 11/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:155356 [6] NCCL INFO Channel 07/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:155358 [5] NCCL INFO Channel 18/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:155356 [6] NCCL INFO Channel 08/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:155358 [5] NCCL INFO Channel 19/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:155355 [4] NCCL INFO Channel 12/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:155356 [6] NCCL INFO Channel 09/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:155360 [3] NCCL INFO Channel 22/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:155358 [5] NCCL INFO Channel 20/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:155355 [4] NCCL INFO Channel 13/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148699:155360 [3] NCCL INFO Channel 23/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:155358 [5] NCCL INFO Channel 21/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:155355 [4] NCCL INFO Channel 14/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:155358 [5] NCCL INFO Channel 22/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:155355 [4] NCCL INFO Channel 15/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:155356 [6] NCCL INFO Channel 10/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148701:155358 [5] NCCL INFO Channel 23/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:155356 [6] NCCL INFO Channel 11/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:155355 [4] NCCL INFO Channel 16/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:155356 [6] NCCL INFO Channel 12/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:155355 [4] NCCL INFO Channel 17/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:155356 [6] NCCL INFO Channel 13/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:155355 [4] NCCL INFO Channel 18/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:155356 [6] NCCL INFO Channel 14/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:155355 [4] NCCL INFO Channel 19/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:155356 [6] NCCL INFO Channel 15/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:155355 [4] NCCL INFO Channel 20/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:155356 [6] NCCL INFO Channel 16/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:155355 [4] NCCL INFO Channel 21/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:155356 [6] NCCL INFO Channel 17/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:155355 [4] NCCL INFO Channel 22/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:155356 [6] NCCL INFO Channel 18/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148700:155355 [4] NCCL INFO Channel 23/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:155356 [6] NCCL INFO Channel 19/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:155356 [6] NCCL INFO Channel 20/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:155356 [6] NCCL INFO Channel 21/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:155356 [6] NCCL INFO Channel 22/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148702:155356 [6] NCCL INFO Channel 23/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:148696:155354 [0] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:148696:155354 [0] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:148696:155354 [0] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:148697:155361 [1] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:148697:155361 [1] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:148697:155361 [1] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:148698:155357 [2] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:148698:155357 [2] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:148698:155357 [2] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:148699:155360 [3] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:148699:155360 [3] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:148699:155360 [3] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:148700:155355 [4] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:148700:155355 [4] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:148700:155355 [4] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:148701:155358 [5] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:148701:155358 [5] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:148701:155358 [5] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:148703:155359 [7] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:148703:155359 [7] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:148703:155359 [7] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:148702:155356 [6] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:148702:155356 [6] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:148702:155356 [6] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:148696:155354 [0] NCCL INFO ncclCommInitRank comm 0x7fabd806abb0 rank 0 nranks 8 cudaDev 0 nvmlDev 0 busId 101c0 commId 0x3fbbf6958192d7b6 - Init COMPLETE +ywang29-vrdb-test1-worker-0:148700:155355 [4] NCCL INFO ncclCommInitRank comm 0x7f03a006a900 rank 4 nranks 8 cudaDev 4 nvmlDev 4 busId 901c0 commId 0x3fbbf6958192d7b6 - Init COMPLETE +ywang29-vrdb-test1-worker-0:148699:155360 [3] NCCL INFO ncclCommInitRank comm 0x7f1b4406b290 rank 3 nranks 8 cudaDev 3 nvmlDev 3 busId 201d0 commId 0x3fbbf6958192d7b6 - Init COMPLETE +ywang29-vrdb-test1-worker-0:148698:155357 [2] NCCL INFO ncclCommInitRank comm 0x7fee6c06afb0 rank 2 nranks 8 cudaDev 2 nvmlDev 2 busId 201c0 commId 0x3fbbf6958192d7b6 - Init COMPLETE +ywang29-vrdb-test1-worker-0:148703:155359 [7] NCCL INFO ncclCommInitRank comm 0x7f250c06ac80 rank 7 nranks 8 cudaDev 7 nvmlDev 7 busId a01d0 commId 0x3fbbf6958192d7b6 - Init COMPLETE +ywang29-vrdb-test1-worker-0:148702:155356 [6] NCCL INFO ncclCommInitRank comm 0x7f9cb806b130 rank 6 nranks 8 cudaDev 6 nvmlDev 6 busId a01c0 commId 0x3fbbf6958192d7b6 - Init COMPLETE +ywang29-vrdb-test1-worker-0:148701:155358 [5] NCCL INFO ncclCommInitRank comm 0x7fb0c806a0b0 rank 5 nranks 8 cudaDev 5 nvmlDev 5 busId 901d0 commId 0x3fbbf6958192d7b6 - Init COMPLETE +ywang29-vrdb-test1-worker-0:148697:155361 [1] NCCL INFO ncclCommInitRank comm 0x7fdb6c06a5d0 rank 1 nranks 8 cudaDev 1 nvmlDev 1 busId 101d0 commId 0x3fbbf6958192d7b6 - Init COMPLETE + 0%| | 1/520 [00:14<2:02:36, 14.18s/it] {'loss': 8.1039, 'grad_norm': 0.42433090644130744, 'learning_rate': 0.0125, 'epoch': 0.0} + 0%| | 1/520 [00:14<2:02:36, 14.18s/it] 0%| | 2/520 [00:17<1:09:23, 8.04s/it] {'loss': 7.3651, 'grad_norm': 0.4381749794680582, 'learning_rate': 0.025, 'epoch': 0.0} + 0%| | 2/520 [00:17<1:09:23, 8.04s/it] 1%| | 3/520 [00:21<51:56, 6.03s/it] {'loss': 6.736, 'grad_norm': 0.2108022866718713, 'learning_rate': 0.037500000000000006, 'epoch': 0.01} + 1%| | 3/520 [00:21<51:56, 6.03s/it] 1%| | 4/520 [00:25<43:45, 5.09s/it] {'loss': 5.545, 'grad_norm': 0.13280182861496026, 'learning_rate': 0.05, 'epoch': 0.01} + 1%| | 4/520 [00:25<43:45, 5.09s/it] 1%| | 5/520 [00:28<39:11, 4.57s/it] {'loss': 4.4774, 'grad_norm': 0.1498079149039122, 'learning_rate': 0.0625, 'epoch': 0.01} + 1%| | 5/520 [00:28<39:11, 4.57s/it] 1%| | 6/520 [00:32<36:30, 4.26s/it] {'loss': 5.0738, 'grad_norm': 0.11763628109551362, 'learning_rate': 0.07500000000000001, 'epoch': 0.01} + 1%| | 6/520 [00:32<36:30, 4.26s/it] 1%|▏ | 7/520 [00:36<34:41, 4.06s/it] {'loss': 3.3731, 'grad_norm': 0.08213078117442542, 'learning_rate': 0.08750000000000001, 'epoch': 0.01} + 1%|▏ | 7/520 [00:36<34:41, 4.06s/it] 2%|▏ | 8/520 [00:40<35:10, 4.12s/it] {'loss': 3.3078, 'grad_norm': 0.1400775994011567, 'learning_rate': 0.1, 'epoch': 0.02} + 2%|▏ | 8/520 [00:40<35:10, 4.12s/it] 2%|▏ | 9/520 [00:44<35:05, 4.12s/it] {'loss': 2.9294, 'grad_norm': 0.034679078688469554, 'learning_rate': 0.1125, 'epoch': 0.02} + 2%|▏ | 9/520 [00:44<35:05, 4.12s/it] 2%|▏ | 10/520 [00:48<33:46, 3.97s/it] {'loss': 2.4825, 'grad_norm': 0.04216044620133159, 'learning_rate': 0.125, 'epoch': 0.02} + 2%|▏ | 10/520 [00:48<33:46, 3.97s/it] 2%|▏ | 11/520 [00:51<33:11, 3.91s/it] {'loss': 2.4639, 'grad_norm': 0.02523437011943625, 'learning_rate': 0.1375, 'epoch': 0.02} + 2%|▏ | 11/520 [00:51<33:11, 3.91s/it] 2%|▏ | 12/520 [00:55<32:27, 3.83s/it] {'loss': 2.6862, 'grad_norm': 0.03207368987406362, 'learning_rate': 0.15000000000000002, 'epoch': 0.02} + 2%|▏ | 12/520 [00:55<32:27, 3.83s/it][2025-10-12 05:30:51,094] [WARNING] [stage3.py:2069:step] 1 pytorch allocator cache flushes since last step. this happens when there is high memory pressure and is detrimental to performance. if this is happening frequently consider adjusting settings to reduce memory consumption. If you are unable to make the cache flushes go away consider adding get_accelerator().empty_cache() calls in your training loop to ensure that all ranks flush their caches at the same time + 2%|▎ | 13/520 [00:59<33:35, 3.98s/it] {'loss': 2.1125, 'grad_norm': 0.014417158626718371, 'learning_rate': 0.1625, 'epoch': 0.03} + 2%|▎ | 13/520 [00:59<33:35, 3.98s/it] 3%|▎ | 14/520 [01:03<32:40, 3.88s/it] {'loss': 2.1392, 'grad_norm': 0.01651237446343107, 'learning_rate': 0.17500000000000002, 'epoch': 0.03} + 3%|▎ | 14/520 [01:03<32:40, 3.88s/it] 3%|▎ | 15/520 [01:07<32:09, 3.82s/it] {'loss': 2.3012, 'grad_norm': 0.018104200375786906, 'learning_rate': 0.1875, 'epoch': 0.03} + 3%|▎ | 15/520 [01:07<32:09, 3.82s/it] 3%|▎ | 16/520 [01:10<31:44, 3.78s/it] {'loss': 2.2609, 'grad_norm': 0.030813171231753502, 'learning_rate': 0.2, 'epoch': 0.03} + 3%|▎ | 16/520 [01:10<31:44, 3.78s/it] 3%|▎ | 17/520 [01:14<31:30, 3.76s/it] {'loss': 2.1473, 'grad_norm': 0.030552630654928456, 'learning_rate': 0.1999980572931538, 'epoch': 0.03} + 3%|▎ | 17/520 [01:14<31:30, 3.76s/it] 3%|▎ | 18/520 [01:18<31:07, 3.72s/it] {'loss': 1.9242, 'grad_norm': 0.0385613411799719, 'learning_rate': 0.19999222924809748, 'epoch': 0.03} + 3%|▎ | 18/520 [01:18<31:07, 3.72s/it] 4%|▎ | 19/520 [01:22<31:13, 3.74s/it] {'loss': 2.5249, 'grad_norm': 0.06206816521322989, 'learning_rate': 0.19998251609127465, 'epoch': 0.04} + 4%|▎ | 19/520 [01:22<31:13, 3.74s/it] 4%|▍ | 20/520 [01:25<31:17, 3.76s/it] {'loss': 1.8615, 'grad_norm': 0.018380546445001537, 'learning_rate': 0.19996891820008164, 'epoch': 0.04} + 4%|▍ | 20/520 [01:25<31:17, 3.76s/it] 4%|▍ | 21/520 [01:29<31:02, 3.73s/it] {'loss': 2.4848, 'grad_norm': 0.04099766384547669, 'learning_rate': 0.19995143610285276, 'epoch': 0.04} + 4%|▍ | 21/520 [01:29<31:02, 3.73s/it] 4%|▍ | 22/520 [01:33<30:47, 3.71s/it] {'loss': 2.112, 'grad_norm': 0.03950287120824162, 'learning_rate': 0.19993007047883987, 'epoch': 0.04} + 4%|▍ | 22/520 [01:33<30:47, 3.71s/it] 4%|▍ | 23/520 [01:36<30:38, 3.70s/it] {'loss': 1.8858, 'grad_norm': 0.010441427083868115, 'learning_rate': 0.1999048221581858, 'epoch': 0.04} + 4%|▍ | 23/520 [01:36<30:38, 3.70s/it] 5%|▍ | 24/520 [01:40<30:25, 3.68s/it] {'loss': 2.153, 'grad_norm': 0.02549818253094147, 'learning_rate': 0.19987569212189224, 'epoch': 0.05} + 5%|▍ | 24/520 [01:40<30:25, 3.68s/it] 5%|▍ | 25/520 [01:44<30:29, 3.70s/it] {'loss': 1.8903, 'grad_norm': 0.014170805355063272, 'learning_rate': 0.19984268150178167, 'epoch': 0.05} + 5%|▍ | 25/520 [01:44<30:29, 3.70s/it] 5%|▌ | 26/520 [01:48<30:43, 3.73s/it] {'loss': 1.8839, 'grad_norm': 0.017686668454284415, 'learning_rate': 0.1998057915804532, 'epoch': 0.05} + 5%|▌ | 26/520 [01:48<30:43, 3.73s/it] 5%|▌ | 27/520 [01:51<30:48, 3.75s/it] {'loss': 1.7064, 'grad_norm': 0.011462304230203306, 'learning_rate': 0.1997650237912329, 'epoch': 0.05} + 5%|▌ | 27/520 [01:51<30:48, 3.75s/it] 5%|▌ | 28/520 [01:55<30:53, 3.77s/it] {'loss': 1.6772, 'grad_norm': 0.007931548434033158, 'learning_rate': 0.199720379718118, 'epoch': 0.05} + 5%|▌ | 28/520 [01:55<30:53, 3.77s/it] 6%|▌ | 29/520 [01:59<30:50, 3.77s/it] {'loss': 1.7147, 'grad_norm': 0.014722137077613833, 'learning_rate': 0.19967186109571552, 'epoch': 0.06} + 6%|▌ | 29/520 [01:59<30:50, 3.77s/it] 6%|▌ | 30/520 [02:03<30:31, 3.74s/it] {'loss': 2.1657, 'grad_norm': 0.019883202181920946, 'learning_rate': 0.19961946980917455, 'epoch': 0.06} + 6%|▌ | 30/520 [02:03<30:31, 3.74s/it] 6%|▌ | 31/520 [02:06<30:21, 3.73s/it] {'loss': 1.6795, 'grad_norm': 0.00851706663566311, 'learning_rate': 0.1995632078941134, 'epoch': 0.06} + 6%|▌ | 31/520 [02:06<30:21, 3.73s/it] 6%|▌ | 32/520 [02:10<30:13, 3.72s/it] {'loss': 2.3215, 'grad_norm': 0.024657067098520717, 'learning_rate': 0.19950307753654017, 'epoch': 0.06} + 6%|▌ | 32/520 [02:10<30:13, 3.72s/it] 6%|▋ | 33/520 [02:14<30:06, 3.71s/it] {'loss': 1.6685, 'grad_norm': 0.009041353734449137, 'learning_rate': 0.19943908107276798, 'epoch': 0.06} + 6%|▋ | 33/520 [02:14<30:06, 3.71s/it] 7%|▋ | 34/520 [02:17<29:53, 3.69s/it] {'loss': 1.657, 'grad_norm': 0.008049083611976118, 'learning_rate': 0.19937122098932428, 'epoch': 0.07} + 7%|▋ | 34/520 [02:17<29:53, 3.69s/it] 7%|▋ | 35/520 [02:21<29:44, 3.68s/it] {'loss': 1.6469, 'grad_norm': 0.007113793121344598, 'learning_rate': 0.19929949992285395, 'epoch': 0.07} + 7%|▋ | 35/520 [02:21<29:44, 3.68s/it] 7%|▋ | 36/520 [02:25<29:42, 3.68s/it] {'loss': 1.7676, 'grad_norm': 0.006293790466978742, 'learning_rate': 0.19922392066001723, 'epoch': 0.07} + 7%|▋ | 36/520 [02:25<29:42, 3.68s/it] 7%|▋ | 37/520 [02:28<29:36, 3.68s/it] {'loss': 2.1297, 'grad_norm': 0.029430933711046568, 'learning_rate': 0.19914448613738106, 'epoch': 0.07} + 7%|▋ | 37/520 [02:28<29:36, 3.68s/it] 7%|▋ | 38/520 [02:32<29:28, 3.67s/it] {'loss': 1.8577, 'grad_norm': 0.014176800962488441, 'learning_rate': 0.1990611994413053, 'epoch': 0.07} + 7%|▋ | 38/520 [02:32<29:28, 3.67s/it] 8%|▊ | 39/520 [02:36<29:27, 3.67s/it] {'loss': 1.6564, 'grad_norm': 0.0069500115171614715, 'learning_rate': 0.19897406380782262, 'epoch': 0.07} + 8%|▊ | 39/520 [02:36<29:27, 3.67s/it] 8%|▊ | 40/520 [02:39<29:19, 3.67s/it] {'loss': 1.6937, 'grad_norm': 0.013172975204579794, 'learning_rate': 0.19888308262251286, 'epoch': 0.08} + 8%|▊ | 40/520 [02:39<29:19, 3.67s/it] 8%|▊ | 41/520 [02:43<29:19, 3.67s/it] {'loss': 1.6571, 'grad_norm': 0.006745836959838009, 'learning_rate': 0.19878825942037148, 'epoch': 0.08} + 8%|▊ | 41/520 [02:43<29:19, 3.67s/it] 8%|▊ | 42/520 [02:47<29:12, 3.67s/it] {'loss': 1.701, 'grad_norm': 0.011499098654907301, 'learning_rate': 0.19868959788567211, 'epoch': 0.08} + 8%|▊ | 42/520 [02:47<29:12, 3.67s/it] 8%|▊ | 43/520 [02:50<29:04, 3.66s/it] {'loss': 1.8922, 'grad_norm': 0.014033882939371045, 'learning_rate': 0.1985871018518236, 'epoch': 0.08} + 8%|▊ | 43/520 [02:50<29:04, 3.66s/it] 8%|▊ | 44/520 [02:54<28:57, 3.65s/it] {'loss': 1.9774, 'grad_norm': 0.009462358944423514, 'learning_rate': 0.19848077530122082, 'epoch': 0.08} + 8%|▊ | 44/520 [02:54<28:57, 3.65s/it] 9%|▊ | 45/520 [02:58<28:53, 3.65s/it] {'loss': 1.6659, 'grad_norm': 0.011695270862822056, 'learning_rate': 0.19837062236509015, 'epoch': 0.09} + 9%|▊ | 45/520 [02:58<28:53, 3.65s/it] 9%|▉ | 46/520 [03:01<28:45, 3.64s/it] {'loss': 2.02, 'grad_norm': 0.007956291482791915, 'learning_rate': 0.19825664732332884, 'epoch': 0.09} + 9%|▉ | 46/520 [03:01<28:45, 3.64s/it] 9%|▉ | 47/520 [03:05<28:39, 3.64s/it] {'loss': 1.6523, 'grad_norm': 0.011696079115600548, 'learning_rate': 0.19813885460433878, 'epoch': 0.09} + 9%|▉ | 47/520 [03:05<28:39, 3.64s/it] 9%|▉ | 48/520 [03:08<28:42, 3.65s/it] {'loss': 1.6486, 'grad_norm': 0.007570622355342347, 'learning_rate': 0.19801724878485438, 'epoch': 0.09} + 9%|▉ | 48/520 [03:08<28:42, 3.65s/it] 9%|▉ | 49/520 [03:12<28:39, 3.65s/it] {'loss': 1.645, 'grad_norm': 0.008380912157900856, 'learning_rate': 0.19789183458976486, 'epoch': 0.09} + 9%|▉ | 49/520 [03:12<28:39, 3.65s/it] 10%|▉ | 50/520 [03:16<28:33, 3.65s/it] {'loss': 1.63, 'grad_norm': 0.007848991917601817, 'learning_rate': 0.19776261689193048, 'epoch': 0.1} + 10%|▉ | 50/520 [03:16<28:33, 3.65s/it] 10%|▉ | 51/520 [03:19<28:31, 3.65s/it] {'loss': 1.5397, 'grad_norm': 0.0055352794864879035, 'learning_rate': 0.19762960071199334, 'epoch': 0.1} + 10%|▉ | 51/520 [03:19<28:31, 3.65s/it] 10%|█ | 52/520 [03:23<28:32, 3.66s/it] {'loss': 1.6957, 'grad_norm': 0.010481141961070717, 'learning_rate': 0.19749279121818236, 'epoch': 0.1} + 10%|█ | 52/520 [03:23<28:32, 3.66s/it] 10%|█ | 53/520 [03:27<28:22, 3.65s/it] {'loss': 1.6942, 'grad_norm': 0.005729284567590582, 'learning_rate': 0.19735219372611235, 'epoch': 0.1} + 10%|█ | 53/520 [03:27<28:22, 3.65s/it] 10%|█ | 54/520 [03:30<28:12, 3.63s/it] {'loss': 1.5497, 'grad_norm': 0.006839240733027484, 'learning_rate': 0.19720781369857746, 'epoch': 0.1} + 10%|█ | 54/520 [03:30<28:12, 3.63s/it] 11%|█ | 55/520 [03:34<28:14, 3.64s/it] {'loss': 1.5365, 'grad_norm': 0.007238876241587366, 'learning_rate': 0.1970596567453391, 'epoch': 0.11} + 11%|█ | 55/520 [03:34<28:14, 3.64s/it] 11%|█ | 56/520 [03:38<28:08, 3.64s/it] {'loss': 1.6926, 'grad_norm': 0.005164643137063366, 'learning_rate': 0.1969077286229078, 'epoch': 0.11} + 11%|█ | 56/520 [03:38<28:08, 3.64s/it] 11%|█ | 57/520 [03:41<28:07, 3.65s/it] {'loss': 1.5413, 'grad_norm': 0.008585943345777476, 'learning_rate': 0.19675203523431964, 'epoch': 0.11} + 11%|█ | 57/520 [03:41<28:07, 3.65s/it] 11%|█ | 58/520 [03:45<28:05, 3.65s/it] {'loss': 1.7089, 'grad_norm': 0.006721539736995277, 'learning_rate': 0.19659258262890683, 'epoch': 0.11} + 11%|█ | 58/520 [03:45<28:05, 3.65s/it] 11%|█▏ | 59/520 [03:49<28:01, 3.65s/it] {'loss': 1.7017, 'grad_norm': 0.009785131872986122, 'learning_rate': 0.19642937700206278, 'epoch': 0.11} + 11%|█▏ | 59/520 [03:49<28:01, 3.65s/it] 12%|█▏ | 60/520 [03:52<27:56, 3.65s/it] {'loss': 1.6078, 'grad_norm': 0.00779753719618451, 'learning_rate': 0.19626242469500121, 'epoch': 0.12} + 12%|█▏ | 60/520 [03:52<27:56, 3.65s/it] 12%|█▏ | 61/520 [03:56<27:49, 3.64s/it] {'loss': 1.8348, 'grad_norm': 0.008801419663724157, 'learning_rate': 0.19609173219450998, 'epoch': 0.12} + 12%|█▏ | 61/520 [03:56<27:49, 3.64s/it] 12%|█▏ | 62/520 [04:00<27:54, 3.66s/it] {'loss': 1.5871, 'grad_norm': 0.006664902457362813, 'learning_rate': 0.19591730613269878, 'epoch': 0.12} + 12%|█▏ | 62/520 [04:00<27:54, 3.66s/it] 12%|█▏ | 63/520 [04:03<27:47, 3.65s/it] {'loss': 1.5842, 'grad_norm': 0.007950849390871987, 'learning_rate': 0.19573915328674182, 'epoch': 0.12} + 12%|█▏ | 63/520 [04:03<27:47, 3.65s/it] 12%|█▏ | 64/520 [04:07<28:01, 3.69s/it] {'loss': 1.6021, 'grad_norm': 0.0060110920199643895, 'learning_rate': 0.1955572805786141, 'epoch': 0.12} + 12%|█▏ | 64/520 [04:07<28:01, 3.69s/it] 12%|█▎ | 65/520 [04:11<27:53, 3.68s/it] {'loss': 1.5996, 'grad_norm': 0.00551098739815152, 'learning_rate': 0.1953716950748227, 'epoch': 0.12} + 12%|█▎ | 65/520 [04:11<27:53, 3.68s/it] 13%|█▎ | 66/520 [04:14<27:41, 3.66s/it] {'loss': 1.5644, 'grad_norm': 0.008182408445785374, 'learning_rate': 0.19518240398613226, 'epoch': 0.13} + 13%|█▎ | 66/520 [04:14<27:41, 3.66s/it] 13%|█▎ | 67/520 [04:18<27:33, 3.65s/it] {'loss': 1.4411, 'grad_norm': 0.006820017738105664, 'learning_rate': 0.1949894146672846, 'epoch': 0.13} + 13%|█▎ | 67/520 [04:18<27:33, 3.65s/it] 13%|█▎ | 68/520 [04:21<27:27, 3.64s/it] {'loss': 1.4689, 'grad_norm': 0.004661711078493773, 'learning_rate': 0.1947927346167132, 'epoch': 0.13} + 13%|█▎ | 68/520 [04:21<27:27, 3.64s/it] 13%|█▎ | 69/520 [04:25<27:27, 3.65s/it] {'loss': 1.4722, 'grad_norm': 0.008973056002093876, 'learning_rate': 0.1945923714762516, 'epoch': 0.13} + 13%|█▎ | 69/520 [04:25<27:27, 3.65s/it] 13%|█▎ | 70/520 [04:29<27:23, 3.65s/it] {'loss': 1.5436, 'grad_norm': 0.008225371242106176, 'learning_rate': 0.19438833303083677, 'epoch': 0.13} + 13%|█▎ | 70/520 [04:29<27:23, 3.65s/it] 14%|█▎ | 71/520 [04:32<27:16, 3.65s/it] {'loss': 1.4411, 'grad_norm': 0.004317117565564573, 'learning_rate': 0.19418062720820636, 'epoch': 0.14} + 14%|█▎ | 71/520 [04:32<27:16, 3.65s/it] 14%|█▍ | 72/520 [04:36<27:07, 3.63s/it] {'loss': 1.6037, 'grad_norm': 0.0068443842357252685, 'learning_rate': 0.19396926207859086, 'epoch': 0.14} + 14%|█▍ | 72/520 [04:36<27:07, 3.63s/it] 14%|█▍ | 73/520 [04:40<27:00, 3.63s/it] {'loss': 1.4263, 'grad_norm': 0.006362801456034968, 'learning_rate': 0.19375424585439993, 'epoch': 0.14} + 14%|█▍ | 73/520 [04:40<27:00, 3.63s/it] 14%|█▍ | 74/520 [04:43<27:01, 3.64s/it] {'loss': 1.5597, 'grad_norm': 0.006023212860131225, 'learning_rate': 0.1935355868899034, 'epoch': 0.14} + 14%|█▍ | 74/520 [04:43<27:01, 3.64s/it] 14%|█▍ | 75/520 [04:47<26:54, 3.63s/it] {'loss': 1.4366, 'grad_norm': 0.005776183630211956, 'learning_rate': 0.19331329368090666, 'epoch': 0.14} + 14%|█▍ | 75/520 [04:47<26:54, 3.63s/it] 15%|█▍ | 76/520 [04:51<27:08, 3.67s/it] {'loss': 1.8689, 'grad_norm': 0.013966336056258855, 'learning_rate': 0.19308737486442043, 'epoch': 0.15} + 15%|█▍ | 76/520 [04:51<27:08, 3.67s/it] 15%|█▍ | 77/520 [04:54<27:03, 3.67s/it] {'loss': 1.3558, 'grad_norm': 0.0056245003069281185, 'learning_rate': 0.19285783921832536, 'epoch': 0.15} + 15%|█▍ | 77/520 [04:54<27:03, 3.67s/it] 15%|█▌ | 78/520 [04:58<27:17, 3.71s/it] {'loss': 1.5032, 'grad_norm': 0.005923966804810904, 'learning_rate': 0.19262469566103088, 'epoch': 0.15} + 15%|█▌ | 78/520 [04:58<27:17, 3.71s/it] 15%|█▌ | 79/520 [05:02<27:24, 3.73s/it] {'loss': 1.477, 'grad_norm': 0.006054781295309957, 'learning_rate': 0.19238795325112867, 'epoch': 0.15} + 15%|█▌ | 79/520 [05:02<27:24, 3.73s/it] 15%|█▌ | 80/520 [05:06<27:30, 3.75s/it] {'loss': 1.8796, 'grad_norm': 0.006400952269648988, 'learning_rate': 0.19214762118704076, 'epoch': 0.15} + 15%|█▌ | 80/520 [05:06<27:30, 3.75s/it] 16%|█▌ | 81/520 [05:09<27:21, 3.74s/it] {'loss': 1.6567, 'grad_norm': 0.007668397091567247, 'learning_rate': 0.19190370880666208, 'epoch': 0.16} + 16%|█▌ | 81/520 [05:09<27:21, 3.74s/it] 16%|█▌ | 82/520 [05:13<27:09, 3.72s/it] {'loss': 1.5445, 'grad_norm': 0.005753806672016272, 'learning_rate': 0.19165622558699763, 'epoch': 0.16} + 16%|█▌ | 82/520 [05:13<27:09, 3.72s/it] 16%|█▌ | 83/520 [05:17<26:56, 3.70s/it] {'loss': 1.6023, 'grad_norm': 0.005283471738564302, 'learning_rate': 0.19140518114379435, 'epoch': 0.16} + 16%|█▌ | 83/520 [05:17<26:56, 3.70s/it] 16%|█▌ | 84/520 [05:21<27:07, 3.73s/it] {'loss': 1.5864, 'grad_norm': 0.0074311516242987225, 'learning_rate': 0.19115058523116735, 'epoch': 0.16} + 16%|█▌ | 84/520 [05:21<27:07, 3.73s/it] 16%|█▋ | 85/520 [05:24<27:04, 3.73s/it] {'loss': 1.5808, 'grad_norm': 0.0044432394835086705, 'learning_rate': 0.1908924477412211, 'epoch': 0.16} + 16%|█▋ | 85/520 [05:24<27:04, 3.73s/it] 17%|█▋ | 86/520 [05:28<27:06, 3.75s/it] {'loss': 1.6371, 'grad_norm': 0.0053895803568757205, 'learning_rate': 0.19063077870366502, 'epoch': 0.17} + 17%|█▋ | 86/520 [05:28<27:06, 3.75s/it] 17%|█▋ | 87/520 [05:32<26:56, 3.73s/it] {'loss': 1.8192, 'grad_norm': 0.008553901983708163, 'learning_rate': 0.1903655882854237, 'epoch': 0.17} + 17%|█▋ | 87/520 [05:32<26:56, 3.73s/it] 17%|█▋ | 88/520 [05:36<27:04, 3.76s/it] {'loss': 1.884, 'grad_norm': 0.008213627378178876, 'learning_rate': 0.19009688679024192, 'epoch': 0.17} + 17%|█▋ | 88/520 [05:36<27:04, 3.76s/it] 17%|█▋ | 89/520 [05:39<26:50, 3.74s/it] {'loss': 1.5341, 'grad_norm': 0.005261248216625808, 'learning_rate': 0.18982468465828442, 'epoch': 0.17} + 17%|█▋ | 89/520 [05:39<26:50, 3.74s/it] 17%|█▋ | 90/520 [05:43<26:34, 3.71s/it] {'loss': 1.4706, 'grad_norm': 0.0052791562250067234, 'learning_rate': 0.1895489924657301, 'epoch': 0.17} + 17%|█▋ | 90/520 [05:43<26:34, 3.71s/it] 18%|█▊ | 91/520 [05:47<26:20, 3.68s/it] {'loss': 1.5605, 'grad_norm': 0.0054839581642244705, 'learning_rate': 0.18926982092436118, 'epoch': 0.17} + 18%|█▊ | 91/520 [05:47<26:20, 3.68s/it] 18%|█▊ | 92/520 [05:50<26:14, 3.68s/it] {'loss': 1.5042, 'grad_norm': 0.005067861591496655, 'learning_rate': 0.18898718088114688, 'epoch': 0.18} + 18%|█▊ | 92/520 [05:50<26:14, 3.68s/it] 18%|█▊ | 93/520 [05:54<26:06, 3.67s/it] {'loss': 1.4975, 'grad_norm': 0.005121500807686858, 'learning_rate': 0.18870108331782218, 'epoch': 0.18} + 18%|█▊ | 93/520 [05:54<26:06, 3.67s/it] 18%|█▊ | 94/520 [05:58<25:58, 3.66s/it] {'loss': 1.6167, 'grad_norm': 0.005369788597203518, 'learning_rate': 0.18841153935046098, 'epoch': 0.18} + 18%|█▊ | 94/520 [05:58<25:58, 3.66s/it] 18%|█▊ | 95/520 [06:01<25:52, 3.65s/it] {'loss': 1.4853, 'grad_norm': 0.005498498881843606, 'learning_rate': 0.18811856022904425, 'epoch': 0.18} + 18%|█▊ | 95/520 [06:01<25:52, 3.65s/it] 18%|█▊ | 96/520 [06:05<25:44, 3.64s/it] {'loss': 1.4814, 'grad_norm': 0.004428686464712456, 'learning_rate': 0.18782215733702287, 'epoch': 0.18} + 18%|█▊ | 96/520 [06:05<25:44, 3.64s/it] 19%|█▊ | 97/520 [06:08<25:39, 3.64s/it] {'loss': 1.4522, 'grad_norm': 0.005787736275547422, 'learning_rate': 0.18752234219087538, 'epoch': 0.19} + 19%|█▊ | 97/520 [06:08<25:39, 3.64s/it] 19%|█▉ | 98/520 [06:12<25:36, 3.64s/it] {'loss': 1.4439, 'grad_norm': 0.0041070590690999035, 'learning_rate': 0.18721912643966054, 'epoch': 0.19} + 19%|█▉ | 98/520 [06:12<25:36, 3.64s/it] 19%|█▉ | 99/520 [06:16<25:31, 3.64s/it] {'loss': 1.4728, 'grad_norm': 0.0045753748139679825, 'learning_rate': 0.18691252186456464, 'epoch': 0.19} + 19%|█▉ | 99/520 [06:16<25:31, 3.64s/it] 19%|█▉ | 100/520 [06:19<25:24, 3.63s/it] {'loss': 1.6325, 'grad_norm': 0.0065199773682989935, 'learning_rate': 0.1866025403784439, 'epoch': 0.19} + 19%|█▉ | 100/520 [06:19<25:24, 3.63s/it] 19%|█▉ | 101/520 [06:23<25:21, 3.63s/it] {'loss': 1.4577, 'grad_norm': 0.004923549344687631, 'learning_rate': 0.18628919402536132, 'epoch': 0.19} + 19%|█▉ | 101/520 [06:23<25:21, 3.63s/it] 20%|█▉ | 102/520 [06:27<25:16, 3.63s/it] {'loss': 1.4843, 'grad_norm': 0.005793074191836701, 'learning_rate': 0.18597249498011903, 'epoch': 0.2} + 20%|█▉ | 102/520 [06:27<25:16, 3.63s/it] 20%|█▉ | 103/520 [06:30<25:11, 3.63s/it] {'loss': 1.3998, 'grad_norm': 0.004378608547889286, 'learning_rate': 0.18565245554778517, 'epoch': 0.2} + 20%|█▉ | 103/520 [06:30<25:11, 3.63s/it] 20%|██ | 104/520 [06:34<25:06, 3.62s/it] {'loss': 1.4929, 'grad_norm': 0.005538297046476515, 'learning_rate': 0.18532908816321558, 'epoch': 0.2} + 20%|██ | 104/520 [06:34<25:06, 3.62s/it] 20%|██ | 105/520 [06:37<25:07, 3.63s/it] {'loss': 1.4792, 'grad_norm': 0.004535249319741433, 'learning_rate': 0.18500240539057092, 'epoch': 0.2} + 20%|██ | 105/520 [06:37<25:07, 3.63s/it] 20%|██ | 106/520 [06:41<25:05, 3.64s/it] {'loss': 1.6118, 'grad_norm': 0.005967574860724672, 'learning_rate': 0.18467241992282843, 'epoch': 0.2} + 20%|██ | 106/520 [06:41<25:05, 3.64s/it] 21%|██ | 107/520 [06:45<25:07, 3.65s/it] {'loss': 1.5949, 'grad_norm': 0.0049202215650445864, 'learning_rate': 0.18433914458128858, 'epoch': 0.21} + 21%|██ | 107/520 [06:45<25:07, 3.65s/it] 21%|██ | 108/520 [06:48<25:03, 3.65s/it] {'loss': 1.4419, 'grad_norm': 0.005305462419740059, 'learning_rate': 0.18400259231507718, 'epoch': 0.21} + 21%|██ | 108/520 [06:48<25:03, 3.65s/it] 21%|██ | 109/520 [06:52<25:01, 3.65s/it] {'loss': 1.581, 'grad_norm': 0.0043822280191243065, 'learning_rate': 0.18366277620064198, 'epoch': 0.21} + 21%|██ | 109/520 [06:52<25:01, 3.65s/it] 21%|██ | 110/520 [06:56<24:52, 3.64s/it] {'loss': 1.6192, 'grad_norm': 0.005028580163616414, 'learning_rate': 0.1833197094412449, 'epoch': 0.21} + 21%|██ | 110/520 [06:56<24:52, 3.64s/it] 21%|██▏ | 111/520 [06:59<24:53, 3.65s/it] {'loss': 1.6459, 'grad_norm': 0.005694648148367141, 'learning_rate': 0.18297340536644877, 'epoch': 0.21} + 21%|██▏ | 111/520 [06:59<24:53, 3.65s/it] 22%|██▏ | 112/520 [07:03<24:52, 3.66s/it] {'loss': 1.5122, 'grad_norm': 0.004466007046211417, 'learning_rate': 0.1826238774315995, 'epoch': 0.22} + 22%|██▏ | 112/520 [07:03<24:52, 3.66s/it] 22%|██▏ | 113/520 [07:07<24:46, 3.65s/it] {'loss': 1.3683, 'grad_norm': 0.004156762378832208, 'learning_rate': 0.18227113921730334, 'epoch': 0.22} + 22%|██▏ | 113/520 [07:07<24:46, 3.65s/it] 22%|██▏ | 114/520 [07:10<24:39, 3.64s/it] {'loss': 1.5012, 'grad_norm': 0.004849900737146681, 'learning_rate': 0.1819152044288992, 'epoch': 0.22} + 22%|██▏ | 114/520 [07:10<24:39, 3.64s/it] 22%|██▏ | 115/520 [07:14<24:34, 3.64s/it] {'loss': 1.6203, 'grad_norm': 0.004936592001006903, 'learning_rate': 0.18155608689592603, 'epoch': 0.22} + 22%|██▏ | 115/520 [07:14<24:34, 3.64s/it] 22%|██▏ | 116/520 [07:18<24:34, 3.65s/it] {'loss': 1.602, 'grad_norm': 0.004094420407642588, 'learning_rate': 0.18119380057158568, 'epoch': 0.22} + 22%|██▏ | 116/520 [07:18<24:34, 3.65s/it] 22%|██▎ | 117/520 [07:21<24:27, 3.64s/it] {'loss': 1.5885, 'grad_norm': 0.004904501588866516, 'learning_rate': 0.18082835953220056, 'epoch': 0.23} + 22%|██▎ | 117/520 [07:21<24:27, 3.64s/it] 23%|██▎ | 118/520 [07:25<24:27, 3.65s/it] {'loss': 1.4557, 'grad_norm': 0.004132948084131639, 'learning_rate': 0.18045977797666685, 'epoch': 0.23} + 23%|██▎ | 118/520 [07:25<24:27, 3.65s/it] 23%|██▎ | 119/520 [07:29<24:22, 3.65s/it] {'loss': 1.4022, 'grad_norm': 0.0039963170329577555, 'learning_rate': 0.1800880702259028, 'epoch': 0.23} + 23%|██▎ | 119/520 [07:29<24:22, 3.65s/it] 23%|██▎ | 120/520 [07:32<24:23, 3.66s/it] {'loss': 1.4367, 'grad_norm': 0.005244533260538644, 'learning_rate': 0.17971325072229227, 'epoch': 0.23} + 23%|██▎ | 120/520 [07:32<24:23, 3.66s/it] 23%|██▎ | 121/520 [07:36<24:20, 3.66s/it] {'loss': 1.4878, 'grad_norm': 0.00507899748604158, 'learning_rate': 0.17933533402912352, 'epoch': 0.23} + 23%|██▎ | 121/520 [07:36<24:20, 3.66s/it] 23%|██▎ | 122/520 [07:40<24:16, 3.66s/it] {'loss': 1.3714, 'grad_norm': 0.004158250746306823, 'learning_rate': 0.17895433483002354, 'epoch': 0.23} + 23%|██▎ | 122/520 [07:40<24:16, 3.66s/it] 24%|██▎ | 123/520 [07:43<24:16, 3.67s/it] {'loss': 1.6762, 'grad_norm': 0.008153167669639699, 'learning_rate': 0.17857026792838737, 'epoch': 0.24} + 24%|██▎ | 123/520 [07:43<24:16, 3.67s/it] 24%|██▍ | 124/520 [07:47<24:11, 3.67s/it] {'loss': 1.4801, 'grad_norm': 0.005402293951243352, 'learning_rate': 0.178183148246803, 'epoch': 0.24} + 24%|██▍ | 124/520 [07:47<24:11, 3.67s/it] 24%|██▍ | 125/520 [07:51<24:05, 3.66s/it] {'loss': 1.4564, 'grad_norm': 0.0050606973933048065, 'learning_rate': 0.1777929908264715, 'epoch': 0.24} + 24%|██▍ | 125/520 [07:51<24:05, 3.66s/it] 24%|██▍ | 126/520 [07:55<25:27, 3.88s/it] {'loss': 1.5553, 'grad_norm': 0.004685767843411746, 'learning_rate': 0.17739981082662276, 'epoch': 0.24} + 24%|██▍ | 126/520 [07:55<25:27, 3.88s/it] 24%|██▍ | 127/520 [07:59<24:59, 3.82s/it] {'loss': 1.434, 'grad_norm': 0.0056725968495671755, 'learning_rate': 0.1770036235239263, 'epoch': 0.24} + 24%|██▍ | 127/520 [07:59<24:59, 3.82s/it] 25%|██▍ | 128/520 [08:02<24:41, 3.78s/it] {'loss': 1.483, 'grad_norm': 0.005332098542395113, 'learning_rate': 0.1766044443118978, 'epoch': 0.25} + 25%|██▍ | 128/520 [08:02<24:41, 3.78s/it] 25%|██▍ | 129/520 [08:06<24:25, 3.75s/it] {'loss': 1.3978, 'grad_norm': 0.003777046368434542, 'learning_rate': 0.17620228870030108, 'epoch': 0.25} + 25%|██▍ | 129/520 [08:06<24:25, 3.75s/it] 25%|██▌ | 130/520 [08:10<24:09, 3.72s/it] {'loss': 1.4733, 'grad_norm': 0.004306056685211168, 'learning_rate': 0.1757971723145453, 'epoch': 0.25} + 25%|██▌ | 130/520 [08:10<24:09, 3.72s/it] 25%|██▌ | 131/520 [08:13<23:59, 3.70s/it] {'loss': 1.5289, 'grad_norm': 0.005000223760916365, 'learning_rate': 0.175389110895078, 'epoch': 0.25} + 25%|██▌ | 131/520 [08:13<23:59, 3.70s/it] 25%|██▌ | 132/520 [08:17<23:46, 3.68s/it] {'loss': 1.5206, 'grad_norm': 0.004558691624474427, 'learning_rate': 0.17497812029677343, 'epoch': 0.25} + 25%|██▌ | 132/520 [08:17<23:46, 3.68s/it] 26%|██▌ | 133/520 [08:21<23:40, 3.67s/it] {'loss': 1.4122, 'grad_norm': 0.004493721737250494, 'learning_rate': 0.17456421648831655, 'epoch': 0.26} + 26%|██▌ | 133/520 [08:21<23:40, 3.67s/it] 26%|██▌ | 134/520 [08:24<23:38, 3.68s/it] {'loss': 1.5008, 'grad_norm': 0.004737976536007403, 'learning_rate': 0.17414741555158267, 'epoch': 0.26} + 26%|██▌ | 134/520 [08:24<23:38, 3.68s/it] 26%|██▌ | 135/520 [08:28<23:38, 3.69s/it] {'loss': 1.5892, 'grad_norm': 0.004412526733712345, 'learning_rate': 0.1737277336810124, 'epoch': 0.26} + 26%|██▌ | 135/520 [08:28<23:38, 3.69s/it] 26%|██▌ | 136/520 [08:32<23:30, 3.67s/it] {'loss': 1.4916, 'grad_norm': 0.004306994111632565, 'learning_rate': 0.17330518718298263, 'epoch': 0.26} + 26%|██▌ | 136/520 [08:32<23:30, 3.67s/it] 26%|██▋ | 137/520 [08:35<23:23, 3.66s/it] {'loss': 1.4097, 'grad_norm': 0.006186820118728552, 'learning_rate': 0.17287979247517285, 'epoch': 0.26} + 26%|██▋ | 137/520 [08:35<23:23, 3.66s/it] 27%|██▋ | 138/520 [08:39<23:14, 3.65s/it] {'loss': 1.4146, 'grad_norm': 0.004067668893125237, 'learning_rate': 0.17245156608592727, 'epoch': 0.27} + 27%|██▋ | 138/520 [08:39<23:14, 3.65s/it] 27%|██▋ | 139/520 [08:42<23:11, 3.65s/it] {'loss': 1.4235, 'grad_norm': 0.0047367695296695, 'learning_rate': 0.17202052465361267, 'epoch': 0.27} + 27%|██▋ | 139/520 [08:42<23:11, 3.65s/it] 27%|██▋ | 140/520 [08:46<23:11, 3.66s/it] {'loss': 1.5766, 'grad_norm': 0.004862192369812503, 'learning_rate': 0.17158668492597184, 'epoch': 0.27} + 27%|██▋ | 140/520 [08:46<23:11, 3.66s/it] 27%|██▋ | 141/520 [08:50<23:07, 3.66s/it] {'loss': 1.5516, 'grad_norm': 0.005176755523009161, 'learning_rate': 0.17115006375947303, 'epoch': 0.27} + 27%|██▋ | 141/520 [08:50<23:07, 3.66s/it] 27%|██▋ | 142/520 [08:54<23:11, 3.68s/it] {'loss': 1.6079, 'grad_norm': 0.004671128383179856, 'learning_rate': 0.17071067811865476, 'epoch': 0.27} + 27%|██▋ | 142/520 [08:54<23:11, 3.68s/it] 28%|██▊ | 143/520 [08:57<23:06, 3.68s/it] {'loss': 1.4617, 'grad_norm': 0.005268114511793845, 'learning_rate': 0.17026854507546693, 'epoch': 0.28} + 28%|██▊ | 143/520 [08:57<23:06, 3.68s/it] 28%|██▊ | 144/520 [09:01<22:59, 3.67s/it] {'loss': 1.4099, 'grad_norm': 0.004973391413356761, 'learning_rate': 0.1698236818086073, 'epoch': 0.28} + 28%|██▊ | 144/520 [09:01<22:59, 3.67s/it] 28%|██▊ | 145/520 [09:05<23:01, 3.68s/it] {'loss': 1.3286, 'grad_norm': 0.0040522182674784055, 'learning_rate': 0.16937610560285418, 'epoch': 0.28} + 28%|██▊ | 145/520 [09:05<23:01, 3.68s/it] 28%|██▊ | 146/520 [09:08<22:58, 3.69s/it] {'loss': 1.6513, 'grad_norm': 0.005119214694230653, 'learning_rate': 0.1689258338483947, 'epoch': 0.28} + 28%|██▊ | 146/520 [09:08<22:58, 3.69s/it] 28%|██▊ | 147/520 [09:12<22:51, 3.68s/it] {'loss': 1.385, 'grad_norm': 0.004803083739246974, 'learning_rate': 0.16847288404014937, 'epoch': 0.28} + 28%|██▊ | 147/520 [09:12<22:51, 3.68s/it] 28%|██▊ | 148/520 [09:16<22:47, 3.68s/it] {'loss': 1.4214, 'grad_norm': 0.004222064707662238, 'learning_rate': 0.16801727377709194, 'epoch': 0.28} + 28%|██▊ | 148/520 [09:16<22:47, 3.68s/it] 29%|██▊ | 149/520 [09:19<22:45, 3.68s/it] {'loss': 1.3683, 'grad_norm': 0.004561873985035059, 'learning_rate': 0.16755902076156604, 'epoch': 0.29} + 29%|██▊ | 149/520 [09:19<22:45, 3.68s/it] 29%|██▉ | 150/520 [09:23<22:39, 3.67s/it] {'loss': 1.6057, 'grad_norm': 0.005183726880212444, 'learning_rate': 0.16709814279859703, 'epoch': 0.29} + 29%|██▉ | 150/520 [09:23<22:39, 3.67s/it] 29%|██▉ | 151/520 [09:27<22:33, 3.67s/it] {'loss': 1.394, 'grad_norm': 0.004285752779454979, 'learning_rate': 0.1666346577952004, 'epoch': 0.29} + 29%|██▉ | 151/520 [09:27<22:33, 3.67s/it] 29%|██▉ | 152/520 [09:30<22:25, 3.66s/it] {'loss': 1.3746, 'grad_norm': 0.004843345180514107, 'learning_rate': 0.16616858375968596, 'epoch': 0.29} + 29%|██▉ | 152/520 [09:30<22:25, 3.66s/it] 29%|██▉ | 153/520 [09:34<22:22, 3.66s/it] {'loss': 1.4179, 'grad_norm': 0.005340130488953743, 'learning_rate': 0.16569993880095807, 'epoch': 0.29} + 29%|██▉ | 153/520 [09:34<22:22, 3.66s/it] 30%|██▉ | 154/520 [09:38<22:19, 3.66s/it] {'loss': 1.4947, 'grad_norm': 0.0040089544617674016, 'learning_rate': 0.16522874112781213, 'epoch': 0.3} + 30%|██▉ | 154/520 [09:38<22:19, 3.66s/it] 30%|██▉ | 155/520 [09:41<22:17, 3.67s/it] {'loss': 1.4067, 'grad_norm': 0.004695084163462336, 'learning_rate': 0.16475500904822704, 'epoch': 0.3} + 30%|██▉ | 155/520 [09:41<22:17, 3.67s/it] 30%|███ | 156/520 [09:45<22:14, 3.67s/it] {'loss': 1.4305, 'grad_norm': 0.00482715777777321, 'learning_rate': 0.16427876096865393, 'epoch': 0.3} + 30%|███ | 156/520 [09:45<22:14, 3.67s/it] 30%|███ | 157/520 [09:49<22:12, 3.67s/it] {'loss': 1.6705, 'grad_norm': 0.004769305026220531, 'learning_rate': 0.16380001539330089, 'epoch': 0.3} + 30%|███ | 157/520 [09:49<22:12, 3.67s/it] 30%|███ | 158/520 [09:52<22:04, 3.66s/it] {'loss': 1.4086, 'grad_norm': 0.005626029315757551, 'learning_rate': 0.163318790923414, 'epoch': 0.3} + 30%|███ | 158/520 [09:52<22:04, 3.66s/it] 31%|███ | 159/520 [09:56<21:57, 3.65s/it] {'loss': 1.4303, 'grad_norm': 0.004061106121258034, 'learning_rate': 0.16283510625655473, 'epoch': 0.31} + 31%|███ | 159/520 [09:56<21:57, 3.65s/it] 31%|███ | 160/520 [09:59<21:51, 3.64s/it] {'loss': 1.4693, 'grad_norm': 0.004805453146891535, 'learning_rate': 0.16234898018587338, 'epoch': 0.31} + 31%|███ | 160/520 [09:59<21:51, 3.64s/it] 31%|███ | 161/520 [10:03<21:48, 3.65s/it] {'loss': 1.4551, 'grad_norm': 0.004505078758054849, 'learning_rate': 0.16186043159937882, 'epoch': 0.31} + 31%|███ | 161/520 [10:03<21:48, 3.65s/it] 31%|███ | 162/520 [10:07<21:41, 3.64s/it] {'loss': 1.5793, 'grad_norm': 0.004734996259537476, 'learning_rate': 0.16136947947920477, 'epoch': 0.31} + 31%|███ | 162/520 [10:07<21:41, 3.64s/it] 31%|███▏ | 163/520 [10:10<21:35, 3.63s/it] {'loss': 1.3156, 'grad_norm': 0.005489137141788543, 'learning_rate': 0.16087614290087207, 'epoch': 0.31} + 31%|███▏ | 163/520 [10:10<21:35, 3.63s/it] 32%|███▏ | 164/520 [10:14<21:32, 3.63s/it] {'loss': 1.2822, 'grad_norm': 0.004336871293235577, 'learning_rate': 0.16038044103254775, 'epoch': 0.32} + 32%|███▏ | 164/520 [10:14<21:32, 3.63s/it] 32%|███▏ | 165/520 [10:18<21:27, 3.63s/it] {'loss': 1.4369, 'grad_norm': 0.004157187815265287, 'learning_rate': 0.15988239313430005, 'epoch': 0.32} + 32%|███▏ | 165/520 [10:18<21:27, 3.63s/it] 32%|███▏ | 166/520 [10:21<21:18, 3.61s/it] {'loss': 1.4358, 'grad_norm': 0.005209907695374024, 'learning_rate': 0.15938201855735015, 'epoch': 0.32} + 32%|███▏ | 166/520 [10:21<21:18, 3.61s/it] 32%|███▏ | 167/520 [10:25<21:20, 3.63s/it] {'loss': 1.4248, 'grad_norm': 0.004890544423713236, 'learning_rate': 0.15887933674332047, 'epoch': 0.32} + 32%|███▏ | 167/520 [10:25<21:20, 3.63s/it] 32%|███▏ | 168/520 [10:29<21:21, 3.64s/it] {'loss': 1.3565, 'grad_norm': 0.004309897856354609, 'learning_rate': 0.158374367223479, 'epoch': 0.32} + 32%|███▏ | 168/520 [10:29<21:21, 3.64s/it] 32%|███▎ | 169/520 [10:32<21:19, 3.65s/it] {'loss': 1.4375, 'grad_norm': 0.004541434721625349, 'learning_rate': 0.1578671296179806, 'epoch': 0.33} + 32%|███▎ | 169/520 [10:32<21:19, 3.65s/it] 33%|███▎ | 170/520 [10:36<21:18, 3.65s/it] {'loss': 1.4993, 'grad_norm': 0.005159711160287193, 'learning_rate': 0.15735764363510463, 'epoch': 0.33} + 33%|███▎ | 170/520 [10:36<21:18, 3.65s/it] 33%|███▎ | 171/520 [10:40<21:18, 3.66s/it] {'loss': 1.3664, 'grad_norm': 0.0042700988742466325, 'learning_rate': 0.15684592907048925, 'epoch': 0.33} + 33%|███▎ | 171/520 [10:40<21:18, 3.66s/it] 33%|███▎ | 172/520 [10:43<21:16, 3.67s/it] {'loss': 1.433, 'grad_norm': 0.004563431013380235, 'learning_rate': 0.1563320058063622, 'epoch': 0.33} + 33%|███▎ | 172/520 [10:43<21:16, 3.67s/it] 33%|███▎ | 173/520 [10:47<21:13, 3.67s/it] {'loss': 1.3616, 'grad_norm': 0.004491155746714079, 'learning_rate': 0.15581589381076844, 'epoch': 0.33} + 33%|███▎ | 173/520 [10:47<21:13, 3.67s/it] 33%|███▎ | 174/520 [10:51<21:03, 3.65s/it] {'loss': 1.4339, 'grad_norm': 0.004607662421022972, 'learning_rate': 0.15529761313679394, 'epoch': 0.33} + 33%|███▎ | 174/520 [10:51<21:03, 3.65s/it] 34%|███▎ | 175/520 [10:54<20:57, 3.64s/it] {'loss': 1.3382, 'grad_norm': 0.0044115482615590665, 'learning_rate': 0.15477718392178716, 'epoch': 0.34} + 34%|███▎ | 175/520 [10:54<20:57, 3.64s/it] 34%|███▍ | 176/520 [10:58<20:55, 3.65s/it] {'loss': 1.5881, 'grad_norm': 0.005352689619586987, 'learning_rate': 0.15425462638657594, 'epoch': 0.34} + 34%|███▍ | 176/520 [10:58<20:55, 3.65s/it] 34%|███▍ | 177/520 [11:01<20:52, 3.65s/it] {'loss': 1.4448, 'grad_norm': 0.004849210288228885, 'learning_rate': 0.1537299608346824, 'epoch': 0.34} + 34%|███▍ | 177/520 [11:01<20:52, 3.65s/it] 34%|███▍ | 178/520 [11:05<20:49, 3.65s/it] {'loss': 1.4187, 'grad_norm': 0.004435483657842682, 'learning_rate': 0.15320320765153367, 'epoch': 0.34} + 34%|███▍ | 178/520 [11:05<20:49, 3.65s/it] 34%|███▍ | 179/520 [11:09<20:46, 3.66s/it] {'loss': 1.5051, 'grad_norm': 0.004376738881180332, 'learning_rate': 0.15267438730367008, 'epoch': 0.34} + 34%|███▍ | 179/520 [11:09<20:46, 3.66s/it] 35%|███▍ | 180/520 [11:12<20:41, 3.65s/it] {'loss': 1.4103, 'grad_norm': 0.00479131740055639, 'learning_rate': 0.1521435203379498, 'epoch': 0.35} + 35%|███▍ | 180/520 [11:12<20:41, 3.65s/it] 35%|███▍ | 181/520 [11:16<20:34, 3.64s/it] {'loss': 1.3748, 'grad_norm': 0.003479249550128572, 'learning_rate': 0.15161062738075068, 'epoch': 0.35} + 35%|███▍ | 181/520 [11:16<20:34, 3.64s/it] 35%|███▌ | 182/520 [11:20<20:30, 3.64s/it] {'loss': 1.3943, 'grad_norm': 0.0038914226072612485, 'learning_rate': 0.1510757291371686, 'epoch': 0.35} + 35%|███▌ | 182/520 [11:20<20:30, 3.64s/it] 35%|███▌ | 183/520 [11:23<20:25, 3.64s/it] {'loss': 1.4202, 'grad_norm': 0.004620779172865361, 'learning_rate': 0.1505388463902131, 'epoch': 0.35} + 35%|███▌ | 183/520 [11:23<20:25, 3.64s/it] 35%|███▌ | 184/520 [11:27<20:21, 3.64s/it] {'loss': 1.3312, 'grad_norm': 0.004561953705907543, 'learning_rate': 0.15000000000000002, 'epoch': 0.35} + 35%|███▌ | 184/520 [11:27<20:21, 3.64s/it] 36%|███▌ | 185/520 [11:31<20:23, 3.65s/it] {'loss': 1.5229, 'grad_norm': 0.004450928316188847, 'learning_rate': 0.14945921090294076, 'epoch': 0.36} + 36%|███▌ | 185/520 [11:31<20:23, 3.65s/it] 36%|███▌ | 186/520 [11:34<20:21, 3.66s/it] {'loss': 1.3406, 'grad_norm': 0.0042319607359221955, 'learning_rate': 0.14891650011092894, 'epoch': 0.36} + 36%|███▌ | 186/520 [11:34<20:21, 3.66s/it] 36%|███▌ | 187/520 [11:38<20:17, 3.66s/it] {'loss': 1.3718, 'grad_norm': 0.00519326365332391, 'learning_rate': 0.14837188871052398, 'epoch': 0.36} + 36%|███▌ | 187/520 [11:38<20:17, 3.66s/it] 36%|███▌ | 188/520 [11:42<20:10, 3.65s/it] {'loss': 1.4378, 'grad_norm': 0.004143667890043469, 'learning_rate': 0.14782539786213184, 'epoch': 0.36} + 36%|███▌ | 188/520 [11:42<20:10, 3.65s/it] 36%|███▋ | 189/520 [11:45<20:11, 3.66s/it] {'loss': 1.4622, 'grad_norm': 0.004064198773718876, 'learning_rate': 0.1472770487991827, 'epoch': 0.36} + 36%|███▋ | 189/520 [11:45<20:11, 3.66s/it] 37%|███▋ | 190/520 [11:49<20:08, 3.66s/it] {'loss': 1.3657, 'grad_norm': 0.004537614959703854, 'learning_rate': 0.1467268628273062, 'epoch': 0.37} + 37%|███▋ | 190/520 [11:49<20:08, 3.66s/it] 37%|███▋ | 191/520 [11:53<20:07, 3.67s/it] {'loss': 1.3217, 'grad_norm': 0.004141666911607498, 'learning_rate': 0.1461748613235034, 'epoch': 0.37} + 37%|███▋ | 191/520 [11:53<20:07, 3.67s/it] 37%|███▋ | 192/520 [11:56<20:03, 3.67s/it] {'loss': 1.4188, 'grad_norm': 0.003943070689096233, 'learning_rate': 0.1456210657353163, 'epoch': 0.37} + 37%|███▋ | 192/520 [11:56<20:03, 3.67s/it] 37%|███▋ | 193/520 [12:00<20:04, 3.68s/it] {'loss': 1.5118, 'grad_norm': 0.004521672384670195, 'learning_rate': 0.14506549757999454, 'epoch': 0.37} + 37%|███▋ | 193/520 [12:00<20:04, 3.68s/it] 37%|███▋ | 194/520 [12:04<20:12, 3.72s/it] {'loss': 1.3707, 'grad_norm': 0.005624311219251103, 'learning_rate': 0.14450817844365924, 'epoch': 0.37} + 37%|███▋ | 194/520 [12:04<20:12, 3.72s/it] 38%|███▊ | 195/520 [12:07<20:02, 3.70s/it] {'loss': 1.4394, 'grad_norm': 0.004120405682411664, 'learning_rate': 0.1439491299804645, 'epoch': 0.38} + 38%|███▊ | 195/520 [12:07<20:02, 3.70s/it] 38%|███▊ | 196/520 [12:11<19:50, 3.67s/it] {'loss': 1.4077, 'grad_norm': 0.0047565830265865, 'learning_rate': 0.14338837391175582, 'epoch': 0.38} + 38%|███▊ | 196/520 [12:11<19:50, 3.67s/it] 38%|███▊ | 197/520 [12:15<19:49, 3.68s/it] {'loss': 1.3571, 'grad_norm': 0.0039974877636474865, 'learning_rate': 0.14282593202522628, 'epoch': 0.38} + 38%|███▊ | 197/520 [12:15<19:49, 3.68s/it] 38%|███▊ | 198/520 [12:18<19:41, 3.67s/it] {'loss': 1.4347, 'grad_norm': 0.004303494338678767, 'learning_rate': 0.14226182617406996, 'epoch': 0.38} + 38%|███▊ | 198/520 [12:18<19:41, 3.67s/it] 38%|███▊ | 199/520 [12:22<19:49, 3.71s/it] {'loss': 1.3432, 'grad_norm': 0.003809974195048834, 'learning_rate': 0.14169607827613281, 'epoch': 0.38} + 38%|███▊ | 199/520 [12:22<19:49, 3.71s/it] 38%|███▊ | 200/520 [12:26<19:56, 3.74s/it] {'loss': 1.4423, 'grad_norm': 0.004488954039476257, 'learning_rate': 0.14112871031306118, 'epoch': 0.38} + 38%|███▊ | 200/520 [12:26<19:56, 3.74s/it] 39%|███▊ | 201/520 [12:30<19:48, 3.72s/it] {'loss': 1.4512, 'grad_norm': 0.004153343725943756, 'learning_rate': 0.1405597443294475, 'epoch': 0.39} + 39%|███▊ | 201/520 [12:30<19:48, 3.72s/it] 39%|███▉ | 202/520 [12:34<19:52, 3.75s/it] {'loss': 1.345, 'grad_norm': 0.004533000505173643, 'learning_rate': 0.13998920243197407, 'epoch': 0.39} + 39%|███▉ | 202/520 [12:34<19:52, 3.75s/it] 39%|███▉ | 203/520 [12:37<19:48, 3.75s/it] {'loss': 1.3986, 'grad_norm': 0.003998664485712691, 'learning_rate': 0.13941710678855396, 'epoch': 0.39} + 39%|███▉ | 203/520 [12:37<19:48, 3.75s/it] 39%|███▉ | 204/520 [12:41<19:45, 3.75s/it] {'loss': 1.4363, 'grad_norm': 0.004241027783834206, 'learning_rate': 0.13884347962746948, 'epoch': 0.39} + 39%|███▉ | 204/520 [12:41<19:45, 3.75s/it] 39%|███▉ | 205/520 [12:45<19:45, 3.76s/it] {'loss': 1.4674, 'grad_norm': 0.0049217034765356125, 'learning_rate': 0.138268343236509, 'epoch': 0.39} + 39%|███▉ | 205/520 [12:45<19:45, 3.76s/it] 40%|███▉ | 206/520 [12:49<19:43, 3.77s/it] {'loss': 1.4792, 'grad_norm': 0.004331692040774826, 'learning_rate': 0.13769171996210053, 'epoch': 0.4} + 40%|███▉ | 206/520 [12:49<19:43, 3.77s/it] 40%|███▉ | 207/520 [12:52<19:42, 3.78s/it] {'loss': 1.4383, 'grad_norm': 0.004407372961024495, 'learning_rate': 0.1371136322084438, 'epoch': 0.4} + 40%|███▉ | 207/520 [12:52<19:42, 3.78s/it] 40%|████ | 208/520 [12:56<19:37, 3.77s/it] {'loss': 1.4354, 'grad_norm': 0.004530789104915021, 'learning_rate': 0.13653410243663952, 'epoch': 0.4} + 40%|████ | 208/520 [12:56<19:37, 3.77s/it] 40%|████ | 209/520 [13:00<19:40, 3.80s/it] {'loss': 1.3548, 'grad_norm': 0.004239157503979015, 'learning_rate': 0.13595315316381676, 'epoch': 0.4} + 40%|████ | 209/520 [13:00<19:40, 3.80s/it] 40%|████ | 210/520 [13:04<19:35, 3.79s/it] {'loss': 1.4369, 'grad_norm': 0.004931136520429039, 'learning_rate': 0.13537080696225814, 'epoch': 0.4} + 40%|████ | 210/520 [13:04<19:35, 3.79s/it] 41%|████ | 211/520 [13:08<19:32, 3.80s/it] {'loss': 1.4418, 'grad_norm': 0.0036476771795101236, 'learning_rate': 0.13478708645852272, 'epoch': 0.41} + 41%|████ | 211/520 [13:08<19:32, 3.80s/it] 41%|████ | 212/520 [13:11<19:29, 3.80s/it] {'loss': 1.4026, 'grad_norm': 0.004015526161587774, 'learning_rate': 0.1342020143325669, 'epoch': 0.41} + 41%|████ | 212/520 [13:11<19:29, 3.80s/it] 41%|████ | 213/520 [13:15<19:27, 3.80s/it] {'loss': 1.374, 'grad_norm': 0.005001077360887777, 'learning_rate': 0.13361561331686309, 'epoch': 0.41} + 41%|████ | 213/520 [13:15<19:27, 3.80s/it] 41%|████ | 214/520 [13:19<19:22, 3.80s/it] {'loss': 1.3669, 'grad_norm': 0.004362461548444781, 'learning_rate': 0.13302790619551672, 'epoch': 0.41} + 41%|████ | 214/520 [13:19<19:22, 3.80s/it] 41%|████▏ | 215/520 [13:23<19:19, 3.80s/it] {'loss': 1.3668, 'grad_norm': 0.004109053524732619, 'learning_rate': 0.1324389158033807, 'epoch': 0.41} + 41%|████▏ | 215/520 [13:23<19:19, 3.80s/it] 42%|████▏ | 216/520 [13:27<19:16, 3.80s/it] {'loss': 1.2699, 'grad_norm': 0.004209458182207956, 'learning_rate': 0.13184866502516845, 'epoch': 0.42} + 42%|████▏ | 216/520 [13:27<19:16, 3.80s/it] 42%|████▏ | 217/520 [13:30<19:11, 3.80s/it] {'loss': 1.4092, 'grad_norm': 0.004179712587055148, 'learning_rate': 0.13125717679456447, 'epoch': 0.42} + 42%|████▏ | 217/520 [13:30<19:11, 3.80s/it] 42%|████▏ | 218/520 [13:34<19:09, 3.81s/it] {'loss': 1.4003, 'grad_norm': 0.0042552111018534945, 'learning_rate': 0.13066447409333345, 'epoch': 0.42} + 42%|████▏ | 218/520 [13:34<19:09, 3.81s/it] 42%|████▏ | 219/520 [13:38<19:03, 3.80s/it] {'loss': 1.3592, 'grad_norm': 0.0035980253215136867, 'learning_rate': 0.1300705799504273, 'epoch': 0.42} + 42%|████▏ | 219/520 [13:38<19:03, 3.80s/it] 42%|████▏ | 220/520 [13:42<18:57, 3.79s/it] {'loss': 1.4183, 'grad_norm': 0.004484686105685769, 'learning_rate': 0.12947551744109043, 'epoch': 0.42} + 42%|████▏ | 220/520 [13:42<18:57, 3.79s/it] 42%|████▎ | 221/520 [13:46<18:56, 3.80s/it] {'loss': 1.4113, 'grad_norm': 0.0040853557718768165, 'learning_rate': 0.128879309685963, 'epoch': 0.42} + 42%|████▎ | 221/520 [13:46<18:56, 3.80s/it] 43%|████▎ | 222/520 [13:49<18:51, 3.80s/it] {'loss': 1.3141, 'grad_norm': 0.004162898778318023, 'learning_rate': 0.12828197985018275, 'epoch': 0.43} + 43%|████▎ | 222/520 [13:49<18:51, 3.80s/it] 43%|████▎ | 223/520 [13:53<18:51, 3.81s/it] {'loss': 1.3002, 'grad_norm': 0.003608423720521303, 'learning_rate': 0.12768355114248495, 'epoch': 0.43} + 43%|████▎ | 223/520 [13:53<18:51, 3.81s/it] 43%|████▎ | 224/520 [13:57<18:46, 3.81s/it] {'loss': 1.6388, 'grad_norm': 0.007959822916173174, 'learning_rate': 0.12708404681430052, 'epoch': 0.43} + 43%|████▎ | 224/520 [13:57<18:46, 3.81s/it] 43%|████▎ | 225/520 [14:01<18:42, 3.80s/it] {'loss': 1.3258, 'grad_norm': 0.004326249487615409, 'learning_rate': 0.1264834901588527, 'epoch': 0.43} + 43%|████▎ | 225/520 [14:01<18:42, 3.80s/it] 43%|████▎ | 226/520 [14:05<18:37, 3.80s/it] {'loss': 1.4307, 'grad_norm': 0.0038160344220189747, 'learning_rate': 0.12588190451025208, 'epoch': 0.43} + 43%|████▎ | 226/520 [14:05<18:37, 3.80s/it] 44%|████▎ | 227/520 [14:08<18:33, 3.80s/it] {'loss': 1.4137, 'grad_norm': 0.0037412052961353637, 'learning_rate': 0.12527931324258976, 'epoch': 0.44} + 44%|████▎ | 227/520 [14:08<18:33, 3.80s/it] 44%|████▍ | 228/520 [14:12<18:31, 3.81s/it] {'loss': 1.5758, 'grad_norm': 0.004190709197473438, 'learning_rate': 0.12467573976902935, 'epoch': 0.44} + 44%|████▍ | 228/520 [14:12<18:31, 3.81s/it] 44%|████▍ | 229/520 [14:16<18:27, 3.81s/it] {'loss': 1.3904, 'grad_norm': 0.003568266037732985, 'learning_rate': 0.12407120754089732, 'epoch': 0.44} + 44%|████▍ | 229/520 [14:16<18:27, 3.81s/it] 44%|████▍ | 230/520 [14:20<18:23, 3.81s/it] {'loss': 1.2584, 'grad_norm': 0.004053927311741186, 'learning_rate': 0.12346574004677154, 'epoch': 0.44} + 44%|████▍ | 230/520 [14:20<18:23, 3.81s/it] 44%|████▍ | 231/520 [14:24<18:10, 3.77s/it] {'loss': 1.3237, 'grad_norm': 0.003482309666307705, 'learning_rate': 0.12285936081156897, 'epoch': 0.44} + 44%|████▍ | 231/520 [14:24<18:10, 3.77s/it] 45%|████▍ | 232/520 [14:27<18:00, 3.75s/it] {'loss': 1.5962, 'grad_norm': 0.004404273132994969, 'learning_rate': 0.12225209339563144, 'epoch': 0.45} + 45%|████▍ | 232/520 [14:27<18:00, 3.75s/it] 45%|████▍ | 233/520 [14:31<17:54, 3.74s/it] {'loss': 1.4713, 'grad_norm': 0.004208822043019374, 'learning_rate': 0.12164396139381028, 'epoch': 0.45} + 45%|████▍ | 233/520 [14:31<17:54, 3.74s/it] 45%|████▌ | 234/520 [14:35<17:48, 3.74s/it] {'loss': 1.273, 'grad_norm': 0.003946973460176254, 'learning_rate': 0.12103498843454959, 'epoch': 0.45} + 45%|████▌ | 234/520 [14:35<17:48, 3.74s/it] 45%|████▌ | 235/520 [14:38<17:45, 3.74s/it] {'loss': 1.3258, 'grad_norm': 0.004394460845539883, 'learning_rate': 0.12042519817896805, 'epoch': 0.45} + 45%|████▌ | 235/520 [14:38<17:45, 3.74s/it] 45%|████▌ | 236/520 [14:42<17:35, 3.72s/it] {'loss': 1.4565, 'grad_norm': 0.0037120965926328315, 'learning_rate': 0.11981461431993977, 'epoch': 0.45} + 45%|████▌ | 236/520 [14:42<17:35, 3.72s/it] 46%|████▌ | 237/520 [14:46<17:29, 3.71s/it] {'loss': 1.4062, 'grad_norm': 0.0038528796881284436, 'learning_rate': 0.11920326058117364, 'epoch': 0.46} + 46%|████▌ | 237/520 [14:46<17:29, 3.71s/it] 46%|████▌ | 238/520 [14:49<17:24, 3.70s/it] {'loss': 1.3414, 'grad_norm': 0.0039329773014513585, 'learning_rate': 0.11859116071629149, 'epoch': 0.46} + 46%|████▌ | 238/520 [14:49<17:24, 3.70s/it] 46%|████▌ | 239/520 [14:53<17:18, 3.69s/it] {'loss': 1.4574, 'grad_norm': 0.004221437526005088, 'learning_rate': 0.11797833850790528, 'epoch': 0.46} + 46%|████▌ | 239/520 [14:53<17:18, 3.69s/it] 46%|████▌ | 240/520 [14:57<17:10, 3.68s/it] {'loss': 1.2078, 'grad_norm': 0.004286700908046019, 'learning_rate': 0.11736481776669305, 'epoch': 0.46} + 46%|████▌ | 240/520 [14:57<17:10, 3.68s/it] 46%|████▋ | 241/520 [15:00<17:03, 3.67s/it] {'loss': 1.3004, 'grad_norm': 0.0039223321010738756, 'learning_rate': 0.11675062233047365, 'epoch': 0.46} + 46%|████▋ | 241/520 [15:00<17:03, 3.67s/it] 47%|████▋ | 242/520 [15:04<17:00, 3.67s/it] {'loss': 1.3203, 'grad_norm': 0.0034744200483160714, 'learning_rate': 0.11613577606328068, 'epoch': 0.47} + 47%|████▋ | 242/520 [15:04<17:00, 3.67s/it] 47%|████▋ | 243/520 [15:08<16:54, 3.66s/it] {'loss': 1.3088, 'grad_norm': 0.003925752731233358, 'learning_rate': 0.11552030285443515, 'epoch': 0.47} + 47%|████▋ | 243/520 [15:08<16:54, 3.66s/it] 47%|████▋ | 244/520 [15:11<16:49, 3.66s/it] {'loss': 1.4518, 'grad_norm': 0.004034291623752221, 'learning_rate': 0.11490422661761744, 'epoch': 0.47} + 47%|████▋ | 244/520 [15:11<16:49, 3.66s/it] 47%|████▋ | 245/520 [15:15<16:44, 3.65s/it] {'loss': 1.2969, 'grad_norm': 0.0037679866441995497, 'learning_rate': 0.11428757128993801, 'epoch': 0.47} + 47%|████▋ | 245/520 [15:15<16:44, 3.65s/it] 47%|████▋ | 246/520 [15:19<16:44, 3.67s/it] {'loss': 1.5922, 'grad_norm': 0.004512903228054606, 'learning_rate': 0.11367036083100734, 'epoch': 0.47} + 47%|████▋ | 246/520 [15:19<16:44, 3.67s/it] 48%|████▊ | 247/520 [15:22<16:41, 3.67s/it] {'loss': 1.4799, 'grad_norm': 0.003909210480884881, 'learning_rate': 0.11305261922200519, 'epoch': 0.47} + 48%|████▊ | 247/520 [15:22<16:41, 3.67s/it] 48%|████▊ | 248/520 [15:26<16:38, 3.67s/it] {'loss': 1.2946, 'grad_norm': 0.003837889627861987, 'learning_rate': 0.11243437046474854, 'epoch': 0.48} + 48%|████▊ | 248/520 [15:26<16:38, 3.67s/it] 48%|████▊ | 249/520 [15:30<16:36, 3.68s/it] {'loss': 1.4096, 'grad_norm': 0.004158882784936612, 'learning_rate': 0.1118156385807593, 'epoch': 0.48} + 48%|████▊ | 249/520 [15:30<16:36, 3.68s/it] 48%|████▊ | 250/520 [15:33<16:30, 3.67s/it] {'loss': 1.3473, 'grad_norm': 0.004287004885149416, 'learning_rate': 0.11119644761033079, 'epoch': 0.48} + 48%|████▊ | 250/520 [15:33<16:30, 3.67s/it] 48%|████▊ | 251/520 [15:37<16:24, 3.66s/it] {'loss': 1.4102, 'grad_norm': 0.003510342312400575, 'learning_rate': 0.1105768216115938, 'epoch': 0.48} + 48%|████▊ | 251/520 [15:37<16:24, 3.66s/it] 48%|████▊ | 252/520 [15:41<16:26, 3.68s/it] {'loss': 1.4416, 'grad_norm': 0.0037428093345233063, 'learning_rate': 0.10995678465958168, 'epoch': 0.48} + 48%|████▊ | 252/520 [15:41<16:26, 3.68s/it] 49%|████▊ | 253/520 [15:45<16:36, 3.73s/it] {'loss': 1.4128, 'grad_norm': 0.004436208817651928, 'learning_rate': 0.10933636084529506, 'epoch': 0.49} + 49%|████▊ | 253/520 [15:45<16:36, 3.73s/it] 49%|████▉ | 254/520 [15:48<16:40, 3.76s/it] {'loss': 1.3154, 'grad_norm': 0.003571309571514644, 'learning_rate': 0.10871557427476584, 'epoch': 0.49} + 49%|████▉ | 254/520 [15:48<16:40, 3.76s/it] 49%|████▉ | 255/520 [15:52<16:42, 3.78s/it] {'loss': 1.3359, 'grad_norm': 0.004043246920304652, 'learning_rate': 0.10809444906812034, 'epoch': 0.49} + 49%|████▉ | 255/520 [15:52<16:42, 3.78s/it] 49%|████▉ | 256/520 [15:56<16:45, 3.81s/it] {'loss': 1.3706, 'grad_norm': 0.004111853254608021, 'learning_rate': 0.10747300935864244, 'epoch': 0.49} + 49%|████▉ | 256/520 [15:56<16:45, 3.81s/it] 49%|████▉ | 257/520 [16:00<16:42, 3.81s/it] {'loss': 1.3611, 'grad_norm': 0.003835498772979485, 'learning_rate': 0.10685127929183567, 'epoch': 0.49} + 49%|████▉ | 257/520 [16:00<16:42, 3.81s/it] 50%|████▉ | 258/520 [16:04<16:40, 3.82s/it] {'loss': 1.3899, 'grad_norm': 0.003403037982306786, 'learning_rate': 0.10622928302448523, 'epoch': 0.5} + 50%|████▉ | 258/520 [16:04<16:40, 3.82s/it] 50%|████▉ | 259/520 [16:08<16:35, 3.81s/it] {'loss': 1.4484, 'grad_norm': 0.004375287167501641, 'learning_rate': 0.10560704472371918, 'epoch': 0.5} + 50%|████▉ | 259/520 [16:08<16:35, 3.81s/it] 50%|█████ | 260/520 [16:11<16:31, 3.81s/it] {'loss': 1.5127, 'grad_norm': 0.003847264524412083, 'learning_rate': 0.10498458856606972, 'epoch': 0.5} + 50%|█████ | 260/520 [16:11<16:31, 3.81s/it] 50%|█████ | 261/520 [16:15<16:17, 3.77s/it] {'loss': 1.4509, 'grad_norm': 0.004164159828797204, 'learning_rate': 0.10436193873653361, 'epoch': 0.5} + 50%|█████ | 261/520 [16:15<16:17, 3.77s/it] 50%|█████ | 262/520 [16:19<16:04, 3.74s/it] {'loss': 1.2891, 'grad_norm': 0.004343396580187692, 'learning_rate': 0.10373911942763259, 'epoch': 0.5} + 50%|█████ | 262/520 [16:19<16:04, 3.74s/it] 51%|█████ | 263/520 [16:22<15:54, 3.71s/it] {'loss': 1.4514, 'grad_norm': 0.004184507308074042, 'learning_rate': 0.10311615483847332, 'epoch': 0.51} + 51%|█████ | 263/520 [16:22<15:54, 3.71s/it] 51%|█████ | 264/520 [16:26<15:48, 3.71s/it] {'loss': 1.414, 'grad_norm': 0.0037977849655339434, 'learning_rate': 0.1024930691738073, 'epoch': 0.51} + 51%|█████ | 264/520 [16:26<15:48, 3.71s/it] 51%|█████ | 265/520 [16:30<15:41, 3.69s/it] {'loss': 1.307, 'grad_norm': 0.005054363124400407, 'learning_rate': 0.10186988664309023, 'epoch': 0.51} + 51%|█████ | 265/520 [16:30<15:41, 3.69s/it] 51%|█████ | 266/520 [16:33<15:33, 3.67s/it] {'loss': 1.1503, 'grad_norm': 0.003339975874844434, 'learning_rate': 0.10124663145954152, 'epoch': 0.51} + 51%|█████ | 266/520 [16:33<15:33, 3.67s/it] 51%|█████▏ | 267/520 [16:37<15:26, 3.66s/it] {'loss': 1.3007, 'grad_norm': 0.0036682186531869997, 'learning_rate': 0.10062332783920336, 'epoch': 0.51} + 51%|█████▏ | 267/520 [16:37<15:26, 3.66s/it] 52%|█████▏ | 268/520 [16:41<15:17, 3.64s/it] {'loss': 1.5897, 'grad_norm': 0.004924843343440044, 'learning_rate': 0.1, 'epoch': 0.52} + 52%|█████▏ | 268/520 [16:41<15:17, 3.64s/it] 52%|█████▏ | 269/520 [16:44<15:18, 3.66s/it] {'loss': 1.4205, 'grad_norm': 0.003935824064545591, 'learning_rate': 0.09937667216079665, 'epoch': 0.52} + 52%|█████▏ | 269/520 [16:44<15:18, 3.66s/it] 52%|█████▏ | 270/520 [16:48<15:14, 3.66s/it] {'loss': 1.3686, 'grad_norm': 0.003858371586993254, 'learning_rate': 0.0987533685404585, 'epoch': 0.52} + 52%|█████▏ | 270/520 [16:48<15:14, 3.66s/it] 52%|█████▏ | 271/520 [16:52<15:07, 3.65s/it] {'loss': 1.4078, 'grad_norm': 0.00393844074371111, 'learning_rate': 0.0981301133569098, 'epoch': 0.52} + 52%|█████▏ | 271/520 [16:52<15:07, 3.65s/it] 52%|█████▏ | 272/520 [16:55<15:04, 3.65s/it] {'loss': 1.3824, 'grad_norm': 0.005023897434851858, 'learning_rate': 0.09750693082619273, 'epoch': 0.52} + 52%|█████▏ | 272/520 [16:55<15:04, 3.65s/it] 52%|█████▎ | 273/520 [16:59<14:59, 3.64s/it] {'loss': 1.5239, 'grad_norm': 0.004876095860203964, 'learning_rate': 0.0968838451615267, 'epoch': 0.53} + 52%|█████▎ | 273/520 [16:59<14:59, 3.64s/it] 53%|█████▎ | 274/520 [17:03<14:59, 3.65s/it] {'loss': 1.3596, 'grad_norm': 0.004022011333993123, 'learning_rate': 0.09626088057236745, 'epoch': 0.53} + 53%|█████▎ | 274/520 [17:03<14:59, 3.65s/it] 53%|█████▎ | 275/520 [17:06<14:56, 3.66s/it] {'loss': 1.2992, 'grad_norm': 0.004711554816380952, 'learning_rate': 0.09563806126346641, 'epoch': 0.53} + 53%|█████▎ | 275/520 [17:06<14:56, 3.66s/it] 53%|█████▎ | 276/520 [17:10<15:05, 3.71s/it] {'loss': 1.3885, 'grad_norm': 0.004350515353949441, 'learning_rate': 0.09501541143393027, 'epoch': 0.53} + 53%|█████▎ | 276/520 [17:10<15:05, 3.71s/it] 53%|█████▎ | 277/520 [17:14<15:18, 3.78s/it] {'loss': 1.5022, 'grad_norm': 0.004881716456403451, 'learning_rate': 0.09439295527628082, 'epoch': 0.53} + 53%|█████▎ | 277/520 [17:14<15:18, 3.78s/it] 53%|█████▎ | 278/520 [17:18<15:18, 3.80s/it] {'loss': 1.2521, 'grad_norm': 0.0035756094835272295, 'learning_rate': 0.0937707169755148, 'epoch': 0.53} + 53%|█████▎ | 278/520 [17:18<15:18, 3.80s/it] 54%|█████▎ | 279/520 [17:22<15:17, 3.81s/it] {'loss': 1.4009, 'grad_norm': 0.0056674353111673195, 'learning_rate': 0.09314872070816434, 'epoch': 0.54} + 54%|█████▎ | 279/520 [17:22<15:17, 3.81s/it] 54%|█████▍ | 280/520 [17:26<15:18, 3.83s/it] {'loss': 1.3027, 'grad_norm': 0.00428045155868802, 'learning_rate': 0.09252699064135758, 'epoch': 0.54} + 54%|█████▍ | 280/520 [17:26<15:18, 3.83s/it] 54%|█████▍ | 281/520 [17:29<15:14, 3.83s/it] {'loss': 1.4201, 'grad_norm': 0.00424558198597029, 'learning_rate': 0.09190555093187967, 'epoch': 0.54} + 54%|█████▍ | 281/520 [17:29<15:14, 3.83s/it] 54%|█████▍ | 282/520 [17:33<15:23, 3.88s/it] {'loss': 1.2683, 'grad_norm': 0.003684998084828612, 'learning_rate': 0.09128442572523418, 'epoch': 0.54} + 54%|█████▍ | 282/520 [17:33<15:23, 3.88s/it] 54%|█████▍ | 283/520 [17:38<15:38, 3.96s/it] {'loss': 1.4443, 'grad_norm': 0.004691866786871056, 'learning_rate': 0.09066363915470495, 'epoch': 0.54} + 54%|█████▍ | 283/520 [17:38<15:38, 3.96s/it] 55%|█████▍ | 284/520 [17:42<15:47, 4.02s/it] {'loss': 1.3781, 'grad_norm': 0.005469531215289677, 'learning_rate': 0.09004321534041836, 'epoch': 0.55} + 55%|█████▍ | 284/520 [17:42<15:47, 4.02s/it] 55%|█████▍ | 285/520 [17:46<15:43, 4.01s/it] {'loss': 1.2922, 'grad_norm': 0.004063761493992122, 'learning_rate': 0.08942317838840624, 'epoch': 0.55} + 55%|█████▍ | 285/520 [17:46<15:43, 4.01s/it] 55%|█████▌ | 286/520 [17:50<15:33, 3.99s/it] {'loss': 1.1549, 'grad_norm': 0.005031112343351182, 'learning_rate': 0.08880355238966922, 'epoch': 0.55} + 55%|█████▌ | 286/520 [17:50<15:33, 3.99s/it] 55%|█████▌ | 287/520 [17:54<15:20, 3.95s/it] {'loss': 1.4117, 'grad_norm': 0.004137953695906106, 'learning_rate': 0.08818436141924073, 'epoch': 0.55} + 55%|█████▌ | 287/520 [17:54<15:20, 3.95s/it] 55%|█████▌ | 288/520 [17:57<15:08, 3.92s/it] {'loss': 1.4672, 'grad_norm': 0.00430055035566648, 'learning_rate': 0.08756562953525152, 'epoch': 0.55} + 55%|█████▌ | 288/520 [17:57<15:08, 3.92s/it] 56%|█████▌ | 289/520 [18:01<14:59, 3.89s/it] {'loss': 1.303, 'grad_norm': 0.00356327731447165, 'learning_rate': 0.08694738077799487, 'epoch': 0.56} + 56%|█████▌ | 289/520 [18:01<14:59, 3.89s/it] 56%|█████▌ | 290/520 [18:05<14:59, 3.91s/it] {'loss': 1.2147, 'grad_norm': 0.003536878886979853, 'learning_rate': 0.08632963916899268, 'epoch': 0.56} + 56%|█████▌ | 290/520 [18:05<14:59, 3.91s/it] 56%|█████▌ | 291/520 [18:09<14:39, 3.84s/it] {'loss': 1.2816, 'grad_norm': 0.005211487074073691, 'learning_rate': 0.08571242871006202, 'epoch': 0.56} + 56%|█████▌ | 291/520 [18:09<14:39, 3.84s/it] 56%|█████▌ | 292/520 [18:12<14:23, 3.79s/it] {'loss': 1.3266, 'grad_norm': 0.0036691972502274355, 'learning_rate': 0.08509577338238256, 'epoch': 0.56} + 56%|█████▌ | 292/520 [18:12<14:23, 3.79s/it] 56%|█████▋ | 293/520 [18:16<14:09, 3.74s/it] {'loss': 1.2706, 'grad_norm': 0.00417745229826909, 'learning_rate': 0.08447969714556484, 'epoch': 0.56} + 56%|█████▋ | 293/520 [18:16<14:09, 3.74s/it] 57%|█████▋ | 294/520 [18:20<14:05, 3.74s/it] {'loss': 1.307, 'grad_norm': 0.004091720774055209, 'learning_rate': 0.08386422393671933, 'epoch': 0.57} + 57%|█████▋ | 294/520 [18:20<14:05, 3.74s/it] 57%|█████▋ | 295/520 [18:24<14:09, 3.78s/it] {'loss': 1.4579, 'grad_norm': 0.005465505333643067, 'learning_rate': 0.08324937766952638, 'epoch': 0.57} + 57%|█████▋ | 295/520 [18:24<14:09, 3.78s/it] 57%|█████▋ | 296/520 [18:28<14:08, 3.79s/it] {'loss': 1.2414, 'grad_norm': 0.004187358450462526, 'learning_rate': 0.08263518223330697, 'epoch': 0.57} + 57%|█████▋ | 296/520 [18:28<14:08, 3.79s/it] 57%|█████▋ | 297/520 [18:31<14:09, 3.81s/it] {'loss': 1.3807, 'grad_norm': 0.00425859752592757, 'learning_rate': 0.08202166149209474, 'epoch': 0.57} + 57%|█████▋ | 297/520 [18:31<14:09, 3.81s/it] 57%|█████▋ | 298/520 [18:35<14:04, 3.81s/it] {'loss': 1.3416, 'grad_norm': 0.003744520386731094, 'learning_rate': 0.08140883928370855, 'epoch': 0.57} + 57%|█████▋ | 298/520 [18:35<14:04, 3.81s/it] 57%|█████▊ | 299/520 [18:39<13:49, 3.76s/it] {'loss': 1.4489, 'grad_norm': 0.0041832554794743034, 'learning_rate': 0.0807967394188264, 'epoch': 0.57} + 57%|█████▊ | 299/520 [18:39<13:49, 3.76s/it] 58%|█████▊ | 300/520 [18:42<13:39, 3.72s/it] {'loss': 1.406, 'grad_norm': 0.004136347058542787, 'learning_rate': 0.08018538568006027, 'epoch': 0.58} + 58%|█████▊ | 300/520 [18:42<13:39, 3.72s/it] 58%|█████▊ | 301/520 [18:46<13:34, 3.72s/it] {'loss': 1.3696, 'grad_norm': 0.0038395604804113137, 'learning_rate': 0.07957480182103199, 'epoch': 0.58} + 58%|█████▊ | 301/520 [18:46<13:34, 3.72s/it] 58%|█████▊ | 302/520 [18:50<13:30, 3.72s/it] {'loss': 1.4629, 'grad_norm': 0.00406959093274051, 'learning_rate': 0.07896501156545044, 'epoch': 0.58} + 58%|█████▊ | 302/520 [18:50<13:30, 3.72s/it] 58%|█████▊ | 303/520 [18:54<13:27, 3.72s/it] {'loss': 1.3041, 'grad_norm': 0.004197241637119344, 'learning_rate': 0.07835603860618973, 'epoch': 0.58} + 58%|█████▊ | 303/520 [18:54<13:27, 3.72s/it] 58%|█████▊ | 304/520 [18:57<13:21, 3.71s/it] {'loss': 1.3958, 'grad_norm': 0.011839603350762489, 'learning_rate': 0.07774790660436857, 'epoch': 0.58} + 58%|█████▊ | 304/520 [18:57<13:21, 3.71s/it] 59%|█████▊ | 305/520 [19:01<13:15, 3.70s/it] {'loss': 1.4252, 'grad_norm': 0.004381565470183298, 'learning_rate': 0.07714063918843106, 'epoch': 0.59} + 59%|█████▊ | 305/520 [19:01<13:15, 3.70s/it] 59%|█████▉ | 306/520 [19:05<13:12, 3.70s/it] {'loss': 1.3461, 'grad_norm': 0.003734938740363352, 'learning_rate': 0.0765342599532285, 'epoch': 0.59} + 59%|█████▉ | 306/520 [19:05<13:12, 3.70s/it] 59%|█████▉ | 307/520 [19:09<13:26, 3.78s/it] {'loss': 1.2887, 'grad_norm': 0.0037108037034591264, 'learning_rate': 0.07592879245910272, 'epoch': 0.59} + 59%|█████▉ | 307/520 [19:09<13:26, 3.78s/it] 59%|█████▉ | 308/520 [19:12<13:22, 3.78s/it] {'loss': 1.4114, 'grad_norm': 0.003950927509552787, 'learning_rate': 0.07532426023097064, 'epoch': 0.59} + 59%|█████▉ | 308/520 [19:12<13:22, 3.78s/it] 59%|█████▉ | 309/520 [19:16<13:10, 3.75s/it] {'loss': 1.279, 'grad_norm': 0.0036387341789725245, 'learning_rate': 0.07472068675741024, 'epoch': 0.59} + 59%|█████▉ | 309/520 [19:16<13:10, 3.75s/it] 60%|█████▉ | 310/520 [19:20<13:01, 3.72s/it] {'loss': 1.2609, 'grad_norm': 0.0038692098103387785, 'learning_rate': 0.07411809548974792, 'epoch': 0.6} + 60%|█████▉ | 310/520 [19:20<13:01, 3.72s/it] 60%|█████▉ | 311/520 [19:23<12:55, 3.71s/it] {'loss': 1.2254, 'grad_norm': 0.003910954991671182, 'learning_rate': 0.07351650984114727, 'epoch': 0.6} + 60%|█████▉ | 311/520 [19:23<12:55, 3.71s/it] 60%|██████ | 312/520 [19:27<12:47, 3.69s/it] {'loss': 1.219, 'grad_norm': 0.00462862077545186, 'learning_rate': 0.0729159531856995, 'epoch': 0.6} + 60%|██████ | 312/520 [19:27<12:47, 3.69s/it] 60%|██████ | 313/520 [19:31<12:41, 3.68s/it] {'loss': 1.2173, 'grad_norm': 0.0038042190258092477, 'learning_rate': 0.07231644885751508, 'epoch': 0.6} + 60%|██████ | 313/520 [19:31<12:41, 3.68s/it] 60%|██████ | 314/520 [19:35<13:08, 3.83s/it] {'loss': 1.247, 'grad_norm': 0.003637689343803788, 'learning_rate': 0.07171802014981725, 'epoch': 0.6} + 60%|██████ | 314/520 [19:35<13:08, 3.83s/it] 61%|██████ | 315/520 [19:39<13:06, 3.83s/it] {'loss': 1.5919, 'grad_norm': 0.012654212378406785, 'learning_rate': 0.07112069031403703, 'epoch': 0.61} + 61%|██████ | 315/520 [19:39<13:06, 3.83s/it] 61%|██████ | 316/520 [19:43<13:32, 3.98s/it] {'loss': 1.2142, 'grad_norm': 0.005006056817228867, 'learning_rate': 0.07052448255890957, 'epoch': 0.61} + 61%|██████ | 316/520 [19:43<13:32, 3.98s/it] 61%|██████ | 317/520 [19:47<13:20, 3.94s/it] {'loss': 1.247, 'grad_norm': 0.00359060308644697, 'learning_rate': 0.0699294200495727, 'epoch': 0.61} + 61%|██████ | 317/520 [19:47<13:20, 3.94s/it] 61%|██████ | 318/520 [19:51<13:11, 3.92s/it] {'loss': 1.3783, 'grad_norm': 0.004187518867734081, 'learning_rate': 0.06933552590666658, 'epoch': 0.61} + 61%|██████ | 318/520 [19:51<13:11, 3.92s/it] 61%|██████▏ | 319/520 [19:55<13:28, 4.02s/it] {'loss': 1.235, 'grad_norm': 0.00407622143530614, 'learning_rate': 0.06874282320543557, 'epoch': 0.61} + 61%|██████▏ | 319/520 [19:55<13:28, 4.02s/it] 62%|██████▏ | 320/520 [19:59<13:14, 3.97s/it] {'loss': 1.174, 'grad_norm': 0.004293642713926176, 'learning_rate': 0.06815133497483157, 'epoch': 0.62} + 62%|██████▏ | 320/520 [19:59<13:14, 3.97s/it] 62%|██████▏ | 321/520 [20:03<12:56, 3.90s/it] {'loss': 1.3846, 'grad_norm': 0.0037912983984800916, 'learning_rate': 0.06756108419661931, 'epoch': 0.62} + 62%|██████▏ | 321/520 [20:03<12:56, 3.90s/it] 62%|██████▏ | 322/520 [20:06<12:38, 3.83s/it] {'loss': 1.3297, 'grad_norm': 0.0057778954791976485, 'learning_rate': 0.06697209380448332, 'epoch': 0.62} + 62%|██████▏ | 322/520 [20:06<12:38, 3.83s/it] 62%|██████▏ | 323/520 [20:10<12:25, 3.78s/it] {'loss': 1.4283, 'grad_norm': 0.009813927904776532, 'learning_rate': 0.06638438668313694, 'epoch': 0.62} + 62%|██████▏ | 323/520 [20:10<12:25, 3.78s/it] 62%|██████▏ | 324/520 [20:14<12:17, 3.76s/it] {'loss': 1.3113, 'grad_norm': 0.0050139257757940495, 'learning_rate': 0.06579798566743314, 'epoch': 0.62} + 62%|██████▏ | 324/520 [20:14<12:17, 3.76s/it] 62%|██████▎ | 325/520 [20:17<12:09, 3.74s/it] {'loss': 1.3273, 'grad_norm': 0.004128392449999362, 'learning_rate': 0.06521291354147728, 'epoch': 0.62} + 62%|██████▎ | 325/520 [20:17<12:09, 3.74s/it] 63%|██████▎ | 326/520 [20:21<12:00, 3.72s/it] {'loss': 1.3062, 'grad_norm': 0.0037909167302595153, 'learning_rate': 0.06462919303774187, 'epoch': 0.63} + 63%|██████▎ | 326/520 [20:21<12:00, 3.72s/it] 63%|██████▎ | 327/520 [20:25<11:53, 3.70s/it] {'loss': 1.4632, 'grad_norm': 0.006007312186758611, 'learning_rate': 0.06404684683618325, 'epoch': 0.63} + 63%|██████▎ | 327/520 [20:25<11:53, 3.70s/it] 63%|██████▎ | 328/520 [20:28<11:46, 3.68s/it] {'loss': 1.386, 'grad_norm': 0.004187350468694303, 'learning_rate': 0.0634658975633605, 'epoch': 0.63} + 63%|██████▎ | 328/520 [20:28<11:46, 3.68s/it] 63%|██████▎ | 329/520 [20:32<11:43, 3.69s/it] {'loss': 1.2209, 'grad_norm': 0.0033580405914609085, 'learning_rate': 0.06288636779155621, 'epoch': 0.63} + 63%|██████▎ | 329/520 [20:32<11:43, 3.69s/it] 63%|██████▎ | 330/520 [20:36<11:37, 3.67s/it] {'loss': 1.297, 'grad_norm': 0.003352565053109525, 'learning_rate': 0.06230828003789948, 'epoch': 0.63} + 63%|██████▎ | 330/520 [20:36<11:37, 3.67s/it] 64%|██████▎ | 331/520 [20:39<11:33, 3.67s/it] {'loss': 1.264, 'grad_norm': 0.003637221347135161, 'learning_rate': 0.06173165676349103, 'epoch': 0.64} + 64%|██████▎ | 331/520 [20:39<11:33, 3.67s/it] 64%|██████▍ | 332/520 [20:43<11:28, 3.66s/it] {'loss': 1.4574, 'grad_norm': 0.003939934504988293, 'learning_rate': 0.06115652037253053, 'epoch': 0.64} + 64%|██████▍ | 332/520 [20:43<11:28, 3.66s/it] 64%|██████▍ | 333/520 [20:47<11:24, 3.66s/it] {'loss': 1.4329, 'grad_norm': 0.004360187162653016, 'learning_rate': 0.06058289321144608, 'epoch': 0.64} + 64%|██████▍ | 333/520 [20:47<11:24, 3.66s/it] 64%|██████▍ | 334/520 [20:50<11:19, 3.65s/it] {'loss': 1.3159, 'grad_norm': 0.004741281083749495, 'learning_rate': 0.06001079756802592, 'epoch': 0.64} + 64%|██████▍ | 334/520 [20:50<11:19, 3.65s/it] 64%|██████▍ | 335/520 [20:54<11:15, 3.65s/it] {'loss': 1.3019, 'grad_norm': 0.003891239858208158, 'learning_rate': 0.059440255670552514, 'epoch': 0.64} + 64%|██████▍ | 335/520 [20:54<11:15, 3.65s/it] 65%|██████▍ | 336/520 [20:58<11:10, 3.64s/it] {'loss': 1.1806, 'grad_norm': 0.00420449737253774, 'learning_rate': 0.05887128968693887, 'epoch': 0.65} + 65%|██████▍ | 336/520 [20:58<11:10, 3.64s/it] 65%|██████▍ | 337/520 [21:01<11:09, 3.66s/it] {'loss': 1.185, 'grad_norm': 0.004274146276348327, 'learning_rate': 0.058303921723867225, 'epoch': 0.65} + 65%|██████▍ | 337/520 [21:01<11:09, 3.66s/it] 65%|██████▌ | 338/520 [21:05<11:05, 3.66s/it] {'loss': 1.3194, 'grad_norm': 0.003930752255427542, 'learning_rate': 0.05773817382593008, 'epoch': 0.65} + 65%|██████▌ | 338/520 [21:05<11:05, 3.66s/it] 65%|██████▌ | 339/520 [21:09<11:02, 3.66s/it] {'loss': 1.2587, 'grad_norm': 0.003762253559199532, 'learning_rate': 0.057174067974773715, 'epoch': 0.65} + 65%|██████▌ | 339/520 [21:09<11:02, 3.66s/it] 65%|██████▌ | 340/520 [21:12<11:00, 3.67s/it] {'loss': 1.2464, 'grad_norm': 0.00368493717970435, 'learning_rate': 0.056611626088244195, 'epoch': 0.65} + 65%|██████▌ | 340/520 [21:12<11:00, 3.67s/it] 66%|██████▌ | 341/520 [21:16<10:55, 3.66s/it] {'loss': 1.2703, 'grad_norm': 0.004024421843260967, 'learning_rate': 0.056050870019535494, 'epoch': 0.66} + 66%|██████▌ | 341/520 [21:16<10:55, 3.66s/it] 66%|██████▌ | 342/520 [21:20<10:50, 3.66s/it] {'loss': 1.45, 'grad_norm': 0.006322263852477994, 'learning_rate': 0.05549182155634076, 'epoch': 0.66} + 66%|██████▌ | 342/520 [21:20<10:50, 3.66s/it] 66%|██████▌ | 343/520 [21:23<10:48, 3.66s/it] {'loss': 1.4263, 'grad_norm': 0.0054973885569834045, 'learning_rate': 0.054934502420005464, 'epoch': 0.66} + 66%|██████▌ | 343/520 [21:23<10:48, 3.66s/it] 66%|██████▌ | 344/520 [21:27<10:44, 3.66s/it] {'loss': 1.2128, 'grad_norm': 0.0039571605811735905, 'learning_rate': 0.0543789342646837, 'epoch': 0.66} + 66%|██████▌ | 344/520 [21:27<10:44, 3.66s/it] 66%|██████▋ | 345/520 [21:31<10:41, 3.67s/it] {'loss': 1.3444, 'grad_norm': 0.00410671033750227, 'learning_rate': 0.05382513867649663, 'epoch': 0.66} + 66%|██████▋ | 345/520 [21:31<10:41, 3.67s/it] 67%|██████▋ | 346/520 [21:34<10:38, 3.67s/it] {'loss': 1.3692, 'grad_norm': 0.004184386381677304, 'learning_rate': 0.0532731371726938, 'epoch': 0.67} + 67%|██████▋ | 346/520 [21:34<10:38, 3.67s/it] 67%|██████▋ | 347/520 [21:38<10:33, 3.66s/it] {'loss': 1.2351, 'grad_norm': 0.003628189721960376, 'learning_rate': 0.05272295120081732, 'epoch': 0.67} + 67%|██████▋ | 347/520 [21:38<10:33, 3.66s/it]Token indices sequence length is longer than the specified maximum sequence length for this model (2778 > 2048). Running this sequence through the model will result in indexing errors + 67%|██████▋ | 348/520 [21:42<10:30, 3.67s/it] {'loss': 1.1938, 'grad_norm': 0.005037870497215469, 'learning_rate': 0.05217460213786821, 'epoch': 0.67} + 67%|██████▋ | 348/520 [21:42<10:30, 3.67s/it] 67%|██████▋ | 349/520 [21:45<10:30, 3.69s/it] {'loss': 1.243, 'grad_norm': 0.004179784197501732, 'learning_rate': 0.051628111289476024, 'epoch': 0.67} + 67%|██████▋ | 349/520 [21:45<10:30, 3.69s/it] 67%|██████▋ | 350/520 [21:49<10:27, 3.69s/it] {'loss': 1.2693, 'grad_norm': 0.0043323133609969875, 'learning_rate': 0.051083499889071106, 'epoch': 0.67} + 67%|██████▋ | 350/520 [21:49<10:27, 3.69s/it] 68%|██████▊ | 351/520 [21:53<10:23, 3.69s/it] {'loss': 1.1736, 'grad_norm': 0.003712037840029458, 'learning_rate': 0.05054078909705926, 'epoch': 0.68} + 68%|██████▊ | 351/520 [21:53<10:23, 3.69s/it] 68%|██████▊ | 352/520 [21:56<10:19, 3.69s/it] {'loss': 1.3062, 'grad_norm': 0.003926124258659014, 'learning_rate': 0.050000000000000024, 'epoch': 0.68} + 68%|██████▊ | 352/520 [21:56<10:19, 3.69s/it] 68%|██████▊ | 353/520 [22:00<10:15, 3.69s/it] {'loss': 1.3048, 'grad_norm': 0.00328108299919681, 'learning_rate': 0.04946115360978696, 'epoch': 0.68} + 68%|██████▊ | 353/520 [22:00<10:15, 3.69s/it] 68%|██████▊ | 354/520 [22:04<10:10, 3.68s/it] {'loss': 1.4764, 'grad_norm': 0.004369742498193127, 'learning_rate': 0.048924270862831465, 'epoch': 0.68} + 68%|██████▊ | 354/520 [22:04<10:10, 3.68s/it] 68%|██████▊ | 355/520 [22:07<10:05, 3.67s/it] {'loss': 1.2463, 'grad_norm': 0.003933485000968477, 'learning_rate': 0.04838937261924933, 'epoch': 0.68} + 68%|██████▊ | 355/520 [22:07<10:05, 3.67s/it] 68%|██████▊ | 356/520 [22:11<10:04, 3.68s/it] {'loss': 1.2461, 'grad_norm': 0.003791610715920293, 'learning_rate': 0.0478564796620502, 'epoch': 0.68} + 68%|██████▊ | 356/520 [22:11<10:04, 3.68s/it] 69%|██████▊ | 357/520 [22:15<10:01, 3.69s/it] {'loss': 1.259, 'grad_norm': 0.0034171556454699126, 'learning_rate': 0.04732561269632993, 'epoch': 0.69} + 69%|██████▊ | 357/520 [22:15<10:01, 3.69s/it] 69%|██████▉ | 358/520 [22:18<09:55, 3.67s/it] {'loss': 1.1866, 'grad_norm': 0.003987532322871887, 'learning_rate': 0.04679679234846636, 'epoch': 0.69} + 69%|██████▉ | 358/520 [22:18<09:55, 3.67s/it] 69%|██████▉ | 359/520 [22:22<09:59, 3.72s/it] {'loss': 1.3804, 'grad_norm': 0.004258226894140306, 'learning_rate': 0.046270039165317606, 'epoch': 0.69} + 69%|██████▉ | 359/520 [22:22<09:59, 3.72s/it] 69%|██████▉ | 360/520 [22:26<10:01, 3.76s/it] {'loss': 1.4018, 'grad_norm': 0.0044146710427047115, 'learning_rate': 0.04574537361342407, 'epoch': 0.69} + 69%|██████▉ | 360/520 [22:26<10:01, 3.76s/it] 69%|██████▉ | 361/520 [22:30<10:02, 3.79s/it] {'loss': 1.3906, 'grad_norm': 0.003573380646932563, 'learning_rate': 0.04522281607821288, 'epoch': 0.69} + 69%|██████▉ | 361/520 [22:30<10:02, 3.79s/it] 70%|██████▉ | 362/520 [22:34<10:02, 3.81s/it] {'loss': 1.2607, 'grad_norm': 0.0041322334093583605, 'learning_rate': 0.04470238686320606, 'epoch': 0.7} + 70%|██████▉ | 362/520 [22:34<10:02, 3.81s/it] 70%|██████▉ | 363/520 [22:38<10:02, 3.84s/it] {'loss': 1.2842, 'grad_norm': 0.00370396069277006, 'learning_rate': 0.044184106189231624, 'epoch': 0.7} + 70%|██████▉ | 363/520 [22:38<10:02, 3.84s/it] 70%|███████ | 364/520 [22:42<10:00, 3.85s/it] {'loss': 1.3953, 'grad_norm': 0.003681990683844659, 'learning_rate': 0.043667994193637795, 'epoch': 0.7} + 70%|███████ | 364/520 [22:42<10:00, 3.85s/it] 70%|███████ | 365/520 [22:46<09:58, 3.86s/it] {'loss': 1.351, 'grad_norm': 0.004134101774237762, 'learning_rate': 0.043154070929510784, 'epoch': 0.7} + 70%|███████ | 365/520 [22:46<09:58, 3.86s/it] 70%|███████ | 366/520 [22:49<09:54, 3.86s/it] {'loss': 1.3059, 'grad_norm': 0.003640364184818505, 'learning_rate': 0.04264235636489542, 'epoch': 0.7} + 70%|███████ | 366/520 [22:49<09:54, 3.86s/it] 71%|███████ | 367/520 [22:53<09:51, 3.87s/it] {'loss': 1.2991, 'grad_norm': 0.0037333591802126086, 'learning_rate': 0.04213287038201943, 'epoch': 0.71} + 71%|███████ | 367/520 [22:53<09:51, 3.87s/it] 71%|███████ | 368/520 [22:57<09:47, 3.87s/it] {'loss': 1.1477, 'grad_norm': 0.0038381280613978045, 'learning_rate': 0.04162563277652104, 'epoch': 0.71} + 71%|███████ | 368/520 [22:57<09:47, 3.87s/it] 71%|███████ | 369/520 [23:01<09:42, 3.86s/it] {'loss': 1.3648, 'grad_norm': 0.0037324034769690003, 'learning_rate': 0.04112066325667954, 'epoch': 0.71} + 71%|███████ | 369/520 [23:01<09:42, 3.86s/it] 71%|███████ | 370/520 [23:05<09:40, 3.87s/it] {'loss': 1.2121, 'grad_norm': 0.0034333979644447054, 'learning_rate': 0.04061798144264986, 'epoch': 0.71} + 71%|███████ | 370/520 [23:05<09:40, 3.87s/it] 71%|███████▏ | 371/520 [23:09<09:36, 3.87s/it] {'loss': 1.2064, 'grad_norm': 0.0037624876696684497, 'learning_rate': 0.04011760686569998, 'epoch': 0.71} + 71%|███████▏ | 371/520 [23:09<09:36, 3.87s/it] 72%|███████▏ | 372/520 [23:13<09:33, 3.87s/it] {'loss': 1.4657, 'grad_norm': 0.0038387845138980895, 'learning_rate': 0.03961955896745224, 'epoch': 0.72} + 72%|███████▏ | 372/520 [23:13<09:33, 3.87s/it] 72%|███████▏ | 373/520 [23:16<09:29, 3.88s/it] {'loss': 1.3209, 'grad_norm': 0.004026403156850908, 'learning_rate': 0.03912385709912794, 'epoch': 0.72} + 72%|███████▏ | 373/520 [23:16<09:29, 3.88s/it] 72%|███████▏ | 374/520 [23:20<09:24, 3.87s/it] {'loss': 1.2889, 'grad_norm': 0.0036693856496086375, 'learning_rate': 0.038630520520795276, 'epoch': 0.72} + 72%|███████▏ | 374/520 [23:20<09:24, 3.87s/it] 72%|███████▏ | 375/520 [23:24<09:21, 3.87s/it] {'loss': 1.1999, 'grad_norm': 0.003713354857804233, 'learning_rate': 0.03813956840062119, 'epoch': 0.72} + 72%|███████▏ | 375/520 [23:24<09:21, 3.87s/it] 72%|███████▏ | 376/520 [23:28<09:16, 3.86s/it] {'loss': 1.3217, 'grad_norm': 0.003609625977824214, 'learning_rate': 0.037651019814126656, 'epoch': 0.72} + 72%|███████▏ | 376/520 [23:28<09:16, 3.86s/it] 72%|███████▎ | 377/520 [23:32<09:13, 3.87s/it] {'loss': 1.2584, 'grad_norm': 0.005134717102415381, 'learning_rate': 0.037164893743445275, 'epoch': 0.72} + 72%|███████▎ | 377/520 [23:32<09:13, 3.87s/it] 73%|███████▎ | 378/520 [23:36<09:08, 3.86s/it] {'loss': 1.3068, 'grad_norm': 0.003949801428636032, 'learning_rate': 0.03668120907658603, 'epoch': 0.73} + 73%|███████▎ | 378/520 [23:36<09:08, 3.86s/it] 73%|███████▎ | 379/520 [23:40<09:03, 3.85s/it] {'loss': 1.2984, 'grad_norm': 0.003555238117575127, 'learning_rate': 0.036199984606699154, 'epoch': 0.73} + 73%|███████▎ | 379/520 [23:40<09:03, 3.85s/it] 73%|███████▎ | 380/520 [23:43<08:58, 3.85s/it] {'loss': 1.4296, 'grad_norm': 0.004120093002198017, 'learning_rate': 0.035721239031346066, 'epoch': 0.73} + 73%|███████▎ | 380/520 [23:43<08:58, 3.85s/it] 73%|███████▎ | 381/520 [23:47<08:54, 3.84s/it] {'loss': 1.2904, 'grad_norm': 0.0036874878827262403, 'learning_rate': 0.03524499095177297, 'epoch': 0.73} + 73%|███████▎ | 381/520 [23:47<08:54, 3.84s/it] 73%|███████▎ | 382/520 [23:51<08:51, 3.85s/it] {'loss': 1.368, 'grad_norm': 0.0037777071891918353, 'learning_rate': 0.03477125887218792, 'epoch': 0.73} + 73%|███████▎ | 382/520 [23:51<08:51, 3.85s/it] 74%|███████▎ | 383/520 [23:55<08:48, 3.86s/it] {'loss': 1.1207, 'grad_norm': 0.0038133770670743168, 'learning_rate': 0.03430006119904196, 'epoch': 0.74} + 74%|███████▎ | 383/520 [23:55<08:48, 3.86s/it] 74%|███████▍ | 384/520 [23:59<08:42, 3.84s/it] {'loss': 1.5027, 'grad_norm': 0.004032739029763675, 'learning_rate': 0.033831416240314084, 'epoch': 0.74} + 74%|███████▍ | 384/520 [23:59<08:42, 3.84s/it] 74%|███████▍ | 385/520 [24:03<08:37, 3.83s/it] {'loss': 1.2687, 'grad_norm': 0.0034785166827651305, 'learning_rate': 0.03336534220479961, 'epoch': 0.74} + 74%|███████▍ | 385/520 [24:03<08:37, 3.83s/it] 74%|███████▍ | 386/520 [24:06<08:27, 3.79s/it] {'loss': 1.2125, 'grad_norm': 0.003404303714802413, 'learning_rate': 0.032901857201403005, 'epoch': 0.74} + 74%|███████▍ | 386/520 [24:06<08:27, 3.79s/it] 74%|███████▍ | 387/520 [24:10<08:16, 3.74s/it] {'loss': 1.4582, 'grad_norm': 0.003975068623855291, 'learning_rate': 0.032440979238433976, 'epoch': 0.74} + 74%|███████▍ | 387/520 [24:10<08:16, 3.74s/it] 75%|███████▍ | 388/520 [24:14<08:09, 3.71s/it] {'loss': 1.1634, 'grad_norm': 0.0034817610546294653, 'learning_rate': 0.03198272622290804, 'epoch': 0.75} + 75%|███████▍ | 388/520 [24:14<08:09, 3.71s/it] 75%|███████▍ | 389/520 [24:17<08:02, 3.68s/it] {'loss': 1.2258, 'grad_norm': 0.004183269526998083, 'learning_rate': 0.03152711595985065, 'epoch': 0.75} + 75%|███████▍ | 389/520 [24:17<08:02, 3.68s/it] 75%|███████▌ | 390/520 [24:21<07:57, 3.67s/it] {'loss': 1.2824, 'grad_norm': 0.0035228167120909142, 'learning_rate': 0.031074166151605298, 'epoch': 0.75} + 75%|███████▌ | 390/520 [24:21<07:57, 3.67s/it] 75%|███████▌ | 391/520 [24:25<07:55, 3.69s/it] {'loss': 1.3774, 'grad_norm': 0.0038642763764863896, 'learning_rate': 0.030623894397145836, 'epoch': 0.75} + 75%|███████▌ | 391/520 [24:25<07:55, 3.69s/it] 75%|███████▌ | 392/520 [24:28<07:50, 3.67s/it] {'loss': 1.1707, 'grad_norm': 0.003568932095019465, 'learning_rate': 0.03017631819139273, 'epoch': 0.75} + 75%|███████▌ | 392/520 [24:28<07:50, 3.67s/it] 76%|███████▌ | 393/520 [24:32<07:45, 3.67s/it] {'loss': 1.2397, 'grad_norm': 0.003516211585185855, 'learning_rate': 0.029731454924533086, 'epoch': 0.76} + 76%|███████▌ | 393/520 [24:32<07:45, 3.67s/it] 76%|███████▌ | 394/520 [24:35<07:40, 3.65s/it] {'loss': 1.2446, 'grad_norm': 0.004166463760341811, 'learning_rate': 0.029289321881345254, 'epoch': 0.76} + 76%|███████▌ | 394/520 [24:35<07:40, 3.65s/it] 76%|███████▌ | 395/520 [24:39<07:35, 3.64s/it] {'loss': 1.2015, 'grad_norm': 0.0038448002889923723, 'learning_rate': 0.028849936240527008, 'epoch': 0.76} + 76%|███████▌ | 395/520 [24:39<07:35, 3.64s/it] 76%|███████▌ | 396/520 [24:43<07:31, 3.64s/it] {'loss': 1.2989, 'grad_norm': 0.0038994440424410326, 'learning_rate': 0.028413315074028157, 'epoch': 0.76} + 76%|███████▌ | 396/520 [24:43<07:31, 3.64s/it] 76%|███████▋ | 397/520 [24:46<07:28, 3.64s/it] {'loss': 1.2753, 'grad_norm': 0.003549561938854292, 'learning_rate': 0.027979475346387363, 'epoch': 0.76} + 76%|███████▋ | 397/520 [24:46<07:28, 3.64s/it] 77%|███████▋ | 398/520 [24:50<07:24, 3.64s/it] {'loss': 1.2671, 'grad_norm': 0.003802668217334711, 'learning_rate': 0.027548433914072735, 'epoch': 0.77} + 77%|███████▋ | 398/520 [24:50<07:24, 3.64s/it] 77%|███████▋ | 399/520 [24:54<07:23, 3.66s/it] {'loss': 1.2974, 'grad_norm': 0.0038190823395144657, 'learning_rate': 0.027120207524827168, 'epoch': 0.77} + 77%|███████▋ | 399/520 [24:54<07:23, 3.66s/it] 77%|███████▋ | 400/520 [24:57<07:18, 3.65s/it] {'loss': 1.3596, 'grad_norm': 0.003617070491665937, 'learning_rate': 0.02669481281701739, 'epoch': 0.77} + 77%|███████▋ | 400/520 [24:57<07:18, 3.65s/it] 77%|███████▋ | 401/520 [25:01<07:16, 3.67s/it] {'loss': 1.0848, 'grad_norm': 0.003679666532480901, 'learning_rate': 0.026272266318987603, 'epoch': 0.77} + 77%|███████▋ | 401/520 [25:01<07:16, 3.67s/it] 77%|███████▋ | 402/520 [25:05<07:12, 3.67s/it] {'loss': 1.1983, 'grad_norm': 0.0037665399794642603, 'learning_rate': 0.02585258444841733, 'epoch': 0.77} + 77%|███████▋ | 402/520 [25:05<07:12, 3.67s/it] 78%|███████▊ | 403/520 [25:08<07:09, 3.67s/it] {'loss': 1.2408, 'grad_norm': 0.003936982172586099, 'learning_rate': 0.025435783511683442, 'epoch': 0.78} + 78%|███████▊ | 403/520 [25:08<07:09, 3.67s/it] 78%|███████▊ | 404/520 [25:12<07:06, 3.68s/it] {'loss': 1.1499, 'grad_norm': 0.004644793958462539, 'learning_rate': 0.02502187970322657, 'epoch': 0.78} + 78%|███████▊ | 404/520 [25:12<07:06, 3.68s/it] 78%|███████▊ | 405/520 [25:16<07:03, 3.68s/it] {'loss': 1.2964, 'grad_norm': 0.0036286757391994382, 'learning_rate': 0.02461088910492202, 'epoch': 0.78} + 78%|███████▊ | 405/520 [25:16<07:03, 3.68s/it] 78%|███████▊ | 406/520 [25:19<06:58, 3.67s/it] {'loss': 1.2281, 'grad_norm': 0.004474150716032522, 'learning_rate': 0.02420282768545469, 'epoch': 0.78} + 78%|███████▊ | 406/520 [25:19<06:58, 3.67s/it] 78%|███████▊ | 407/520 [25:23<06:55, 3.67s/it] {'loss': 1.3432, 'grad_norm': 0.003718290799794689, 'learning_rate': 0.02379771129969892, 'epoch': 0.78} + 78%|███████▊ | 407/520 [25:23<06:55, 3.67s/it] 78%|███████▊ | 408/520 [25:27<06:50, 3.67s/it] {'loss': 1.2221, 'grad_norm': 0.003918412843390288, 'learning_rate': 0.023395555688102213, 'epoch': 0.78} + 78%|███████▊ | 408/520 [25:27<06:50, 3.67s/it] 79%|███████▊ | 409/520 [25:30<06:46, 3.66s/it] {'loss': 1.3557, 'grad_norm': 0.004170366063822139, 'learning_rate': 0.02299637647607372, 'epoch': 0.79} + 79%|███████▊ | 409/520 [25:30<06:46, 3.66s/it] 79%|███████▉ | 410/520 [25:34<06:42, 3.66s/it] {'loss': 1.0645, 'grad_norm': 0.003663775490087835, 'learning_rate': 0.022600189173377264, 'epoch': 0.79} + 79%|███████▉ | 410/520 [25:34<06:42, 3.66s/it] 79%|███████▉ | 411/520 [25:38<06:38, 3.65s/it] {'loss': 1.3267, 'grad_norm': 0.0040361141372475245, 'learning_rate': 0.022207009173528525, 'epoch': 0.79} + 79%|███████▉ | 411/520 [25:38<06:38, 3.65s/it] 79%|███████▉ | 412/520 [25:41<06:36, 3.67s/it] {'loss': 1.246, 'grad_norm': 0.00370842556047127, 'learning_rate': 0.02181685175319702, 'epoch': 0.79} + 79%|███████▉ | 412/520 [25:41<06:36, 3.67s/it] 79%|███████▉ | 413/520 [25:45<06:32, 3.67s/it] {'loss': 1.3349, 'grad_norm': 0.003938172801259814, 'learning_rate': 0.021429732071612653, 'epoch': 0.79} + 79%|███████▉ | 413/520 [25:45<06:32, 3.67s/it] 80%|███████▉ | 414/520 [25:49<06:29, 3.67s/it] {'loss': 1.1159, 'grad_norm': 0.003590859259622594, 'learning_rate': 0.02104566516997647, 'epoch': 0.8} + 80%|███████▉ | 414/520 [25:49<06:29, 3.67s/it] 80%|███████▉ | 415/520 [25:52<06:26, 3.68s/it] {'loss': 1.2165, 'grad_norm': 0.003561032171386618, 'learning_rate': 0.020664665970876496, 'epoch': 0.8} + 80%|███████▉ | 415/520 [25:52<06:26, 3.68s/it] 80%|████████ | 416/520 [25:56<06:21, 3.67s/it] {'loss': 1.1376, 'grad_norm': 0.004117209399397182, 'learning_rate': 0.020286749277707784, 'epoch': 0.8} + 80%|████████ | 416/520 [25:56<06:21, 3.67s/it] 80%|████████ | 417/520 [26:00<06:18, 3.67s/it] {'loss': 1.3058, 'grad_norm': 0.004196897745556433, 'learning_rate': 0.019911929774097215, 'epoch': 0.8} + 80%|████████ | 417/520 [26:00<06:18, 3.67s/it] 80%|████████ | 418/520 [26:03<06:15, 3.68s/it] {'loss': 1.2822, 'grad_norm': 0.0034407633291377096, 'learning_rate': 0.019540222023333165, 'epoch': 0.8} + 80%|████████ | 418/520 [26:03<06:15, 3.68s/it] 81%|████████ | 419/520 [26:07<06:11, 3.68s/it] {'loss': 1.2707, 'grad_norm': 0.004266228443002762, 'learning_rate': 0.01917164046779948, 'epoch': 0.81} + 81%|████████ | 419/520 [26:07<06:11, 3.68s/it] 81%|████████ | 420/520 [26:11<06:07, 3.67s/it] {'loss': 1.1485, 'grad_norm': 0.003929495434643994, 'learning_rate': 0.018806199428414352, 'epoch': 0.81} + 81%|████████ | 420/520 [26:11<06:07, 3.67s/it] 81%|████████ | 421/520 [26:15<06:03, 3.67s/it] {'loss': 1.0797, 'grad_norm': 0.003889486292389075, 'learning_rate': 0.018443913104073985, 'epoch': 0.81} + 81%|████████ | 421/520 [26:15<06:03, 3.67s/it] 81%|████████ | 422/520 [26:18<06:00, 3.67s/it] {'loss': 1.2157, 'grad_norm': 0.003953367091798379, 'learning_rate': 0.01808479557110081, 'epoch': 0.81} + 81%|████████ | 422/520 [26:18<06:00, 3.67s/it] 81%|████████▏ | 423/520 [26:22<05:56, 3.68s/it] {'loss': 1.2077, 'grad_norm': 0.004314750242206981, 'learning_rate': 0.017728860782696667, 'epoch': 0.81} + 81%|████████▏ | 423/520 [26:22<05:56, 3.68s/it] 82%|████████▏ | 424/520 [26:26<05:53, 3.68s/it] {'loss': 1.417, 'grad_norm': 0.004541884265172922, 'learning_rate': 0.017376122568400532, 'epoch': 0.82} + 82%|████████▏ | 424/520 [26:26<05:53, 3.68s/it] 82%|████████▏ | 425/520 [26:29<05:49, 3.68s/it] {'loss': 1.2054, 'grad_norm': 0.00363609653373172, 'learning_rate': 0.017026594633551252, 'epoch': 0.82} + 82%|████████▏ | 425/520 [26:29<05:49, 3.68s/it] 82%|████████▏ | 426/520 [26:33<05:46, 3.68s/it] {'loss': 1.2404, 'grad_norm': 0.0049974249512140904, 'learning_rate': 0.01668029055875512, 'epoch': 0.82} + 82%|████████▏ | 426/520 [26:33<05:46, 3.68s/it] 82%|████████▏ | 427/520 [26:37<05:42, 3.68s/it] {'loss': 1.1384, 'grad_norm': 0.003730077007158367, 'learning_rate': 0.016337223799358026, 'epoch': 0.82} + 82%|████████▏ | 427/520 [26:37<05:42, 3.68s/it] 82%|████████▏ | 428/520 [26:40<05:37, 3.67s/it] {'loss': 1.117, 'grad_norm': 0.0038174469447528224, 'learning_rate': 0.01599740768492286, 'epoch': 0.82} + 82%|████████▏ | 428/520 [26:40<05:37, 3.67s/it] 82%|████████▎ | 429/520 [26:44<05:34, 3.68s/it] {'loss': 1.2219, 'grad_norm': 0.003709451969764192, 'learning_rate': 0.015660855418711452, 'epoch': 0.82} + 82%|████████▎ | 429/520 [26:44<05:34, 3.68s/it]Token indices sequence length is longer than the specified maximum sequence length for this model (2076 > 2048). Running this sequence through the model will result in indexing errors + 83%|████████▎ | 430/520 [26:48<05:30, 3.67s/it] {'loss': 1.2155, 'grad_norm': 0.0034435233067643674, 'learning_rate': 0.015327580077171589, 'epoch': 0.83} + 83%|████████▎ | 430/520 [26:48<05:30, 3.67s/it] 83%|████████▎ | 431/520 [26:51<05:30, 3.72s/it] {'loss': 1.2955, 'grad_norm': 0.004094132159118526, 'learning_rate': 0.014997594609429088, 'epoch': 0.83} + 83%|████████▎ | 431/520 [26:51<05:30, 3.72s/it] 83%|████████▎ | 432/520 [26:55<05:26, 3.71s/it] {'loss': 1.1239, 'grad_norm': 0.0041405648078959205, 'learning_rate': 0.01467091183678444, 'epoch': 0.83} + 83%|████████▎ | 432/520 [26:55<05:26, 3.71s/it] 83%|████████▎ | 433/520 [26:59<05:24, 3.74s/it] {'loss': 1.264, 'grad_norm': 0.0037033594605585507, 'learning_rate': 0.014347544452214867, 'epoch': 0.83} + 83%|████████▎ | 433/520 [26:59<05:24, 3.74s/it] 83%|████████▎ | 434/520 [27:03<05:21, 3.74s/it] {'loss': 1.0012, 'grad_norm': 0.0036241361700265796, 'learning_rate': 0.014027505019880971, 'epoch': 0.83} + 83%|████████▎ | 434/520 [27:03<05:21, 3.74s/it] 84%|████████▎ | 435/520 [27:06<05:16, 3.73s/it] {'loss': 1.2939, 'grad_norm': 0.004049820336061321, 'learning_rate': 0.013710805974638696, 'epoch': 0.84} + 84%|████████▎ | 435/520 [27:06<05:16, 3.73s/it] 84%|████████▍ | 436/520 [27:10<05:11, 3.71s/it] {'loss': 1.0831, 'grad_norm': 0.0037159346996533645, 'learning_rate': 0.01339745962155613, 'epoch': 0.84} + 84%|████████▍ | 436/520 [27:10<05:11, 3.71s/it] 84%|████████▍ | 437/520 [27:14<05:07, 3.71s/it] {'loss': 1.3302, 'grad_norm': 0.003750253558694144, 'learning_rate': 0.01308747813543536, 'epoch': 0.84} + 84%|████████▍ | 437/520 [27:14<05:07, 3.71s/it] 84%|████████▍ | 438/520 [27:17<05:03, 3.70s/it] {'loss': 1.1194, 'grad_norm': 0.0034575596880653332, 'learning_rate': 0.012780873560339467, 'epoch': 0.84} + 84%|████████▍ | 438/520 [27:17<05:03, 3.70s/it] 84%|████████▍ | 439/520 [27:21<05:02, 3.74s/it] {'loss': 1.2556, 'grad_norm': 0.0032714746443627163, 'learning_rate': 0.012477657809124632, 'epoch': 0.84} + 84%|████████▍ | 439/520 [27:21<05:02, 3.74s/it] 85%|████████▍ | 440/520 [27:25<05:02, 3.78s/it] {'loss': 1.1784, 'grad_norm': 0.0036093054736708948, 'learning_rate': 0.012177842662977134, 'epoch': 0.85} + 85%|████████▍ | 440/520 [27:25<05:02, 3.78s/it] 85%|████████▍ | 441/520 [27:29<05:00, 3.80s/it] {'loss': 1.2929, 'grad_norm': 0.0035718987955125903, 'learning_rate': 0.01188143977095576, 'epoch': 0.85} + 85%|████████▍ | 441/520 [27:29<05:00, 3.80s/it] 85%|████████▌ | 442/520 [27:33<04:58, 3.82s/it] {'loss': 1.2347, 'grad_norm': 0.004391565020584776, 'learning_rate': 0.011588460649539035, 'epoch': 0.85} + 85%|████████▌ | 442/520 [27:33<04:58, 3.82s/it] 85%|████████▌ | 443/520 [27:37<04:55, 3.83s/it] {'loss': 1.2593, 'grad_norm': 0.004051212718673459, 'learning_rate': 0.011298916682177829, 'epoch': 0.85} + 85%|████████▌ | 443/520 [27:37<04:55, 3.83s/it] 85%|████████▌ | 444/520 [27:41<04:50, 3.83s/it] {'loss': 1.2235, 'grad_norm': 0.0033106577471923673, 'learning_rate': 0.011012819118853146, 'epoch': 0.85} + 85%|████████▌ | 444/520 [27:41<04:50, 3.83s/it] 86%|████████▌ | 445/520 [27:44<04:47, 3.84s/it] {'loss': 1.1378, 'grad_norm': 0.003683208942223845, 'learning_rate': 0.01073017907563887, 'epoch': 0.86} + 86%|████████▌ | 445/520 [27:44<04:47, 3.84s/it] 86%|████████▌ | 446/520 [27:48<04:44, 3.85s/it] {'loss': 1.3692, 'grad_norm': 0.003671331288303237, 'learning_rate': 0.010451007534269908, 'epoch': 0.86} + 86%|████████▌ | 446/520 [27:48<04:44, 3.85s/it] 86%|████████▌ | 447/520 [27:52<04:41, 3.86s/it] {'loss': 1.2306, 'grad_norm': 0.0038226500817417843, 'learning_rate': 0.010175315341715598, 'epoch': 0.86} + 86%|████████▌ | 447/520 [27:52<04:41, 3.86s/it] 86%|████████▌ | 448/520 [27:56<04:38, 3.87s/it] {'loss': 1.2048, 'grad_norm': 0.004035631876607452, 'learning_rate': 0.009903113209758098, 'epoch': 0.86} + 86%|████████▌ | 448/520 [27:56<04:38, 3.87s/it] 86%|████████▋ | 449/520 [28:00<04:33, 3.86s/it] {'loss': 1.3243, 'grad_norm': 0.0037792601516340742, 'learning_rate': 0.009634411714576352, 'epoch': 0.86} + 86%|████████▋ | 449/520 [28:00<04:33, 3.86s/it] 87%|████████▋ | 450/520 [28:04<04:28, 3.84s/it] {'loss': 1.2546, 'grad_norm': 0.003644610656082428, 'learning_rate': 0.009369221296335007, 'epoch': 0.87} + 87%|████████▋ | 450/520 [28:04<04:28, 3.84s/it] 87%|████████▋ | 451/520 [28:07<04:21, 3.79s/it] {'loss': 1.2429, 'grad_norm': 0.0038269284861162413, 'learning_rate': 0.009107552258778906, 'epoch': 0.87} + 87%|████████▋ | 451/520 [28:07<04:21, 3.79s/it] 87%|████████▋ | 452/520 [28:11<04:15, 3.75s/it] {'loss': 1.3555, 'grad_norm': 0.0036045786189368933, 'learning_rate': 0.008849414768832687, 'epoch': 0.87} + 87%|████████▋ | 452/520 [28:11<04:15, 3.75s/it] 87%|████████▋ | 453/520 [28:15<04:09, 3.73s/it] {'loss': 1.3358, 'grad_norm': 0.0037662607270266568, 'learning_rate': 0.008594818856205699, 'epoch': 0.87} + 87%|████████▋ | 453/520 [28:15<04:09, 3.73s/it] 87%|████████▋ | 454/520 [28:18<04:05, 3.72s/it] {'loss': 1.156, 'grad_norm': 0.004072884811107638, 'learning_rate': 0.00834377441300238, 'epoch': 0.87} + 87%|████████▋ | 454/520 [28:18<04:05, 3.72s/it] 88%|████████▊ | 455/520 [28:22<04:03, 3.75s/it] {'loss': 1.2827, 'grad_norm': 0.003662397250151896, 'learning_rate': 0.008096291193337934, 'epoch': 0.88} + 88%|████████▊ | 455/520 [28:22<04:03, 3.75s/it] 88%|████████▊ | 456/520 [28:26<03:58, 3.72s/it] {'loss': 1.2049, 'grad_norm': 0.003664379088293592, 'learning_rate': 0.007852378812959226, 'epoch': 0.88} + 88%|████████▊ | 456/520 [28:26<03:58, 3.72s/it] 88%|████████▊ | 457/520 [28:30<03:53, 3.70s/it] {'loss': 1.3098, 'grad_norm': 0.003499144110021664, 'learning_rate': 0.007612046748871327, 'epoch': 0.88} + 88%|████████▊ | 457/520 [28:30<03:53, 3.70s/it] 88%|████████▊ | 458/520 [28:33<03:49, 3.70s/it] {'loss': 1.3565, 'grad_norm': 0.0038858845248454124, 'learning_rate': 0.007375304338969136, 'epoch': 0.88} + 88%|████████▊ | 458/520 [28:33<03:49, 3.70s/it] 88%|████████▊ | 459/520 [28:37<03:45, 3.70s/it] {'loss': 1.2802, 'grad_norm': 0.003637785162436585, 'learning_rate': 0.007142160781674645, 'epoch': 0.88} + 88%|████████▊ | 459/520 [28:37<03:45, 3.70s/it] 88%|████████▊ | 460/520 [28:41<03:41, 3.69s/it] {'loss': 1.1408, 'grad_norm': 0.0034774035112600634, 'learning_rate': 0.006912625135579587, 'epoch': 0.88} + 88%|████████▊ | 460/520 [28:41<03:41, 3.69s/it] 89%|████████▊ | 461/520 [28:44<03:38, 3.70s/it] {'loss': 1.3969, 'grad_norm': 0.0031562046631692835, 'learning_rate': 0.0066867063190933496, 'epoch': 0.89} + 89%|████████▊ | 461/520 [28:44<03:38, 3.70s/it] 89%|████████▉ | 462/520 [28:48<03:33, 3.69s/it] {'loss': 1.4068, 'grad_norm': 0.0036622507257248163, 'learning_rate': 0.006464413110096601, 'epoch': 0.89} + 89%|████████▉ | 462/520 [28:48<03:33, 3.69s/it] 89%|████████▉ | 463/520 [28:52<03:29, 3.68s/it] {'loss': 1.112, 'grad_norm': 0.003906112877692824, 'learning_rate': 0.006245754145600091, 'epoch': 0.89} + 89%|████████▉ | 463/520 [28:52<03:29, 3.68s/it] 89%|████████▉ | 464/520 [28:55<03:25, 3.67s/it] {'loss': 1.2603, 'grad_norm': 0.0038413280135929334, 'learning_rate': 0.006030737921409169, 'epoch': 0.89} + 89%|████████▉ | 464/520 [28:55<03:25, 3.67s/it] 89%|████████▉ | 465/520 [28:59<03:22, 3.68s/it] {'loss': 1.3806, 'grad_norm': 0.004564613262088114, 'learning_rate': 0.005819372791793654, 'epoch': 0.89} + 89%|████████▉ | 465/520 [28:59<03:22, 3.68s/it] 90%|████████▉ | 466/520 [29:03<03:19, 3.69s/it] {'loss': 1.2475, 'grad_norm': 0.003292293441514522, 'learning_rate': 0.005611666969163243, 'epoch': 0.9} + 90%|████████▉ | 466/520 [29:03<03:19, 3.69s/it] 90%|████████▉ | 467/520 [29:06<03:15, 3.69s/it] {'loss': 1.2849, 'grad_norm': 0.0035499734982648284, 'learning_rate': 0.005407628523748398, 'epoch': 0.9} + 90%|████████▉ | 467/520 [29:06<03:15, 3.69s/it] 90%|█████████ | 468/520 [29:10<03:11, 3.68s/it] {'loss': 1.2282, 'grad_norm': 0.004152918883474852, 'learning_rate': 0.00520726538328683, 'epoch': 0.9} + 90%|█████████ | 468/520 [29:10<03:11, 3.68s/it] 90%|█████████ | 469/520 [29:14<03:07, 3.68s/it] {'loss': 1.2778, 'grad_norm': 0.0038930533059577115, 'learning_rate': 0.005010585332715401, 'epoch': 0.9} + 90%|█████████ | 469/520 [29:14<03:07, 3.68s/it] 90%|█████████ | 470/520 [29:17<03:03, 3.68s/it] {'loss': 1.1575, 'grad_norm': 0.003182126575501721, 'learning_rate': 0.004817596013867765, 'epoch': 0.9} + 90%|█████████ | 470/520 [29:17<03:03, 3.68s/it] 91%|█████████ | 471/520 [29:21<03:00, 3.67s/it] {'loss': 1.1754, 'grad_norm': 0.0039765198637101715, 'learning_rate': 0.004628304925177318, 'epoch': 0.91} + 91%|█████████ | 471/520 [29:21<03:00, 3.67s/it] 91%|█████████ | 472/520 [29:25<02:57, 3.69s/it] {'loss': 1.1462, 'grad_norm': 0.0035815494817038602, 'learning_rate': 0.004442719421385921, 'epoch': 0.91} + 91%|█████████ | 472/520 [29:25<02:57, 3.69s/it] 91%|█████████ | 473/520 [29:28<02:53, 3.69s/it] {'loss': 1.202, 'grad_norm': 0.003624585950693076, 'learning_rate': 0.004260846713258193, 'epoch': 0.91} + 91%|█████████ | 473/520 [29:28<02:53, 3.69s/it] 91%|█████████ | 474/520 [29:32<02:49, 3.68s/it] {'loss': 1.326, 'grad_norm': 0.0035020111663480826, 'learning_rate': 0.004082693867301224, 'epoch': 0.91} + 91%|█████████ | 474/520 [29:32<02:49, 3.68s/it] 91%|█████████▏| 475/520 [29:36<02:45, 3.69s/it] {'loss': 1.2324, 'grad_norm': 0.003518719315804924, 'learning_rate': 0.003908267805490051, 'epoch': 0.91} + 91%|█████████▏| 475/520 [29:36<02:45, 3.69s/it] 92%|█████████▏| 476/520 [29:39<02:41, 3.67s/it] {'loss': 1.2059, 'grad_norm': 0.003783061844457606, 'learning_rate': 0.003737575304998797, 'epoch': 0.92} + 92%|█████████▏| 476/520 [29:39<02:41, 3.67s/it] 92%|█████████▏| 477/520 [29:43<02:38, 3.67s/it] {'loss': 1.186, 'grad_norm': 0.004446614763531052, 'learning_rate': 0.003570622997937234, 'epoch': 0.92} + 92%|█████████▏| 477/520 [29:43<02:38, 3.67s/it] 92%|█████████▏| 478/520 [29:47<02:33, 3.66s/it] {'loss': 1.1487, 'grad_norm': 0.0034359440350924843, 'learning_rate': 0.00340741737109318, 'epoch': 0.92} + 92%|█████████▏| 478/520 [29:47<02:33, 3.66s/it] 92%|█████████▏| 479/520 [29:50<02:30, 3.67s/it] {'loss': 1.291, 'grad_norm': 0.00396701824774457, 'learning_rate': 0.003247964765680389, 'epoch': 0.92} + 92%|█████████▏| 479/520 [29:50<02:30, 3.67s/it] 92%|█████████▏| 480/520 [29:54<02:26, 3.67s/it] {'loss': 1.3223, 'grad_norm': 0.0035568599255448737, 'learning_rate': 0.0030922713770922153, 'epoch': 0.92} + 92%|█████████▏| 480/520 [29:54<02:26, 3.67s/it] 92%|█████████▎| 481/520 [29:58<02:22, 3.66s/it] {'loss': 1.3393, 'grad_norm': 0.003925508302823024, 'learning_rate': 0.0029403432546609046, 'epoch': 0.93} + 92%|█████████▎| 481/520 [29:58<02:22, 3.66s/it] 93%|█████████▎| 482/520 [30:01<02:19, 3.66s/it] {'loss': 1.3369, 'grad_norm': 0.0037667662510194218, 'learning_rate': 0.0027921863014225504, 'epoch': 0.93} + 93%|█████████▎| 482/520 [30:01<02:19, 3.66s/it] 93%|█████████▎| 483/520 [30:05<02:15, 3.67s/it] {'loss': 1.2226, 'grad_norm': 0.003922196542772685, 'learning_rate': 0.002647806273887665, 'epoch': 0.93} + 93%|█████████▎| 483/520 [30:05<02:15, 3.67s/it] 93%|█████████▎| 484/520 [30:09<02:12, 3.68s/it] {'loss': 1.2258, 'grad_norm': 0.00363594880813777, 'learning_rate': 0.0025072087818176383, 'epoch': 0.93} + 93%|█████████▎| 484/520 [30:09<02:12, 3.68s/it] 93%|█████████▎| 485/520 [30:13<02:10, 3.73s/it] {'loss': 1.1683, 'grad_norm': 0.0035714154934252543, 'learning_rate': 0.002370399288006664, 'epoch': 0.93} + 93%|█████████▎| 485/520 [30:13<02:10, 3.73s/it] 93%|█████████▎| 486/520 [30:16<02:07, 3.76s/it] {'loss': 1.2967, 'grad_norm': 0.003906064664511428, 'learning_rate': 0.0022373831080695463, 'epoch': 0.93} + 93%|█████████▎| 486/520 [30:16<02:07, 3.76s/it] 94%|█████████▎| 487/520 [30:20<02:04, 3.79s/it] {'loss': 1.145, 'grad_norm': 0.0034506086139734983, 'learning_rate': 0.0021081654102351635, 'epoch': 0.94} + 94%|█████████▎| 487/520 [30:20<02:04, 3.79s/it] 94%|█████████▍| 488/520 [30:24<02:02, 3.81s/it] {'loss': 1.0841, 'grad_norm': 0.003799902078501977, 'learning_rate': 0.0019827512151456175, 'epoch': 0.94} + 94%|█████████▍| 488/520 [30:24<02:02, 3.81s/it] 94%|█████████▍| 489/520 [30:28<01:56, 3.77s/it] {'loss': 1.3173, 'grad_norm': 0.0031640005598716956, 'learning_rate': 0.0018611453956612345, 'epoch': 0.94} + 94%|█████████▍| 489/520 [30:28<01:56, 3.77s/it] 94%|█████████▍| 490/520 [30:32<01:51, 3.73s/it] {'loss': 1.2142, 'grad_norm': 0.003721197265505607, 'learning_rate': 0.0017433526766711727, 'epoch': 0.94} + 94%|█████████▍| 490/520 [30:32<01:51, 3.73s/it] 94%|█████████▍| 491/520 [30:35<01:47, 3.70s/it] {'loss': 1.1662, 'grad_norm': 0.0037465868202714663, 'learning_rate': 0.0016293776349098677, 'epoch': 0.94} + 94%|█████████▍| 491/520 [30:35<01:47, 3.70s/it] 95%|█████████▍| 492/520 [30:39<01:43, 3.69s/it] {'loss': 1.3014, 'grad_norm': 0.0038220958575908436, 'learning_rate': 0.0015192246987791981, 'epoch': 0.95} + 95%|█████████▍| 492/520 [30:39<01:43, 3.69s/it] 95%|█████████▍| 493/520 [30:42<01:39, 3.67s/it] {'loss': 1.3781, 'grad_norm': 0.004140765344975031, 'learning_rate': 0.0014128981481764114, 'epoch': 0.95} + 95%|█████████▍| 493/520 [30:42<01:39, 3.67s/it] 95%|█████████▌| 494/520 [30:46<01:35, 3.67s/it] {'loss': 1.2308, 'grad_norm': 0.0033057063183924537, 'learning_rate': 0.0013104021143278911, 'epoch': 0.95} + 95%|█████████▌| 494/520 [30:46<01:35, 3.67s/it] 95%|█████████▌| 495/520 [30:50<01:32, 3.71s/it] {'loss': 1.1844, 'grad_norm': 0.003669382032874386, 'learning_rate': 0.0012117405796285285, 'epoch': 0.95} + 95%|█████████▌| 495/520 [30:50<01:32, 3.71s/it] 95%|█████████▌| 496/520 [30:54<01:29, 3.73s/it] {'loss': 1.1103, 'grad_norm': 0.003941613613576805, 'learning_rate': 0.0011169173774871477, 'epoch': 0.95} + 95%|█████████▌| 496/520 [30:54<01:29, 3.73s/it] 96%|█████████▌| 497/520 [30:57<01:25, 3.72s/it] {'loss': 1.2458, 'grad_norm': 0.003331095150146318, 'learning_rate': 0.0010259361921774012, 'epoch': 0.96} + 96%|█████████▌| 497/520 [30:57<01:25, 3.72s/it] 96%|█████████▌| 498/520 [31:01<01:21, 3.70s/it] {'loss': 1.194, 'grad_norm': 0.0039197157207667655, 'learning_rate': 0.000938800558694719, 'epoch': 0.96} + 96%|█████████▌| 498/520 [31:01<01:21, 3.70s/it] 96%|█████████▌| 499/520 [31:05<01:17, 3.70s/it] {'loss': 1.3913, 'grad_norm': 0.003780884150315754, 'learning_rate': 0.0008555138626189618, 'epoch': 0.96} + 96%|█████████▌| 499/520 [31:05<01:17, 3.70s/it] 96%|█████████▌| 500/520 [31:08<01:13, 3.68s/it] {'loss': 1.3081, 'grad_norm': 0.004333313692030883, 'learning_rate': 0.0007760793399827937, 'epoch': 0.96} + 96%|█████████▌| 500/520 [31:08<01:13, 3.68s/it] 96%|█████████▋| 501/520 [31:12<01:10, 3.70s/it] {'loss': 1.3103, 'grad_norm': 0.004496598282713003, 'learning_rate': 0.000700500077146038, 'epoch': 0.96} + 96%|█████████▋| 501/520 [31:12<01:10, 3.70s/it] 97%|█████████▋| 502/520 [31:16<01:06, 3.69s/it] {'loss': 1.2289, 'grad_norm': 0.0034901138091323544, 'learning_rate': 0.0006287790106757397, 'epoch': 0.97} + 97%|█████████▋| 502/520 [31:16<01:06, 3.69s/it] 97%|█████████▋| 503/520 [31:19<01:02, 3.68s/it] {'loss': 1.2784, 'grad_norm': 0.0037876645553322247, 'learning_rate': 0.0005609189272320237, 'epoch': 0.97} + 97%|█████████▋| 503/520 [31:19<01:02, 3.68s/it] 97%|█████████▋| 504/520 [31:23<00:58, 3.68s/it] {'loss': 1.2286, 'grad_norm': 0.0041005078835258234, 'learning_rate': 0.000496922463459859, 'epoch': 0.97} + 97%|█████████▋| 504/520 [31:23<00:58, 3.68s/it] 97%|█████████▋| 505/520 [31:27<00:55, 3.67s/it] {'loss': 1.2716, 'grad_norm': 0.003741980490486888, 'learning_rate': 0.0004367921058866187, 'epoch': 0.97} + 97%|█████████▋| 505/520 [31:27<00:55, 3.67s/it] 97%|█████████▋| 506/520 [31:30<00:51, 3.65s/it] {'loss': 1.1811, 'grad_norm': 0.004132315782632253, 'learning_rate': 0.0003805301908254455, 'epoch': 0.97} + 97%|█████████▋| 506/520 [31:30<00:51, 3.65s/it] 98%|█████████▊| 507/520 [31:34<00:47, 3.66s/it] {'loss': 1.4335, 'grad_norm': 0.0035013767460028984, 'learning_rate': 0.0003281389042844918, 'epoch': 0.97} + 98%|█████████▊| 507/520 [31:34<00:47, 3.66s/it] 98%|█████████▊| 508/520 [31:38<00:43, 3.65s/it] {'loss': 1.2987, 'grad_norm': 0.003652717797341307, 'learning_rate': 0.00027962028188198705, 'epoch': 0.98} + 98%|█████████▊| 508/520 [31:38<00:43, 3.65s/it] 98%|█████████▊| 509/520 [31:41<00:40, 3.65s/it] {'loss': 1.2687, 'grad_norm': 0.0035947054461122063, 'learning_rate': 0.00023497620876711256, 'epoch': 0.98} + 98%|█████████▊| 509/520 [31:41<00:40, 3.65s/it] 98%|█████████▊| 510/520 [31:45<00:36, 3.66s/it] {'loss': 1.2206, 'grad_norm': 0.0035475529230502858, 'learning_rate': 0.00019420841954681523, 'epoch': 0.98} + 98%|█████████▊| 510/520 [31:45<00:36, 3.66s/it] 98%|█████████▊| 511/520 [31:49<00:32, 3.66s/it] {'loss': 1.1938, 'grad_norm': 0.0034868275793018145, 'learning_rate': 0.00015731849821833956, 'epoch': 0.98} + 98%|█████████▊| 511/520 [31:49<00:32, 3.66s/it] 98%|█████████▊| 512/520 [31:52<00:29, 3.67s/it] {'loss': 1.0762, 'grad_norm': 0.003562080161244628, 'learning_rate': 0.00012430787810776555, 'epoch': 0.98} + 98%|█████████▊| 512/520 [31:52<00:29, 3.67s/it] 99%|█████████▊| 513/520 [31:56<00:25, 3.66s/it] {'loss': 1.2848, 'grad_norm': 0.00414016431030969, 'learning_rate': 9.517784181422018e-05, 'epoch': 0.99} + 99%|█████████▊| 513/520 [31:56<00:25, 3.66s/it] 99%|█████████▉| 514/520 [32:00<00:22, 3.68s/it] {'loss': 1.2581, 'grad_norm': 0.0033324911609113462, 'learning_rate': 6.992952116013917e-05, 'epoch': 0.99} + 99%|█████████▉| 514/520 [32:00<00:22, 3.68s/it] 99%|█████████▉| 515/520 [32:03<00:18, 3.70s/it] {'loss': 1.3084, 'grad_norm': 0.004251612204057862, 'learning_rate': 4.856389714723575e-05, 'epoch': 0.99} + 99%|█████████▉| 515/520 [32:04<00:18, 3.70s/it] 99%|█████████▉| 516/520 [32:07<00:14, 3.69s/it] {'loss': 1.1894, 'grad_norm': 0.003577681121339615, 'learning_rate': 3.108179991837545e-05, 'epoch': 0.99} + 99%|█████████▉| 516/520 [32:07<00:14, 3.69s/it] 99%|█████████▉| 517/520 [32:11<00:10, 3.66s/it] {'loss': 1.3488, 'grad_norm': 0.003873114962018393, 'learning_rate': 1.7483908725357544e-05, 'epoch': 0.99} + 99%|█████████▉| 517/520 [32:11<00:10, 3.66s/it] 100%|█████████▉| 518/520 [32:14<00:07, 3.65s/it] {'loss': 1.2145, 'grad_norm': 0.003834622816594602, 'learning_rate': 7.770751902513862e-06, 'epoch': 1.0} + 100%|█████████▉| 518/520 [32:14<00:07, 3.65s/it] 100%|█████████▉| 519/520 [32:18<00:03, 3.64s/it] {'loss': 1.2829, 'grad_norm': 0.0036139802245810783, 'learning_rate': 1.9427068461808086e-06, 'epoch': 1.0} + 100%|█████████▉| 519/520 [32:18<00:03, 3.64s/it] 100%|██████████| 520/520 [32:22<00:00, 3.89s/it] {'loss': 1.3722, 'grad_norm': 0.0040171741260657595, 'learning_rate': 0.0, 'epoch': 1.0} + 100%|██████████| 520/520 [32:22<00:00, 3.89s/it] {'train_runtime': 1942.9926, 'train_samples_per_second': 34.24, 'train_steps_per_second': 0.268, 'train_loss': 1.4795227156235622, 'epoch': 1.0} + 100%|██████████| 520/520 [32:22<00:00, 3.89s/it] 100%|██████████| 520/520 [32:22<00:00, 3.74s/it] +[2025-10-12 06:02:20,718] [INFO] [launch.py:348:main] Process 148702 exits successfully. +[2025-10-12 06:02:20,719] [INFO] [launch.py:348:main] Process 148699 exits successfully. +[2025-10-12 06:02:20,719] [INFO] [launch.py:348:main] Process 148703 exits successfully. +[2025-10-12 06:02:20,720] [INFO] [launch.py:348:main] Process 148698 exits successfully. +[2025-10-12 06:02:21,722] [INFO] [launch.py:348:main] Process 148701 exits successfully. +[2025-10-12 06:02:21,722] [INFO] [launch.py:348:main] Process 148697 exits successfully. +[2025-10-12 06:02:21,723] [INFO] [launch.py:348:main] Process 148700 exits successfully. +[2025-10-12 06:02:24,727] [INFO] [launch.py:348:main] Process 148696 exits successfully. +==== EXPERIMENT COMPLETED: qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_1.9_2e-1_connector-1.0_1.9_2e-1_ablation ==== +Log File: qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_1.9_2e-1_connector-1.0_1.9_2e-1_ablation_20251012_052820.log +Timestamp: 2025-10-12 06:02:27 +===================================== diff --git a/logs_oct11/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_1.9_2e-1_connector-1.0_1.9_2e-1_ablation_20251012_155158.log b/logs_oct11/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_1.9_2e-1_connector-1.0_1.9_2e-1_ablation_20251012_155158.log new file mode 100644 index 0000000000000000000000000000000000000000..c69d15f3a8fe8e845af29ad5a7e4fb12759963f1 --- /dev/null +++ b/logs_oct11/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_1.9_2e-1_connector-1.0_1.9_2e-1_ablation_20251012_155158.log @@ -0,0 +1,6 @@ +==== STARTING EXPERIMENT: qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_1.9_2e-1_connector-1.0_1.9_2e-1_ablation ==== +Log File: qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_1.9_2e-1_connector-1.0_1.9_2e-1_ablation_20251012_155158.log +Timestamp: 2025-10-12 15:51:58 +===================================== +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] diff --git a/logs_oct11/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.1_2e-1_connector-1.0_2.1_2e-1_ablation_20251012_060227.log b/logs_oct11/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.1_2e-1_connector-1.0_2.1_2e-1_ablation_20251012_060227.log new file mode 100644 index 0000000000000000000000000000000000000000..6192582346eb0e82527e362d029ec9e7d8ebd71b --- /dev/null +++ b/logs_oct11/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.1_2e-1_connector-1.0_2.1_2e-1_ablation_20251012_060227.log @@ -0,0 +1,2312 @@ +==== STARTING EXPERIMENT: qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.1_2e-1_connector-1.0_2.1_2e-1_ablation ==== +Log File: qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.1_2e-1_connector-1.0_2.1_2e-1_ablation_20251012_060227.log +Timestamp: 2025-10-12 06:02:27 +===================================== +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 06:02:30,157] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 06:02:32,905] [WARNING] [runner.py:202:fetch_hostfile] Unable to find hostfile, will proceed with training with local resources only. +[2025-10-12 06:02:32,906] [INFO] [runner.py:568:main] cmd = /opt/conda/envs/tinyllava/bin/python3.10 -u -m deepspeed.launcher.launch --world_info=eyJsb2NhbGhvc3QiOiBbMCwgMSwgMiwgMywgNCwgNSwgNiwgN119 --master_addr=127.0.0.1 --master_port=29501 --enable_each_rank_log=None tinyllava/train/train.py --deepspeed ./scripts/zero3.json --data_path /s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json --image_folder /s3-code/ywang29/datasets/tinyllava --is_multimodal True --conv_version qwen2_base --model_name_or_path Qwen/Qwen2.5-0.5B --vision_tower google/siglip-so400m-patch14-384 --vision_tower2 --connector_type mlp2x_gelu --mm_vision_select_layer -2 --image_aspect_ratio square --attn_implementation flash_attention_2 --bf16 True --training_recipe common --tune_type_llm full --tune_type_vision_tower frozen --tune_vision_tower_from_layer 0 --tune_type_connector full --group_by_modality_length True --pretrained_model_path /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain --output_dir /nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.1_2e-1_connector-1.0_2.1_2e-1_ablation --num_train_epochs 1 --per_device_train_batch_size 4 --per_device_eval_batch_size 4 --gradient_accumulation_steps 4 --evaluation_strategy no --learning_rate 2e-1 --weight_decay 0. --warmup_ratio 0.03 --lr_scheduler_type cosine --logging_steps 1 --tf32 False --model_max_length 2048 --gradient_checkpointing True --dataloader_num_workers 8 --lazy_preprocess True --report_to tensorboard --tokenizer_use_fast False --run_name tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune --subnet_mode_text both --subnet_type_text None --mask_type_text soft --init_mean_text 1.0 --temperature_attn_text 2.1 --temperature_mlp_text 2.1 --backward_type_text normal --masked_layers_text all --subnet_mode_vision both --subnet_type_vision None --mask_type_vision soft --init_mean_vision 1.0 --temperature_attn_vision 2.1 --temperature_mlp_vision 2.1 --backward_type_vision normal --masked_layers_vision all --subnet_type_connector global --mask_type_connector soft --init_mean_connector 1.0 --temperature_connector 2.1 --backward_type_connector normal --mm_projector_lr 2e-1 --seed 42 --mask_model llm-connector --save_strategy steps --save_steps 50000 --save_total_limit 1 --train_data_ratio 0.1 +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 06:02:35,475] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 06:02:36,572] [INFO] [launch.py:138:main] 0 NCCL_VERSION=2.21.5 +[2025-10-12 06:02:36,572] [INFO] [launch.py:138:main] 0 NCCL_SOCKET_IFNAME=eth +[2025-10-12 06:02:36,572] [INFO] [launch.py:145:main] WORLD INFO DICT: {'localhost': [0, 1, 2, 3, 4, 5, 6, 7]} +[2025-10-12 06:02:36,572] [INFO] [launch.py:151:main] nnodes=1, num_local_procs=8, node_rank=0 +[2025-10-12 06:02:36,572] [INFO] [launch.py:162:main] global_rank_mapping=defaultdict(, {'localhost': [0, 1, 2, 3, 4, 5, 6, 7]}) +[2025-10-12 06:02:36,572] [INFO] [launch.py:163:main] dist_world_size=8 +[2025-10-12 06:02:36,572] [INFO] [launch.py:165:main] Setting CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 +[2025-10-12 06:02:36,575] [INFO] [launch.py:253:main] process 170055 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=0', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.1_2e-1_connector-1.0_2.1_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '1.0', '--temperature_attn_text', '2.1', '--temperature_mlp_text', '2.1', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '1.0', '--temperature_attn_vision', '2.1', '--temperature_mlp_vision', '2.1', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '1.0', '--temperature_connector', '2.1', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 06:02:36,577] [INFO] [launch.py:253:main] process 170056 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=1', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.1_2e-1_connector-1.0_2.1_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '1.0', '--temperature_attn_text', '2.1', '--temperature_mlp_text', '2.1', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '1.0', '--temperature_attn_vision', '2.1', '--temperature_mlp_vision', '2.1', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '1.0', '--temperature_connector', '2.1', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 06:02:36,579] [INFO] [launch.py:253:main] process 170057 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=2', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.1_2e-1_connector-1.0_2.1_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '1.0', '--temperature_attn_text', '2.1', '--temperature_mlp_text', '2.1', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '1.0', '--temperature_attn_vision', '2.1', '--temperature_mlp_vision', '2.1', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '1.0', '--temperature_connector', '2.1', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 06:02:36,581] [INFO] [launch.py:253:main] process 170058 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=3', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.1_2e-1_connector-1.0_2.1_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '1.0', '--temperature_attn_text', '2.1', '--temperature_mlp_text', '2.1', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '1.0', '--temperature_attn_vision', '2.1', '--temperature_mlp_vision', '2.1', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '1.0', '--temperature_connector', '2.1', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 06:02:36,583] [INFO] [launch.py:253:main] process 170059 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=4', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.1_2e-1_connector-1.0_2.1_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '1.0', '--temperature_attn_text', '2.1', '--temperature_mlp_text', '2.1', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '1.0', '--temperature_attn_vision', '2.1', '--temperature_mlp_vision', '2.1', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '1.0', '--temperature_connector', '2.1', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 06:02:36,586] [INFO] [launch.py:253:main] process 170060 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=5', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.1_2e-1_connector-1.0_2.1_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '1.0', '--temperature_attn_text', '2.1', '--temperature_mlp_text', '2.1', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '1.0', '--temperature_attn_vision', '2.1', '--temperature_mlp_vision', '2.1', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '1.0', '--temperature_connector', '2.1', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 06:02:36,588] [INFO] [launch.py:253:main] process 170061 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=6', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.1_2e-1_connector-1.0_2.1_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '1.0', '--temperature_attn_text', '2.1', '--temperature_mlp_text', '2.1', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '1.0', '--temperature_attn_vision', '2.1', '--temperature_mlp_vision', '2.1', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '1.0', '--temperature_connector', '2.1', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 06:02:36,590] [INFO] [launch.py:253:main] process 170062 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=7', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.1_2e-1_connector-1.0_2.1_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '1.0', '--temperature_attn_text', '2.1', '--temperature_mlp_text', '2.1', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '1.0', '--temperature_attn_vision', '2.1', '--temperature_mlp_vision', '2.1', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '1.0', '--temperature_connector', '2.1', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 06:02:43,188] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 06:02:43,301] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 06:02:43,351] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 06:02:43,351] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 06:02:43,405] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 06:02:43,415] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 06:02:43,418] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 06:02:43,421] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 06:02:43,594] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 06:02:43,594] [INFO] [comm.py:668:init_distributed] Initializing TorchBackend in DeepSpeed with backend nccl +[2025-10-12 06:02:43,721] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 06:02:43,757] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 06:02:43,757] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 06:02:43,808] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 06:02:43,822] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 06:02:43,825] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 06:02:43,835] [INFO] [comm.py:637:init_distributed] cdb=None +Apply masks for the following modules: ['llm', 'connector'] +{'llm': {'model_name_or_path': 'Qwen/Qwen2.5-0.5B', 'cache_dir': None, 'attn_implementation': 'flash_attention_2', 'subnet_mode': 'both', 'subnet_type': 'None', 'sparsity_attn': None, 'sparsity_mlp': None, 'threshold_attn': None, 'threshold_mlp': None, 'temperature_attn': 2.1, 'temperature_mlp': 2.1, 'masked_layers': 'all', 'mask_type': 'soft', 'backward_type': 'normal'}, 'vision_tower': {'model_name_or_path': 'google/siglip-so400m-patch14-384'}, 'connector': {'connector_type': 'mlp2x_gelu', 'subnet_type': 'global', 'threshold': None, 'sparsity': None, 'temperature': 2.1, 'mask_type': 'soft', 'backward_type': 'normal'}} +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +Apply masks for the following modules: ['llm', 'connector'] +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +TinyLlavaConfig { + "backward_type_connector": "normal", + "cache_dir": null, + "connector_type": "mlp2x_gelu", + "hidden_size": 896, + "ignore_index": -100, + "image_aspect_ratio": "square", + "image_token_index": -200, + "llm_model_name_or_path": "Qwen/Qwen2.5-0.5B", + "mask_model": [ + "llm", + "connector" + ], + "mask_type_connector": "soft", + "model_type": "tinyllava", + "num_queries": 128, + "num_resampler_layers": 3, + "pad_token": null, + "resampler_hidden_size": 768, + "sparsity_connector": null, + "subnet_type_connector": "global", + "temperature_connector": 2.1, + "text_config": { + "_name_or_path": "Qwen/Qwen2.5-0.5B", + "architectures": [ + "Qwen2ForCausalLM" + ], + "backward_type": "normal", + "bos_token_id": 151643, + "eos_token_id": 151643, + "hidden_size": 896, + "intermediate_size": 4864, + "mask_type": "soft", + "masked_layers": "all", + "max_position_embeddings": 32768, + "max_window_layers": 24, + "model_type": "qwen2", + "num_attention_heads": 14, + "num_hidden_layers": 24, + "num_key_value_heads": 2, + "rope_theta": 1000000.0, + "sliding_window": 32768, + "subnet_mode": "both", + "subnet_type": "None", + "temperature_attn": 2.1, + "temperature_mlp": 2.1, + "tie_word_embeddings": true, + "torch_dtype": "bfloat16", + "use_mrope": false, + "use_sliding_window": false, + "vocab_size": 151936 + }, + "threshold_connector": null, + "tokenizer_model_max_length": 2048, + "tokenizer_name_or_path": "Qwen/Qwen2.5-0.5B", + "tokenizer_padding_side": "right", + "tokenizer_use_fast": false, + "transformers_version": "4.40.1", + "tune_type_connector": "frozen", + "tune_type_llm": "frozen", + "tune_type_vision_tower": "frozen", + "tune_vision_tower_from_layer": -1, + "use_cache": false, + "vision_config": { + "hidden_act": "gelu_pytorch_tanh", + "hidden_size": 1152, + "image_size": 384, + "intermediate_size": 4304, + "layer_norm_eps": 1e-06, + "model_name_or_path": "google/siglip-so400m-patch14-384", + "model_name_or_path2": "", + "model_type": "siglip_vision_model", + "num_attention_heads": 16, + "num_hidden_layers": 27, + "patch_size": 14 + }, + "vision_feature_layer": -2, + "vision_feature_select_strategy": "patch", + "vision_hidden_size": 1152, + "vision_model_name_or_path": "google/siglip-so400m-patch14-384", + "vision_model_name_or_path2": "", + "vocab_size": 151936 +} + +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test1-worker-0:170055:170055 [0] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:170055:170055 [0] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:170055:170055 [0] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:170055:170055 [0] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:170055:170055 [0] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test1-worker-0:170055:170055 [0] NCCL INFO cudaDriverVersion 12040 +NCCL version 2.21.5+cuda12.1 +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test1-worker-0:170058:170058 [3] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test1-worker-0:170058:170058 [3] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:170058:170058 [3] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:170058:170058 [3] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:170058:170058 [3] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:170058:170058 [3] NCCL INFO NET/Plugin: Using internal network plugin. +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +ywang29-vrdb-test1-worker-0:170055:171623 [0] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:170055:171623 [0] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:170055:171623 [0] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:170055:171623 [0] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:170055:171623 [0] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:170055:171623 [0] NCCL INFO Using network Socket +ywang29-vrdb-test1-worker-0:170058:171624 [3] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:170058:171624 [3] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:170058:171624 [3] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:170058:171624 [3] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:170058:171624 [3] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:170058:171624 [3] NCCL INFO Using network Socket +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test1-worker-0:170062:170062 [7] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test1-worker-0:170062:170062 [7] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:170062:170062 [7] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:170062:170062 [7] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:170062:170062 [7] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:170062:170062 [7] NCCL INFO NET/Plugin: Using internal network plugin. +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +ywang29-vrdb-test1-worker-0:170062:171641 [7] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:170062:171641 [7] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:170062:171641 [7] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:170062:171641 [7] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:170062:171641 [7] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:170062:171641 [7] NCCL INFO Using network Socket +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test1-worker-0:170057:170057 [2] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test1-worker-0:170057:170057 [2] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:170057:170057 [2] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:170057:170057 [2] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:170057:170057 [2] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:170057:170057 [2] NCCL INFO NET/Plugin: Using internal network plugin. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test1-worker-0:170061:170061 [6] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test1-worker-0:170061:170061 [6] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:170061:170061 [6] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:170061:170061 [6] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:170061:170061 [6] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:170061:170061 [6] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test1-worker-0:170057:171642 [2] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:170057:171642 [2] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:170057:171642 [2] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:170057:171642 [2] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:170057:171642 [2] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:170057:171642 [2] NCCL INFO Using network Socket +ywang29-vrdb-test1-worker-0:170061:171643 [6] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:170061:171643 [6] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:170061:171643 [6] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:170061:171643 [6] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:170061:171643 [6] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:170061:171643 [6] NCCL INFO Using network Socket +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test1-worker-0:170060:170060 [5] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test1-worker-0:170060:170060 [5] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:170060:170060 [5] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:170060:170060 [5] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:170060:170060 [5] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:170060:170060 [5] NCCL INFO NET/Plugin: Using internal network plugin. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test1-worker-0:170056:170056 [1] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test1-worker-0:170056:170056 [1] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:170056:170056 [1] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:170056:170056 [1] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:170056:170056 [1] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:170056:170056 [1] NCCL INFO NET/Plugin: Using internal network plugin. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test1-worker-0:170059:170059 [4] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test1-worker-0:170059:170059 [4] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:170059:170059 [4] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:170059:170059 [4] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:170059:170059 [4] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:170059:170059 [4] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test1-worker-0:170060:171644 [5] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:170060:171644 [5] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:170060:171644 [5] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:170060:171644 [5] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:170060:171644 [5] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:170060:171644 [5] NCCL INFO Using network Socket +ywang29-vrdb-test1-worker-0:170056:171645 [1] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:170056:171645 [1] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:170056:171645 [1] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:170056:171645 [1] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:170056:171645 [1] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:170056:171645 [1] NCCL INFO Using network Socket +ywang29-vrdb-test1-worker-0:170059:171646 [4] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:170059:171646 [4] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:170059:171646 [4] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:170059:171646 [4] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:170059:171646 [4] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:170059:171646 [4] NCCL INFO Using network Socket +ywang29-vrdb-test1-worker-0:170059:171646 [4] NCCL INFO ncclCommInitRank comm 0x55b1f111fd40 rank 4 nranks 8 cudaDev 4 nvmlDev 4 busId 901c0 commId 0x28491918895aaf6d - Init START +ywang29-vrdb-test1-worker-0:170060:171644 [5] NCCL INFO ncclCommInitRank comm 0x565418f0a9d0 rank 5 nranks 8 cudaDev 5 nvmlDev 5 busId 901d0 commId 0x28491918895aaf6d - Init START +ywang29-vrdb-test1-worker-0:170061:171643 [6] NCCL INFO ncclCommInitRank comm 0x558889ff5290 rank 6 nranks 8 cudaDev 6 nvmlDev 6 busId a01c0 commId 0x28491918895aaf6d - Init START +ywang29-vrdb-test1-worker-0:170058:171624 [3] NCCL INFO ncclCommInitRank comm 0x55ac282e7ef0 rank 3 nranks 8 cudaDev 3 nvmlDev 3 busId 201d0 commId 0x28491918895aaf6d - Init START +ywang29-vrdb-test1-worker-0:170057:171642 [2] NCCL INFO ncclCommInitRank comm 0x55cd112baa80 rank 2 nranks 8 cudaDev 2 nvmlDev 2 busId 201c0 commId 0x28491918895aaf6d - Init START +ywang29-vrdb-test1-worker-0:170056:171645 [1] NCCL INFO ncclCommInitRank comm 0x5653182ddb60 rank 1 nranks 8 cudaDev 1 nvmlDev 1 busId 101d0 commId 0x28491918895aaf6d - Init START +ywang29-vrdb-test1-worker-0:170055:171623 [0] NCCL INFO ncclCommInitRank comm 0x558d4d4ea250 rank 0 nranks 8 cudaDev 0 nvmlDev 0 busId 101c0 commId 0x28491918895aaf6d - Init START +ywang29-vrdb-test1-worker-0:170062:171641 [7] NCCL INFO ncclCommInitRank comm 0x562599d97c90 rank 7 nranks 8 cudaDev 7 nvmlDev 7 busId a01d0 commId 0x28491918895aaf6d - Init START +ywang29-vrdb-test1-worker-0:170062:171641 [7] NCCL INFO Setting affinity for GPU 7 to ffffff00,0000ffff,ff000000 +ywang29-vrdb-test1-worker-0:170062:171641 [7] NCCL INFO NVLS multicast support is not available on dev 7 +ywang29-vrdb-test1-worker-0:170057:171642 [2] NCCL INFO Setting affinity for GPU 2 to ff,ffff0000,00ffffff +ywang29-vrdb-test1-worker-0:170057:171642 [2] NCCL INFO NVLS multicast support is not available on dev 2 +ywang29-vrdb-test1-worker-0:170058:171624 [3] NCCL INFO Setting affinity for GPU 3 to ff,ffff0000,00ffffff +ywang29-vrdb-test1-worker-0:170058:171624 [3] NCCL INFO NVLS multicast support is not available on dev 3 +ywang29-vrdb-test1-worker-0:170059:171646 [4] NCCL INFO Setting affinity for GPU 4 to ffffff00,0000ffff,ff000000 +ywang29-vrdb-test1-worker-0:170059:171646 [4] NCCL INFO NVLS multicast support is not available on dev 4 +ywang29-vrdb-test1-worker-0:170060:171644 [5] NCCL INFO Setting affinity for GPU 5 to ffffff00,0000ffff,ff000000 +ywang29-vrdb-test1-worker-0:170060:171644 [5] NCCL INFO NVLS multicast support is not available on dev 5 +ywang29-vrdb-test1-worker-0:170055:171623 [0] NCCL INFO Setting affinity for GPU 0 to ff,ffff0000,00ffffff +ywang29-vrdb-test1-worker-0:170055:171623 [0] NCCL INFO NVLS multicast support is not available on dev 0 +ywang29-vrdb-test1-worker-0:170061:171643 [6] NCCL INFO Setting affinity for GPU 6 to ffffff00,0000ffff,ff000000 +ywang29-vrdb-test1-worker-0:170056:171645 [1] NCCL INFO Setting affinity for GPU 1 to ff,ffff0000,00ffffff +ywang29-vrdb-test1-worker-0:170056:171645 [1] NCCL INFO NVLS multicast support is not available on dev 1 +ywang29-vrdb-test1-worker-0:170061:171643 [6] NCCL INFO NVLS multicast support is not available on dev 6 +ywang29-vrdb-test1-worker-0:170061:171643 [6] NCCL INFO comm 0x558889ff5290 rank 6 nRanks 8 nNodes 1 localRanks 8 localRank 6 MNNVL 0 +ywang29-vrdb-test1-worker-0:170060:171644 [5] NCCL INFO comm 0x565418f0a9d0 rank 5 nRanks 8 nNodes 1 localRanks 8 localRank 5 MNNVL 0 +ywang29-vrdb-test1-worker-0:170059:171646 [4] NCCL INFO comm 0x55b1f111fd40 rank 4 nRanks 8 nNodes 1 localRanks 8 localRank 4 MNNVL 0 +ywang29-vrdb-test1-worker-0:170058:171624 [3] NCCL INFO comm 0x55ac282e7ef0 rank 3 nRanks 8 nNodes 1 localRanks 8 localRank 3 MNNVL 0 +ywang29-vrdb-test1-worker-0:170057:171642 [2] NCCL INFO comm 0x55cd112baa80 rank 2 nRanks 8 nNodes 1 localRanks 8 localRank 2 MNNVL 0 +ywang29-vrdb-test1-worker-0:170056:171645 [1] NCCL INFO comm 0x5653182ddb60 rank 1 nRanks 8 nNodes 1 localRanks 8 localRank 1 MNNVL 0 +ywang29-vrdb-test1-worker-0:170055:171623 [0] NCCL INFO comm 0x558d4d4ea250 rank 0 nRanks 8 nNodes 1 localRanks 8 localRank 0 MNNVL 0 +ywang29-vrdb-test1-worker-0:170061:171643 [6] NCCL INFO Trees [0] 7/-1/-1->6->5 [1] 7/-1/-1->6->5 [2] 7/-1/-1->6->5 [3] 7/-1/-1->6->5 [4] 7/-1/-1->6->5 [5] 7/-1/-1->6->5 [6] 7/-1/-1->6->5 [7] 7/-1/-1->6->5 [8] 7/-1/-1->6->5 [9] 7/-1/-1->6->5 [10] 7/-1/-1->6->5 [11] 7/-1/-1->6->5 [12] 7/-1/-1->6->5 [13] 7/-1/-1->6->5 [14] 7/-1/-1->6->5 [15] 7/-1/-1->6->5 [16] 7/-1/-1->6->5 [17] 7/-1/-1->6->5 [18] 7/-1/-1->6->5 [19] 7/-1/-1->6->5 [20] 7/-1/-1->6->5 [21] 7/-1/-1->6->5 [22] 7/-1/-1->6->5 [23] 7/-1/-1->6->5 +ywang29-vrdb-test1-worker-0:170062:171641 [7] NCCL INFO comm 0x562599d97c90 rank 7 nRanks 8 nNodes 1 localRanks 8 localRank 7 MNNVL 0 +ywang29-vrdb-test1-worker-0:170055:171623 [0] NCCL INFO Channel 00/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:170061:171643 [6] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:170055:171623 [0] NCCL INFO Channel 01/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:170059:171646 [4] NCCL INFO Trees [0] 5/-1/-1->4->3 [1] 5/-1/-1->4->3 [2] 5/-1/-1->4->3 [3] 5/-1/-1->4->3 [4] 5/-1/-1->4->3 [5] 5/-1/-1->4->3 [6] 5/-1/-1->4->3 [7] 5/-1/-1->4->3 [8] 5/-1/-1->4->3 [9] 5/-1/-1->4->3 [10] 5/-1/-1->4->3 [11] 5/-1/-1->4->3 [12] 5/-1/-1->4->3 [13] 5/-1/-1->4->3 [14] 5/-1/-1->4->3 [15] 5/-1/-1->4->3 [16] 5/-1/-1->4->3 [17] 5/-1/-1->4->3 [18] 5/-1/-1->4->3 [19] 5/-1/-1->4->3 [20] 5/-1/-1->4->3 [21] 5/-1/-1->4->3 [22] 5/-1/-1->4->3 [23] 5/-1/-1->4->3 +ywang29-vrdb-test1-worker-0:170055:171623 [0] NCCL INFO Channel 02/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:170055:171623 [0] NCCL INFO Channel 03/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:170059:171646 [4] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:170058:171624 [3] NCCL INFO Trees [0] 4/-1/-1->3->2 [1] 4/-1/-1->3->2 [2] 4/-1/-1->3->2 [3] 4/-1/-1->3->2 [4] 4/-1/-1->3->2 [5] 4/-1/-1->3->2 [6] 4/-1/-1->3->2 [7] 4/-1/-1->3->2 [8] 4/-1/-1->3->2 [9] 4/-1/-1->3->2 [10] 4/-1/-1->3->2 [11] 4/-1/-1->3->2 [12] 4/-1/-1->3->2 [13] 4/-1/-1->3->2 [14] 4/-1/-1->3->2 [15] 4/-1/-1->3->2 [16] 4/-1/-1->3->2 [17] 4/-1/-1->3->2 [18] 4/-1/-1->3->2 [19] 4/-1/-1->3->2 [20] 4/-1/-1->3->2 [21] 4/-1/-1->3->2 [22] 4/-1/-1->3->2 [23] 4/-1/-1->3->2 +ywang29-vrdb-test1-worker-0:170055:171623 [0] NCCL INFO Channel 04/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:170055:171623 [0] NCCL INFO Channel 05/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:170057:171642 [2] NCCL INFO Trees [0] 3/-1/-1->2->1 [1] 3/-1/-1->2->1 [2] 3/-1/-1->2->1 [3] 3/-1/-1->2->1 [4] 3/-1/-1->2->1 [5] 3/-1/-1->2->1 [6] 3/-1/-1->2->1 [7] 3/-1/-1->2->1 [8] 3/-1/-1->2->1 [9] 3/-1/-1->2->1 [10] 3/-1/-1->2->1 [11] 3/-1/-1->2->1 [12] 3/-1/-1->2->1 [13] 3/-1/-1->2->1 [14] 3/-1/-1->2->1 [15] 3/-1/-1->2->1 [16] 3/-1/-1->2->1 [17] 3/-1/-1->2->1 [18] 3/-1/-1->2->1 [19] 3/-1/-1->2->1 [20] 3/-1/-1->2->1 [21] 3/-1/-1->2->1 [22] 3/-1/-1->2->1 [23] 3/-1/-1->2->1 +ywang29-vrdb-test1-worker-0:170055:171623 [0] NCCL INFO Channel 06/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:170058:171624 [3] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:170056:171645 [1] NCCL INFO Trees [0] 2/-1/-1->1->0 [1] 2/-1/-1->1->0 [2] 2/-1/-1->1->0 [3] 2/-1/-1->1->0 [4] 2/-1/-1->1->0 [5] 2/-1/-1->1->0 [6] 2/-1/-1->1->0 [7] 2/-1/-1->1->0 [8] 2/-1/-1->1->0 [9] 2/-1/-1->1->0 [10] 2/-1/-1->1->0 [11] 2/-1/-1->1->0 [12] 2/-1/-1->1->0 [13] 2/-1/-1->1->0 [14] 2/-1/-1->1->0 [15] 2/-1/-1->1->0 [16] 2/-1/-1->1->0 [17] 2/-1/-1->1->0 [18] 2/-1/-1->1->0 [19] 2/-1/-1->1->0 [20] 2/-1/-1->1->0 [21] 2/-1/-1->1->0 [22] 2/-1/-1->1->0 [23] 2/-1/-1->1->0 +ywang29-vrdb-test1-worker-0:170055:171623 [0] NCCL INFO Channel 07/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:170057:171642 [2] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:170055:171623 [0] NCCL INFO Channel 08/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:170056:171645 [1] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:170060:171644 [5] NCCL INFO Trees [0] 6/-1/-1->5->4 [1] 6/-1/-1->5->4 [2] 6/-1/-1->5->4 [3] 6/-1/-1->5->4 [4] 6/-1/-1->5->4 [5] 6/-1/-1->5->4 [6] 6/-1/-1->5->4 [7] 6/-1/-1->5->4 [8] 6/-1/-1->5->4 [9] 6/-1/-1->5->4 [10] 6/-1/-1->5->4 [11] 6/-1/-1->5->4 [12] 6/-1/-1->5->4 [13] 6/-1/-1->5->4 [14] 6/-1/-1->5->4 [15] 6/-1/-1->5->4 [16] 6/-1/-1->5->4 [17] 6/-1/-1->5->4 [18] 6/-1/-1->5->4 [19] 6/-1/-1->5->4 [20] 6/-1/-1->5->4 [21] 6/-1/-1->5->4 [22] 6/-1/-1->5->4 [23] 6/-1/-1->5->4 +ywang29-vrdb-test1-worker-0:170055:171623 [0] NCCL INFO Channel 09/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:170055:171623 [0] NCCL INFO Channel 10/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:170055:171623 [0] NCCL INFO Channel 11/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:170060:171644 [5] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:170055:171623 [0] NCCL INFO Channel 12/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:170055:171623 [0] NCCL INFO Channel 13/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:170055:171623 [0] NCCL INFO Channel 14/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:170055:171623 [0] NCCL INFO Channel 15/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:170055:171623 [0] NCCL INFO Channel 16/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:170055:171623 [0] NCCL INFO Channel 17/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:170055:171623 [0] NCCL INFO Channel 18/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:170055:171623 [0] NCCL INFO Channel 19/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:170062:171641 [7] NCCL INFO Trees [0] -1/-1/-1->7->6 [1] -1/-1/-1->7->6 [2] -1/-1/-1->7->6 [3] -1/-1/-1->7->6 [4] -1/-1/-1->7->6 [5] -1/-1/-1->7->6 [6] -1/-1/-1->7->6 [7] -1/-1/-1->7->6 [8] -1/-1/-1->7->6 [9] -1/-1/-1->7->6 [10] -1/-1/-1->7->6 [11] -1/-1/-1->7->6 [12] -1/-1/-1->7->6 [13] -1/-1/-1->7->6 [14] -1/-1/-1->7->6 [15] -1/-1/-1->7->6 [16] -1/-1/-1->7->6 [17] -1/-1/-1->7->6 [18] -1/-1/-1->7->6 [19] -1/-1/-1->7->6 [20] -1/-1/-1->7->6 [21] -1/-1/-1->7->6 [22] -1/-1/-1->7->6 [23] -1/-1/-1->7->6 +ywang29-vrdb-test1-worker-0:170055:171623 [0] NCCL INFO Channel 20/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:170055:171623 [0] NCCL INFO Channel 21/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:170055:171623 [0] NCCL INFO Channel 22/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:170055:171623 [0] NCCL INFO Channel 23/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:170062:171641 [7] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:170055:171623 [0] NCCL INFO Trees [0] 1/-1/-1->0->-1 [1] 1/-1/-1->0->-1 [2] 1/-1/-1->0->-1 [3] 1/-1/-1->0->-1 [4] 1/-1/-1->0->-1 [5] 1/-1/-1->0->-1 [6] 1/-1/-1->0->-1 [7] 1/-1/-1->0->-1 [8] 1/-1/-1->0->-1 [9] 1/-1/-1->0->-1 [10] 1/-1/-1->0->-1 [11] 1/-1/-1->0->-1 [12] 1/-1/-1->0->-1 [13] 1/-1/-1->0->-1 [14] 1/-1/-1->0->-1 [15] 1/-1/-1->0->-1 [16] 1/-1/-1->0->-1 [17] 1/-1/-1->0->-1 [18] 1/-1/-1->0->-1 [19] 1/-1/-1->0->-1 [20] 1/-1/-1->0->-1 [21] 1/-1/-1->0->-1 [22] 1/-1/-1->0->-1 [23] 1/-1/-1->0->-1 +ywang29-vrdb-test1-worker-0:170055:171623 [0] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:170055:171623 [0] NCCL INFO Channel 00/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:171643 [6] NCCL INFO Channel 00/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170055:171623 [0] NCCL INFO Channel 01/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:171643 [6] NCCL INFO Channel 01/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170055:171623 [0] NCCL INFO Channel 02/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:171643 [6] NCCL INFO Channel 02/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170055:171623 [0] NCCL INFO Channel 03/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:171643 [6] NCCL INFO Channel 03/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:171644 [5] NCCL INFO Channel 00/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170055:171623 [0] NCCL INFO Channel 04/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:171643 [6] NCCL INFO Channel 04/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:171644 [5] NCCL INFO Channel 01/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170055:171623 [0] NCCL INFO Channel 05/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:171643 [6] NCCL INFO Channel 05/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:171644 [5] NCCL INFO Channel 02/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170055:171623 [0] NCCL INFO Channel 06/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:171643 [6] NCCL INFO Channel 06/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:171644 [5] NCCL INFO Channel 03/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:171643 [6] NCCL INFO Channel 07/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170055:171623 [0] NCCL INFO Channel 07/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:171644 [5] NCCL INFO Channel 04/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:171643 [6] NCCL INFO Channel 08/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170055:171623 [0] NCCL INFO Channel 08/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:171644 [5] NCCL INFO Channel 05/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:171643 [6] NCCL INFO Channel 09/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170055:171623 [0] NCCL INFO Channel 09/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:171644 [5] NCCL INFO Channel 06/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:171643 [6] NCCL INFO Channel 10/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170055:171623 [0] NCCL INFO Channel 10/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:171644 [5] NCCL INFO Channel 07/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:171643 [6] NCCL INFO Channel 11/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170055:171623 [0] NCCL INFO Channel 11/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:171644 [5] NCCL INFO Channel 08/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:171643 [6] NCCL INFO Channel 12/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170055:171623 [0] NCCL INFO Channel 12/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:171644 [5] NCCL INFO Channel 09/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:171643 [6] NCCL INFO Channel 13/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170055:171623 [0] NCCL INFO Channel 13/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:171644 [5] NCCL INFO Channel 10/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:171643 [6] NCCL INFO Channel 14/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170055:171623 [0] NCCL INFO Channel 14/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:171645 [1] NCCL INFO Channel 00/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:171624 [3] NCCL INFO Channel 00/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:171644 [5] NCCL INFO Channel 11/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:171643 [6] NCCL INFO Channel 15/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170055:171623 [0] NCCL INFO Channel 15/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:171645 [1] NCCL INFO Channel 01/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:171624 [3] NCCL INFO Channel 01/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:171644 [5] NCCL INFO Channel 12/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170055:171623 [0] NCCL INFO Channel 16/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:171643 [6] NCCL INFO Channel 16/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:171645 [1] NCCL INFO Channel 02/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:171624 [3] NCCL INFO Channel 02/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170055:171623 [0] NCCL INFO Channel 17/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:171644 [5] NCCL INFO Channel 13/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:171643 [6] NCCL INFO Channel 17/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:171645 [1] NCCL INFO Channel 03/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:171624 [3] NCCL INFO Channel 03/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170055:171623 [0] NCCL INFO Channel 18/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:171644 [5] NCCL INFO Channel 14/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:171643 [6] NCCL INFO Channel 18/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:171645 [1] NCCL INFO Channel 04/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:171624 [3] NCCL INFO Channel 04/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:171643 [6] NCCL INFO Channel 19/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170055:171623 [0] NCCL INFO Channel 19/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:171644 [5] NCCL INFO Channel 15/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:171645 [1] NCCL INFO Channel 05/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:171624 [3] NCCL INFO Channel 05/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:171643 [6] NCCL INFO Channel 20/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170055:171623 [0] NCCL INFO Channel 20/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:171644 [5] NCCL INFO Channel 16/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:171645 [1] NCCL INFO Channel 06/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:171624 [3] NCCL INFO Channel 06/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:171643 [6] NCCL INFO Channel 21/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170055:171623 [0] NCCL INFO Channel 21/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:171644 [5] NCCL INFO Channel 17/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:171645 [1] NCCL INFO Channel 07/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:171624 [3] NCCL INFO Channel 07/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:171643 [6] NCCL INFO Channel 22/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170055:171623 [0] NCCL INFO Channel 22/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:171644 [5] NCCL INFO Channel 18/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:171645 [1] NCCL INFO Channel 08/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:171624 [3] NCCL INFO Channel 08/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:171646 [4] NCCL INFO Channel 00/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:171643 [6] NCCL INFO Channel 23/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170055:171623 [0] NCCL INFO Channel 23/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:171644 [5] NCCL INFO Channel 19/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:171645 [1] NCCL INFO Channel 09/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:171624 [3] NCCL INFO Channel 09/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:171646 [4] NCCL INFO Channel 01/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:171644 [5] NCCL INFO Channel 20/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:171642 [2] NCCL INFO Channel 00/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:171645 [1] NCCL INFO Channel 10/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:171624 [3] NCCL INFO Channel 10/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:171646 [4] NCCL INFO Channel 02/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:171644 [5] NCCL INFO Channel 21/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:171642 [2] NCCL INFO Channel 01/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:171645 [1] NCCL INFO Channel 11/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:171624 [3] NCCL INFO Channel 11/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:171646 [4] NCCL INFO Channel 03/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:171644 [5] NCCL INFO Channel 22/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:171642 [2] NCCL INFO Channel 02/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:171645 [1] NCCL INFO Channel 12/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:171624 [3] NCCL INFO Channel 12/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:171646 [4] NCCL INFO Channel 04/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:171644 [5] NCCL INFO Channel 23/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:171642 [2] NCCL INFO Channel 03/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:171645 [1] NCCL INFO Channel 13/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:171624 [3] NCCL INFO Channel 13/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:171646 [4] NCCL INFO Channel 05/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:171641 [7] NCCL INFO Channel 00/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:171642 [2] NCCL INFO Channel 04/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:171624 [3] NCCL INFO Channel 14/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:171645 [1] NCCL INFO Channel 14/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:171646 [4] NCCL INFO Channel 06/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:171641 [7] NCCL INFO Channel 01/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:171642 [2] NCCL INFO Channel 05/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:171624 [3] NCCL INFO Channel 15/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:171645 [1] NCCL INFO Channel 15/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:171646 [4] NCCL INFO Channel 07/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:171641 [7] NCCL INFO Channel 02/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:171642 [2] NCCL INFO Channel 06/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:171624 [3] NCCL INFO Channel 16/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:171645 [1] NCCL INFO Channel 16/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:171646 [4] NCCL INFO Channel 08/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:171641 [7] NCCL INFO Channel 03/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:171642 [2] NCCL INFO Channel 07/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:171624 [3] NCCL INFO Channel 17/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:171645 [1] NCCL INFO Channel 17/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:171646 [4] NCCL INFO Channel 09/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:171641 [7] NCCL INFO Channel 04/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:171642 [2] NCCL INFO Channel 08/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:171624 [3] NCCL INFO Channel 18/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:171645 [1] NCCL INFO Channel 18/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:171646 [4] NCCL INFO Channel 10/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:171641 [7] NCCL INFO Channel 05/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:171642 [2] NCCL INFO Channel 09/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:171624 [3] NCCL INFO Channel 19/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:171645 [1] NCCL INFO Channel 19/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:171646 [4] NCCL INFO Channel 11/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:171641 [7] NCCL INFO Channel 06/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:171642 [2] NCCL INFO Channel 10/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:171624 [3] NCCL INFO Channel 20/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:171645 [1] NCCL INFO Channel 20/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:171646 [4] NCCL INFO Channel 12/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:171641 [7] NCCL INFO Channel 07/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:171642 [2] NCCL INFO Channel 11/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:171624 [3] NCCL INFO Channel 21/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:171645 [1] NCCL INFO Channel 21/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:171646 [4] NCCL INFO Channel 13/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:171641 [7] NCCL INFO Channel 08/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:171642 [2] NCCL INFO Channel 12/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:171624 [3] NCCL INFO Channel 22/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:171645 [1] NCCL INFO Channel 22/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:171646 [4] NCCL INFO Channel 14/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:171641 [7] NCCL INFO Channel 09/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:171642 [2] NCCL INFO Channel 13/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:171624 [3] NCCL INFO Channel 23/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:171645 [1] NCCL INFO Channel 23/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:171646 [4] NCCL INFO Channel 15/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:171641 [7] NCCL INFO Channel 10/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:171642 [2] NCCL INFO Channel 14/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:171646 [4] NCCL INFO Channel 16/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:171641 [7] NCCL INFO Channel 11/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:171642 [2] NCCL INFO Channel 15/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:171646 [4] NCCL INFO Channel 17/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:171642 [2] NCCL INFO Channel 16/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:171641 [7] NCCL INFO Channel 12/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:171646 [4] NCCL INFO Channel 18/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:171642 [2] NCCL INFO Channel 17/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:171641 [7] NCCL INFO Channel 13/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:171646 [4] NCCL INFO Channel 19/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:171642 [2] NCCL INFO Channel 18/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:171641 [7] NCCL INFO Channel 14/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:171646 [4] NCCL INFO Channel 20/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:171642 [2] NCCL INFO Channel 19/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:171646 [4] NCCL INFO Channel 21/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:171641 [7] NCCL INFO Channel 15/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:171642 [2] NCCL INFO Channel 20/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:171646 [4] NCCL INFO Channel 22/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:171641 [7] NCCL INFO Channel 16/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:171642 [2] NCCL INFO Channel 21/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:171646 [4] NCCL INFO Channel 23/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:171642 [2] NCCL INFO Channel 22/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:171641 [7] NCCL INFO Channel 17/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:171642 [2] NCCL INFO Channel 23/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:171641 [7] NCCL INFO Channel 18/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:171641 [7] NCCL INFO Channel 19/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:171641 [7] NCCL INFO Channel 20/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:171641 [7] NCCL INFO Channel 21/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:171641 [7] NCCL INFO Channel 22/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:171641 [7] NCCL INFO Channel 23/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:171642 [2] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:170058:171624 [3] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:170056:171645 [1] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:170055:171623 [0] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:170059:171646 [4] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:170062:171641 [7] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:170062:171641 [7] NCCL INFO Channel 00/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:171644 [5] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:170061:171643 [6] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:170062:171641 [7] NCCL INFO Channel 01/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:171641 [7] NCCL INFO Channel 02/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:171641 [7] NCCL INFO Channel 03/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:171641 [7] NCCL INFO Channel 04/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:171641 [7] NCCL INFO Channel 05/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:171641 [7] NCCL INFO Channel 06/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:171641 [7] NCCL INFO Channel 07/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:171641 [7] NCCL INFO Channel 08/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:171641 [7] NCCL INFO Channel 09/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:171641 [7] NCCL INFO Channel 10/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:171641 [7] NCCL INFO Channel 11/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:171642 [2] NCCL INFO Channel 00/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:171624 [3] NCCL INFO Channel 00/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:171641 [7] NCCL INFO Channel 12/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:171642 [2] NCCL INFO Channel 01/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:171624 [3] NCCL INFO Channel 01/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:171642 [2] NCCL INFO Channel 02/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:171624 [3] NCCL INFO Channel 02/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:171642 [2] NCCL INFO Channel 03/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:171624 [3] NCCL INFO Channel 03/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:171642 [2] NCCL INFO Channel 04/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:171624 [3] NCCL INFO Channel 04/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:171642 [2] NCCL INFO Channel 05/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:171624 [3] NCCL INFO Channel 05/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:171642 [2] NCCL INFO Channel 06/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:171624 [3] NCCL INFO Channel 06/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:171642 [2] NCCL INFO Channel 07/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:171624 [3] NCCL INFO Channel 07/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:171642 [2] NCCL INFO Channel 08/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:171645 [1] NCCL INFO Channel 00/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:171624 [3] NCCL INFO Channel 08/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:171642 [2] NCCL INFO Channel 09/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:171645 [1] NCCL INFO Channel 01/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:171624 [3] NCCL INFO Channel 09/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:171642 [2] NCCL INFO Channel 10/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:171645 [1] NCCL INFO Channel 02/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:171624 [3] NCCL INFO Channel 10/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:171642 [2] NCCL INFO Channel 11/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:171645 [1] NCCL INFO Channel 03/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:171624 [3] NCCL INFO Channel 11/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:171641 [7] NCCL INFO Channel 13/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:171642 [2] NCCL INFO Channel 12/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:171645 [1] NCCL INFO Channel 04/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:171624 [3] NCCL INFO Channel 12/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:171641 [7] NCCL INFO Channel 14/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:171643 [6] NCCL INFO Channel 00/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:171642 [2] NCCL INFO Channel 13/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:171644 [5] NCCL INFO Channel 00/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:171645 [1] NCCL INFO Channel 05/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:171624 [3] NCCL INFO Channel 13/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:171641 [7] NCCL INFO Channel 15/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:171643 [6] NCCL INFO Channel 01/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:171642 [2] NCCL INFO Channel 14/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:171644 [5] NCCL INFO Channel 01/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:171645 [1] NCCL INFO Channel 06/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:171624 [3] NCCL INFO Channel 14/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:171641 [7] NCCL INFO Channel 16/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:171643 [6] NCCL INFO Channel 02/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:171642 [2] NCCL INFO Channel 15/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:171644 [5] NCCL INFO Channel 02/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:171645 [1] NCCL INFO Channel 07/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:171624 [3] NCCL INFO Channel 15/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:171641 [7] NCCL INFO Channel 17/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:171643 [6] NCCL INFO Channel 03/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:171642 [2] NCCL INFO Channel 16/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:171644 [5] NCCL INFO Channel 03/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:171645 [1] NCCL INFO Channel 08/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:171624 [3] NCCL INFO Channel 16/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:171643 [6] NCCL INFO Channel 04/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:171641 [7] NCCL INFO Channel 18/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:171642 [2] NCCL INFO Channel 17/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:171644 [5] NCCL INFO Channel 04/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:171645 [1] NCCL INFO Channel 09/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:171624 [3] NCCL INFO Channel 17/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:171643 [6] NCCL INFO Channel 05/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:171642 [2] NCCL INFO Channel 18/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:171645 [1] NCCL INFO Channel 10/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:171624 [3] NCCL INFO Channel 18/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:171643 [6] NCCL INFO Channel 06/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:171646 [4] NCCL INFO Channel 00/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:171642 [2] NCCL INFO Channel 19/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:171645 [1] NCCL INFO Channel 11/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:171624 [3] NCCL INFO Channel 19/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:171643 [6] NCCL INFO Channel 07/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:171646 [4] NCCL INFO Channel 01/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:171642 [2] NCCL INFO Channel 20/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:171645 [1] NCCL INFO Channel 12/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:171624 [3] NCCL INFO Channel 20/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:171643 [6] NCCL INFO Channel 08/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:171646 [4] NCCL INFO Channel 02/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:171642 [2] NCCL INFO Channel 21/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:171645 [1] NCCL INFO Channel 13/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:171641 [7] NCCL INFO Channel 19/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:171624 [3] NCCL INFO Channel 21/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:171643 [6] NCCL INFO Channel 09/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:171646 [4] NCCL INFO Channel 03/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:171642 [2] NCCL INFO Channel 22/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:171645 [1] NCCL INFO Channel 14/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:171641 [7] NCCL INFO Channel 20/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:171624 [3] NCCL INFO Channel 22/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:171643 [6] NCCL INFO Channel 10/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:171642 [2] NCCL INFO Channel 23/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:171646 [4] NCCL INFO Channel 04/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:171645 [1] NCCL INFO Channel 15/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:171641 [7] NCCL INFO Channel 21/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:171624 [3] NCCL INFO Channel 23/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:171643 [6] NCCL INFO Channel 11/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:171644 [5] NCCL INFO Channel 05/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:171646 [4] NCCL INFO Channel 05/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:171645 [1] NCCL INFO Channel 16/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:171641 [7] NCCL INFO Channel 22/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:171643 [6] NCCL INFO Channel 12/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:171644 [5] NCCL INFO Channel 06/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:171646 [4] NCCL INFO Channel 06/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:171645 [1] NCCL INFO Channel 17/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:171641 [7] NCCL INFO Channel 23/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:171643 [6] NCCL INFO Channel 13/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:171644 [5] NCCL INFO Channel 07/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:171646 [4] NCCL INFO Channel 07/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:171645 [1] NCCL INFO Channel 18/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:171643 [6] NCCL INFO Channel 14/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:171644 [5] NCCL INFO Channel 08/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:171646 [4] NCCL INFO Channel 08/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:171645 [1] NCCL INFO Channel 19/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:171643 [6] NCCL INFO Channel 15/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:171644 [5] NCCL INFO Channel 09/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:171646 [4] NCCL INFO Channel 09/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:171645 [1] NCCL INFO Channel 20/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:171643 [6] NCCL INFO Channel 16/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:171644 [5] NCCL INFO Channel 10/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:171646 [4] NCCL INFO Channel 10/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:171645 [1] NCCL INFO Channel 21/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:171643 [6] NCCL INFO Channel 17/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:171644 [5] NCCL INFO Channel 11/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:171646 [4] NCCL INFO Channel 11/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:171645 [1] NCCL INFO Channel 22/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:171643 [6] NCCL INFO Channel 18/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:171644 [5] NCCL INFO Channel 12/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:171646 [4] NCCL INFO Channel 12/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:171645 [1] NCCL INFO Channel 23/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:171643 [6] NCCL INFO Channel 19/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:171644 [5] NCCL INFO Channel 13/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:171646 [4] NCCL INFO Channel 13/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:171643 [6] NCCL INFO Channel 20/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:171644 [5] NCCL INFO Channel 14/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:171646 [4] NCCL INFO Channel 14/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:171643 [6] NCCL INFO Channel 21/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:171644 [5] NCCL INFO Channel 15/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:171646 [4] NCCL INFO Channel 15/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:171643 [6] NCCL INFO Channel 22/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:171644 [5] NCCL INFO Channel 16/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:171646 [4] NCCL INFO Channel 16/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:171643 [6] NCCL INFO Channel 23/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:171644 [5] NCCL INFO Channel 17/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:171646 [4] NCCL INFO Channel 17/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:171646 [4] NCCL INFO Channel 18/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:171644 [5] NCCL INFO Channel 18/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:171644 [5] NCCL INFO Channel 19/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:171646 [4] NCCL INFO Channel 19/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:171644 [5] NCCL INFO Channel 20/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:171646 [4] NCCL INFO Channel 20/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:171646 [4] NCCL INFO Channel 21/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:171644 [5] NCCL INFO Channel 21/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:171644 [5] NCCL INFO Channel 22/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:171646 [4] NCCL INFO Channel 22/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:171646 [4] NCCL INFO Channel 23/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:171644 [5] NCCL INFO Channel 23/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170055:171623 [0] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:170055:171623 [0] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:170055:171623 [0] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:170056:171645 [1] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:170056:171645 [1] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:170056:171645 [1] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:170057:171642 [2] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:170057:171642 [2] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:170057:171642 [2] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:170059:171646 [4] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:170059:171646 [4] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:170058:171624 [3] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:170059:171646 [4] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:170058:171624 [3] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:170058:171624 [3] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:170060:171644 [5] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:170060:171644 [5] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:170060:171644 [5] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:170061:171643 [6] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:170061:171643 [6] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:170061:171643 [6] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:170062:171641 [7] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:170062:171641 [7] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:170062:171641 [7] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:170058:171624 [3] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:170058:171624 [3] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:170058:171624 [3] NCCL INFO ncclCommInitRank comm 0x55ac282e7ef0 rank 3 nranks 8 cudaDev 3 nvmlDev 3 busId 201d0 commId 0x28491918895aaf6d - Init COMPLETE +ywang29-vrdb-test1-worker-0:170062:171641 [7] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:170062:171641 [7] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:170062:171641 [7] NCCL INFO ncclCommInitRank comm 0x562599d97c90 rank 7 nranks 8 cudaDev 7 nvmlDev 7 busId a01d0 commId 0x28491918895aaf6d - Init COMPLETE +ywang29-vrdb-test1-worker-0:170060:171644 [5] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:170060:171644 [5] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:170060:171644 [5] NCCL INFO ncclCommInitRank comm 0x565418f0a9d0 rank 5 nranks 8 cudaDev 5 nvmlDev 5 busId 901d0 commId 0x28491918895aaf6d - Init COMPLETE +ywang29-vrdb-test1-worker-0:170056:171645 [1] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:170056:171645 [1] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:170056:171645 [1] NCCL INFO ncclCommInitRank comm 0x5653182ddb60 rank 1 nranks 8 cudaDev 1 nvmlDev 1 busId 101d0 commId 0x28491918895aaf6d - Init COMPLETE +ywang29-vrdb-test1-worker-0:170059:171646 [4] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:170059:171646 [4] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:170059:171646 [4] NCCL INFO ncclCommInitRank comm 0x55b1f111fd40 rank 4 nranks 8 cudaDev 4 nvmlDev 4 busId 901c0 commId 0x28491918895aaf6d - Init COMPLETE +ywang29-vrdb-test1-worker-0:170055:171623 [0] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:170055:171623 [0] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:170055:171623 [0] NCCL INFO ncclCommInitRank comm 0x558d4d4ea250 rank 0 nranks 8 cudaDev 0 nvmlDev 0 busId 101c0 commId 0x28491918895aaf6d - Init COMPLETE +ywang29-vrdb-test1-worker-0:170057:171642 [2] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:170057:171642 [2] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:170057:171642 [2] NCCL INFO ncclCommInitRank comm 0x55cd112baa80 rank 2 nranks 8 cudaDev 2 nvmlDev 2 busId 201c0 commId 0x28491918895aaf6d - Init COMPLETE +ywang29-vrdb-test1-worker-0:170061:171643 [6] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:170061:171643 [6] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:170061:171643 [6] NCCL INFO ncclCommInitRank comm 0x558889ff5290 rank 6 nranks 8 cudaDev 6 nvmlDev 6 busId a01c0 commId 0x28491918895aaf6d - Init COMPLETE +[2025-10-12 06:03:28,859] [INFO] [partition_parameters.py:343:__exit__] finished initializing model - num_params = 459, num_elems = 0.99B +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.laSome weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +loading language model from yers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. + /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +[2025-10-12 07:32:49,794] [INFO] [partition_parameters.py:343:__exit__] finished initializing model - num_params = 907, num_elems = 1.42B +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +TinyLlavaForConditionalGeneration( + (language_model): Qwen2ForCausalLM( + (model): Qwen2Model( + (embed_tokens): Embedding(151936, 896) + (layers): ModuleList( + (0-23): 24 x Qwen2DecoderLayer( + (self_attn): Qwen2FlashAttention2( + (q_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=True) + (k_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=128, bias=True) + (v_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=128, bias=True) + (o_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=False) + (rotary_emb): Qwen2RotaryEmbedding() + ) + (mlp): Qwen2MLP( + (gate_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=4864, bias=False) + (up_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=4864, bias=False) + (down_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=4864, out_features=896, bias=False) + (act_fn): SiLU() + ) + (input_layernorm): Qwen2RMSNorm() + (post_attention_layernorm): Qwen2RMSNorm() + ) + ) + (norm): Qwen2RMSNorm() + ) + (lm_head): Linear(in_features=896, out_features=151936, bias=False) + ) + (vision_tower): SIGLIPVisionTower( + (_vision_tower): SiglipVisionModel( + (vision_model): SiglipVisionTransformer( + (embeddings): SiglipVisionEmbeddings( + (patch_embedding): Conv2d(3, 1152, kernel_size=(14, 14), stride=(14, 14), padding=valid) + (position_embedding): Embedding(729, 1152) + ) + (encoder): SiglipEncoder( + (layers): ModuleList( + (0-26): 27 x SiglipEncoderLayer( + (self_attn): SiglipAttention( + (k_proj): Linear(in_features=1152, out_features=1152, bias=True) + (v_proj): Linear(in_features=1152, out_features=1152, bias=True) + (q_proj): Linear(in_features=1152, out_features=1152, bias=True) + (out_proj): Linear(in_features=1152, out_features=1152, bias=True) + ) + (layer_norm1): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + (layer_norm2): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + ) + ) + ) + (post_layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (head): SiglipMultiheadAttentionPoolingHead( + (attention): MultiheadAttention( + (out_proj): NonDynamicallyQuantizableLinear(in_features=1152, out_features=1152, bias=True) + ) + (layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + ) + ) + ) + ) + (connector): MLPConnector( + (_connector): Sequential( + (0): SupermaskLinearSparsity_SoftForward_Normal(in_features=1152, out_features=896, bias=True) + (1): GELU(approximate='none') + (2): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=True) + ) + ) +) +Pre-training init language_model.model.layers.0.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.0.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.0.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.0.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.0.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.0.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.0.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.1.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.1.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.1.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.1.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.1.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.1.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.1.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.2.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.2.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.2.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.2.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.2.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.2.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.2.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.3.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.3.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.3.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.3.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.3.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.3.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.3.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.4.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.4.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.4.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.4.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.4.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.4.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.4.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.5.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.5.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.5.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.5.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.5.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.5.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.5.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.6.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.6.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.6.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.6.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.6.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.6.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.6.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.7.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.7.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.7.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.7.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.7.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.7.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.7.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.8.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.8.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.8.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.8.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.8.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.8.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.8.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.9.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.9.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.9.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.9.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.9.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.9.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.9.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.10.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.10.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.10.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.10.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.10.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.10.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.10.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.11.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.11.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.11.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.11.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.11.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.11.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.11.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.12.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.12.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.12.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.12.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.12.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.12.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.12.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.13.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.13.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.13.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.13.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.13.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.13.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.13.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.14.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.14.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.14.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.14.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.14.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.14.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.14.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.15.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.15.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.15.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.15.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.15.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.15.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.15.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.16.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.16.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.16.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.16.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.16.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.16.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.16.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.17.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.17.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.17.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.17.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.17.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.17.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.17.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.18.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.18.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.18.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.18.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.18.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.18.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.18.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.19.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.19.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.19.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.19.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.19.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.19.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.19.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.20.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.20.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.20.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.20.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.20.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.20.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.20.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.21.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.21.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.21.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.21.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.21.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.21.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.21.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.22.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.22.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.22.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.22.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.22.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.22.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.22.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.23.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.23.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.23.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.23.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.23.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.23.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.23.mlp.down_proj.scores: Mean=1.000000 +Pre-training init connector._connector.0.scores: Mean=1.000005 +Pre-training init connector._connector.2.scores: Mean=0.999970 +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +2025-10-12 07:33:07,626 | INFO: Total Parameters: 1283756736, Total Trainable Parameters: 359661568 +2025-10-12 07:33:07,630 | INFO: Trainable Parameters: +language_model.model.layers.0.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.0.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.0.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.0.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.0.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.0.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.0.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.1.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.1.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.1.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.1.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.1.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.1.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.1.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.2.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.2.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.2.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.2.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.2.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.2.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.2.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.3.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.3.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.3.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.3.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.3.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.3.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.3.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.4.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.4.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.4.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.4.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.4.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.4.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.4.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.5.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.5.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.5.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.5.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.5.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.5.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.5.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.6.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.6.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.6.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.6.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.6.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.6.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.6.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.7.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.7.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.7.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.7.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.7.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.7.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.7.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.8.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.8.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.8.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.8.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.8.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.8.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.8.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.9.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.9.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.9.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.9.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.9.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.9.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.9.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.10.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.10.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.10.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.10.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.10.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.10.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.10.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.11.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.11.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.11.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.11.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.11.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.11.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.11.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.12.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.12.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.12.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.12.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.12.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.12.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.12.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.13.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.13.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.13.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.13.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.13.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.13.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.13.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.14.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.14.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.14.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.14.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.14.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.14.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.14.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.15.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.15.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.15.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.15.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.15.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.15.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.15.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.16.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.16.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.16.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.16.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.16.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.16.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.16.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.17.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.17.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.17.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.17.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.17.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.17.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.17.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.18.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.18.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.18.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.18.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.18.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.18.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.18.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.19.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.19.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.19.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.19.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.19.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.19.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.19.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.20.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.20.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.20.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.20.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.20.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.20.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.20.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.21.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.21.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.21.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.21.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.21.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.21.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.21.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.22.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.22.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.22.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.22.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.22.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.22.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.22.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.23.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.23.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.23.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.23.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.23.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.23.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.23.mlp.down_proj.scores: 4358144 parameters +connector._connector.0.scores: 1032192 parameters +connector._connector.2.scores: 802816 parameters +Parameter Offload: Total persistent parameters: 486464 in 403 params + 0%| | 0/520 [00:003->2 [1] 4/-1/-1->3->2 [2] 4/-1/-1->3->2 [3] 4/-1/-1->3->2 [4] 4/-1/-1->3->2 [5] 4/-1/-1->3->2 [6] 4/-1/-1->3->2 [7] 4/-1/-1->3->2 [8] 4/-1/-1->3->2 [9] 4/-1/-1->3->2 [10] 4/-1/-1->3->2 [11] 4/-1/-1->3->2 [12] 4/-1/-1->3->2 [13] 4/-1/-1->3->2 [14] 4/-1/-1->3->2 [15] 4/-1/-1->3->2 [16] 4/-1/-1->3->2 [17] 4/-1/-1->3->2 [18] 4/-1/-1->3->2 [19] 4/-1/-1->3->2 [20] 4/-1/-1->3->2 [21] 4/-1/-1->3->2 [22] 4/-1/-1->3->2 [23] 4/-1/-1->3->2 +ywang29-vrdb-test1-worker-0:170058:178022 [3] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:170055:178021 [0] NCCL INFO Channel 00/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:170057:178023 [2] NCCL INFO Trees [0] 3/-1/-1->2->1 [1] 3/-1/-1->2->1 [2] 3/-1/-1->2->1 [3] 3/-1/-1->2->1 [4] 3/-1/-1->2->1 [5] 3/-1/-1->2->1 [6] 3/-1/-1->2->1 [7] 3/-1/-1->2->1 [8] 3/-1/-1->2->1 [9] 3/-1/-1->2->1 [10] 3/-1/-1->2->1 [11] 3/-1/-1->2->1 [12] 3/-1/-1->2->1 [13] 3/-1/-1->2->1 [14] 3/-1/-1->2->1 [15] 3/-1/-1->2->1 [16] 3/-1/-1->2->1 [17] 3/-1/-1->2->1 [18] 3/-1/-1->2->1 [19] 3/-1/-1->2->1 [20] 3/-1/-1->2->1 [21] 3/-1/-1->2->1 [22] 3/-1/-1->2->1 [23] 3/-1/-1->2->1 +ywang29-vrdb-test1-worker-0:170055:178021 [0] NCCL INFO Channel 01/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:170057:178023 [2] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:170055:178021 [0] NCCL INFO Channel 02/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:170055:178021 [0] NCCL INFO Channel 03/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:170061:178025 [6] NCCL INFO Trees [0] 7/-1/-1->6->5 [1] 7/-1/-1->6->5 [2] 7/-1/-1->6->5 [3] 7/-1/-1->6->5 [4] 7/-1/-1->6->5 [5] 7/-1/-1->6->5 [6] 7/-1/-1->6->5 [7] 7/-1/-1->6->5 [8] 7/-1/-1->6->5 [9] 7/-1/-1->6->5 [10] 7/-1/-1->6->5 [11] 7/-1/-1->6->5 [12] 7/-1/-1->6->5 [13] 7/-1/-1->6->5 [14] 7/-1/-1->6->5 [15] 7/-1/-1->6->5 [16] 7/-1/-1->6->5 [17] 7/-1/-1->6->5 [18] 7/-1/-1->6->5 [19] 7/-1/-1->6->5 [20] 7/-1/-1->6->5 [21] 7/-1/-1->6->5 [22] 7/-1/-1->6->5 [23] 7/-1/-1->6->5 +ywang29-vrdb-test1-worker-0:170055:178021 [0] NCCL INFO Channel 04/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:170060:178026 [5] NCCL INFO Trees [0] 6/-1/-1->5->4 [1] 6/-1/-1->5->4 [2] 6/-1/-1->5->4 [3] 6/-1/-1->5->4 [4] 6/-1/-1->5->4 [5] 6/-1/-1->5->4 [6] 6/-1/-1->5->4 [7] 6/-1/-1->5->4 [8] 6/-1/-1->5->4 [9] 6/-1/-1->5->4 [10] 6/-1/-1->5->4 [11] 6/-1/-1->5->4 [12] 6/-1/-1->5->4 [13] 6/-1/-1->5->4 [14] 6/-1/-1->5->4 [15] 6/-1/-1->5->4 [16] 6/-1/-1->5->4 [17] 6/-1/-1->5->4 [18] 6/-1/-1->5->4 [19] 6/-1/-1->5->4 [20] 6/-1/-1->5->4 [21] 6/-1/-1->5->4 [22] 6/-1/-1->5->4 [23] 6/-1/-1->5->4 +ywang29-vrdb-test1-worker-0:170056:178028 [1] NCCL INFO Trees [0] 2/-1/-1->1->0 [1] 2/-1/-1->1->0 [2] 2/-1/-1->1->0 [3] 2/-1/-1->1->0 [4] 2/-1/-1->1->0 [5] 2/-1/-1->1->0 [6] 2/-1/-1->1->0 [7] 2/-1/-1->1->0 [8] 2/-1/-1->1->0 [9] 2/-1/-1->1->0 [10] 2/-1/-1->1->0 [11] 2/-1/-1->1->0 [12] 2/-1/-1->1->0 [13] 2/-1/-1->1->0 [14] 2/-1/-1->1->0 [15] 2/-1/-1->1->0 [16] 2/-1/-1->1->0 [17] 2/-1/-1->1->0 [18] 2/-1/-1->1->0 [19] 2/-1/-1->1->0 [20] 2/-1/-1->1->0 [21] 2/-1/-1->1->0 [22] 2/-1/-1->1->0 [23] 2/-1/-1->1->0 +ywang29-vrdb-test1-worker-0:170061:178025 [6] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:170055:178021 [0] NCCL INFO Channel 05/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:170060:178026 [5] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:170056:178028 [1] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:170055:178021 [0] NCCL INFO Channel 06/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:170055:178021 [0] NCCL INFO Channel 07/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:170055:178021 [0] NCCL INFO Channel 08/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:170055:178021 [0] NCCL INFO Channel 09/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:170055:178021 [0] NCCL INFO Channel 10/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:170055:178021 [0] NCCL INFO Channel 11/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:170055:178021 [0] NCCL INFO Channel 12/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:170055:178021 [0] NCCL INFO Channel 13/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:170055:178021 [0] NCCL INFO Channel 14/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:170059:178024 [4] NCCL INFO Trees [0] 5/-1/-1->4->3 [1] 5/-1/-1->4->3 [2] 5/-1/-1->4->3 [3] 5/-1/-1->4->3 [4] 5/-1/-1->4->3 [5] 5/-1/-1->4->3 [6] 5/-1/-1->4->3 [7] 5/-1/-1->4->3 [8] 5/-1/-1->4->3 [9] 5/-1/-1->4->3 [10] 5/-1/-1->4->3 [11] 5/-1/-1->4->3 [12] 5/-1/-1->4->3 [13] 5/-1/-1->4->3 [14] 5/-1/-1->4->3 [15] 5/-1/-1->4->3 [16] 5/-1/-1->4->3 [17] 5/-1/-1->4->3 [18] 5/-1/-1->4->3 [19] 5/-1/-1->4->3 [20] 5/-1/-1->4->3 [21] 5/-1/-1->4->3 [22] 5/-1/-1->4->3 [23] 5/-1/-1->4->3 +ywang29-vrdb-test1-worker-0:170055:178021 [0] NCCL INFO Channel 15/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:170055:178021 [0] NCCL INFO Channel 16/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:170059:178024 [4] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:170062:178027 [7] NCCL INFO Trees [0] -1/-1/-1->7->6 [1] -1/-1/-1->7->6 [2] -1/-1/-1->7->6 [3] -1/-1/-1->7->6 [4] -1/-1/-1->7->6 [5] -1/-1/-1->7->6 [6] -1/-1/-1->7->6 [7] -1/-1/-1->7->6 [8] -1/-1/-1->7->6 [9] -1/-1/-1->7->6 [10] -1/-1/-1->7->6 [11] -1/-1/-1->7->6 [12] -1/-1/-1->7->6 [13] -1/-1/-1->7->6 [14] -1/-1/-1->7->6 [15] -1/-1/-1->7->6 [16] -1/-1/-1->7->6 [17] -1/-1/-1->7->6 [18] -1/-1/-1->7->6 [19] -1/-1/-1->7->6 [20] -1/-1/-1->7->6 [21] -1/-1/-1->7->6 [22] -1/-1/-1->7->6 [23] -1/-1/-1->7->6 +ywang29-vrdb-test1-worker-0:170055:178021 [0] NCCL INFO Channel 17/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:170055:178021 [0] NCCL INFO Channel 18/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:170062:178027 [7] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:170055:178021 [0] NCCL INFO Channel 19/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:170055:178021 [0] NCCL INFO Channel 20/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:170055:178021 [0] NCCL INFO Channel 21/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:170055:178021 [0] NCCL INFO Channel 22/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:170055:178021 [0] NCCL INFO Channel 23/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:170055:178021 [0] NCCL INFO Trees [0] 1/-1/-1->0->-1 [1] 1/-1/-1->0->-1 [2] 1/-1/-1->0->-1 [3] 1/-1/-1->0->-1 [4] 1/-1/-1->0->-1 [5] 1/-1/-1->0->-1 [6] 1/-1/-1->0->-1 [7] 1/-1/-1->0->-1 [8] 1/-1/-1->0->-1 [9] 1/-1/-1->0->-1 [10] 1/-1/-1->0->-1 [11] 1/-1/-1->0->-1 [12] 1/-1/-1->0->-1 [13] 1/-1/-1->0->-1 [14] 1/-1/-1->0->-1 [15] 1/-1/-1->0->-1 [16] 1/-1/-1->0->-1 [17] 1/-1/-1->0->-1 [18] 1/-1/-1->0->-1 [19] 1/-1/-1->0->-1 [20] 1/-1/-1->0->-1 [21] 1/-1/-1->0->-1 [22] 1/-1/-1->0->-1 [23] 1/-1/-1->0->-1 +ywang29-vrdb-test1-worker-0:170055:178021 [0] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:170057:178023 [2] NCCL INFO Channel 00/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:178026 [5] NCCL INFO Channel 00/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:178028 [1] NCCL INFO Channel 00/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:178025 [6] NCCL INFO Channel 00/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:178023 [2] NCCL INFO Channel 01/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:178026 [5] NCCL INFO Channel 01/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170055:178021 [0] NCCL INFO Channel 00/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:178028 [1] NCCL INFO Channel 01/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:178025 [6] NCCL INFO Channel 01/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:178023 [2] NCCL INFO Channel 02/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:178026 [5] NCCL INFO Channel 02/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170055:178021 [0] NCCL INFO Channel 01/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:178028 [1] NCCL INFO Channel 02/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:178025 [6] NCCL INFO Channel 02/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:178023 [2] NCCL INFO Channel 03/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:178026 [5] NCCL INFO Channel 03/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170055:178021 [0] NCCL INFO Channel 02/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:178028 [1] NCCL INFO Channel 03/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:178025 [6] NCCL INFO Channel 03/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:178023 [2] NCCL INFO Channel 04/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:178026 [5] NCCL INFO Channel 04/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:178024 [4] NCCL INFO Channel 00/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170055:178021 [0] NCCL INFO Channel 03/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:178027 [7] NCCL INFO Channel 00/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:178028 [1] NCCL INFO Channel 04/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:178025 [6] NCCL INFO Channel 04/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:178023 [2] NCCL INFO Channel 05/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:178026 [5] NCCL INFO Channel 05/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:178024 [4] NCCL INFO Channel 01/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170055:178021 [0] NCCL INFO Channel 04/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:178028 [1] NCCL INFO Channel 05/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:178027 [7] NCCL INFO Channel 01/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:178025 [6] NCCL INFO Channel 05/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:178023 [2] NCCL INFO Channel 06/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:178026 [5] NCCL INFO Channel 06/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:178024 [4] NCCL INFO Channel 02/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170055:178021 [0] NCCL INFO Channel 05/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:178028 [1] NCCL INFO Channel 06/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:178025 [6] NCCL INFO Channel 06/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:178027 [7] NCCL INFO Channel 02/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:178023 [2] NCCL INFO Channel 07/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:178026 [5] NCCL INFO Channel 07/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:178024 [4] NCCL INFO Channel 03/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170055:178021 [0] NCCL INFO Channel 06/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:178028 [1] NCCL INFO Channel 07/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:178025 [6] NCCL INFO Channel 07/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:178027 [7] NCCL INFO Channel 03/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:178023 [2] NCCL INFO Channel 08/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:178026 [5] NCCL INFO Channel 08/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:178024 [4] NCCL INFO Channel 04/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170055:178021 [0] NCCL INFO Channel 07/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:178028 [1] NCCL INFO Channel 08/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:178025 [6] NCCL INFO Channel 08/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:178027 [7] NCCL INFO Channel 04/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:178023 [2] NCCL INFO Channel 09/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:178026 [5] NCCL INFO Channel 09/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:178024 [4] NCCL INFO Channel 05/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170055:178021 [0] NCCL INFO Channel 08/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:178028 [1] NCCL INFO Channel 09/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:178025 [6] NCCL INFO Channel 09/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:178027 [7] NCCL INFO Channel 05/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:178023 [2] NCCL INFO Channel 10/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:178026 [5] NCCL INFO Channel 10/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170055:178021 [0] NCCL INFO Channel 09/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:178024 [4] NCCL INFO Channel 06/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:178028 [1] NCCL INFO Channel 10/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:178025 [6] NCCL INFO Channel 10/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:178023 [2] NCCL INFO Channel 11/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:178027 [7] NCCL INFO Channel 06/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:178026 [5] NCCL INFO Channel 11/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170055:178021 [0] NCCL INFO Channel 10/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:178024 [4] NCCL INFO Channel 07/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:178028 [1] NCCL INFO Channel 11/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:178025 [6] NCCL INFO Channel 11/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:178023 [2] NCCL INFO Channel 12/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:178027 [7] NCCL INFO Channel 07/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:178026 [5] NCCL INFO Channel 12/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170055:178021 [0] NCCL INFO Channel 11/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:178024 [4] NCCL INFO Channel 08/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:178028 [1] NCCL INFO Channel 12/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:178025 [6] NCCL INFO Channel 12/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:178023 [2] NCCL INFO Channel 13/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:178026 [5] NCCL INFO Channel 13/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:178027 [7] NCCL INFO Channel 08/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170055:178021 [0] NCCL INFO Channel 12/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:178024 [4] NCCL INFO Channel 09/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:178025 [6] NCCL INFO Channel 13/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:178028 [1] NCCL INFO Channel 13/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:178023 [2] NCCL INFO Channel 14/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:178026 [5] NCCL INFO Channel 14/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:178027 [7] NCCL INFO Channel 09/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170055:178021 [0] NCCL INFO Channel 13/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:178024 [4] NCCL INFO Channel 10/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:178025 [6] NCCL INFO Channel 14/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:178028 [1] NCCL INFO Channel 14/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:178026 [5] NCCL INFO Channel 15/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:178027 [7] NCCL INFO Channel 10/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:178023 [2] NCCL INFO Channel 15/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170055:178021 [0] NCCL INFO Channel 14/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:178024 [4] NCCL INFO Channel 11/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:178025 [6] NCCL INFO Channel 15/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:178028 [1] NCCL INFO Channel 15/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:178026 [5] NCCL INFO Channel 16/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:178027 [7] NCCL INFO Channel 11/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:178023 [2] NCCL INFO Channel 16/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170055:178021 [0] NCCL INFO Channel 15/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:178022 [3] NCCL INFO Channel 00/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:178024 [4] NCCL INFO Channel 12/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:178025 [6] NCCL INFO Channel 16/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:178028 [1] NCCL INFO Channel 16/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:178026 [5] NCCL INFO Channel 17/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:178023 [2] NCCL INFO Channel 17/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:178027 [7] NCCL INFO Channel 12/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170055:178021 [0] NCCL INFO Channel 16/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:178022 [3] NCCL INFO Channel 01/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:178024 [4] NCCL INFO Channel 13/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:178025 [6] NCCL INFO Channel 17/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:178028 [1] NCCL INFO Channel 17/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:178026 [5] NCCL INFO Channel 18/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:178023 [2] NCCL INFO Channel 18/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:178027 [7] NCCL INFO Channel 13/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170055:178021 [0] NCCL INFO Channel 17/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:178022 [3] NCCL INFO Channel 02/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:178025 [6] NCCL INFO Channel 18/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:178024 [4] NCCL INFO Channel 14/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:178028 [1] NCCL INFO Channel 18/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:178026 [5] NCCL INFO Channel 19/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:178023 [2] NCCL INFO Channel 19/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170055:178021 [0] NCCL INFO Channel 18/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:178027 [7] NCCL INFO Channel 14/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:178022 [3] NCCL INFO Channel 03/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:178025 [6] NCCL INFO Channel 19/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:178024 [4] NCCL INFO Channel 15/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:178028 [1] NCCL INFO Channel 19/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:178026 [5] NCCL INFO Channel 20/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:178023 [2] NCCL INFO Channel 20/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170055:178021 [0] NCCL INFO Channel 19/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:178027 [7] NCCL INFO Channel 15/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:178022 [3] NCCL INFO Channel 04/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:178024 [4] NCCL INFO Channel 16/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:178028 [1] NCCL INFO Channel 20/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:178025 [6] NCCL INFO Channel 20/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:178026 [5] NCCL INFO Channel 21/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:178023 [2] NCCL INFO Channel 21/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170055:178021 [0] NCCL INFO Channel 20/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:178027 [7] NCCL INFO Channel 16/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:178022 [3] NCCL INFO Channel 05/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:178024 [4] NCCL INFO Channel 17/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:178028 [1] NCCL INFO Channel 21/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:178025 [6] NCCL INFO Channel 21/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:178026 [5] NCCL INFO Channel 22/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:178023 [2] NCCL INFO Channel 22/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170055:178021 [0] NCCL INFO Channel 21/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:178022 [3] NCCL INFO Channel 06/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:178027 [7] NCCL INFO Channel 17/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:178024 [4] NCCL INFO Channel 18/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:178028 [1] NCCL INFO Channel 22/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:178025 [6] NCCL INFO Channel 22/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:178026 [5] NCCL INFO Channel 23/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:178023 [2] NCCL INFO Channel 23/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170055:178021 [0] NCCL INFO Channel 22/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:178022 [3] NCCL INFO Channel 07/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:178024 [4] NCCL INFO Channel 19/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:178027 [7] NCCL INFO Channel 18/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:178025 [6] NCCL INFO Channel 23/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:178028 [1] NCCL INFO Channel 23/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170055:178021 [0] NCCL INFO Channel 23/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:178022 [3] NCCL INFO Channel 08/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:178024 [4] NCCL INFO Channel 20/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:178027 [7] NCCL INFO Channel 19/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:178022 [3] NCCL INFO Channel 09/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:178024 [4] NCCL INFO Channel 21/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:178027 [7] NCCL INFO Channel 20/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:178022 [3] NCCL INFO Channel 10/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:178024 [4] NCCL INFO Channel 22/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:178027 [7] NCCL INFO Channel 21/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:178022 [3] NCCL INFO Channel 11/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:178024 [4] NCCL INFO Channel 23/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:178027 [7] NCCL INFO Channel 22/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:178022 [3] NCCL INFO Channel 12/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:178027 [7] NCCL INFO Channel 23/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:178022 [3] NCCL INFO Channel 13/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:178022 [3] NCCL INFO Channel 14/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:178022 [3] NCCL INFO Channel 15/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:178022 [3] NCCL INFO Channel 16/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:178022 [3] NCCL INFO Channel 17/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:178022 [3] NCCL INFO Channel 18/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:178022 [3] NCCL INFO Channel 19/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:178022 [3] NCCL INFO Channel 20/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:178022 [3] NCCL INFO Channel 21/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:178022 [3] NCCL INFO Channel 22/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:178022 [3] NCCL INFO Channel 23/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:178023 [2] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:170058:178022 [3] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:170059:178024 [4] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:170056:178028 [1] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:170055:178021 [0] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:170062:178027 [7] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:170062:178027 [7] NCCL INFO Channel 00/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:178026 [5] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:170061:178025 [6] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:170062:178027 [7] NCCL INFO Channel 01/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:178027 [7] NCCL INFO Channel 02/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:178027 [7] NCCL INFO Channel 03/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:178027 [7] NCCL INFO Channel 04/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:178027 [7] NCCL INFO Channel 05/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:178027 [7] NCCL INFO Channel 06/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:178027 [7] NCCL INFO Channel 07/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:178027 [7] NCCL INFO Channel 08/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:178027 [7] NCCL INFO Channel 09/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:178027 [7] NCCL INFO Channel 10/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:178027 [7] NCCL INFO Channel 11/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:178027 [7] NCCL INFO Channel 12/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:178027 [7] NCCL INFO Channel 13/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:178024 [4] NCCL INFO Channel 00/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:178027 [7] NCCL INFO Channel 14/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:178024 [4] NCCL INFO Channel 01/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:178027 [7] NCCL INFO Channel 15/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:178023 [2] NCCL INFO Channel 00/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:178024 [4] NCCL INFO Channel 02/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:178027 [7] NCCL INFO Channel 16/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:178023 [2] NCCL INFO Channel 01/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:178024 [4] NCCL INFO Channel 03/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:178027 [7] NCCL INFO Channel 17/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:178023 [2] NCCL INFO Channel 02/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:178024 [4] NCCL INFO Channel 04/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:178027 [7] NCCL INFO Channel 18/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:178023 [2] NCCL INFO Channel 03/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:178024 [4] NCCL INFO Channel 05/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:178027 [7] NCCL INFO Channel 19/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:178023 [2] NCCL INFO Channel 04/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:178024 [4] NCCL INFO Channel 06/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:178027 [7] NCCL INFO Channel 20/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:178023 [2] NCCL INFO Channel 05/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:178024 [4] NCCL INFO Channel 07/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:178022 [3] NCCL INFO Channel 00/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:178027 [7] NCCL INFO Channel 21/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:178023 [2] NCCL INFO Channel 06/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:178024 [4] NCCL INFO Channel 08/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:178022 [3] NCCL INFO Channel 01/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:178027 [7] NCCL INFO Channel 22/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:178023 [2] NCCL INFO Channel 07/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:178024 [4] NCCL INFO Channel 09/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:178022 [3] NCCL INFO Channel 02/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170062:178027 [7] NCCL INFO Channel 23/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:178023 [2] NCCL INFO Channel 08/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:178024 [4] NCCL INFO Channel 10/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:178022 [3] NCCL INFO Channel 03/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:178025 [6] NCCL INFO Channel 00/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:178028 [1] NCCL INFO Channel 00/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:178023 [2] NCCL INFO Channel 09/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:178024 [4] NCCL INFO Channel 11/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:178022 [3] NCCL INFO Channel 04/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:178025 [6] NCCL INFO Channel 01/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:178028 [1] NCCL INFO Channel 01/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:178023 [2] NCCL INFO Channel 10/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:178024 [4] NCCL INFO Channel 12/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:178022 [3] NCCL INFO Channel 05/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:178028 [1] NCCL INFO Channel 02/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:178025 [6] NCCL INFO Channel 02/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:178023 [2] NCCL INFO Channel 11/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:178024 [4] NCCL INFO Channel 13/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:178022 [3] NCCL INFO Channel 06/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:178028 [1] NCCL INFO Channel 03/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:178025 [6] NCCL INFO Channel 03/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:178023 [2] NCCL INFO Channel 12/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:178024 [4] NCCL INFO Channel 14/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:178022 [3] NCCL INFO Channel 07/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:178028 [1] NCCL INFO Channel 04/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:178025 [6] NCCL INFO Channel 04/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:178023 [2] NCCL INFO Channel 13/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:178026 [5] NCCL INFO Channel 00/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:178022 [3] NCCL INFO Channel 08/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:178024 [4] NCCL INFO Channel 15/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:178028 [1] NCCL INFO Channel 05/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:178025 [6] NCCL INFO Channel 05/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:178023 [2] NCCL INFO Channel 14/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:178026 [5] NCCL INFO Channel 01/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:178022 [3] NCCL INFO Channel 09/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:178024 [4] NCCL INFO Channel 16/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:178028 [1] NCCL INFO Channel 06/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:178025 [6] NCCL INFO Channel 06/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:178023 [2] NCCL INFO Channel 15/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:178026 [5] NCCL INFO Channel 02/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:178022 [3] NCCL INFO Channel 10/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:178024 [4] NCCL INFO Channel 17/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:178028 [1] NCCL INFO Channel 07/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:178025 [6] NCCL INFO Channel 07/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:178023 [2] NCCL INFO Channel 16/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:178026 [5] NCCL INFO Channel 03/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:178022 [3] NCCL INFO Channel 11/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:178024 [4] NCCL INFO Channel 18/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:178028 [1] NCCL INFO Channel 08/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:178025 [6] NCCL INFO Channel 08/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:178023 [2] NCCL INFO Channel 17/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:178026 [5] NCCL INFO Channel 04/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:178022 [3] NCCL INFO Channel 12/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:178024 [4] NCCL INFO Channel 19/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:178028 [1] NCCL INFO Channel 09/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:178025 [6] NCCL INFO Channel 09/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:178023 [2] NCCL INFO Channel 18/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:178026 [5] NCCL INFO Channel 05/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:178022 [3] NCCL INFO Channel 13/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:178024 [4] NCCL INFO Channel 20/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:178028 [1] NCCL INFO Channel 10/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:178025 [6] NCCL INFO Channel 10/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:178023 [2] NCCL INFO Channel 19/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:178026 [5] NCCL INFO Channel 06/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:178022 [3] NCCL INFO Channel 14/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:178024 [4] NCCL INFO Channel 21/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:178028 [1] NCCL INFO Channel 11/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:178025 [6] NCCL INFO Channel 11/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:178023 [2] NCCL INFO Channel 20/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:178026 [5] NCCL INFO Channel 07/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:178024 [4] NCCL INFO Channel 22/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:178025 [6] NCCL INFO Channel 12/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:178023 [2] NCCL INFO Channel 21/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170059:178024 [4] NCCL INFO Channel 23/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:178025 [6] NCCL INFO Channel 13/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:178023 [2] NCCL INFO Channel 22/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:178025 [6] NCCL INFO Channel 14/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170057:178023 [2] NCCL INFO Channel 23/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:178025 [6] NCCL INFO Channel 15/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:178025 [6] NCCL INFO Channel 16/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:178025 [6] NCCL INFO Channel 17/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:178025 [6] NCCL INFO Channel 18/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:178025 [6] NCCL INFO Channel 19/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:178025 [6] NCCL INFO Channel 20/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:178022 [3] NCCL INFO Channel 15/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:178025 [6] NCCL INFO Channel 21/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:178022 [3] NCCL INFO Channel 16/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:178025 [6] NCCL INFO Channel 22/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:178028 [1] NCCL INFO Channel 12/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:178022 [3] NCCL INFO Channel 17/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170061:178025 [6] NCCL INFO Channel 23/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:178028 [1] NCCL INFO Channel 13/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:178022 [3] NCCL INFO Channel 18/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:178028 [1] NCCL INFO Channel 14/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:178022 [3] NCCL INFO Channel 19/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:178028 [1] NCCL INFO Channel 15/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:178022 [3] NCCL INFO Channel 20/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:178028 [1] NCCL INFO Channel 16/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:178022 [3] NCCL INFO Channel 21/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:178026 [5] NCCL INFO Channel 08/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:178028 [1] NCCL INFO Channel 17/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:178022 [3] NCCL INFO Channel 22/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:178026 [5] NCCL INFO Channel 09/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:178028 [1] NCCL INFO Channel 18/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170058:178022 [3] NCCL INFO Channel 23/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:178026 [5] NCCL INFO Channel 10/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:178028 [1] NCCL INFO Channel 19/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:178028 [1] NCCL INFO Channel 20/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:178026 [5] NCCL INFO Channel 11/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:178028 [1] NCCL INFO Channel 21/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:178026 [5] NCCL INFO Channel 12/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:178028 [1] NCCL INFO Channel 22/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:178026 [5] NCCL INFO Channel 13/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170056:178028 [1] NCCL INFO Channel 23/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:178026 [5] NCCL INFO Channel 14/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:178026 [5] NCCL INFO Channel 15/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:178026 [5] NCCL INFO Channel 16/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:178026 [5] NCCL INFO Channel 17/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:178026 [5] NCCL INFO Channel 18/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:178026 [5] NCCL INFO Channel 19/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:178026 [5] NCCL INFO Channel 20/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:178026 [5] NCCL INFO Channel 21/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:178026 [5] NCCL INFO Channel 22/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170060:178026 [5] NCCL INFO Channel 23/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:170055:178021 [0] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:170055:178021 [0] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:170055:178021 [0] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:170056:178028 [1] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:170056:178028 [1] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:170056:178028 [1] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:170057:178023 [2] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:170057:178023 [2] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:170057:178023 [2] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:170058:178022 [3] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:170058:178022 [3] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:170058:178022 [3] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:170059:178024 [4] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:170059:178024 [4] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:170059:178024 [4] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:170060:178026 [5] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:170060:178026 [5] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:170060:178026 [5] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:170062:178027 [7] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:170061:178025 [6] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:170062:178027 [7] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:170062:178027 [7] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:170061:178025 [6] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:170061:178025 [6] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:170062:178027 [7] NCCL INFO ncclCommInitRank comm 0x7f892c06aca0 rank 7 nranks 8 cudaDev 7 nvmlDev 7 busId a01d0 commId 0x84b7cee2ede89cd2 - Init COMPLETE +ywang29-vrdb-test1-worker-0:170058:178022 [3] NCCL INFO ncclCommInitRank comm 0x7faf0c06ae80 rank 3 nranks 8 cudaDev 3 nvmlDev 3 busId 201d0 commId 0x84b7cee2ede89cd2 - Init COMPLETE +ywang29-vrdb-test1-worker-0:170056:178028 [1] NCCL INFO ncclCommInitRank comm 0x7fbdf806a8c0 rank 1 nranks 8 cudaDev 1 nvmlDev 1 busId 101d0 commId 0x84b7cee2ede89cd2 - Init COMPLETE +ywang29-vrdb-test1-worker-0:170060:178026 [5] NCCL INFO ncclCommInitRank comm 0x7f619006aef0 rank 5 nranks 8 cudaDev 5 nvmlDev 5 busId 901d0 commId 0x84b7cee2ede89cd2 - Init COMPLETE +ywang29-vrdb-test1-worker-0:170055:178021 [0] NCCL INFO ncclCommInitRank comm 0x7f915c06b3b0 rank 0 nranks 8 cudaDev 0 nvmlDev 0 busId 101c0 commId 0x84b7cee2ede89cd2 - Init COMPLETE +ywang29-vrdb-test1-worker-0:170061:178025 [6] NCCL INFO ncclCommInitRank comm 0x7ff2ec06b040 rank 6 nranks 8 cudaDev 6 nvmlDev 6 busId a01c0 commId 0x84b7cee2ede89cd2 - Init COMPLETE +ywang29-vrdb-test1-worker-0:170057:178023 [2] NCCL INFO ncclCommInitRank comm 0x7f8a4c06ac50 rank 2 nranks 8 cudaDev 2 nvmlDev 2 busId 201c0 commId 0x84b7cee2ede89cd2 - Init COMPLETE +ywang29-vrdb-test1-worker-0:170059:178024 [4] NCCL INFO ncclCommInitRank comm 0x7f9da806ade0 rank 4 nranks 8 cudaDev 4 nvmlDev 4 busId 901c0 commId 0x84b7cee2ede89cd2 - Init COMPLETE + 0%| | 1/520 [00:13<2:00:23, 13.92s/it] {'loss': 8.3465, 'grad_norm': 0.4095218949714936, 'learning_rate': 0.0125, 'epoch': 0.0} + 0%| | 1/520 [00:13<2:00:23, 13.92s/it] 0%| | 2/520 [00:17<1:09:25, 8.04s/it] {'loss': 7.601, 'grad_norm': 0.420334987693839, 'learning_rate': 0.025, 'epoch': 0.0} + 0%| | 2/520 [00:17<1:09:25, 8.04s/it] 1%| | 3/520 [00:21<53:22, 6.19s/it] {'loss': 7.0416, 'grad_norm': 0.2431154089143593, 'learning_rate': 0.037500000000000006, 'epoch': 0.01} + 1%| | 3/520 [00:21<53:22, 6.19s/it] 1%| | 4/520 [00:25<45:31, 5.29s/it] {'loss': 6.1125, 'grad_norm': 0.13083792018773896, 'learning_rate': 0.05, 'epoch': 0.01} + 1%| | 4/520 [00:25<45:31, 5.29s/it] 1%| | 5/520 [00:29<41:03, 4.78s/it] {'loss': 4.9778, 'grad_norm': 0.0994167145107128, 'learning_rate': 0.0625, 'epoch': 0.01} + 1%| | 5/520 [00:29<41:03, 4.78s/it] 1%| | 6/520 [00:33<38:26, 4.49s/it] {'loss': 5.8231, 'grad_norm': 0.33942902855426754, 'learning_rate': 0.07500000000000001, 'epoch': 0.01} + 1%| | 6/520 [00:33<38:26, 4.49s/it] 1%|▏ | 7/520 [00:37<36:41, 4.29s/it] {'loss': 3.8165, 'grad_norm': 0.10176231580875067, 'learning_rate': 0.08750000000000001, 'epoch': 0.01} + 1%|▏ | 7/520 [00:37<36:41, 4.29s/it] 2%|▏ | 8/520 [00:41<36:51, 4.32s/it] {'loss': 4.115, 'grad_norm': 0.1169109106994234, 'learning_rate': 0.1, 'epoch': 0.02} + 2%|▏ | 8/520 [00:41<36:51, 4.32s/it] 2%|▏ | 9/520 [00:45<36:23, 4.27s/it] {'loss': 3.7968, 'grad_norm': 0.05976546076300262, 'learning_rate': 0.1125, 'epoch': 0.02} + 2%|▏ | 9/520 [00:45<36:23, 4.27s/it] 2%|▏ | 10/520 [00:49<34:34, 4.07s/it] {'loss': 2.99, 'grad_norm': 0.048582473439670856, 'learning_rate': 0.125, 'epoch': 0.02} + 2%|▏ | 10/520 [00:49<34:34, 4.07s/it] 2%|▏ | 11/520 [00:53<33:56, 4.00s/it] {'loss': 2.9915, 'grad_norm': 0.03313206593589166, 'learning_rate': 0.1375, 'epoch': 0.02} + 2%|▏ | 11/520 [00:53<33:56, 4.00s/it] 2%|▏ | 12/520 [00:57<33:02, 3.90s/it] {'loss': 3.8536, 'grad_norm': 0.08394661400522642, 'learning_rate': 0.15000000000000002, 'epoch': 0.02} + 2%|▏ | 12/520 [00:57<33:02, 3.90s/it][2025-10-12 07:34:13,408] [WARNING] [stage3.py:2069:step] 1 pytorch allocator cache flushes since last step. this happens when there is high memory pressure and is detrimental to performance. if this is happening frequently consider adjusting settings to reduce memory consumption. If you are unable to make the cache flushes go away consider adding get_accelerator().empty_cache() calls in your training loop to ensure that all ranks flush their caches at the same time + 2%|▎ | 13/520 [01:01<34:11, 4.05s/it] {'loss': 2.5774, 'grad_norm': 0.023720321449072795, 'learning_rate': 0.1625, 'epoch': 0.03} + 2%|▎ | 13/520 [01:01<34:11, 4.05s/it] 3%|▎ | 14/520 [01:05<33:11, 3.94s/it] {'loss': 2.5145, 'grad_norm': 0.018619655399983753, 'learning_rate': 0.17500000000000002, 'epoch': 0.03} + 3%|▎ | 14/520 [01:05<33:11, 3.94s/it] 3%|▎ | 15/520 [01:08<32:37, 3.88s/it] {'loss': 3.226, 'grad_norm': 0.0442950972052187, 'learning_rate': 0.1875, 'epoch': 0.03} + 3%|▎ | 15/520 [01:08<32:37, 3.88s/it] 3%|▎ | 16/520 [01:12<32:12, 3.83s/it] {'loss': 3.1364, 'grad_norm': 0.09412603808257099, 'learning_rate': 0.2, 'epoch': 0.03} + 3%|▎ | 16/520 [01:12<32:12, 3.83s/it] 3%|▎ | 17/520 [01:16<31:45, 3.79s/it] {'loss': 2.8816, 'grad_norm': 0.0708257830825393, 'learning_rate': 0.1999980572931538, 'epoch': 0.03} + 3%|▎ | 17/520 [01:16<31:45, 3.79s/it] 3%|▎ | 18/520 [01:19<31:16, 3.74s/it] {'loss': 3.0863, 'grad_norm': 0.15114893687953818, 'learning_rate': 0.19999222924809748, 'epoch': 0.03} + 3%|▎ | 18/520 [01:19<31:16, 3.74s/it] 4%|▎ | 19/520 [01:23<31:07, 3.73s/it] {'loss': 4.1568, 'grad_norm': 0.24660813183039088, 'learning_rate': 0.19998251609127465, 'epoch': 0.04} + 4%|▎ | 19/520 [01:23<31:07, 3.73s/it] 4%|▍ | 20/520 [01:27<30:58, 3.72s/it] {'loss': 2.4934, 'grad_norm': 0.02609401070859364, 'learning_rate': 0.19996891820008164, 'epoch': 0.04} + 4%|▍ | 20/520 [01:27<30:58, 3.72s/it] 4%|▍ | 21/520 [01:31<30:55, 3.72s/it] {'loss': 4.2274, 'grad_norm': 0.06431563639437621, 'learning_rate': 0.19995143610285276, 'epoch': 0.04} + 4%|▍ | 21/520 [01:31<30:55, 3.72s/it] 4%|▍ | 22/520 [01:34<30:45, 3.71s/it] {'loss': 2.5574, 'grad_norm': 0.0234284508262777, 'learning_rate': 0.19993007047883987, 'epoch': 0.04} + 4%|▍ | 22/520 [01:34<30:45, 3.71s/it] 4%|▍ | 23/520 [01:38<30:31, 3.69s/it] {'loss': 2.3173, 'grad_norm': 0.017229973843154507, 'learning_rate': 0.1999048221581858, 'epoch': 0.04} + 4%|▍ | 23/520 [01:38<30:31, 3.69s/it] 5%|▍ | 24/520 [01:42<30:31, 3.69s/it] {'loss': 3.4996, 'grad_norm': 0.0704220121873399, 'learning_rate': 0.19987569212189224, 'epoch': 0.05} + 5%|▍ | 24/520 [01:42<30:31, 3.69s/it] 5%|▍ | 25/520 [01:45<30:21, 3.68s/it] {'loss': 2.7759, 'grad_norm': 0.10742201513235065, 'learning_rate': 0.19984268150178167, 'epoch': 0.05} + 5%|▍ | 25/520 [01:45<30:21, 3.68s/it] 5%|▌ | 26/520 [01:49<30:09, 3.66s/it] {'loss': 2.4133, 'grad_norm': 0.014883237559003832, 'learning_rate': 0.1998057915804532, 'epoch': 0.05} + 5%|▌ | 26/520 [01:49<30:09, 3.66s/it] 5%|▌ | 27/520 [01:52<30:00, 3.65s/it] {'loss': 2.1622, 'grad_norm': 0.01721540610534889, 'learning_rate': 0.1997650237912329, 'epoch': 0.05} + 5%|▌ | 27/520 [01:53<30:00, 3.65s/it] 5%|▌ | 28/520 [01:56<29:54, 3.65s/it] {'loss': 2.0612, 'grad_norm': 0.013802910157225689, 'learning_rate': 0.199720379718118, 'epoch': 0.05} + 5%|▌ | 28/520 [01:56<29:54, 3.65s/it] 6%|▌ | 29/520 [02:00<29:51, 3.65s/it] {'loss': 2.0418, 'grad_norm': 0.009605235625535117, 'learning_rate': 0.19967186109571552, 'epoch': 0.06} + 6%|▌ | 29/520 [02:00<29:51, 3.65s/it] 6%|▌ | 30/520 [02:03<29:46, 3.65s/it] {'loss': 2.8853, 'grad_norm': 0.022645142328090184, 'learning_rate': 0.19961946980917455, 'epoch': 0.06} + 6%|▌ | 30/520 [02:03<29:46, 3.65s/it] 6%|▌ | 31/520 [02:07<29:44, 3.65s/it] {'loss': 1.9776, 'grad_norm': 0.00767891143235087, 'learning_rate': 0.1995632078941134, 'epoch': 0.06} + 6%|▌ | 31/520 [02:07<29:44, 3.65s/it] 6%|▌ | 32/520 [02:11<29:51, 3.67s/it] {'loss': 3.2592, 'grad_norm': 0.034799112762604174, 'learning_rate': 0.19950307753654017, 'epoch': 0.06} + 6%|▌ | 32/520 [02:11<29:51, 3.67s/it] 6%|▋ | 33/520 [02:15<30:16, 3.73s/it] {'loss': 1.9783, 'grad_norm': 0.016602160100858375, 'learning_rate': 0.19943908107276798, 'epoch': 0.06} + 6%|▋ | 33/520 [02:15<30:16, 3.73s/it] 7%|▋ | 34/520 [02:18<30:09, 3.72s/it] {'loss': 1.8627, 'grad_norm': 0.00896407915483864, 'learning_rate': 0.19937122098932428, 'epoch': 0.07} + 7%|▋ | 34/520 [02:18<30:09, 3.72s/it] 7%|▋ | 35/520 [02:22<29:58, 3.71s/it] {'loss': 1.886, 'grad_norm': 0.012727170832706783, 'learning_rate': 0.19929949992285395, 'epoch': 0.07} + 7%|▋ | 35/520 [02:22<29:58, 3.71s/it] 7%|▋ | 36/520 [02:26<29:47, 3.69s/it] {'loss': 1.9983, 'grad_norm': 0.0066317773190545, 'learning_rate': 0.19922392066001723, 'epoch': 0.07} + 7%|▋ | 36/520 [02:26<29:47, 3.69s/it] 7%|▋ | 37/520 [02:29<29:45, 3.70s/it] {'loss': 4.03, 'grad_norm': 0.1402202938624636, 'learning_rate': 0.19914448613738106, 'epoch': 0.07} + 7%|▋ | 37/520 [02:29<29:45, 3.70s/it] 7%|▋ | 38/520 [02:33<29:39, 3.69s/it] {'loss': 2.085, 'grad_norm': 0.010710432448901868, 'learning_rate': 0.1990611994413053, 'epoch': 0.07} + 7%|▋ | 38/520 [02:33<29:39, 3.69s/it] 8%|▊ | 39/520 [02:37<29:35, 3.69s/it] {'loss': 1.9701, 'grad_norm': 0.023469219936968447, 'learning_rate': 0.19897406380782262, 'epoch': 0.07} + 8%|▊ | 39/520 [02:37<29:35, 3.69s/it] 8%|▊ | 40/520 [02:40<29:28, 3.69s/it] {'loss': 1.9585, 'grad_norm': 0.013133445663148186, 'learning_rate': 0.19888308262251286, 'epoch': 0.08} + 8%|▊ | 40/520 [02:40<29:28, 3.69s/it] 8%|▊ | 41/520 [02:44<29:25, 3.69s/it] {'loss': 1.8839, 'grad_norm': 0.009090510343279278, 'learning_rate': 0.19878825942037148, 'epoch': 0.08} + 8%|▊ | 41/520 [02:44<29:25, 3.69s/it] 8%|▊ | 42/520 [02:48<29:14, 3.67s/it] {'loss': 1.9677, 'grad_norm': 0.013058827969696159, 'learning_rate': 0.19868959788567211, 'epoch': 0.08} + 8%|▊ | 42/520 [02:48<29:14, 3.67s/it] 8%|▊ | 43/520 [02:51<29:13, 3.68s/it] {'loss': 2.8595, 'grad_norm': 0.062339591356042064, 'learning_rate': 0.1985871018518236, 'epoch': 0.08} + 8%|▊ | 43/520 [02:51<29:13, 3.68s/it] 8%|▊ | 44/520 [02:55<29:05, 3.67s/it] {'loss': 2.6381, 'grad_norm': 0.013944007467241577, 'learning_rate': 0.19848077530122082, 'epoch': 0.08} + 8%|▊ | 44/520 [02:55<29:05, 3.67s/it] 9%|▊ | 45/520 [02:59<29:03, 3.67s/it] {'loss': 1.8904, 'grad_norm': 0.013839349781256918, 'learning_rate': 0.19837062236509015, 'epoch': 0.09} + 9%|▊ | 45/520 [02:59<29:03, 3.67s/it] 9%|▉ | 46/520 [03:03<29:22, 3.72s/it] {'loss': 2.4963, 'grad_norm': 0.012165484319512004, 'learning_rate': 0.19825664732332884, 'epoch': 0.09} + 9%|▉ | 46/520 [03:03<29:22, 3.72s/it] 9%|▉ | 47/520 [03:06<29:18, 3.72s/it] {'loss': 1.834, 'grad_norm': 0.010239012281875348, 'learning_rate': 0.19813885460433878, 'epoch': 0.09} + 9%|▉ | 47/520 [03:06<29:18, 3.72s/it] 9%|▉ | 48/520 [03:10<29:06, 3.70s/it] {'loss': 1.7937, 'grad_norm': 0.007188646305683641, 'learning_rate': 0.19801724878485438, 'epoch': 0.09} + 9%|▉ | 48/520 [03:10<29:06, 3.70s/it] 9%|▉ | 49/520 [03:14<29:09, 3.71s/it] {'loss': 1.8011, 'grad_norm': 0.00929666564402266, 'learning_rate': 0.19789183458976486, 'epoch': 0.09} + 9%|▉ | 49/520 [03:14<29:09, 3.71s/it] 10%|▉ | 50/520 [03:17<29:04, 3.71s/it] {'loss': 1.7624, 'grad_norm': 0.006471494846791832, 'learning_rate': 0.19776261689193048, 'epoch': 0.1} + 10%|▉ | 50/520 [03:17<29:04, 3.71s/it] 10%|▉ | 51/520 [03:21<29:08, 3.73s/it] {'loss': 1.667, 'grad_norm': 0.0070456661084031215, 'learning_rate': 0.19762960071199334, 'epoch': 0.1} + 10%|▉ | 51/520 [03:21<29:08, 3.73s/it] 10%|█ | 52/520 [03:25<29:04, 3.73s/it] {'loss': 1.817, 'grad_norm': 0.005931173128865797, 'learning_rate': 0.19749279121818236, 'epoch': 0.1} + 10%|█ | 52/520 [03:25<29:04, 3.73s/it] 10%|█ | 53/520 [03:29<28:52, 3.71s/it] {'loss': 1.8224, 'grad_norm': 0.006587021590213661, 'learning_rate': 0.19735219372611235, 'epoch': 0.1} + 10%|█ | 53/520 [03:29<28:52, 3.71s/it] 10%|█ | 54/520 [03:32<29:09, 3.76s/it] {'loss': 1.6408, 'grad_norm': 0.006048973168317885, 'learning_rate': 0.19720781369857746, 'epoch': 0.1} + 10%|█ | 54/520 [03:32<29:09, 3.76s/it] 11%|█ | 55/520 [03:36<29:20, 3.79s/it] {'loss': 1.6677, 'grad_norm': 0.008000833836794638, 'learning_rate': 0.1970596567453391, 'epoch': 0.11} + 11%|█ | 55/520 [03:36<29:20, 3.79s/it] 11%|█ | 56/520 [03:40<29:35, 3.83s/it] {'loss': 1.8043, 'grad_norm': 0.005805983379886498, 'learning_rate': 0.1969077286229078, 'epoch': 0.11} + 11%|█ | 56/520 [03:40<29:35, 3.83s/it] 11%|█ | 57/520 [03:44<29:31, 3.83s/it] {'loss': 1.6324, 'grad_norm': 0.008102794982146765, 'learning_rate': 0.19675203523431964, 'epoch': 0.11} + 11%|█ | 57/520 [03:44<29:31, 3.83s/it] 11%|█ | 58/520 [03:48<29:30, 3.83s/it] {'loss': 1.7936, 'grad_norm': 0.006365134784268623, 'learning_rate': 0.19659258262890683, 'epoch': 0.11} + 11%|█ | 58/520 [03:48<29:30, 3.83s/it] 11%|█▏ | 59/520 [03:52<29:30, 3.84s/it] {'loss': 2.056, 'grad_norm': 0.014012700780947608, 'learning_rate': 0.19642937700206278, 'epoch': 0.11} + 11%|█▏ | 59/520 [03:52<29:30, 3.84s/it] 12%|█▏ | 60/520 [03:56<29:18, 3.82s/it] {'loss': 1.7086, 'grad_norm': 0.006962311137194432, 'learning_rate': 0.19626242469500121, 'epoch': 0.12} + 12%|█▏ | 60/520 [03:56<29:18, 3.82s/it] 12%|█▏ | 61/520 [03:59<28:49, 3.77s/it] {'loss': 2.1793, 'grad_norm': 0.009243729553074554, 'learning_rate': 0.19609173219450998, 'epoch': 0.12} + 12%|█▏ | 61/520 [03:59<28:49, 3.77s/it] 12%|█▏ | 62/520 [04:03<28:30, 3.73s/it] {'loss': 1.6667, 'grad_norm': 0.007736231013823276, 'learning_rate': 0.19591730613269878, 'epoch': 0.12} + 12%|█▏ | 62/520 [04:03<28:30, 3.73s/it] 12%|█▏ | 63/520 [04:07<28:18, 3.72s/it] {'loss': 1.6619, 'grad_norm': 0.0053867270303991155, 'learning_rate': 0.19573915328674182, 'epoch': 0.12} + 12%|█▏ | 63/520 [04:07<28:18, 3.72s/it] 12%|█▏ | 64/520 [04:10<28:14, 3.72s/it] {'loss': 1.6853, 'grad_norm': 0.004839820393407215, 'learning_rate': 0.1955572805786141, 'epoch': 0.12} + 12%|█▏ | 64/520 [04:10<28:14, 3.72s/it] 12%|█▎ | 65/520 [04:14<28:00, 3.69s/it] {'loss': 1.6989, 'grad_norm': 0.00561141139476775, 'learning_rate': 0.1953716950748227, 'epoch': 0.12} + 12%|█▎ | 65/520 [04:14<28:00, 3.69s/it] 13%|█▎ | 66/520 [04:18<27:48, 3.68s/it] {'loss': 1.6441, 'grad_norm': 0.005013369976598125, 'learning_rate': 0.19518240398613226, 'epoch': 0.13} + 13%|█▎ | 66/520 [04:18<27:48, 3.68s/it] 13%|█▎ | 67/520 [04:21<27:46, 3.68s/it] {'loss': 1.5073, 'grad_norm': 0.004588526262239831, 'learning_rate': 0.1949894146672846, 'epoch': 0.13} + 13%|█▎ | 67/520 [04:21<27:46, 3.68s/it] 13%|█▎ | 68/520 [04:25<27:36, 3.66s/it] {'loss': 1.5448, 'grad_norm': 0.005086778866701793, 'learning_rate': 0.1947927346167132, 'epoch': 0.13} + 13%|█▎ | 68/520 [04:25<27:36, 3.66s/it] 13%|█▎ | 69/520 [04:28<27:30, 3.66s/it] {'loss': 1.5278, 'grad_norm': 0.006288746060320267, 'learning_rate': 0.1945923714762516, 'epoch': 0.13} + 13%|█▎ | 69/520 [04:29<27:30, 3.66s/it] 13%|█▎ | 70/520 [04:32<27:28, 3.66s/it] {'loss': 1.6057, 'grad_norm': 0.005297970318394406, 'learning_rate': 0.19438833303083677, 'epoch': 0.13} + 13%|█▎ | 70/520 [04:32<27:28, 3.66s/it] 14%|█▎ | 71/520 [04:36<27:27, 3.67s/it] {'loss': 1.4891, 'grad_norm': 0.00547186036453552, 'learning_rate': 0.19418062720820636, 'epoch': 0.14} + 14%|█▎ | 71/520 [04:36<27:27, 3.67s/it] 14%|█▍ | 72/520 [04:39<27:15, 3.65s/it] {'loss': 1.666, 'grad_norm': 0.005441101946850768, 'learning_rate': 0.19396926207859086, 'epoch': 0.14} + 14%|█▍ | 72/520 [04:39<27:15, 3.65s/it] 14%|█▍ | 73/520 [04:43<27:14, 3.66s/it] {'loss': 1.4752, 'grad_norm': 0.004964764647522464, 'learning_rate': 0.19375424585439993, 'epoch': 0.14} + 14%|█▍ | 73/520 [04:43<27:14, 3.66s/it] 14%|█▍ | 74/520 [04:47<27:13, 3.66s/it] {'loss': 1.5987, 'grad_norm': 0.005491719981301707, 'learning_rate': 0.1935355868899034, 'epoch': 0.14} + 14%|█▍ | 74/520 [04:47<27:13, 3.66s/it] 14%|█▍ | 75/520 [04:50<27:09, 3.66s/it] {'loss': 1.4893, 'grad_norm': 0.004677283488840504, 'learning_rate': 0.19331329368090666, 'epoch': 0.14} + 14%|█▍ | 75/520 [04:50<27:09, 3.66s/it] 15%|█▍ | 76/520 [04:54<27:03, 3.66s/it] {'loss': 2.061, 'grad_norm': 0.007800817864138159, 'learning_rate': 0.19308737486442043, 'epoch': 0.15} + 15%|█▍ | 76/520 [04:54<27:03, 3.66s/it] 15%|█▍ | 77/520 [04:58<27:01, 3.66s/it] {'loss': 1.418, 'grad_norm': 0.005926366497902714, 'learning_rate': 0.19285783921832536, 'epoch': 0.15} + 15%|█▍ | 77/520 [04:58<27:01, 3.66s/it] 15%|█▌ | 78/520 [05:01<26:59, 3.66s/it] {'loss': 1.5456, 'grad_norm': 0.005331553614584305, 'learning_rate': 0.19262469566103088, 'epoch': 0.15} + 15%|█▌ | 78/520 [05:01<26:59, 3.66s/it] 15%|█▌ | 79/520 [05:05<26:54, 3.66s/it] {'loss': 1.5209, 'grad_norm': 0.0048004646994902235, 'learning_rate': 0.19238795325112867, 'epoch': 0.15} + 15%|█▌ | 79/520 [05:05<26:54, 3.66s/it] 15%|█▌ | 80/520 [05:09<26:49, 3.66s/it] {'loss': 2.0674, 'grad_norm': 0.006868717390157598, 'learning_rate': 0.19214762118704076, 'epoch': 0.15} + 15%|█▌ | 80/520 [05:09<26:49, 3.66s/it] 16%|█▌ | 81/520 [05:12<26:52, 3.67s/it] {'loss': 1.6972, 'grad_norm': 0.006839594614495101, 'learning_rate': 0.19190370880666208, 'epoch': 0.16} + 16%|█▌ | 81/520 [05:12<26:52, 3.67s/it] 16%|█▌ | 82/520 [05:16<26:45, 3.67s/it] {'loss': 1.5851, 'grad_norm': 0.004615900787774756, 'learning_rate': 0.19165622558699763, 'epoch': 0.16} + 16%|█▌ | 82/520 [05:16<26:45, 3.67s/it] 16%|█▌ | 83/520 [05:20<26:34, 3.65s/it] {'loss': 1.6318, 'grad_norm': 0.005149365785690455, 'learning_rate': 0.19140518114379435, 'epoch': 0.16} + 16%|█▌ | 83/520 [05:20<26:34, 3.65s/it] 16%|█▌ | 84/520 [05:23<26:35, 3.66s/it] {'loss': 1.6172, 'grad_norm': 0.005932122708870783, 'learning_rate': 0.19115058523116735, 'epoch': 0.16} + 16%|█▌ | 84/520 [05:23<26:35, 3.66s/it] 16%|█▋ | 85/520 [05:27<26:28, 3.65s/it] {'loss': 1.618, 'grad_norm': 0.005509012514504756, 'learning_rate': 0.1908924477412211, 'epoch': 0.16} + 16%|█▋ | 85/520 [05:27<26:28, 3.65s/it] 17%|█▋ | 86/520 [05:31<26:23, 3.65s/it] {'loss': 1.6826, 'grad_norm': 0.005475591504361661, 'learning_rate': 0.19063077870366502, 'epoch': 0.17} + 17%|█▋ | 86/520 [05:31<26:23, 3.65s/it] 17%|█▋ | 87/520 [05:34<26:21, 3.65s/it] {'loss': 1.9571, 'grad_norm': 0.008206728071855773, 'learning_rate': 0.1903655882854237, 'epoch': 0.17} + 17%|█▋ | 87/520 [05:34<26:21, 3.65s/it] 17%|█▋ | 88/520 [05:38<26:21, 3.66s/it] {'loss': 2.1123, 'grad_norm': 0.007802617974405128, 'learning_rate': 0.19009688679024192, 'epoch': 0.17} + 17%|█▋ | 88/520 [05:38<26:21, 3.66s/it] 17%|█▋ | 89/520 [05:42<26:17, 3.66s/it] {'loss': 1.5918, 'grad_norm': 0.005317241188993234, 'learning_rate': 0.18982468465828442, 'epoch': 0.17} + 17%|█▋ | 89/520 [05:42<26:17, 3.66s/it] 17%|█▋ | 90/520 [05:45<26:12, 3.66s/it] {'loss': 1.5102, 'grad_norm': 0.005224753135271097, 'learning_rate': 0.1895489924657301, 'epoch': 0.17} + 17%|█▋ | 90/520 [05:45<26:12, 3.66s/it] 18%|█▊ | 91/520 [05:49<26:05, 3.65s/it] {'loss': 1.5921, 'grad_norm': 0.005207933775390539, 'learning_rate': 0.18926982092436118, 'epoch': 0.17} + 18%|█▊ | 91/520 [05:49<26:05, 3.65s/it] 18%|█▊ | 92/520 [05:53<26:04, 3.66s/it] {'loss': 1.5245, 'grad_norm': 0.004552885331556028, 'learning_rate': 0.18898718088114688, 'epoch': 0.18} + 18%|█▊ | 92/520 [05:53<26:04, 3.66s/it] 18%|█▊ | 93/520 [05:56<26:02, 3.66s/it] {'loss': 1.5182, 'grad_norm': 0.005090028260876908, 'learning_rate': 0.18870108331782218, 'epoch': 0.18} + 18%|█▊ | 93/520 [05:56<26:02, 3.66s/it] 18%|█▊ | 94/520 [06:00<25:57, 3.66s/it] {'loss': 1.6302, 'grad_norm': 0.005326018851314604, 'learning_rate': 0.18841153935046098, 'epoch': 0.18} + 18%|█▊ | 94/520 [06:00<25:57, 3.66s/it] 18%|█▊ | 95/520 [06:04<25:57, 3.66s/it] {'loss': 1.5054, 'grad_norm': 0.005330392346628984, 'learning_rate': 0.18811856022904425, 'epoch': 0.18} + 18%|█▊ | 95/520 [06:04<25:57, 3.66s/it] 18%|█▊ | 96/520 [06:07<25:50, 3.66s/it] {'loss': 1.4968, 'grad_norm': 0.004611266373478052, 'learning_rate': 0.18782215733702287, 'epoch': 0.18} + 18%|█▊ | 96/520 [06:07<25:50, 3.66s/it] 19%|█▊ | 97/520 [06:11<25:50, 3.67s/it] {'loss': 1.4691, 'grad_norm': 0.005511862335674179, 'learning_rate': 0.18752234219087538, 'epoch': 0.19} + 19%|█▊ | 97/520 [06:11<25:50, 3.67s/it] 19%|█▉ | 98/520 [06:15<25:45, 3.66s/it] {'loss': 1.4733, 'grad_norm': 0.0041958501067330825, 'learning_rate': 0.18721912643966054, 'epoch': 0.19} + 19%|█▉ | 98/520 [06:15<25:45, 3.66s/it] 19%|█▉ | 99/520 [06:18<25:51, 3.69s/it] {'loss': 1.4962, 'grad_norm': 0.004684650388422069, 'learning_rate': 0.18691252186456464, 'epoch': 0.19} + 19%|█▉ | 99/520 [06:18<25:51, 3.69s/it] 19%|█▉ | 100/520 [06:22<26:06, 3.73s/it] {'loss': 1.7809, 'grad_norm': 0.007837719310482365, 'learning_rate': 0.1866025403784439, 'epoch': 0.19} + 19%|█▉ | 100/520 [06:22<26:06, 3.73s/it] 19%|█▉ | 101/520 [06:26<25:52, 3.71s/it] {'loss': 1.486, 'grad_norm': 0.0047933422418251544, 'learning_rate': 0.18628919402536132, 'epoch': 0.19} + 19%|█▉ | 101/520 [06:26<25:52, 3.71s/it] 20%|█▉ | 102/520 [06:29<25:37, 3.68s/it] {'loss': 1.5002, 'grad_norm': 0.0053714162758819566, 'learning_rate': 0.18597249498011903, 'epoch': 0.2} + 20%|█▉ | 102/520 [06:29<25:37, 3.68s/it] 20%|█▉ | 103/520 [06:33<25:31, 3.67s/it] {'loss': 1.417, 'grad_norm': 0.0039492822148137095, 'learning_rate': 0.18565245554778517, 'epoch': 0.2} + 20%|█▉ | 103/520 [06:33<25:31, 3.67s/it] 20%|██ | 104/520 [06:37<25:25, 3.67s/it] {'loss': 1.5003, 'grad_norm': 0.004791298582761909, 'learning_rate': 0.18532908816321558, 'epoch': 0.2} + 20%|██ | 104/520 [06:37<25:25, 3.67s/it] 20%|██ | 105/520 [06:40<25:23, 3.67s/it] {'loss': 1.5021, 'grad_norm': 0.004262230275886522, 'learning_rate': 0.18500240539057092, 'epoch': 0.2} + 20%|██ | 105/520 [06:40<25:23, 3.67s/it] 20%|██ | 106/520 [06:44<25:20, 3.67s/it] {'loss': 1.7186, 'grad_norm': 0.005188924321757355, 'learning_rate': 0.18467241992282843, 'epoch': 0.2} + 20%|██ | 106/520 [06:44<25:20, 3.67s/it] 21%|██ | 107/520 [06:48<25:16, 3.67s/it] {'loss': 1.7064, 'grad_norm': 0.006371742146741625, 'learning_rate': 0.18433914458128858, 'epoch': 0.21} + 21%|██ | 107/520 [06:48<25:16, 3.67s/it] 21%|██ | 108/520 [06:51<25:13, 3.67s/it] {'loss': 1.4523, 'grad_norm': 0.005096600858631069, 'learning_rate': 0.18400259231507718, 'epoch': 0.21} + 21%|██ | 108/520 [06:51<25:13, 3.67s/it] 21%|██ | 109/520 [06:55<25:05, 3.66s/it] {'loss': 1.694, 'grad_norm': 0.005203580606684685, 'learning_rate': 0.18366277620064198, 'epoch': 0.21} + 21%|██ | 109/520 [06:55<25:05, 3.66s/it] 21%|██ | 110/520 [06:59<24:58, 3.66s/it] {'loss': 1.6389, 'grad_norm': 0.0047928768461746535, 'learning_rate': 0.1833197094412449, 'epoch': 0.21} + 21%|██ | 110/520 [06:59<24:58, 3.66s/it] 21%|██▏ | 111/520 [07:02<24:54, 3.65s/it] {'loss': 1.661, 'grad_norm': 0.005623095983837976, 'learning_rate': 0.18297340536644877, 'epoch': 0.21} + 21%|██▏ | 111/520 [07:02<24:54, 3.65s/it] 22%|██▏ | 112/520 [07:06<24:51, 3.66s/it] {'loss': 1.5293, 'grad_norm': 0.004444181766650605, 'learning_rate': 0.1826238774315995, 'epoch': 0.22} + 22%|██▏ | 112/520 [07:06<24:51, 3.66s/it] 22%|██▏ | 113/520 [07:10<24:48, 3.66s/it] {'loss': 1.3795, 'grad_norm': 0.00415748644473943, 'learning_rate': 0.18227113921730334, 'epoch': 0.22} + 22%|██▏ | 113/520 [07:10<24:48, 3.66s/it] 22%|██▏ | 114/520 [07:13<24:54, 3.68s/it] {'loss': 1.4914, 'grad_norm': 0.0046274075868010635, 'learning_rate': 0.1819152044288992, 'epoch': 0.22} + 22%|██▏ | 114/520 [07:13<24:54, 3.68s/it] 22%|██▏ | 115/520 [07:17<24:49, 3.68s/it] {'loss': 1.6335, 'grad_norm': 0.0045590093718447825, 'learning_rate': 0.18155608689592603, 'epoch': 0.22} + 22%|██▏ | 115/520 [07:17<24:49, 3.68s/it] 22%|██▏ | 116/520 [07:21<24:47, 3.68s/it] {'loss': 1.6083, 'grad_norm': 0.00400337806920208, 'learning_rate': 0.18119380057158568, 'epoch': 0.22} + 22%|██▏ | 116/520 [07:21<24:47, 3.68s/it] 22%|██▎ | 117/520 [07:24<24:38, 3.67s/it] {'loss': 1.5899, 'grad_norm': 0.005090893250506973, 'learning_rate': 0.18082835953220056, 'epoch': 0.23} + 22%|██▎ | 117/520 [07:24<24:38, 3.67s/it] 23%|██▎ | 118/520 [07:28<24:29, 3.66s/it] {'loss': 1.4689, 'grad_norm': 0.0038182588400489373, 'learning_rate': 0.18045977797666685, 'epoch': 0.23} + 23%|██▎ | 118/520 [07:28<24:29, 3.66s/it] 23%|██▎ | 119/520 [07:32<24:23, 3.65s/it] {'loss': 1.4072, 'grad_norm': 0.004224143478151947, 'learning_rate': 0.1800880702259028, 'epoch': 0.23} + 23%|██▎ | 119/520 [07:32<24:23, 3.65s/it] 23%|██▎ | 120/520 [07:35<24:16, 3.64s/it] {'loss': 1.4431, 'grad_norm': 0.004875068273108048, 'learning_rate': 0.17971325072229227, 'epoch': 0.23} + 23%|██▎ | 120/520 [07:35<24:16, 3.64s/it] 23%|██▎ | 121/520 [07:39<24:13, 3.64s/it] {'loss': 1.5023, 'grad_norm': 0.005448437323600828, 'learning_rate': 0.17933533402912352, 'epoch': 0.23} + 23%|██▎ | 121/520 [07:39<24:13, 3.64s/it] 23%|██▎ | 122/520 [07:43<24:13, 3.65s/it] {'loss': 1.3811, 'grad_norm': 0.004157185288590129, 'learning_rate': 0.17895433483002354, 'epoch': 0.23} + 23%|██▎ | 122/520 [07:43<24:13, 3.65s/it] 24%|██▎ | 123/520 [07:46<24:12, 3.66s/it] {'loss': 1.7425, 'grad_norm': 0.006447613968172008, 'learning_rate': 0.17857026792838737, 'epoch': 0.24} + 24%|██▎ | 123/520 [07:46<24:12, 3.66s/it] 24%|██▍ | 124/520 [07:50<24:13, 3.67s/it] {'loss': 1.4827, 'grad_norm': 0.004959298200850979, 'learning_rate': 0.178183148246803, 'epoch': 0.24} + 24%|██▍ | 124/520 [07:50<24:13, 3.67s/it] 24%|██▍ | 125/520 [07:54<24:16, 3.69s/it] {'loss': 1.4651, 'grad_norm': 0.00430646054749742, 'learning_rate': 0.1777929908264715, 'epoch': 0.24} + 24%|██▍ | 125/520 [07:54<24:16, 3.69s/it] 24%|██▍ | 126/520 [07:58<25:32, 3.89s/it] {'loss': 1.6168, 'grad_norm': 0.0046111409231265575, 'learning_rate': 0.17739981082662276, 'epoch': 0.24} + 24%|██▍ | 126/520 [07:58<25:32, 3.89s/it] 24%|██▍ | 127/520 [08:02<25:02, 3.82s/it] {'loss': 1.439, 'grad_norm': 0.006007621805791378, 'learning_rate': 0.1770036235239263, 'epoch': 0.24} + 24%|██▍ | 127/520 [08:02<25:02, 3.82s/it] 25%|██▍ | 128/520 [08:05<24:42, 3.78s/it] {'loss': 1.4879, 'grad_norm': 0.004370946815309512, 'learning_rate': 0.1766044443118978, 'epoch': 0.25} + 25%|██▍ | 128/520 [08:05<24:42, 3.78s/it] 25%|██▍ | 129/520 [08:09<24:25, 3.75s/it] {'loss': 1.4007, 'grad_norm': 0.00387713916563295, 'learning_rate': 0.17620228870030108, 'epoch': 0.25} + 25%|██▍ | 129/520 [08:09<24:25, 3.75s/it] 25%|██▌ | 130/520 [08:13<24:13, 3.73s/it] {'loss': 1.4759, 'grad_norm': 0.00497173592481853, 'learning_rate': 0.1757971723145453, 'epoch': 0.25} + 25%|██▌ | 130/520 [08:13<24:13, 3.73s/it] 25%|██▌ | 131/520 [08:16<24:03, 3.71s/it] {'loss': 1.6002, 'grad_norm': 0.005538240540358096, 'learning_rate': 0.175389110895078, 'epoch': 0.25} + 25%|██▌ | 131/520 [08:16<24:03, 3.71s/it] 25%|██▌ | 132/520 [08:20<23:55, 3.70s/it] {'loss': 1.518, 'grad_norm': 0.004914505526662321, 'learning_rate': 0.17497812029677343, 'epoch': 0.25} + 25%|██▌ | 132/520 [08:20<23:55, 3.70s/it] 26%|██▌ | 133/520 [08:24<23:48, 3.69s/it] {'loss': 1.4236, 'grad_norm': 0.004401528147456331, 'learning_rate': 0.17456421648831655, 'epoch': 0.26} + 26%|██▌ | 133/520 [08:24<23:48, 3.69s/it] 26%|██▌ | 134/520 [08:28<23:42, 3.68s/it] {'loss': 1.5126, 'grad_norm': 0.00426087481391123, 'learning_rate': 0.17414741555158267, 'epoch': 0.26} + 26%|██▌ | 134/520 [08:28<23:42, 3.68s/it] 26%|██▌ | 135/520 [08:31<23:34, 3.67s/it] {'loss': 1.5873, 'grad_norm': 0.004429021419014588, 'learning_rate': 0.1737277336810124, 'epoch': 0.26} + 26%|██▌ | 135/520 [08:31<23:34, 3.67s/it] 26%|██▌ | 136/520 [08:35<23:28, 3.67s/it] {'loss': 1.4924, 'grad_norm': 0.004391465006621195, 'learning_rate': 0.17330518718298263, 'epoch': 0.26} + 26%|██▌ | 136/520 [08:35<23:28, 3.67s/it] 26%|██▋ | 137/520 [08:39<23:30, 3.68s/it] {'loss': 1.4145, 'grad_norm': 0.004997805558083794, 'learning_rate': 0.17287979247517285, 'epoch': 0.26} + 26%|██▋ | 137/520 [08:39<23:30, 3.68s/it] 27%|██▋ | 138/520 [08:42<23:25, 3.68s/it] {'loss': 1.4189, 'grad_norm': 0.004375883458461362, 'learning_rate': 0.17245156608592727, 'epoch': 0.27} + 27%|██▋ | 138/520 [08:42<23:25, 3.68s/it] 27%|██▋ | 139/520 [08:46<23:24, 3.69s/it] {'loss': 1.4816, 'grad_norm': 0.004697637215938851, 'learning_rate': 0.17202052465361267, 'epoch': 0.27} + 27%|██▋ | 139/520 [08:46<23:24, 3.69s/it] 27%|██▋ | 140/520 [08:50<23:22, 3.69s/it] {'loss': 1.6204, 'grad_norm': 0.005039556713604907, 'learning_rate': 0.17158668492597184, 'epoch': 0.27} + 27%|██▋ | 140/520 [08:50<23:22, 3.69s/it] 27%|██▋ | 141/520 [08:53<23:22, 3.70s/it] {'loss': 1.5447, 'grad_norm': 0.004708141907312232, 'learning_rate': 0.17115006375947303, 'epoch': 0.27} + 27%|██▋ | 141/520 [08:53<23:22, 3.70s/it] 27%|██▋ | 142/520 [08:57<23:18, 3.70s/it] {'loss': 1.6567, 'grad_norm': 0.005376095053698829, 'learning_rate': 0.17071067811865476, 'epoch': 0.27} + 27%|██▋ | 142/520 [08:57<23:18, 3.70s/it] 28%|██▊ | 143/520 [09:01<23:10, 3.69s/it] {'loss': 1.4672, 'grad_norm': 0.005184876064779801, 'learning_rate': 0.17026854507546693, 'epoch': 0.28} + 28%|██▊ | 143/520 [09:01<23:10, 3.69s/it] 28%|██▊ | 144/520 [09:04<23:06, 3.69s/it] {'loss': 1.4166, 'grad_norm': 0.004957223356497621, 'learning_rate': 0.1698236818086073, 'epoch': 0.28} + 28%|██▊ | 144/520 [09:04<23:06, 3.69s/it] 28%|██▊ | 145/520 [09:08<23:09, 3.70s/it] {'loss': 1.3402, 'grad_norm': 0.004004851578575198, 'learning_rate': 0.16937610560285418, 'epoch': 0.28} + 28%|██▊ | 145/520 [09:08<23:09, 3.70s/it] 28%|██▊ | 146/520 [09:12<23:12, 3.72s/it] {'loss': 1.6933, 'grad_norm': 0.005350965996937834, 'learning_rate': 0.1689258338483947, 'epoch': 0.28} + 28%|██▊ | 146/520 [09:12<23:12, 3.72s/it] 28%|██▊ | 147/520 [09:16<23:02, 3.71s/it] {'loss': 1.3819, 'grad_norm': 0.004187862154810301, 'learning_rate': 0.16847288404014937, 'epoch': 0.28} + 28%|██▊ | 147/520 [09:16<23:02, 3.71s/it] 28%|██▊ | 148/520 [09:19<22:53, 3.69s/it] {'loss': 1.4286, 'grad_norm': 0.003949572656623808, 'learning_rate': 0.16801727377709194, 'epoch': 0.28} + 28%|██▊ | 148/520 [09:19<22:53, 3.69s/it] 29%|██▊ | 149/520 [09:23<22:55, 3.71s/it] {'loss': 1.3774, 'grad_norm': 0.004293558392434266, 'learning_rate': 0.16755902076156604, 'epoch': 0.29} + 29%|██▊ | 149/520 [09:23<22:55, 3.71s/it] 29%|██▉ | 150/520 [09:27<22:50, 3.70s/it] {'loss': 1.5959, 'grad_norm': 0.00481357026896033, 'learning_rate': 0.16709814279859703, 'epoch': 0.29} + 29%|██▉ | 150/520 [09:27<22:50, 3.70s/it] 29%|██▉ | 151/520 [09:30<22:46, 3.70s/it] {'loss': 1.4032, 'grad_norm': 0.004139747216934926, 'learning_rate': 0.1666346577952004, 'epoch': 0.29} + 29%|██▉ | 151/520 [09:30<22:46, 3.70s/it] 29%|██▉ | 152/520 [09:34<22:47, 3.72s/it] {'loss': 1.3723, 'grad_norm': 0.0043237512540015, 'learning_rate': 0.16616858375968596, 'epoch': 0.29} + 29%|██▉ | 152/520 [09:34<22:47, 3.72s/it] 29%|██▉ | 153/520 [09:38<22:38, 3.70s/it] {'loss': 1.4198, 'grad_norm': 0.004886852336410126, 'learning_rate': 0.16569993880095807, 'epoch': 0.29} + 29%|██▉ | 153/520 [09:38<22:38, 3.70s/it] 30%|██▉ | 154/520 [09:41<22:32, 3.70s/it] {'loss': 1.4914, 'grad_norm': 0.003958399254803175, 'learning_rate': 0.16522874112781213, 'epoch': 0.3} + 30%|██▉ | 154/520 [09:41<22:32, 3.70s/it] 30%|██▉ | 155/520 [09:45<22:37, 3.72s/it] {'loss': 1.3945, 'grad_norm': 0.004323334077677175, 'learning_rate': 0.16475500904822704, 'epoch': 0.3} + 30%|██▉ | 155/520 [09:45<22:37, 3.72s/it] 30%|███ | 156/520 [09:49<22:31, 3.71s/it] {'loss': 1.4311, 'grad_norm': 0.004695767871195654, 'learning_rate': 0.16427876096865393, 'epoch': 0.3} + 30%|███ | 156/520 [09:49<22:31, 3.71s/it] 30%|███ | 157/520 [09:53<22:21, 3.70s/it] {'loss': 1.7057, 'grad_norm': 0.005633966293137499, 'learning_rate': 0.16380001539330089, 'epoch': 0.3} + 30%|███ | 157/520 [09:53<22:21, 3.70s/it] 30%|███ | 158/520 [09:56<22:12, 3.68s/it] {'loss': 1.4093, 'grad_norm': 0.005324968987012723, 'learning_rate': 0.163318790923414, 'epoch': 0.3} + 30%|███ | 158/520 [09:56<22:12, 3.68s/it] 31%|███ | 159/520 [10:00<22:08, 3.68s/it] {'loss': 1.4314, 'grad_norm': 0.003821431641607167, 'learning_rate': 0.16283510625655473, 'epoch': 0.31} + 31%|███ | 159/520 [10:00<22:08, 3.68s/it] 31%|███ | 160/520 [10:04<22:14, 3.71s/it] {'loss': 1.4719, 'grad_norm': 0.004639520723779572, 'learning_rate': 0.16234898018587338, 'epoch': 0.31} + 31%|███ | 160/520 [10:04<22:14, 3.71s/it] 31%|███ | 161/520 [10:07<22:14, 3.72s/it] {'loss': 1.4546, 'grad_norm': 0.004619772788975194, 'learning_rate': 0.16186043159937882, 'epoch': 0.31} + 31%|███ | 161/520 [10:07<22:14, 3.72s/it] 31%|███ | 162/520 [10:11<22:12, 3.72s/it] {'loss': 1.602, 'grad_norm': 0.0046094929778499995, 'learning_rate': 0.16136947947920477, 'epoch': 0.31} + 31%|███ | 162/520 [10:11<22:12, 3.72s/it] 31%|███▏ | 163/520 [10:15<22:02, 3.71s/it] {'loss': 1.3165, 'grad_norm': 0.004930569151517009, 'learning_rate': 0.16087614290087207, 'epoch': 0.31} + 31%|███▏ | 163/520 [10:15<22:02, 3.71s/it] 32%|███▏ | 164/520 [10:18<21:54, 3.69s/it] {'loss': 1.2881, 'grad_norm': 0.0037911848240668837, 'learning_rate': 0.16038044103254775, 'epoch': 0.32} + 32%|███▏ | 164/520 [10:18<21:54, 3.69s/it] 32%|███▏ | 165/520 [10:22<21:48, 3.69s/it] {'loss': 1.428, 'grad_norm': 0.003880447857698146, 'learning_rate': 0.15988239313430005, 'epoch': 0.32} + 32%|███▏ | 165/520 [10:22<21:48, 3.69s/it] 32%|███▏ | 166/520 [10:26<21:38, 3.67s/it] {'loss': 1.4178, 'grad_norm': 0.004461301659789694, 'learning_rate': 0.15938201855735015, 'epoch': 0.32} + 32%|███▏ | 166/520 [10:26<21:38, 3.67s/it] 32%|███▏ | 167/520 [10:29<21:37, 3.68s/it] {'loss': 1.4149, 'grad_norm': 0.004251015780618185, 'learning_rate': 0.15887933674332047, 'epoch': 0.32} + 32%|███▏ | 167/520 [10:29<21:37, 3.68s/it] 32%|███▏ | 168/520 [10:33<21:33, 3.67s/it] {'loss': 1.3577, 'grad_norm': 0.004299023978987326, 'learning_rate': 0.158374367223479, 'epoch': 0.32} + 32%|███▏ | 168/520 [10:33<21:33, 3.67s/it] 32%|███▎ | 169/520 [10:37<21:37, 3.70s/it] {'loss': 1.4254, 'grad_norm': 0.004099790636266085, 'learning_rate': 0.1578671296179806, 'epoch': 0.33} + 32%|███▎ | 169/520 [10:37<21:37, 3.70s/it] 33%|███▎ | 170/520 [10:41<21:53, 3.75s/it] {'loss': 1.5112, 'grad_norm': 0.0045696607573151886, 'learning_rate': 0.15735764363510463, 'epoch': 0.33} + 33%|███▎ | 170/520 [10:41<21:53, 3.75s/it] 33%|███▎ | 171/520 [10:45<22:07, 3.80s/it] {'loss': 1.3621, 'grad_norm': 0.004405719581736167, 'learning_rate': 0.15684592907048925, 'epoch': 0.33} + 33%|███▎ | 171/520 [10:45<22:07, 3.80s/it] 33%|███▎ | 172/520 [10:49<22:10, 3.82s/it] {'loss': 1.4248, 'grad_norm': 0.004354814397142415, 'learning_rate': 0.1563320058063622, 'epoch': 0.33} + 33%|███▎ | 172/520 [10:49<22:10, 3.82s/it] 33%|███▎ | 173/520 [10:53<22:18, 3.86s/it] {'loss': 1.3633, 'grad_norm': 0.0041125234701265305, 'learning_rate': 0.15581589381076844, 'epoch': 0.33} + 33%|███▎ | 173/520 [10:53<22:18, 3.86s/it] 33%|███▎ | 174/520 [10:56<22:21, 3.88s/it] {'loss': 1.4437, 'grad_norm': 0.004729625662166209, 'learning_rate': 0.15529761313679394, 'epoch': 0.33} + 33%|███▎ | 174/520 [10:56<22:21, 3.88s/it] 34%|███▎ | 175/520 [11:00<22:14, 3.87s/it] {'loss': 1.3387, 'grad_norm': 0.004109406899378858, 'learning_rate': 0.15477718392178716, 'epoch': 0.34} + 34%|███▎ | 175/520 [11:00<22:14, 3.87s/it] 34%|███▍ | 176/520 [11:04<22:22, 3.90s/it] {'loss': 1.6098, 'grad_norm': 0.004850828274905143, 'learning_rate': 0.15425462638657594, 'epoch': 0.34} + 34%|███▍ | 176/520 [11:04<22:22, 3.90s/it] 34%|███▍ | 177/520 [11:08<22:15, 3.89s/it] {'loss': 1.4664, 'grad_norm': 0.004962041437958554, 'learning_rate': 0.1537299608346824, 'epoch': 0.34} + 34%|███▍ | 177/520 [11:08<22:15, 3.89s/it] 34%|███▍ | 178/520 [11:12<22:09, 3.89s/it] {'loss': 1.4184, 'grad_norm': 0.004509160776756174, 'learning_rate': 0.15320320765153367, 'epoch': 0.34} + 34%|███▍ | 178/520 [11:12<22:09, 3.89s/it] 34%|███▍ | 179/520 [11:16<22:02, 3.88s/it] {'loss': 1.4993, 'grad_norm': 0.0038887466818058746, 'learning_rate': 0.15267438730367008, 'epoch': 0.34} + 34%|███▍ | 179/520 [11:16<22:02, 3.88s/it] 35%|███▍ | 180/520 [11:20<21:58, 3.88s/it] {'loss': 1.4015, 'grad_norm': 0.004452828883590338, 'learning_rate': 0.1521435203379498, 'epoch': 0.35} + 35%|███▍ | 180/520 [11:20<21:58, 3.88s/it] 35%|███▍ | 181/520 [11:24<21:55, 3.88s/it] {'loss': 1.3712, 'grad_norm': 0.004064551869785976, 'learning_rate': 0.15161062738075068, 'epoch': 0.35} + 35%|███▍ | 181/520 [11:24<21:55, 3.88s/it] 35%|███▌ | 182/520 [11:28<21:52, 3.88s/it] {'loss': 1.3927, 'grad_norm': 0.004102298081781246, 'learning_rate': 0.1510757291371686, 'epoch': 0.35} + 35%|███▌ | 182/520 [11:28<21:52, 3.88s/it] 35%|███▌ | 183/520 [11:31<21:47, 3.88s/it] {'loss': 1.426, 'grad_norm': 0.00421110890199113, 'learning_rate': 0.1505388463902131, 'epoch': 0.35} + 35%|███▌ | 183/520 [11:31<21:47, 3.88s/it] 35%|███▌ | 184/520 [11:35<21:40, 3.87s/it] {'loss': 1.3333, 'grad_norm': 0.004373433384107478, 'learning_rate': 0.15000000000000002, 'epoch': 0.35} + 35%|███▌ | 184/520 [11:35<21:40, 3.87s/it] 36%|███▌ | 185/520 [11:39<21:36, 3.87s/it] {'loss': 1.519, 'grad_norm': 0.00416397753528565, 'learning_rate': 0.14945921090294076, 'epoch': 0.36} + 36%|███▌ | 185/520 [11:39<21:36, 3.87s/it] 36%|███▌ | 186/520 [11:43<21:31, 3.87s/it] {'loss': 1.3427, 'grad_norm': 0.0039043971110320104, 'learning_rate': 0.14891650011092894, 'epoch': 0.36} + 36%|███▌ | 186/520 [11:43<21:31, 3.87s/it] 36%|███▌ | 187/520 [11:47<21:27, 3.87s/it] {'loss': 1.3655, 'grad_norm': 0.00527975233936064, 'learning_rate': 0.14837188871052398, 'epoch': 0.36} + 36%|███▌ | 187/520 [11:47<21:27, 3.87s/it] 36%|███▌ | 188/520 [11:51<21:22, 3.86s/it] {'loss': 1.4302, 'grad_norm': 0.004107129961254606, 'learning_rate': 0.14782539786213184, 'epoch': 0.36} + 36%|███▌ | 188/520 [11:51<21:22, 3.86s/it] 36%|███▋ | 189/520 [11:55<21:21, 3.87s/it] {'loss': 1.4563, 'grad_norm': 0.0037073251873678474, 'learning_rate': 0.1472770487991827, 'epoch': 0.36} + 36%|███▋ | 189/520 [11:55<21:21, 3.87s/it] 37%|███▋ | 190/520 [11:58<21:18, 3.88s/it] {'loss': 1.3635, 'grad_norm': 0.004243657206603321, 'learning_rate': 0.1467268628273062, 'epoch': 0.37} + 37%|███▋ | 190/520 [11:58<21:18, 3.88s/it] 37%|███▋ | 191/520 [12:02<21:14, 3.87s/it] {'loss': 1.3216, 'grad_norm': 0.0038059811227130454, 'learning_rate': 0.1461748613235034, 'epoch': 0.37} + 37%|███▋ | 191/520 [12:02<21:14, 3.87s/it] 37%|███▋ | 192/520 [12:06<21:10, 3.87s/it] {'loss': 1.4159, 'grad_norm': 0.003817915647264265, 'learning_rate': 0.1456210657353163, 'epoch': 0.37} + 37%|███▋ | 192/520 [12:06<21:10, 3.87s/it] 37%|███▋ | 193/520 [12:10<21:04, 3.87s/it] {'loss': 1.5492, 'grad_norm': 0.0051787141461298, 'learning_rate': 0.14506549757999454, 'epoch': 0.37} + 37%|███▋ | 193/520 [12:10<21:04, 3.87s/it] 37%|███▋ | 194/520 [12:14<21:02, 3.87s/it] {'loss': 1.3922, 'grad_norm': 0.004266785709745161, 'learning_rate': 0.14450817844365924, 'epoch': 0.37} + 37%|███▋ | 194/520 [12:14<21:02, 3.87s/it] 38%|███▊ | 195/520 [12:18<20:58, 3.87s/it] {'loss': 1.4345, 'grad_norm': 0.003900820012857035, 'learning_rate': 0.1439491299804645, 'epoch': 0.38} + 38%|███▊ | 195/520 [12:18<20:58, 3.87s/it] 38%|███▊ | 196/520 [12:22<20:53, 3.87s/it] {'loss': 1.4041, 'grad_norm': 0.004577688891111747, 'learning_rate': 0.14338837391175582, 'epoch': 0.38} + 38%|███▊ | 196/520 [12:22<20:53, 3.87s/it] 38%|███▊ | 197/520 [12:26<20:47, 3.86s/it] {'loss': 1.3568, 'grad_norm': 0.003798623831610301, 'learning_rate': 0.14282593202522628, 'epoch': 0.38} + 38%|███▊ | 197/520 [12:26<20:47, 3.86s/it] 38%|███▊ | 198/520 [12:29<20:44, 3.87s/it] {'loss': 1.4352, 'grad_norm': 0.004106738889405534, 'learning_rate': 0.14226182617406996, 'epoch': 0.38} + 38%|███▊ | 198/520 [12:29<20:44, 3.87s/it] 38%|███▊ | 199/520 [12:33<20:33, 3.84s/it] {'loss': 1.3443, 'grad_norm': 0.0038517778758432953, 'learning_rate': 0.14169607827613281, 'epoch': 0.38} + 38%|███▊ | 199/520 [12:33<20:33, 3.84s/it] 38%|███▊ | 200/520 [12:37<20:22, 3.82s/it] {'loss': 1.4584, 'grad_norm': 0.004496041902089933, 'learning_rate': 0.14112871031306118, 'epoch': 0.38} + 38%|███▊ | 200/520 [12:37<20:22, 3.82s/it] 39%|███▊ | 201/520 [12:41<20:09, 3.79s/it] {'loss': 1.4657, 'grad_norm': 0.003756081986016175, 'learning_rate': 0.1405597443294475, 'epoch': 0.39} + 39%|███▊ | 201/520 [12:41<20:09, 3.79s/it] 39%|███▉ | 202/520 [12:44<20:05, 3.79s/it] {'loss': 1.3288, 'grad_norm': 0.0037808653632759783, 'learning_rate': 0.13998920243197407, 'epoch': 0.39} + 39%|███▉ | 202/520 [12:44<20:05, 3.79s/it] 39%|███▉ | 203/520 [12:48<19:52, 3.76s/it] {'loss': 1.392, 'grad_norm': 0.0041996761791189266, 'learning_rate': 0.13941710678855396, 'epoch': 0.39} + 39%|███▉ | 203/520 [12:48<19:52, 3.76s/it] 39%|███▉ | 204/520 [12:52<19:44, 3.75s/it] {'loss': 1.4395, 'grad_norm': 0.004274012758882297, 'learning_rate': 0.13884347962746948, 'epoch': 0.39} + 39%|███▉ | 204/520 [12:52<19:44, 3.75s/it] 39%|███▉ | 205/520 [12:56<19:31, 3.72s/it] {'loss': 1.4855, 'grad_norm': 0.004308733958591908, 'learning_rate': 0.138268343236509, 'epoch': 0.39} + 39%|███▉ | 205/520 [12:56<19:31, 3.72s/it] 40%|███▉ | 206/520 [12:59<19:26, 3.72s/it] {'loss': 1.4614, 'grad_norm': 0.004085640429127447, 'learning_rate': 0.13769171996210053, 'epoch': 0.4} + 40%|███▉ | 206/520 [12:59<19:26, 3.72s/it] 40%|███▉ | 207/520 [13:03<19:27, 3.73s/it] {'loss': 1.4594, 'grad_norm': 0.00421792160759696, 'learning_rate': 0.1371136322084438, 'epoch': 0.4} + 40%|███▉ | 207/520 [13:03<19:27, 3.73s/it] 40%|████ | 208/520 [13:07<19:25, 3.74s/it] {'loss': 1.4334, 'grad_norm': 0.004408887359749318, 'learning_rate': 0.13653410243663952, 'epoch': 0.4} + 40%|████ | 208/520 [13:07<19:25, 3.74s/it] 40%|████ | 209/520 [13:11<19:23, 3.74s/it] {'loss': 1.3558, 'grad_norm': 0.003999946807501211, 'learning_rate': 0.13595315316381676, 'epoch': 0.4} + 40%|████ | 209/520 [13:11<19:23, 3.74s/it] 40%|████ | 210/520 [13:14<19:18, 3.74s/it] {'loss': 1.4353, 'grad_norm': 0.004302553602603873, 'learning_rate': 0.13537080696225814, 'epoch': 0.4} + 40%|████ | 210/520 [13:14<19:18, 3.74s/it] 41%|████ | 211/520 [13:18<19:13, 3.73s/it] {'loss': 1.4326, 'grad_norm': 0.0036030248215361475, 'learning_rate': 0.13478708645852272, 'epoch': 0.41} + 41%|████ | 211/520 [13:18<19:13, 3.73s/it] 41%|████ | 212/520 [13:22<19:10, 3.74s/it] {'loss': 1.3993, 'grad_norm': 0.003902399397085315, 'learning_rate': 0.1342020143325669, 'epoch': 0.41} + 41%|████ | 212/520 [13:22<19:10, 3.74s/it] 41%|████ | 213/520 [13:25<19:06, 3.73s/it] {'loss': 1.3667, 'grad_norm': 0.004435299988922411, 'learning_rate': 0.13361561331686309, 'epoch': 0.41} + 41%|████ | 213/520 [13:25<19:06, 3.73s/it] 41%|████ | 214/520 [13:29<19:00, 3.73s/it] {'loss': 1.359, 'grad_norm': 0.004159474527062423, 'learning_rate': 0.13302790619551672, 'epoch': 0.41} + 41%|████ | 214/520 [13:29<19:00, 3.73s/it] 41%|████▏ | 215/520 [13:33<18:56, 3.73s/it] {'loss': 1.3778, 'grad_norm': 0.004068326658052392, 'learning_rate': 0.1324389158033807, 'epoch': 0.41} + 41%|████▏ | 215/520 [13:33<18:56, 3.73s/it] 42%|████▏ | 216/520 [13:37<18:52, 3.73s/it] {'loss': 1.2751, 'grad_norm': 0.0037663757267419697, 'learning_rate': 0.13184866502516845, 'epoch': 0.42} + 42%|████▏ | 216/520 [13:37<18:52, 3.73s/it] 42%|████▏ | 217/520 [13:40<18:48, 3.72s/it] {'loss': 1.3997, 'grad_norm': 0.003920965800208431, 'learning_rate': 0.13125717679456447, 'epoch': 0.42} + 42%|████▏ | 217/520 [13:40<18:48, 3.72s/it] 42%|████▏ | 218/520 [13:44<18:45, 3.73s/it] {'loss': 1.4002, 'grad_norm': 0.004179152835552029, 'learning_rate': 0.13066447409333345, 'epoch': 0.42} + 42%|████▏ | 218/520 [13:44<18:45, 3.73s/it] 42%|████▏ | 219/520 [13:48<18:38, 3.72s/it] {'loss': 1.3547, 'grad_norm': 0.003532918840287616, 'learning_rate': 0.1300705799504273, 'epoch': 0.42} + 42%|████▏ | 219/520 [13:48<18:38, 3.72s/it] 42%|████▏ | 220/520 [13:51<18:33, 3.71s/it] {'loss': 1.426, 'grad_norm': 0.004096597826282207, 'learning_rate': 0.12947551744109043, 'epoch': 0.42} + 42%|████▏ | 220/520 [13:51<18:33, 3.71s/it] 42%|████▎ | 221/520 [13:55<18:30, 3.72s/it] {'loss': 1.4011, 'grad_norm': 0.004187173203555469, 'learning_rate': 0.128879309685963, 'epoch': 0.42} + 42%|████▎ | 221/520 [13:55<18:30, 3.72s/it] 43%|████▎ | 222/520 [13:59<18:27, 3.72s/it] {'loss': 1.3033, 'grad_norm': 0.003983711184427142, 'learning_rate': 0.12828197985018275, 'epoch': 0.43} + 43%|████▎ | 222/520 [13:59<18:27, 3.72s/it] 43%|████▎ | 223/520 [14:03<18:24, 3.72s/it] {'loss': 1.295, 'grad_norm': 0.003955758123154568, 'learning_rate': 0.12768355114248495, 'epoch': 0.43} + 43%|████▎ | 223/520 [14:03<18:24, 3.72s/it] 43%|████▎ | 224/520 [14:06<18:24, 3.73s/it] {'loss': 1.6666, 'grad_norm': 0.005392280066934744, 'learning_rate': 0.12708404681430052, 'epoch': 0.43} + 43%|████▎ | 224/520 [14:06<18:24, 3.73s/it] 43%|████▎ | 225/520 [14:10<18:19, 3.73s/it] {'loss': 1.3238, 'grad_norm': 0.004303291809602955, 'learning_rate': 0.1264834901588527, 'epoch': 0.43} + 43%|████▎ | 225/520 [14:10<18:19, 3.73s/it] 43%|████▎ | 226/520 [14:14<18:17, 3.73s/it] {'loss': 1.4223, 'grad_norm': 0.0037373691868304764, 'learning_rate': 0.12588190451025208, 'epoch': 0.43} + 43%|████▎ | 226/520 [14:14<18:17, 3.73s/it] 44%|████▎ | 227/520 [14:18<18:15, 3.74s/it] {'loss': 1.4093, 'grad_norm': 0.003767405392158434, 'learning_rate': 0.12527931324258976, 'epoch': 0.44} + 44%|████▎ | 227/520 [14:18<18:15, 3.74s/it] 44%|████▍ | 228/520 [14:21<18:15, 3.75s/it] {'loss': 1.5892, 'grad_norm': 0.004515131629033444, 'learning_rate': 0.12467573976902935, 'epoch': 0.44} + 44%|████▍ | 228/520 [14:21<18:15, 3.75s/it] 44%|████▍ | 229/520 [14:25<18:04, 3.73s/it] {'loss': 1.39, 'grad_norm': 0.0036595198496152306, 'learning_rate': 0.12407120754089732, 'epoch': 0.44} + 44%|████▍ | 229/520 [14:25<18:04, 3.73s/it] 44%|████▍ | 230/520 [14:29<18:02, 3.73s/it] {'loss': 1.2595, 'grad_norm': 0.004222718560975486, 'learning_rate': 0.12346574004677154, 'epoch': 0.44} + 44%|████▍ | 230/520 [14:29<18:02, 3.73s/it] 44%|████▍ | 231/520 [14:32<17:54, 3.72s/it] {'loss': 1.3136, 'grad_norm': 0.003437693017371007, 'learning_rate': 0.12285936081156897, 'epoch': 0.44} + 44%|████▍ | 231/520 [14:32<17:54, 3.72s/it] 45%|████▍ | 232/520 [14:36<17:45, 3.70s/it] {'loss': 1.6011, 'grad_norm': 0.004488328809803731, 'learning_rate': 0.12225209339563144, 'epoch': 0.45} + 45%|████▍ | 232/520 [14:36<17:45, 3.70s/it] 45%|████▍ | 233/520 [14:40<17:40, 3.70s/it] {'loss': 1.4836, 'grad_norm': 0.004497235887389779, 'learning_rate': 0.12164396139381028, 'epoch': 0.45} + 45%|████▍ | 233/520 [14:40<17:40, 3.70s/it] 45%|████▌ | 234/520 [14:44<17:50, 3.74s/it] {'loss': 1.2755, 'grad_norm': 0.004057694389790942, 'learning_rate': 0.12103498843454959, 'epoch': 0.45} + 45%|████▌ | 234/520 [14:44<17:50, 3.74s/it] 45%|████▌ | 235/520 [14:47<17:36, 3.71s/it] {'loss': 1.3224, 'grad_norm': 0.0043114320882115245, 'learning_rate': 0.12042519817896805, 'epoch': 0.45} + 45%|████▌ | 235/520 [14:47<17:36, 3.71s/it] 45%|████▌ | 236/520 [14:51<17:27, 3.69s/it] {'loss': 1.4521, 'grad_norm': 0.003766262660871739, 'learning_rate': 0.11981461431993977, 'epoch': 0.45} + 45%|████▌ | 236/520 [14:51<17:27, 3.69s/it] 46%|████▌ | 237/520 [14:55<17:20, 3.68s/it] {'loss': 1.4047, 'grad_norm': 0.003873195307782829, 'learning_rate': 0.11920326058117364, 'epoch': 0.46} + 46%|████▌ | 237/520 [14:55<17:20, 3.68s/it] 46%|████▌ | 238/520 [14:58<17:15, 3.67s/it] {'loss': 1.347, 'grad_norm': 0.0039424829847193405, 'learning_rate': 0.11859116071629149, 'epoch': 0.46} + 46%|████▌ | 238/520 [14:58<17:15, 3.67s/it] 46%|████▌ | 239/520 [15:02<17:10, 3.67s/it] {'loss': 1.4502, 'grad_norm': 0.004017456469709424, 'learning_rate': 0.11797833850790528, 'epoch': 0.46} + 46%|████▌ | 239/520 [15:02<17:10, 3.67s/it] 46%|████▌ | 240/520 [15:06<17:08, 3.67s/it] {'loss': 1.2093, 'grad_norm': 0.003822993544853384, 'learning_rate': 0.11736481776669305, 'epoch': 0.46} + 46%|████▌ | 240/520 [15:06<17:08, 3.67s/it] 46%|████▋ | 241/520 [15:09<17:06, 3.68s/it] {'loss': 1.2982, 'grad_norm': 0.00390924819775839, 'learning_rate': 0.11675062233047365, 'epoch': 0.46} + 46%|████▋ | 241/520 [15:09<17:06, 3.68s/it] 47%|████▋ | 242/520 [15:13<17:05, 3.69s/it] {'loss': 1.3173, 'grad_norm': 0.0035115529503933194, 'learning_rate': 0.11613577606328068, 'epoch': 0.47} + 47%|████▋ | 242/520 [15:13<17:05, 3.69s/it] 47%|████▋ | 243/520 [15:17<16:56, 3.67s/it] {'loss': 1.3084, 'grad_norm': 0.003912730758655926, 'learning_rate': 0.11552030285443515, 'epoch': 0.47} + 47%|████▋ | 243/520 [15:17<16:56, 3.67s/it] 47%|████▋ | 244/520 [15:20<16:52, 3.67s/it] {'loss': 1.4448, 'grad_norm': 0.0040838290791207965, 'learning_rate': 0.11490422661761744, 'epoch': 0.47} + 47%|████▋ | 244/520 [15:20<16:52, 3.67s/it] 47%|████▋ | 245/520 [15:24<16:48, 3.67s/it] {'loss': 1.2961, 'grad_norm': 0.00391046312503565, 'learning_rate': 0.11428757128993801, 'epoch': 0.47} + 47%|████▋ | 245/520 [15:24<16:48, 3.67s/it] 47%|████▋ | 246/520 [15:28<16:45, 3.67s/it] {'loss': 1.5799, 'grad_norm': 0.0044828409112120295, 'learning_rate': 0.11367036083100734, 'epoch': 0.47} + 47%|████▋ | 246/520 [15:28<16:45, 3.67s/it] 48%|████▊ | 247/520 [15:31<16:45, 3.68s/it] {'loss': 1.4787, 'grad_norm': 0.004096177873219058, 'learning_rate': 0.11305261922200519, 'epoch': 0.47} + 48%|████▊ | 247/520 [15:31<16:45, 3.68s/it] 48%|████▊ | 248/520 [15:35<16:40, 3.68s/it] {'loss': 1.289, 'grad_norm': 0.0038328837246215954, 'learning_rate': 0.11243437046474854, 'epoch': 0.48} + 48%|████▊ | 248/520 [15:35<16:40, 3.68s/it] 48%|████▊ | 249/520 [15:39<16:36, 3.68s/it] {'loss': 1.4096, 'grad_norm': 0.0039793749585519764, 'learning_rate': 0.1118156385807593, 'epoch': 0.48} + 48%|████▊ | 249/520 [15:39<16:36, 3.68s/it] 48%|████▊ | 250/520 [15:42<16:35, 3.69s/it] {'loss': 1.3467, 'grad_norm': 0.00406760918977298, 'learning_rate': 0.11119644761033079, 'epoch': 0.48} + 48%|████▊ | 250/520 [15:42<16:35, 3.69s/it] 48%|████▊ | 251/520 [15:46<16:32, 3.69s/it] {'loss': 1.4064, 'grad_norm': 0.0035856605539323658, 'learning_rate': 0.1105768216115938, 'epoch': 0.48} + 48%|████▊ | 251/520 [15:46<16:32, 3.69s/it] 48%|████▊ | 252/520 [15:50<16:28, 3.69s/it] {'loss': 1.4498, 'grad_norm': 0.003949759657704636, 'learning_rate': 0.10995678465958168, 'epoch': 0.48} + 48%|████▊ | 252/520 [15:50<16:28, 3.69s/it] 49%|████▊ | 253/520 [15:53<16:24, 3.69s/it] {'loss': 1.4002, 'grad_norm': 0.004030209169223663, 'learning_rate': 0.10933636084529506, 'epoch': 0.49} + 49%|████▊ | 253/520 [15:53<16:24, 3.69s/it] 49%|████▉ | 254/520 [15:57<16:17, 3.68s/it] {'loss': 1.3131, 'grad_norm': 0.0035045853579638775, 'learning_rate': 0.10871557427476584, 'epoch': 0.49} + 49%|████▉ | 254/520 [15:57<16:17, 3.68s/it] 49%|████▉ | 255/520 [16:01<16:13, 3.67s/it] {'loss': 1.3355, 'grad_norm': 0.004011101629227388, 'learning_rate': 0.10809444906812034, 'epoch': 0.49} + 49%|████▉ | 255/520 [16:01<16:13, 3.67s/it] 49%|████▉ | 256/520 [16:04<16:11, 3.68s/it] {'loss': 1.3759, 'grad_norm': 0.0039029773318284785, 'learning_rate': 0.10747300935864244, 'epoch': 0.49} + 49%|████▉ | 256/520 [16:04<16:11, 3.68s/it] 49%|████▉ | 257/520 [16:08<16:05, 3.67s/it] {'loss': 1.3751, 'grad_norm': 0.003792015870505506, 'learning_rate': 0.10685127929183567, 'epoch': 0.49} + 49%|████▉ | 257/520 [16:08<16:05, 3.67s/it] 50%|████▉ | 258/520 [16:12<16:01, 3.67s/it] {'loss': 1.386, 'grad_norm': 0.0033259949969045082, 'learning_rate': 0.10622928302448523, 'epoch': 0.5} + 50%|████▉ | 258/520 [16:12<16:01, 3.67s/it] 50%|████▉ | 259/520 [16:15<15:57, 3.67s/it] {'loss': 1.4547, 'grad_norm': 0.004278029232527598, 'learning_rate': 0.10560704472371918, 'epoch': 0.5} + 50%|████▉ | 259/520 [16:15<15:57, 3.67s/it] 50%|█████ | 260/520 [16:19<15:55, 3.68s/it] {'loss': 1.5522, 'grad_norm': 0.003989376567048945, 'learning_rate': 0.10498458856606972, 'epoch': 0.5} + 50%|█████ | 260/520 [16:19<15:55, 3.68s/it] 50%|█████ | 261/520 [16:23<15:51, 3.67s/it] {'loss': 1.4665, 'grad_norm': 0.004087457570295507, 'learning_rate': 0.10436193873653361, 'epoch': 0.5} + 50%|█████ | 261/520 [16:23<15:51, 3.67s/it] 50%|█████ | 262/520 [16:26<15:46, 3.67s/it] {'loss': 1.2883, 'grad_norm': 0.0037757434980154564, 'learning_rate': 0.10373911942763259, 'epoch': 0.5} + 50%|█████ | 262/520 [16:26<15:46, 3.67s/it] 51%|█████ | 263/520 [16:30<15:41, 3.66s/it] {'loss': 1.4682, 'grad_norm': 0.0041922395270319514, 'learning_rate': 0.10311615483847332, 'epoch': 0.51} + 51%|█████ | 263/520 [16:30<15:41, 3.66s/it] 51%|█████ | 264/520 [16:34<15:37, 3.66s/it] {'loss': 1.413, 'grad_norm': 0.0037370735231788605, 'learning_rate': 0.1024930691738073, 'epoch': 0.51} + 51%|█████ | 264/520 [16:34<15:37, 3.66s/it] 51%|█████ | 265/520 [16:37<15:32, 3.66s/it] {'loss': 1.3027, 'grad_norm': 0.00435036487640513, 'learning_rate': 0.10186988664309023, 'epoch': 0.51} + 51%|█████ | 265/520 [16:37<15:32, 3.66s/it] 51%|█████ | 266/520 [16:41<15:27, 3.65s/it] {'loss': 1.1482, 'grad_norm': 0.003258381757716239, 'learning_rate': 0.10124663145954152, 'epoch': 0.51} + 51%|█████ | 266/520 [16:41<15:27, 3.65s/it] 51%|█████▏ | 267/520 [16:45<15:39, 3.71s/it] {'loss': 1.2981, 'grad_norm': 0.0036659327175111024, 'learning_rate': 0.10062332783920336, 'epoch': 0.51} + 51%|█████▏ | 267/520 [16:45<15:39, 3.71s/it] 52%|█████▏ | 268/520 [16:49<15:44, 3.75s/it] {'loss': 1.6013, 'grad_norm': 0.004637436954191689, 'learning_rate': 0.1, 'epoch': 0.52} + 52%|█████▏ | 268/520 [16:49<15:44, 3.75s/it] 52%|█████▏ | 269/520 [16:53<15:45, 3.77s/it] {'loss': 1.4172, 'grad_norm': 0.003966642754001942, 'learning_rate': 0.09937667216079665, 'epoch': 0.52} + 52%|█████▏ | 269/520 [16:53<15:45, 3.77s/it] 52%|█████▏ | 270/520 [16:56<15:46, 3.79s/it] {'loss': 1.3837, 'grad_norm': 0.003946880161346753, 'learning_rate': 0.0987533685404585, 'epoch': 0.52} + 52%|█████▏ | 270/520 [16:56<15:46, 3.79s/it] 52%|█████▏ | 271/520 [17:00<15:46, 3.80s/it] {'loss': 1.3979, 'grad_norm': 0.003853456446003649, 'learning_rate': 0.0981301133569098, 'epoch': 0.52} + 52%|█████▏ | 271/520 [17:00<15:46, 3.80s/it] 52%|█████▏ | 272/520 [17:04<15:44, 3.81s/it] {'loss': 1.3885, 'grad_norm': 0.004207679465484306, 'learning_rate': 0.09750693082619273, 'epoch': 0.52} + 52%|█████▏ | 272/520 [17:04<15:44, 3.81s/it] 52%|█████▎ | 273/520 [17:08<15:38, 3.80s/it] {'loss': 1.5197, 'grad_norm': 0.004244254248369329, 'learning_rate': 0.0968838451615267, 'epoch': 0.53} + 52%|█████▎ | 273/520 [17:08<15:38, 3.80s/it] 53%|█████▎ | 274/520 [17:12<15:35, 3.80s/it] {'loss': 1.3583, 'grad_norm': 0.0040761315571065555, 'learning_rate': 0.09626088057236745, 'epoch': 0.53} + 53%|█████▎ | 274/520 [17:12<15:35, 3.80s/it] 53%|█████▎ | 275/520 [17:15<15:33, 3.81s/it] {'loss': 1.2943, 'grad_norm': 0.004665514638963983, 'learning_rate': 0.09563806126346641, 'epoch': 0.53} + 53%|█████▎ | 275/520 [17:15<15:33, 3.81s/it] 53%|█████▎ | 276/520 [17:19<15:30, 3.81s/it] {'loss': 1.39, 'grad_norm': 0.004220348097989554, 'learning_rate': 0.09501541143393027, 'epoch': 0.53} + 53%|█████▎ | 276/520 [17:19<15:30, 3.81s/it] 53%|█████▎ | 277/520 [17:23<15:29, 3.83s/it] {'loss': 1.5146, 'grad_norm': 0.0046452403465732294, 'learning_rate': 0.09439295527628082, 'epoch': 0.53} + 53%|█████▎ | 277/520 [17:23<15:29, 3.83s/it] 53%|█████▎ | 278/520 [17:27<15:24, 3.82s/it] {'loss': 1.2476, 'grad_norm': 0.003488519336103931, 'learning_rate': 0.0937707169755148, 'epoch': 0.53} + 53%|█████▎ | 278/520 [17:27<15:24, 3.82s/it] 54%|█████▎ | 279/520 [17:31<15:18, 3.81s/it] {'loss': 1.4155, 'grad_norm': 0.004641959794284071, 'learning_rate': 0.09314872070816434, 'epoch': 0.54} + 54%|█████▎ | 279/520 [17:31<15:18, 3.81s/it] 54%|█████▍ | 280/520 [17:35<15:17, 3.82s/it] {'loss': 1.3133, 'grad_norm': 0.004330124822384814, 'learning_rate': 0.09252699064135758, 'epoch': 0.54} + 54%|█████▍ | 280/520 [17:35<15:17, 3.82s/it] 54%|█████▍ | 281/520 [17:38<15:13, 3.82s/it] {'loss': 1.4208, 'grad_norm': 0.00417369438487931, 'learning_rate': 0.09190555093187967, 'epoch': 0.54} + 54%|█████▍ | 281/520 [17:38<15:13, 3.82s/it] 54%|█████▍ | 282/520 [17:42<15:10, 3.82s/it] {'loss': 1.2631, 'grad_norm': 0.003690631067024653, 'learning_rate': 0.09128442572523418, 'epoch': 0.54} + 54%|█████▍ | 282/520 [17:42<15:10, 3.82s/it] 54%|█████▍ | 283/520 [17:46<14:57, 3.79s/it] {'loss': 1.4409, 'grad_norm': 0.004292726317065164, 'learning_rate': 0.09066363915470495, 'epoch': 0.54} + 54%|█████▍ | 283/520 [17:46<14:57, 3.79s/it] 55%|█████▍ | 284/520 [17:50<14:48, 3.76s/it] {'loss': 1.3827, 'grad_norm': 0.004247851596264772, 'learning_rate': 0.09004321534041836, 'epoch': 0.55} + 55%|█████▍ | 284/520 [17:50<14:48, 3.76s/it] 55%|█████▍ | 285/520 [17:53<14:37, 3.74s/it] {'loss': 1.2909, 'grad_norm': 0.003905494226716533, 'learning_rate': 0.08942317838840624, 'epoch': 0.55} + 55%|█████▍ | 285/520 [17:53<14:37, 3.74s/it] 55%|█████▌ | 286/520 [17:57<14:29, 3.71s/it] {'loss': 1.157, 'grad_norm': 0.003940332418573554, 'learning_rate': 0.08880355238966922, 'epoch': 0.55} + 55%|█████▌ | 286/520 [17:57<14:29, 3.71s/it] 55%|█████▌ | 287/520 [18:01<14:20, 3.70s/it] {'loss': 1.4142, 'grad_norm': 0.003998276779253748, 'learning_rate': 0.08818436141924073, 'epoch': 0.55} + 55%|█████▌ | 287/520 [18:01<14:20, 3.70s/it] 55%|█████▌ | 288/520 [18:04<14:16, 3.69s/it] {'loss': 1.4618, 'grad_norm': 0.00433547095724735, 'learning_rate': 0.08756562953525152, 'epoch': 0.55} + 55%|█████▌ | 288/520 [18:04<14:16, 3.69s/it] 56%|█████▌ | 289/520 [18:08<14:09, 3.68s/it] {'loss': 1.3114, 'grad_norm': 0.003630855260574458, 'learning_rate': 0.08694738077799487, 'epoch': 0.56} + 56%|█████▌ | 289/520 [18:08<14:09, 3.68s/it] 56%|█████▌ | 290/520 [18:12<14:07, 3.68s/it] {'loss': 1.2145, 'grad_norm': 0.003449028348013516, 'learning_rate': 0.08632963916899268, 'epoch': 0.56} + 56%|█████▌ | 290/520 [18:12<14:07, 3.68s/it] 56%|█████▌ | 291/520 [18:15<14:04, 3.69s/it] {'loss': 1.2803, 'grad_norm': 0.004342751203104316, 'learning_rate': 0.08571242871006202, 'epoch': 0.56} + 56%|█████▌ | 291/520 [18:15<14:04, 3.69s/it] 56%|█████▌ | 292/520 [18:19<13:57, 3.67s/it] {'loss': 1.3221, 'grad_norm': 0.003738362273853977, 'learning_rate': 0.08509577338238256, 'epoch': 0.56} + 56%|█████▌ | 292/520 [18:19<13:57, 3.67s/it] 56%|█████▋ | 293/520 [18:23<13:52, 3.67s/it] {'loss': 1.274, 'grad_norm': 0.004239647706512972, 'learning_rate': 0.08447969714556484, 'epoch': 0.56} + 56%|█████▋ | 293/520 [18:23<13:52, 3.67s/it] 57%|█████▋ | 294/520 [18:26<13:47, 3.66s/it] {'loss': 1.2989, 'grad_norm': 0.004126574405480324, 'learning_rate': 0.08386422393671933, 'epoch': 0.57} + 57%|█████▋ | 294/520 [18:26<13:47, 3.66s/it] 57%|█████▋ | 295/520 [18:30<13:42, 3.65s/it] {'loss': 1.4612, 'grad_norm': 0.004795650115337895, 'learning_rate': 0.08324937766952638, 'epoch': 0.57} + 57%|█████▋ | 295/520 [18:30<13:42, 3.65s/it] 57%|█████▋ | 296/520 [18:34<13:46, 3.69s/it] {'loss': 1.2438, 'grad_norm': 0.00424652063705415, 'learning_rate': 0.08263518223330697, 'epoch': 0.57} + 57%|█████▋ | 296/520 [18:34<13:46, 3.69s/it] 57%|█████▋ | 297/520 [18:38<13:52, 3.73s/it] {'loss': 1.3799, 'grad_norm': 0.004413605247207849, 'learning_rate': 0.08202166149209474, 'epoch': 0.57} + 57%|█████▋ | 297/520 [18:38<13:52, 3.73s/it] 57%|█████▋ | 298/520 [18:41<13:56, 3.77s/it] {'loss': 1.3418, 'grad_norm': 0.0035017142530602693, 'learning_rate': 0.08140883928370855, 'epoch': 0.57} + 57%|█████▋ | 298/520 [18:41<13:56, 3.77s/it] 57%|█████▊ | 299/520 [18:45<13:57, 3.79s/it] {'loss': 1.4463, 'grad_norm': 0.0038201331895595303, 'learning_rate': 0.0807967394188264, 'epoch': 0.57} + 57%|█████▊ | 299/520 [18:45<13:57, 3.79s/it] 58%|█████▊ | 300/520 [18:49<13:57, 3.81s/it] {'loss': 1.4032, 'grad_norm': 0.004056891977514212, 'learning_rate': 0.08018538568006027, 'epoch': 0.58} + 58%|█████▊ | 300/520 [18:49<13:57, 3.81s/it] 58%|█████▊ | 301/520 [18:53<13:56, 3.82s/it] {'loss': 1.3653, 'grad_norm': 0.003859711139887944, 'learning_rate': 0.07957480182103199, 'epoch': 0.58} + 58%|█████▊ | 301/520 [18:53<13:56, 3.82s/it] 58%|█████▊ | 302/520 [18:57<13:53, 3.82s/it] {'loss': 1.4736, 'grad_norm': 0.0038463143540589046, 'learning_rate': 0.07896501156545044, 'epoch': 0.58} + 58%|█████▊ | 302/520 [18:57<13:53, 3.82s/it] 58%|█████▊ | 303/520 [19:01<13:52, 3.84s/it] {'loss': 1.3038, 'grad_norm': 0.004160523572422393, 'learning_rate': 0.07835603860618973, 'epoch': 0.58} + 58%|█████▊ | 303/520 [19:01<13:52, 3.84s/it] 58%|█████▊ | 304/520 [19:04<13:45, 3.82s/it] {'loss': 1.3662, 'grad_norm': 0.004337598970698155, 'learning_rate': 0.07774790660436857, 'epoch': 0.58} + 58%|█████▊ | 304/520 [19:04<13:45, 3.82s/it] 59%|█████▊ | 305/520 [19:08<13:31, 3.77s/it] {'loss': 1.4212, 'grad_norm': 0.004461962559637543, 'learning_rate': 0.07714063918843106, 'epoch': 0.59} + 59%|█████▊ | 305/520 [19:08<13:31, 3.77s/it] 59%|█████▉ | 306/520 [19:12<13:25, 3.77s/it] {'loss': 1.34, 'grad_norm': 0.00395264334345113, 'learning_rate': 0.0765342599532285, 'epoch': 0.59} + 59%|█████▉ | 306/520 [19:12<13:25, 3.77s/it] 59%|█████▉ | 307/520 [19:16<13:49, 3.89s/it] {'loss': 1.2896, 'grad_norm': 0.0036011712666764737, 'learning_rate': 0.07592879245910272, 'epoch': 0.59} + 59%|█████▉ | 307/520 [19:16<13:49, 3.89s/it] 59%|█████▉ | 308/520 [19:20<13:39, 3.87s/it] {'loss': 1.4054, 'grad_norm': 0.0038479416873348532, 'learning_rate': 0.07532426023097064, 'epoch': 0.59} + 59%|█████▉ | 308/520 [19:20<13:39, 3.87s/it] 59%|█████▉ | 309/520 [19:23<13:23, 3.81s/it] {'loss': 1.2799, 'grad_norm': 0.0037274903075944876, 'learning_rate': 0.07472068675741024, 'epoch': 0.59} + 59%|█████▉ | 309/520 [19:23<13:23, 3.81s/it] 60%|█████▉ | 310/520 [19:27<13:09, 3.76s/it] {'loss': 1.2614, 'grad_norm': 0.0037710756744785243, 'learning_rate': 0.07411809548974792, 'epoch': 0.6} + 60%|█████▉ | 310/520 [19:27<13:09, 3.76s/it] 60%|█████▉ | 311/520 [19:31<12:59, 3.73s/it] {'loss': 1.2257, 'grad_norm': 0.003733641030205925, 'learning_rate': 0.07351650984114727, 'epoch': 0.6} + 60%|█████▉ | 311/520 [19:31<12:59, 3.73s/it] 60%|██████ | 312/520 [19:34<12:53, 3.72s/it] {'loss': 1.2189, 'grad_norm': 0.004382253313725207, 'learning_rate': 0.0729159531856995, 'epoch': 0.6} + 60%|██████ | 312/520 [19:34<12:53, 3.72s/it] 60%|██████ | 313/520 [19:38<12:45, 3.70s/it] {'loss': 1.2114, 'grad_norm': 0.0033585023109627826, 'learning_rate': 0.07231644885751508, 'epoch': 0.6} + 60%|██████ | 313/520 [19:38<12:45, 3.70s/it] 60%|██████ | 314/520 [19:42<13:11, 3.84s/it] {'loss': 1.247, 'grad_norm': 0.003464013230295214, 'learning_rate': 0.07171802014981725, 'epoch': 0.6} + 60%|██████ | 314/520 [19:42<13:11, 3.84s/it] 61%|██████ | 315/520 [19:46<13:10, 3.86s/it] {'loss': 1.4575, 'grad_norm': 0.00480257098361353, 'learning_rate': 0.07112069031403703, 'epoch': 0.61} + 61%|██████ | 315/520 [19:46<13:10, 3.86s/it] 61%|██████ | 316/520 [19:51<13:36, 4.00s/it] {'loss': 1.218, 'grad_norm': 0.004323575401060492, 'learning_rate': 0.07052448255890957, 'epoch': 0.61} + 61%|██████ | 316/520 [19:51<13:36, 4.00s/it] 61%|██████ | 317/520 [19:54<13:26, 3.97s/it] {'loss': 1.2448, 'grad_norm': 0.0034886257423520743, 'learning_rate': 0.0699294200495727, 'epoch': 0.61} + 61%|██████ | 317/520 [33:24<13:26, 3.97s/it] 61%|██████ | 318/520 [33:28<13:50:53, 246.80s/it] {'loss': 1.3757, 'grad_norm': 0.004144810999560052, 'learning_rate': 0.06933552590666658, 'epoch': 0.61} + 61%|██████ | 318/520 [33:28<13:50:53, 246.80s/it] 61%|██████▏ | 319/520 [33:32<9:42:51, 173.99s/it] {'loss': 1.2268, 'grad_norm': 0.0035664080611272795, 'learning_rate': 0.06874282320543557, 'epoch': 0.61} + 61%|██████▏ | 319/520 [33:32<9:42:51, 173.99s/it] 62%|██████▏ | 320/520 [33:36<6:49:38, 122.89s/it] {'loss': 1.1755, 'grad_norm': 0.003991537010960863, 'learning_rate': 0.06815133497483157, 'epoch': 0.62} + 62%|██████▏ | 320/520 [33:36<6:49:38, 122.89s/it] 62%|██████▏ | 321/520 [33:39<4:48:59, 87.13s/it] {'loss': 1.3809, 'grad_norm': 0.00388643586337685, 'learning_rate': 0.06756108419661931, 'epoch': 0.62} + 62%|██████▏ | 321/520 [33:39<4:48:59, 87.13s/it] 62%|██████▏ | 322/520 [33:43<3:25:00, 62.13s/it] {'loss': 1.287, 'grad_norm': 0.004000324427260342, 'learning_rate': 0.06697209380448332, 'epoch': 0.62} + 62%|██████▏ | 322/520 [33:43<3:25:00, 62.13s/it] 62%|██████▏ | 323/520 [33:47<2:26:24, 44.59s/it] {'loss': 1.3806, 'grad_norm': 0.004245022903895392, 'learning_rate': 0.06638438668313694, 'epoch': 0.62} + 62%|██████▏ | 323/520 [33:47<2:26:24, 44.59s/it] 62%|██████▏ | 324/520 [33:50<1:45:35, 32.32s/it] {'loss': 1.3067, 'grad_norm': 0.004136711390002879, 'learning_rate': 0.06579798566743314, 'epoch': 0.62} + 62%|██████▏ | 324/520 [33:50<1:45:35, 32.32s/it] 62%|██████▎ | 325/520 [33:54<1:17:05, 23.72s/it] {'loss': 1.3193, 'grad_norm': 0.004134895647358979, 'learning_rate': 0.06521291354147728, 'epoch': 0.62} + 62%|██████▎ | 325/520 [33:54<1:17:05, 23.72s/it] 63%|██████▎ | 326/520 [33:58<57:17, 17.72s/it] {'loss': 1.2973, 'grad_norm': 0.0038516584700617443, 'learning_rate': 0.06462919303774187, 'epoch': 0.63} + 63%|██████▎ | 326/520 [33:58<57:17, 17.72s/it] 63%|██████▎ | 327/520 [34:02<43:27, 13.51s/it] {'loss': 1.46, 'grad_norm': 0.004998801322214398, 'learning_rate': 0.06404684683618325, 'epoch': 0.63} + 63%|██████▎ | 327/520 [34:02<43:27, 13.51s/it] 63%|██████▎ | 328/520 [34:05<33:48, 10.56s/it] {'loss': 1.3719, 'grad_norm': 0.003944838822641901, 'learning_rate': 0.0634658975633605, 'epoch': 0.63} + 63%|██████▎ | 328/520 [34:05<33:48, 10.56s/it] 63%|██████▎ | 329/520 [34:09<27:03, 8.50s/it] {'loss': 1.22, 'grad_norm': 0.0033295867124965312, 'learning_rate': 0.06288636779155621, 'epoch': 0.63} + 63%|██████▎ | 329/520 [34:09<27:03, 8.50s/it] 63%|██████▎ | 330/520 [34:13<22:22, 7.07s/it] {'loss': 1.2951, 'grad_norm': 0.003452344288059599, 'learning_rate': 0.06230828003789948, 'epoch': 0.63} + 63%|██████▎ | 330/520 [34:13<22:22, 7.07s/it] 64%|██████▎ | 331/520 [34:16<19:04, 6.06s/it] {'loss': 1.2575, 'grad_norm': 0.0037520626601595824, 'learning_rate': 0.06173165676349103, 'epoch': 0.64} + 64%|██████▎ | 331/520 [34:16<19:04, 6.06s/it] 64%|██████▍ | 332/520 [34:20<16:46, 5.35s/it] {'loss': 1.4489, 'grad_norm': 0.00404125226848044, 'learning_rate': 0.06115652037253053, 'epoch': 0.64} + 64%|██████▍ | 332/520 [34:20<16:46, 5.35s/it] 64%|██████▍ | 333/520 [34:24<15:08, 4.86s/it] {'loss': 1.4356, 'grad_norm': 0.004174572657162631, 'learning_rate': 0.06058289321144608, 'epoch': 0.64} + 64%|██████▍ | 333/520 [34:24<15:08, 4.86s/it] 64%|██████▍ | 334/520 [34:27<13:57, 4.50s/it] {'loss': 1.3078, 'grad_norm': 0.004276572887017878, 'learning_rate': 0.06001079756802592, 'epoch': 0.64} + 64%|██████▍ | 334/520 [34:27<13:57, 4.50s/it] 64%|██████▍ | 335/520 [34:31<13:06, 4.25s/it] {'loss': 1.3063, 'grad_norm': 0.0035750049667495473, 'learning_rate': 0.059440255670552514, 'epoch': 0.64} + 64%|██████▍ | 335/520 [34:31<13:06, 4.25s/it] 65%|██████▍ | 336/520 [34:35<12:31, 4.08s/it] {'loss': 1.1898, 'grad_norm': 0.004081001357223521, 'learning_rate': 0.05887128968693887, 'epoch': 0.65} + 65%|██████▍ | 336/520 [34:35<12:31, 4.08s/it] 65%|██████▍ | 337/520 [34:38<12:04, 3.96s/it] {'loss': 1.1889, 'grad_norm': 0.0036347224949424984, 'learning_rate': 0.058303921723867225, 'epoch': 0.65} + 65%|██████▍ | 337/520 [34:38<12:04, 3.96s/it] 65%|██████▌ | 338/520 [34:42<11:47, 3.89s/it] {'loss': 1.3233, 'grad_norm': 0.003933914122177299, 'learning_rate': 0.05773817382593008, 'epoch': 0.65} + 65%|██████▌ | 338/520 [34:42<11:47, 3.89s/it] 65%|██████▌ | 339/520 [34:46<11:31, 3.82s/it] {'loss': 1.255, 'grad_norm': 0.0037005328842886496, 'learning_rate': 0.057174067974773715, 'epoch': 0.65} + 65%|██████▌ | 339/520 [34:46<11:31, 3.82s/it] 65%|██████▌ | 340/520 [34:50<11:22, 3.79s/it] {'loss': 1.2449, 'grad_norm': 0.003802399816401456, 'learning_rate': 0.056611626088244195, 'epoch': 0.65} + 65%|██████▌ | 340/520 [34:50<11:22, 3.79s/it] 66%|██████▌ | 341/520 [34:53<11:13, 3.76s/it] {'loss': 1.2727, 'grad_norm': 0.004066461192248553, 'learning_rate': 0.056050870019535494, 'epoch': 0.66} + 66%|██████▌ | 341/520 [34:53<11:13, 3.76s/it] 66%|██████▌ | 342/520 [34:57<11:06, 3.75s/it] {'loss': 1.4188, 'grad_norm': 0.004343947258468996, 'learning_rate': 0.05549182155634076, 'epoch': 0.66} + 66%|██████▌ | 342/520 [34:57<11:06, 3.75s/it] 66%|██████▌ | 343/520 [35:01<11:05, 3.76s/it] {'loss': 1.3933, 'grad_norm': 0.004365198103041345, 'learning_rate': 0.054934502420005464, 'epoch': 0.66} + 66%|██████▌ | 343/520 [35:01<11:05, 3.76s/it] 66%|██████▌ | 344/520 [35:04<10:59, 3.75s/it] {'loss': 1.2208, 'grad_norm': 0.0038586901865206476, 'learning_rate': 0.0543789342646837, 'epoch': 0.66} + 66%|██████▌ | 344/520 [35:04<10:59, 3.75s/it] 66%|██████▋ | 345/520 [35:08<10:51, 3.72s/it] {'loss': 1.3381, 'grad_norm': 0.0044404589844918695, 'learning_rate': 0.05382513867649663, 'epoch': 0.66} + 66%|██████▋ | 345/520 [35:08<10:51, 3.72s/it] 67%|██████▋ | 346/520 [35:12<10:45, 3.71s/it] {'loss': 1.3707, 'grad_norm': 0.003816787599974208, 'learning_rate': 0.0532731371726938, 'epoch': 0.67} + 67%|██████▋ | 346/520 [35:12<10:45, 3.71s/it] 67%|██████▋ | 347/520 [35:15<10:38, 3.69s/it] {'loss': 1.2379, 'grad_norm': 0.003635332786677169, 'learning_rate': 0.05272295120081732, 'epoch': 0.67} + 67%|██████▋ | 347/520 [35:15<10:38, 3.69s/it]Token indices sequence length is longer than the specified maximum sequence length for this model (2778 > 2048). Running this sequence through the model will result in indexing errors + 67%|██████▋ | 348/520 [35:19<10:33, 3.69s/it] {'loss': 1.2014, 'grad_norm': 0.00458977509542213, 'learning_rate': 0.05217460213786821, 'epoch': 0.67} + 67%|██████▋ | 348/520 [35:19<10:33, 3.69s/it] 67%|██████▋ | 349/520 [35:23<10:30, 3.68s/it] {'loss': 1.2428, 'grad_norm': 0.0040120174278233086, 'learning_rate': 0.051628111289476024, 'epoch': 0.67} + 67%|██████▋ | 349/520 [35:23<10:30, 3.68s/it] 67%|██████▋ | 350/520 [35:26<10:25, 3.68s/it] {'loss': 1.2751, 'grad_norm': 0.004019807813157819, 'learning_rate': 0.051083499889071106, 'epoch': 0.67} + 67%|██████▋ | 350/520 [35:26<10:25, 3.68s/it] 68%|██████▊ | 351/520 [35:30<10:35, 3.76s/it] {'loss': 1.1797, 'grad_norm': 0.0036225404276690057, 'learning_rate': 0.05054078909705926, 'epoch': 0.68} + 68%|██████▊ | 351/520 [35:30<10:35, 3.76s/it] 68%|██████▊ | 352/520 [35:34<10:40, 3.81s/it] {'loss': 1.3067, 'grad_norm': 0.003637844632022396, 'learning_rate': 0.050000000000000024, 'epoch': 0.68} + 68%|██████▊ | 352/520 [35:34<10:40, 3.81s/it] 68%|██████▊ | 353/520 [35:38<10:44, 3.86s/it] {'loss': 1.3181, 'grad_norm': 0.0034501690932364697, 'learning_rate': 0.04946115360978696, 'epoch': 0.68} + 68%|██████▊ | 353/520 [35:38<10:44, 3.86s/it] 68%|██████▊ | 354/520 [35:42<10:43, 3.88s/it] {'loss': 1.4546, 'grad_norm': 0.00415363625431768, 'learning_rate': 0.048924270862831465, 'epoch': 0.68} + 68%|██████▊ | 354/520 [35:42<10:43, 3.88s/it] 68%|██████▊ | 355/520 [35:46<10:31, 3.83s/it] {'loss': 1.2457, 'grad_norm': 0.0038004196450125926, 'learning_rate': 0.04838937261924933, 'epoch': 0.68} + 68%|██████▊ | 355/520 [35:46<10:31, 3.83s/it] 68%|██████▊ | 356/520 [35:50<10:19, 3.77s/it] {'loss': 1.2437, 'grad_norm': 0.003950396518526643, 'learning_rate': 0.0478564796620502, 'epoch': 0.68} + 68%|██████▊ | 356/520 [35:50<10:19, 3.77s/it] 69%|██████▊ | 357/520 [35:53<10:10, 3.75s/it] {'loss': 1.2626, 'grad_norm': 0.003405570473696854, 'learning_rate': 0.04732561269632993, 'epoch': 0.69} + 69%|██████▊ | 357/520 [35:53<10:10, 3.75s/it] 69%|██████▉ | 358/520 [35:57<10:02, 3.72s/it] {'loss': 1.1952, 'grad_norm': 0.003750337182143702, 'learning_rate': 0.04679679234846636, 'epoch': 0.69} + 69%|██████▉ | 358/520 [35:57<10:02, 3.72s/it] 69%|██████▉ | 359/520 [36:01<10:05, 3.76s/it] {'loss': 1.3797, 'grad_norm': 0.003883867568516182, 'learning_rate': 0.046270039165317606, 'epoch': 0.69} + 69%|██████▉ | 359/520 [36:01<10:05, 3.76s/it] 69%|██████▉ | 360/520 [36:05<10:06, 3.79s/it] {'loss': 1.3965, 'grad_norm': 0.004398294755137541, 'learning_rate': 0.04574537361342407, 'epoch': 0.69} + 69%|██████▉ | 360/520 [36:05<10:06, 3.79s/it] 69%|██████▉ | 361/520 [36:09<10:06, 3.81s/it] {'loss': 1.3868, 'grad_norm': 0.0037923697322435057, 'learning_rate': 0.04522281607821288, 'epoch': 0.69} + 69%|██████▉ | 361/520 [36:09<10:06, 3.81s/it] 70%|██████▉ | 362/520 [36:12<10:04, 3.83s/it] {'loss': 1.2594, 'grad_norm': 0.004051798837894101, 'learning_rate': 0.04470238686320606, 'epoch': 0.7} + 70%|██████▉ | 362/520 [36:12<10:04, 3.83s/it] 70%|██████▉ | 363/520 [36:16<10:01, 3.83s/it] {'loss': 1.2899, 'grad_norm': 0.0037138788461406396, 'learning_rate': 0.044184106189231624, 'epoch': 0.7} + 70%|██████▉ | 363/520 [36:16<10:01, 3.83s/it] 70%|███████ | 364/520 [36:20<10:01, 3.85s/it] {'loss': 1.3986, 'grad_norm': 0.003962902019779408, 'learning_rate': 0.043667994193637795, 'epoch': 0.7} + 70%|███████ | 364/520 [36:20<10:01, 3.85s/it] 70%|███████ | 365/520 [36:24<09:59, 3.87s/it] {'loss': 1.3526, 'grad_norm': 0.0038866450322866393, 'learning_rate': 0.043154070929510784, 'epoch': 0.7} + 70%|███████ | 365/520 [36:24<09:59, 3.87s/it] 70%|███████ | 366/520 [36:28<10:00, 3.90s/it] {'loss': 1.3013, 'grad_norm': 0.003492476072972362, 'learning_rate': 0.04264235636489542, 'epoch': 0.7} + 70%|███████ | 366/520 [36:28<10:00, 3.90s/it] 71%|███████ | 367/520 [36:32<10:01, 3.93s/it] {'loss': 1.3056, 'grad_norm': 0.003780948111684492, 'learning_rate': 0.04213287038201943, 'epoch': 0.71} + 71%|███████ | 367/520 [36:32<10:01, 3.93s/it] 71%|███████ | 368/520 [36:36<10:02, 3.96s/it] {'loss': 1.1464, 'grad_norm': 0.003867257123575523, 'learning_rate': 0.04162563277652104, 'epoch': 0.71} + 71%|███████ | 368/520 [36:36<10:02, 3.96s/it] 71%|███████ | 369/520 [36:40<09:56, 3.95s/it] {'loss': 1.3632, 'grad_norm': 0.0035886433018776074, 'learning_rate': 0.04112066325667954, 'epoch': 0.71} + 71%|███████ | 369/520 [36:40<09:56, 3.95s/it] 71%|███████ | 370/520 [36:44<09:50, 3.94s/it] {'loss': 1.2121, 'grad_norm': 0.003440339519862252, 'learning_rate': 0.04061798144264986, 'epoch': 0.71} + 71%|███████ | 370/520 [36:44<09:50, 3.94s/it] 71%|███████▏ | 371/520 [36:48<09:41, 3.91s/it] {'loss': 1.204, 'grad_norm': 0.00382467024044372, 'learning_rate': 0.04011760686569998, 'epoch': 0.71} + 71%|███████▏ | 371/520 [36:48<09:41, 3.91s/it] 72%|███████▏ | 372/520 [36:52<09:43, 3.94s/it] {'loss': 1.46, 'grad_norm': 0.003563971661864781, 'learning_rate': 0.03961955896745224, 'epoch': 0.72} + 72%|███████▏ | 372/520 [36:52<09:43, 3.94s/it] 72%|███████▏ | 373/520 [36:56<09:36, 3.93s/it] {'loss': 1.3328, 'grad_norm': 0.004056986838851181, 'learning_rate': 0.03912385709912794, 'epoch': 0.72} + 72%|███████▏ | 373/520 [36:56<09:36, 3.93s/it] 72%|███████▏ | 374/520 [37:00<09:31, 3.91s/it] {'loss': 1.3003, 'grad_norm': 0.0036909899506458897, 'learning_rate': 0.038630520520795276, 'epoch': 0.72} + 72%|███████▏ | 374/520 [37:00<09:31, 3.91s/it] 72%|███████▏ | 375/520 [37:03<09:19, 3.86s/it] {'loss': 1.1977, 'grad_norm': 0.0038892292938861556, 'learning_rate': 0.03813956840062119, 'epoch': 0.72} + 72%|███████▏ | 375/520 [37:03<09:19, 3.86s/it] 72%|███████▏ | 376/520 [37:07<09:11, 3.83s/it] {'loss': 1.3251, 'grad_norm': 0.0037487396651421307, 'learning_rate': 0.037651019814126656, 'epoch': 0.72} + 72%|███████▏ | 376/520 [37:07<09:11, 3.83s/it] 72%|███████▎ | 377/520 [37:11<09:05, 3.81s/it] {'loss': 1.2626, 'grad_norm': 0.00393528879706923, 'learning_rate': 0.037164893743445275, 'epoch': 0.72} + 72%|███████▎ | 377/520 [37:11<09:05, 3.81s/it] 73%|███████▎ | 378/520 [37:15<09:02, 3.82s/it] {'loss': 1.3118, 'grad_norm': 0.0035575342455691737, 'learning_rate': 0.03668120907658603, 'epoch': 0.73} + 73%|███████▎ | 378/520 [37:15<09:02, 3.82s/it] 73%|███████▎ | 379/520 [37:18<08:59, 3.82s/it] {'loss': 1.2968, 'grad_norm': 0.0035161309225948136, 'learning_rate': 0.036199984606699154, 'epoch': 0.73} + 73%|███████▎ | 379/520 [37:18<08:59, 3.82s/it] 73%|███████▎ | 380/520 [37:22<08:54, 3.82s/it] {'loss': 1.4355, 'grad_norm': 0.004369071700388306, 'learning_rate': 0.035721239031346066, 'epoch': 0.73} + 73%|███████▎ | 380/520 [37:22<08:54, 3.82s/it] 73%|███████▎ | 381/520 [37:26<08:51, 3.83s/it] {'loss': 1.2902, 'grad_norm': 0.0036831347659696144, 'learning_rate': 0.03524499095177297, 'epoch': 0.73} + 73%|███████▎ | 381/520 [37:26<08:51, 3.83s/it] 73%|███████▎ | 382/520 [37:30<08:51, 3.85s/it] {'loss': 1.3683, 'grad_norm': 0.0036642469205876126, 'learning_rate': 0.03477125887218792, 'epoch': 0.73} + 73%|███████▎ | 382/520 [37:30<08:51, 3.85s/it] 74%|███████▎ | 383/520 [37:34<08:48, 3.85s/it] {'loss': 1.1235, 'grad_norm': 0.00402124541861871, 'learning_rate': 0.03430006119904196, 'epoch': 0.74} + 74%|███████▎ | 383/520 [37:34<08:48, 3.85s/it] 74%|███████▍ | 384/520 [37:38<08:43, 3.85s/it] {'loss': 1.5071, 'grad_norm': 0.0037222524324438304, 'learning_rate': 0.033831416240314084, 'epoch': 0.74} + 74%|███████▍ | 384/520 [37:38<08:43, 3.85s/it] 74%|███████▍ | 385/520 [37:42<08:40, 3.85s/it] {'loss': 1.269, 'grad_norm': 0.0035653622915625635, 'learning_rate': 0.03336534220479961, 'epoch': 0.74} + 74%|███████▍ | 385/520 [37:42<08:40, 3.85s/it] 74%|███████▍ | 386/520 [37:45<08:37, 3.86s/it] {'loss': 1.2168, 'grad_norm': 0.0033290606440244293, 'learning_rate': 0.032901857201403005, 'epoch': 0.74} + 74%|███████▍ | 386/520 [37:45<08:37, 3.86s/it] 74%|███████▍ | 387/520 [37:49<08:29, 3.83s/it] {'loss': 1.453, 'grad_norm': 0.0038610601061042143, 'learning_rate': 0.032440979238433976, 'epoch': 0.74} + 74%|███████▍ | 387/520 [37:49<08:29, 3.83s/it] 75%|███████▍ | 388/520 [37:53<08:20, 3.79s/it] {'loss': 1.167, 'grad_norm': 0.0034285787774464074, 'learning_rate': 0.03198272622290804, 'epoch': 0.75} + 75%|███████▍ | 388/520 [37:53<08:20, 3.79s/it] 75%|███████▍ | 389/520 [37:57<08:11, 3.75s/it] {'loss': 1.2417, 'grad_norm': 0.004227288702019738, 'learning_rate': 0.03152711595985065, 'epoch': 0.75} + 75%|███████▍ | 389/520 [37:57<08:11, 3.75s/it] 75%|███████▌ | 390/520 [38:00<08:05, 3.73s/it] {'loss': 1.2886, 'grad_norm': 0.0034359721087200523, 'learning_rate': 0.031074166151605298, 'epoch': 0.75} + 75%|███████▌ | 390/520 [38:00<08:05, 3.73s/it] 75%|███████▌ | 391/520 [38:04<07:59, 3.72s/it] {'loss': 1.3737, 'grad_norm': 0.003739398793945416, 'learning_rate': 0.030623894397145836, 'epoch': 0.75} + 75%|███████▌ | 391/520 [38:04<07:59, 3.72s/it] 75%|███████▌ | 392/520 [38:08<07:54, 3.71s/it] {'loss': 1.1786, 'grad_norm': 0.0035307622177173387, 'learning_rate': 0.03017631819139273, 'epoch': 0.75} + 75%|███████▌ | 392/520 [38:08<07:54, 3.71s/it] 76%|███████▌ | 393/520 [38:11<07:50, 3.70s/it] {'loss': 1.2488, 'grad_norm': 0.003362768478792506, 'learning_rate': 0.029731454924533086, 'epoch': 0.76} + 76%|███████▌ | 393/520 [38:11<07:50, 3.70s/it] 76%|███████▌ | 394/520 [38:15<07:46, 3.70s/it] {'loss': 1.2421, 'grad_norm': 0.004030290028469471, 'learning_rate': 0.029289321881345254, 'epoch': 0.76} + 76%|███████▌ | 394/520 [38:15<07:46, 3.70s/it] 76%|███████▌ | 395/520 [38:19<07:42, 3.70s/it] {'loss': 1.2051, 'grad_norm': 0.003948816496285298, 'learning_rate': 0.028849936240527008, 'epoch': 0.76} + 76%|███████▌ | 395/520 [38:19<07:42, 3.70s/it] 76%|███████▌ | 396/520 [38:22<07:36, 3.68s/it] {'loss': 1.2888, 'grad_norm': 0.003822047885301867, 'learning_rate': 0.028413315074028157, 'epoch': 0.76} + 76%|███████▌ | 396/520 [38:22<07:36, 3.68s/it] 76%|███████▋ | 397/520 [38:26<07:33, 3.69s/it] {'loss': 1.2795, 'grad_norm': 0.0035612582606693627, 'learning_rate': 0.027979475346387363, 'epoch': 0.76} + 76%|███████▋ | 397/520 [38:26<07:33, 3.69s/it] 77%|███████▋ | 398/520 [38:30<07:32, 3.71s/it] {'loss': 1.2666, 'grad_norm': 0.003947387967798251, 'learning_rate': 0.027548433914072735, 'epoch': 0.77} + 77%|███████▋ | 398/520 [38:30<07:32, 3.71s/it] 77%|███████▋ | 399/520 [38:34<07:29, 3.71s/it] {'loss': 1.3049, 'grad_norm': 0.0037366433101590837, 'learning_rate': 0.027120207524827168, 'epoch': 0.77} + 77%|███████▋ | 399/520 [38:34<07:29, 3.71s/it] 77%|███████▋ | 400/520 [38:37<07:24, 3.70s/it] {'loss': 1.3642, 'grad_norm': 0.0037201844278592733, 'learning_rate': 0.02669481281701739, 'epoch': 0.77} + 77%|███████▋ | 400/520 [38:37<07:24, 3.70s/it] 77%|███████▋ | 401/520 [38:41<07:18, 3.69s/it] {'loss': 1.0858, 'grad_norm': 0.003910737756648084, 'learning_rate': 0.026272266318987603, 'epoch': 0.77} + 77%|███████▋ | 401/520 [38:41<07:18, 3.69s/it] 77%|███████▋ | 402/520 [38:45<07:15, 3.69s/it] {'loss': 1.2083, 'grad_norm': 0.0038073140462726814, 'learning_rate': 0.02585258444841733, 'epoch': 0.77} + 77%|███████▋ | 402/520 [38:45<07:15, 3.69s/it] 78%|███████▊ | 403/520 [38:48<07:08, 3.67s/it] {'loss': 1.2453, 'grad_norm': 0.0040273071858792, 'learning_rate': 0.025435783511683442, 'epoch': 0.78} + 78%|███████▊ | 403/520 [38:48<07:08, 3.67s/it] 78%|███████▊ | 404/520 [38:52<07:04, 3.66s/it] {'loss': 1.1562, 'grad_norm': 0.004528247966618409, 'learning_rate': 0.02502187970322657, 'epoch': 0.78} + 78%|███████▊ | 404/520 [38:52<07:04, 3.66s/it] 78%|███████▊ | 405/520 [38:55<07:00, 3.66s/it] {'loss': 1.3028, 'grad_norm': 0.003591145814600703, 'learning_rate': 0.02461088910492202, 'epoch': 0.78} + 78%|███████▊ | 405/520 [38:55<07:00, 3.66s/it] 78%|███████▊ | 406/520 [38:59<06:56, 3.66s/it] {'loss': 1.2458, 'grad_norm': 0.004515275608086695, 'learning_rate': 0.02420282768545469, 'epoch': 0.78} + 78%|███████▊ | 406/520 [38:59<06:56, 3.66s/it] 78%|███████▊ | 407/520 [39:03<06:52, 3.65s/it] {'loss': 1.3393, 'grad_norm': 0.0038945124508647363, 'learning_rate': 0.02379771129969892, 'epoch': 0.78} + 78%|███████▊ | 407/520 [39:03<06:52, 3.65s/it] 78%|███████▊ | 408/520 [39:06<06:47, 3.64s/it] {'loss': 1.2307, 'grad_norm': 0.003955167868933628, 'learning_rate': 0.023395555688102213, 'epoch': 0.78} + 78%|███████▊ | 408/520 [39:06<06:47, 3.64s/it] 79%|███████▊ | 409/520 [39:10<06:44, 3.65s/it] {'loss': 1.3667, 'grad_norm': 0.00419715670620588, 'learning_rate': 0.02299637647607372, 'epoch': 0.79} + 79%|███████▊ | 409/520 [39:10<06:44, 3.65s/it] 79%|███████▉ | 410/520 [39:14<06:42, 3.66s/it] {'loss': 1.0766, 'grad_norm': 0.0037114910679811423, 'learning_rate': 0.022600189173377264, 'epoch': 0.79} + 79%|███████▉ | 410/520 [39:14<06:42, 3.66s/it] 79%|███████▉ | 411/520 [39:18<06:43, 3.70s/it] {'loss': 1.3309, 'grad_norm': 0.003987684691709063, 'learning_rate': 0.022207009173528525, 'epoch': 0.79} + 79%|███████▉ | 411/520 [39:18<06:43, 3.70s/it] 79%|███████▉ | 412/520 [39:21<06:46, 3.76s/it] {'loss': 1.2508, 'grad_norm': 0.0037057792998212292, 'learning_rate': 0.02181685175319702, 'epoch': 0.79} + 79%|███████▉ | 412/520 [39:21<06:46, 3.76s/it] 79%|███████▉ | 413/520 [39:25<06:48, 3.81s/it] {'loss': 1.349, 'grad_norm': 0.003724328329675125, 'learning_rate': 0.021429732071612653, 'epoch': 0.79} + 79%|███████▉ | 413/520 [39:25<06:48, 3.81s/it] 80%|███████▉ | 414/520 [39:29<06:48, 3.86s/it] {'loss': 1.1304, 'grad_norm': 0.0035208978504955245, 'learning_rate': 0.02104566516997647, 'epoch': 0.8} + 80%|███████▉ | 414/520 [39:29<06:48, 3.86s/it] 80%|███████▉ | 415/520 [39:33<06:49, 3.90s/it] {'loss': 1.222, 'grad_norm': 0.003447383930218272, 'learning_rate': 0.020664665970876496, 'epoch': 0.8} + 80%|███████▉ | 415/520 [39:33<06:49, 3.90s/it] 80%|████████ | 416/520 [39:37<06:45, 3.90s/it] {'loss': 1.1377, 'grad_norm': 0.004350722030650917, 'learning_rate': 0.020286749277707784, 'epoch': 0.8} + 80%|████████ | 416/520 [39:37<06:45, 3.90s/it] 80%|████████ | 417/520 [39:41<06:42, 3.90s/it] {'loss': 1.3112, 'grad_norm': 0.00418987559937938, 'learning_rate': 0.019911929774097215, 'epoch': 0.8} + 80%|████████ | 417/520 [39:41<06:42, 3.90s/it] 80%|████████ | 418/520 [39:45<06:38, 3.90s/it] {'loss': 1.2924, 'grad_norm': 0.0035348942687395183, 'learning_rate': 0.019540222023333165, 'epoch': 0.8} + 80%|████████ | 418/520 [39:45<06:38, 3.90s/it] 81%|████████ | 419/520 [39:49<06:34, 3.90s/it] {'loss': 1.2794, 'grad_norm': 0.0038931540223786794, 'learning_rate': 0.01917164046779948, 'epoch': 0.81} + 81%|████████ | 419/520 [39:49<06:34, 3.90s/it] 81%|████████ | 420/520 [39:53<06:30, 3.90s/it] {'loss': 1.149, 'grad_norm': 0.0037193298125260317, 'learning_rate': 0.018806199428414352, 'epoch': 0.81} + 81%|████████ | 420/520 [39:53<06:30, 3.90s/it] 81%|████████ | 421/520 [39:56<06:19, 3.83s/it] {'loss': 1.0863, 'grad_norm': 0.003929698628327199, 'learning_rate': 0.018443913104073985, 'epoch': 0.81} + 81%|████████ | 421/520 [39:57<06:19, 3.83s/it] 81%|████████ | 422/520 [40:00<06:10, 3.78s/it] {'loss': 1.214, 'grad_norm': 0.003935122086323532, 'learning_rate': 0.01808479557110081, 'epoch': 0.81} + 81%|████████ | 422/520 [40:00<06:10, 3.78s/it] 81%|████████▏ | 423/520 [40:04<06:04, 3.76s/it] {'loss': 1.206, 'grad_norm': 0.004334252023525384, 'learning_rate': 0.017728860782696667, 'epoch': 0.81} + 81%|████████▏ | 423/520 [40:04<06:04, 3.76s/it] 82%|████████▏ | 424/520 [40:08<05:57, 3.73s/it] {'loss': 1.4262, 'grad_norm': 0.004079740999241218, 'learning_rate': 0.017376122568400532, 'epoch': 0.82} + 82%|████████▏ | 424/520 [40:08<05:57, 3.73s/it] 82%|████████▏ | 425/520 [40:11<05:52, 3.71s/it] {'loss': 1.2088, 'grad_norm': 0.003581456370739972, 'learning_rate': 0.017026594633551252, 'epoch': 0.82} + 82%|████████▏ | 425/520 [40:11<05:52, 3.71s/it] 82%|████████▏ | 426/520 [40:15<05:46, 3.68s/it] {'loss': 1.2378, 'grad_norm': 0.00525114391511483, 'learning_rate': 0.01668029055875512, 'epoch': 0.82} + 82%|████████▏ | 426/520 [40:15<05:46, 3.68s/it] 82%|████████▏ | 427/520 [40:18<05:41, 3.67s/it] {'loss': 1.1513, 'grad_norm': 0.003829058039577684, 'learning_rate': 0.016337223799358026, 'epoch': 0.82} + 82%|████████▏ | 427/520 [40:18<05:41, 3.67s/it] 82%|████████▏ | 428/520 [40:22<05:38, 3.68s/it] {'loss': 1.1197, 'grad_norm': 0.0037818208888582257, 'learning_rate': 0.01599740768492286, 'epoch': 0.82} + 82%|████████▏ | 428/520 [40:22<05:38, 3.68s/it] 82%|████████▎ | 429/520 [40:26<05:34, 3.68s/it] {'loss': 1.2272, 'grad_norm': 0.003587157939894783, 'learning_rate': 0.015660855418711452, 'epoch': 0.82} + 82%|████████▎ | 429/520 [40:26<05:34, 3.68s/it]Token indices sequence length is longer than the specified maximum sequence length for this model (2076 > 2048). Running this sequence through the model will result in indexing errors + 83%|████████▎ | 430/520 [40:30<05:31, 3.68s/it] {'loss': 1.2235, 'grad_norm': 0.003405767161408271, 'learning_rate': 0.015327580077171589, 'epoch': 0.83} + 83%|████████▎ | 430/520 [40:30<05:31, 3.68s/it] 83%|████████▎ | 431/520 [40:33<05:27, 3.68s/it] {'loss': 1.3062, 'grad_norm': 0.003891030601922335, 'learning_rate': 0.014997594609429088, 'epoch': 0.83} + 83%|████████▎ | 431/520 [40:33<05:27, 3.68s/it] 83%|████████▎ | 432/520 [40:37<05:23, 3.68s/it] {'loss': 1.1283, 'grad_norm': 0.004170585824830143, 'learning_rate': 0.01467091183678444, 'epoch': 0.83} + 83%|████████▎ | 432/520 [40:37<05:23, 3.68s/it] 83%|████████▎ | 433/520 [40:41<05:20, 3.68s/it] {'loss': 1.2718, 'grad_norm': 0.003677373270139028, 'learning_rate': 0.014347544452214867, 'epoch': 0.83} + 83%|████████▎ | 433/520 [40:41<05:20, 3.68s/it] 83%|████████▎ | 434/520 [40:44<05:17, 3.69s/it] {'loss': 1.0053, 'grad_norm': 0.0034521664246291942, 'learning_rate': 0.014027505019880971, 'epoch': 0.83} + 83%|████████▎ | 434/520 [40:44<05:17, 3.69s/it] 84%|████████▎ | 435/520 [40:48<05:12, 3.67s/it] {'loss': 1.2998, 'grad_norm': 0.004171690910040855, 'learning_rate': 0.013710805974638696, 'epoch': 0.84} + 84%|████████▎ | 435/520 [40:48<05:12, 3.67s/it] 84%|████████▍ | 436/520 [40:52<05:08, 3.68s/it] {'loss': 1.0945, 'grad_norm': 0.003709642783237724, 'learning_rate': 0.01339745962155613, 'epoch': 0.84} + 84%|████████▍ | 436/520 [40:52<05:08, 3.68s/it] 84%|████████▍ | 437/520 [40:55<05:05, 3.68s/it] {'loss': 1.3296, 'grad_norm': 0.0037954478062005538, 'learning_rate': 0.01308747813543536, 'epoch': 0.84} + 84%|████████▍ | 437/520 [40:55<05:05, 3.68s/it] 84%|████████▍ | 438/520 [40:59<05:01, 3.68s/it] {'loss': 1.1288, 'grad_norm': 0.0036311371835896634, 'learning_rate': 0.012780873560339467, 'epoch': 0.84} + 84%|████████▍ | 438/520 [40:59<05:01, 3.68s/it] 84%|████████▍ | 439/520 [41:03<04:58, 3.68s/it] {'loss': 1.2687, 'grad_norm': 0.0031323095883406147, 'learning_rate': 0.012477657809124632, 'epoch': 0.84} + 84%|████████▍ | 439/520 [41:03<04:58, 3.68s/it] 85%|████████▍ | 440/520 [41:06<04:54, 3.68s/it] {'loss': 1.1879, 'grad_norm': 0.003802981942781014, 'learning_rate': 0.012177842662977134, 'epoch': 0.85} + 85%|████████▍ | 440/520 [41:06<04:54, 3.68s/it] 85%|████████▍ | 441/520 [41:10<04:50, 3.67s/it] {'loss': 1.31, 'grad_norm': 0.00374785402423887, 'learning_rate': 0.01188143977095576, 'epoch': 0.85} + 85%|████████▍ | 441/520 [41:10<04:50, 3.67s/it] 85%|████████▌ | 442/520 [41:14<04:46, 3.67s/it] {'loss': 1.2403, 'grad_norm': 0.004345643180649251, 'learning_rate': 0.011588460649539035, 'epoch': 0.85} + 85%|████████▌ | 442/520 [41:14<04:46, 3.67s/it] 85%|████████▌ | 443/520 [41:17<04:41, 3.65s/it] {'loss': 1.258, 'grad_norm': 0.0037859499092624998, 'learning_rate': 0.011298916682177829, 'epoch': 0.85} + 85%|████████▌ | 443/520 [41:17<04:41, 3.65s/it] 85%|████████▌ | 444/520 [41:21<04:37, 3.66s/it] {'loss': 1.2285, 'grad_norm': 0.003454122198701185, 'learning_rate': 0.011012819118853146, 'epoch': 0.85} + 85%|████████▌ | 444/520 [41:21<04:37, 3.66s/it] 86%|████████▌ | 445/520 [41:25<04:36, 3.69s/it] {'loss': 1.141, 'grad_norm': 0.0036658771238219775, 'learning_rate': 0.01073017907563887, 'epoch': 0.86} + 86%|████████▌ | 445/520 [41:25<04:36, 3.69s/it] 86%|████████▌ | 446/520 [41:29<04:37, 3.75s/it] {'loss': 1.3751, 'grad_norm': 0.0035745336507995587, 'learning_rate': 0.010451007534269908, 'epoch': 0.86} + 86%|████████▌ | 446/520 [41:29<04:37, 3.75s/it] 86%|████████▌ | 447/520 [41:32<04:36, 3.79s/it] {'loss': 1.2367, 'grad_norm': 0.0037414504309027374, 'learning_rate': 0.010175315341715598, 'epoch': 0.86} + 86%|████████▌ | 447/520 [41:32<04:36, 3.79s/it] 86%|████████▌ | 448/520 [41:36<04:34, 3.81s/it] {'loss': 1.21, 'grad_norm': 0.0039012965285577313, 'learning_rate': 0.009903113209758098, 'epoch': 0.86} + 86%|████████▌ | 448/520 [41:36<04:34, 3.81s/it] 86%|████████▋ | 449/520 [41:40<04:31, 3.82s/it] {'loss': 1.3297, 'grad_norm': 0.003803248775731238, 'learning_rate': 0.009634411714576352, 'epoch': 0.86} + 86%|████████▋ | 449/520 [41:40<04:31, 3.82s/it] 87%|████████▋ | 450/520 [41:44<04:27, 3.83s/it] {'loss': 1.2528, 'grad_norm': 0.0037130929202643906, 'learning_rate': 0.009369221296335007, 'epoch': 0.87} + 87%|████████▋ | 450/520 [41:44<04:27, 3.83s/it] 87%|████████▋ | 451/520 [41:48<04:24, 3.83s/it] {'loss': 1.2468, 'grad_norm': 0.0038384260907729137, 'learning_rate': 0.009107552258778906, 'epoch': 0.87} + 87%|████████▋ | 451/520 [41:48<04:24, 3.83s/it] 87%|████████▋ | 452/520 [41:52<04:22, 3.86s/it] {'loss': 1.3618, 'grad_norm': 0.0037728372387698926, 'learning_rate': 0.008849414768832687, 'epoch': 0.87} + 87%|████████▋ | 452/520 [41:52<04:22, 3.86s/it] 87%|████████▋ | 453/520 [41:56<04:18, 3.86s/it] {'loss': 1.3451, 'grad_norm': 0.0038352981014633395, 'learning_rate': 0.008594818856205699, 'epoch': 0.87} + 87%|████████▋ | 453/520 [41:56<04:18, 3.86s/it] 87%|████████▋ | 454/520 [41:59<04:14, 3.85s/it] {'loss': 1.1582, 'grad_norm': 0.004008771570975883, 'learning_rate': 0.00834377441300238, 'epoch': 0.87} + 87%|████████▋ | 454/520 [41:59<04:14, 3.85s/it] 88%|████████▊ | 455/520 [42:03<04:10, 3.85s/it] {'loss': 1.2904, 'grad_norm': 0.003689950695856545, 'learning_rate': 0.008096291193337934, 'epoch': 0.88} + 88%|████████▊ | 455/520 [42:03<04:10, 3.85s/it] 88%|████████▊ | 456/520 [42:07<04:05, 3.84s/it] {'loss': 1.2117, 'grad_norm': 0.0036875563289462078, 'learning_rate': 0.007852378812959226, 'epoch': 0.88} + 88%|████████▊ | 456/520 [42:07<04:05, 3.84s/it] 88%|████████▊ | 457/520 [42:11<04:02, 3.85s/it] {'loss': 1.3373, 'grad_norm': 0.0035015173012342858, 'learning_rate': 0.007612046748871327, 'epoch': 0.88} + 88%|████████▊ | 457/520 [42:11<04:02, 3.85s/it] 88%|████████▊ | 458/520 [42:15<04:00, 3.88s/it] {'loss': 1.3536, 'grad_norm': 0.0039483836353978626, 'learning_rate': 0.007375304338969136, 'epoch': 0.88} + 88%|████████▊ | 458/520 [42:15<04:00, 3.88s/it] 88%|████████▊ | 459/520 [42:19<03:56, 3.87s/it] {'loss': 1.2835, 'grad_norm': 0.0036644930936721855, 'learning_rate': 0.007142160781674645, 'epoch': 0.88} + 88%|████████▊ | 459/520 [42:19<03:56, 3.87s/it] 88%|████████▊ | 460/520 [42:23<03:49, 3.83s/it] {'loss': 1.157, 'grad_norm': 0.003578083501747229, 'learning_rate': 0.006912625135579587, 'epoch': 0.88} + 88%|████████▊ | 460/520 [42:23<03:49, 3.83s/it] 89%|████████▊ | 461/520 [42:26<03:42, 3.78s/it] {'loss': 1.412, 'grad_norm': 0.003237098346573943, 'learning_rate': 0.0066867063190933496, 'epoch': 0.89} + 89%|████████▊ | 461/520 [42:26<03:42, 3.78s/it] 89%|████████▉ | 462/520 [42:30<03:37, 3.74s/it] {'loss': 1.4168, 'grad_norm': 0.003699279922147121, 'learning_rate': 0.006464413110096601, 'epoch': 0.89} + 89%|████████▉ | 462/520 [42:30<03:37, 3.74s/it] 89%|████████▉ | 463/520 [42:34<03:31, 3.71s/it] {'loss': 1.1218, 'grad_norm': 0.0040101230107318225, 'learning_rate': 0.006245754145600091, 'epoch': 0.89} + 89%|████████▉ | 463/520 [42:34<03:31, 3.71s/it] 89%|████████▉ | 464/520 [42:37<03:27, 3.70s/it] {'loss': 1.2705, 'grad_norm': 0.003821385485694576, 'learning_rate': 0.006030737921409169, 'epoch': 0.89} + 89%|████████▉ | 464/520 [42:37<03:27, 3.70s/it] 89%|████████▉ | 465/520 [42:41<03:22, 3.69s/it] {'loss': 1.3801, 'grad_norm': 0.00425963578056506, 'learning_rate': 0.005819372791793654, 'epoch': 0.89} + 89%|████████▉ | 465/520 [42:41<03:22, 3.69s/it] 90%|████████▉ | 466/520 [42:44<03:18, 3.67s/it] {'loss': 1.2472, 'grad_norm': 0.003319016246272143, 'learning_rate': 0.005611666969163243, 'epoch': 0.9} + 90%|████████▉ | 466/520 [42:44<03:18, 3.67s/it] 90%|████████▉ | 467/520 [42:48<03:14, 3.67s/it] {'loss': 1.2926, 'grad_norm': 0.0035630988843583446, 'learning_rate': 0.005407628523748398, 'epoch': 0.9} + 90%|████████▉ | 467/520 [42:48<03:14, 3.67s/it] 90%|█████████ | 468/520 [42:52<03:10, 3.67s/it] {'loss': 1.2265, 'grad_norm': 0.004066632143702232, 'learning_rate': 0.00520726538328683, 'epoch': 0.9} + 90%|█████████ | 468/520 [42:52<03:10, 3.67s/it] 90%|█████████ | 469/520 [42:55<03:07, 3.67s/it] {'loss': 1.295, 'grad_norm': 0.0040338773174477685, 'learning_rate': 0.005010585332715401, 'epoch': 0.9} + 90%|█████████ | 469/520 [42:55<03:07, 3.67s/it] 90%|█████████ | 470/520 [42:59<03:03, 3.66s/it] {'loss': 1.1632, 'grad_norm': 0.0032657218405383456, 'learning_rate': 0.004817596013867765, 'epoch': 0.9} + 90%|█████████ | 470/520 [42:59<03:03, 3.66s/it] 91%|█████████ | 471/520 [43:03<02:59, 3.66s/it] {'loss': 1.1821, 'grad_norm': 0.0037477977694796334, 'learning_rate': 0.004628304925177318, 'epoch': 0.91} + 91%|█████████ | 471/520 [43:03<02:59, 3.66s/it] 91%|█████████ | 472/520 [43:06<02:55, 3.66s/it] {'loss': 1.1539, 'grad_norm': 0.003678400159196095, 'learning_rate': 0.004442719421385921, 'epoch': 0.91} + 91%|█████████ | 472/520 [43:06<02:55, 3.66s/it] 91%|█████████ | 473/520 [43:10<02:51, 3.66s/it] {'loss': 1.2105, 'grad_norm': 0.003634660503951996, 'learning_rate': 0.004260846713258193, 'epoch': 0.91} + 91%|█████████ | 473/520 [43:10<02:51, 3.66s/it] 91%|█████████ | 474/520 [43:14<02:47, 3.65s/it] {'loss': 1.3312, 'grad_norm': 0.00353979070313434, 'learning_rate': 0.004082693867301224, 'epoch': 0.91} + 91%|█████████ | 474/520 [43:14<02:47, 3.65s/it] 91%|█████████▏| 475/520 [43:17<02:44, 3.66s/it] {'loss': 1.2437, 'grad_norm': 0.0035268092705061133, 'learning_rate': 0.003908267805490051, 'epoch': 0.91} + 91%|█████████▏| 475/520 [43:17<02:44, 3.66s/it] 92%|█████████▏| 476/520 [43:21<02:44, 3.73s/it] {'loss': 1.2023, 'grad_norm': 0.0039234023501603455, 'learning_rate': 0.003737575304998797, 'epoch': 0.92} + 92%|█████████▏| 476/520 [43:21<02:44, 3.73s/it] 92%|█████████▏| 477/520 [43:25<02:41, 3.76s/it] {'loss': 1.1991, 'grad_norm': 0.004171533678244563, 'learning_rate': 0.003570622997937234, 'epoch': 0.92} + 92%|█████████▏| 477/520 [43:25<02:41, 3.76s/it] 92%|█████████▏| 478/520 [43:29<02:36, 3.72s/it] {'loss': 1.1542, 'grad_norm': 0.0036395881322330696, 'learning_rate': 0.00340741737109318, 'epoch': 0.92} + 92%|█████████▏| 478/520 [43:29<02:36, 3.72s/it] 92%|█████████▏| 479/520 [43:32<02:31, 3.71s/it] {'loss': 1.3106, 'grad_norm': 0.003958980326619334, 'learning_rate': 0.003247964765680389, 'epoch': 0.92} + 92%|█████████▏| 479/520 [43:32<02:31, 3.71s/it] 92%|█████████▏| 480/520 [43:36<02:27, 3.69s/it] {'loss': 1.3268, 'grad_norm': 0.0035536442378239328, 'learning_rate': 0.0030922713770922153, 'epoch': 0.92} + 92%|█████████▏| 480/520 [43:36<02:27, 3.69s/it] 92%|█████████▎| 481/520 [43:40<02:24, 3.70s/it] {'loss': 1.3413, 'grad_norm': 0.0035431376593947086, 'learning_rate': 0.0029403432546609046, 'epoch': 0.93} + 92%|█████████▎| 481/520 [43:40<02:24, 3.70s/it] 93%|█████████▎| 482/520 [43:43<02:20, 3.70s/it] {'loss': 1.3378, 'grad_norm': 0.003664400563355644, 'learning_rate': 0.0027921863014225504, 'epoch': 0.93} + 93%|█████████▎| 482/520 [43:43<02:20, 3.70s/it] 93%|█████████▎| 483/520 [43:47<02:16, 3.70s/it] {'loss': 1.2295, 'grad_norm': 0.0038715530095995284, 'learning_rate': 0.002647806273887665, 'epoch': 0.93} + 93%|█████████▎| 483/520 [43:47<02:16, 3.70s/it] 93%|█████████▎| 484/520 [43:51<02:13, 3.71s/it] {'loss': 1.231, 'grad_norm': 0.003678956445979777, 'learning_rate': 0.0025072087818176383, 'epoch': 0.93} + 93%|█████████▎| 484/520 [43:51<02:13, 3.71s/it] 93%|█████████▎| 485/520 [43:55<02:09, 3.71s/it] {'loss': 1.176, 'grad_norm': 0.003558667860759577, 'learning_rate': 0.002370399288006664, 'epoch': 0.93} + 93%|█████████▎| 485/520 [43:55<02:09, 3.71s/it] 93%|█████████▎| 486/520 [43:58<02:05, 3.70s/it] {'loss': 1.3005, 'grad_norm': 0.0038974075223755706, 'learning_rate': 0.0022373831080695463, 'epoch': 0.93} + 93%|█████████▎| 486/520 [43:58<02:05, 3.70s/it] 94%|█████████▎| 487/520 [44:02<02:02, 3.71s/it] {'loss': 1.1549, 'grad_norm': 0.003559779692649507, 'learning_rate': 0.0021081654102351635, 'epoch': 0.94} + 94%|█████████▎| 487/520 [44:02<02:02, 3.71s/it] 94%|█████████▍| 488/520 [44:06<01:58, 3.70s/it] {'loss': 1.0907, 'grad_norm': 0.003743889229463068, 'learning_rate': 0.0019827512151456175, 'epoch': 0.94} + 94%|█████████▍| 488/520 [44:06<01:58, 3.70s/it] 94%|█████████▍| 489/520 [44:09<01:54, 3.69s/it] {'loss': 1.3289, 'grad_norm': 0.0032458894563488602, 'learning_rate': 0.0018611453956612345, 'epoch': 0.94} + 94%|█████████▍| 489/520 [44:09<01:54, 3.69s/it] 94%|█████████▍| 490/520 [44:13<01:50, 3.68s/it] {'loss': 1.2206, 'grad_norm': 0.0038098155431594482, 'learning_rate': 0.0017433526766711727, 'epoch': 0.94} + 94%|█████████▍| 490/520 [44:13<01:50, 3.68s/it] 94%|█████████▍| 491/520 [44:17<01:46, 3.68s/it] {'loss': 1.178, 'grad_norm': 0.0038537999884300147, 'learning_rate': 0.0016293776349098677, 'epoch': 0.94} + 94%|█████████▍| 491/520 [44:17<01:46, 3.68s/it] 95%|█████████▍| 492/520 [44:20<01:43, 3.69s/it] {'loss': 1.3037, 'grad_norm': 0.003917956354864113, 'learning_rate': 0.0015192246987791981, 'epoch': 0.95} + 95%|█████████▍| 492/520 [44:20<01:43, 3.69s/it] 95%|█████████▍| 493/520 [44:24<01:39, 3.68s/it] {'loss': 1.3845, 'grad_norm': 0.00399396089929038, 'learning_rate': 0.0014128981481764114, 'epoch': 0.95} + 95%|█████████▍| 493/520 [44:24<01:39, 3.68s/it] 95%|█████████▌| 494/520 [44:28<01:35, 3.67s/it] {'loss': 1.2343, 'grad_norm': 0.003420750140524696, 'learning_rate': 0.0013104021143278911, 'epoch': 0.95} + 95%|█████████▌| 494/520 [44:28<01:35, 3.67s/it] 95%|█████████▌| 495/520 [44:31<01:32, 3.69s/it] {'loss': 1.194, 'grad_norm': 0.003587387047602343, 'learning_rate': 0.0012117405796285285, 'epoch': 0.95} + 95%|█████████▌| 495/520 [44:31<01:32, 3.69s/it] 95%|█████████▌| 496/520 [44:35<01:28, 3.69s/it] {'loss': 1.1145, 'grad_norm': 0.0038265816151914676, 'learning_rate': 0.0011169173774871477, 'epoch': 0.95} + 95%|█████████▌| 496/520 [44:35<01:28, 3.69s/it] 96%|█████████▌| 497/520 [44:39<01:24, 3.67s/it] {'loss': 1.2566, 'grad_norm': 0.003403761888968341, 'learning_rate': 0.0010259361921774012, 'epoch': 0.96} + 96%|█████████▌| 497/520 [44:39<01:24, 3.67s/it] 96%|█████████▌| 498/520 [44:43<01:21, 3.69s/it] {'loss': 1.1983, 'grad_norm': 0.0038743369231452316, 'learning_rate': 0.000938800558694719, 'epoch': 0.96} + 96%|█████████▌| 498/520 [44:43<01:21, 3.69s/it] 96%|█████████▌| 499/520 [44:46<01:19, 3.77s/it] {'loss': 1.4037, 'grad_norm': 0.0038093532104932827, 'learning_rate': 0.0008555138626189618, 'epoch': 0.96} + 96%|█████████▌| 499/520 [44:46<01:19, 3.77s/it] 96%|█████████▌| 500/520 [44:50<01:15, 3.78s/it] {'loss': 1.3189, 'grad_norm': 0.004303957010204251, 'learning_rate': 0.0007760793399827937, 'epoch': 0.96} + 96%|█████████▌| 500/520 [44:50<01:15, 3.78s/it] 96%|█████████▋| 501/520 [44:54<01:12, 3.80s/it] {'loss': 1.3227, 'grad_norm': 0.004910590449132182, 'learning_rate': 0.000700500077146038, 'epoch': 0.96} + 96%|█████████▋| 501/520 [44:54<01:12, 3.80s/it] 97%|█████████▋| 502/520 [44:58<01:08, 3.79s/it] {'loss': 1.2359, 'grad_norm': 0.0035705255384588827, 'learning_rate': 0.0006287790106757397, 'epoch': 0.97} + 97%|█████████▋| 502/520 [44:58<01:08, 3.79s/it] 97%|█████████▋| 503/520 [45:02<01:04, 3.79s/it] {'loss': 1.288, 'grad_norm': 0.0037945954211907173, 'learning_rate': 0.0005609189272320237, 'epoch': 0.97} + 97%|█████████▋| 503/520 [45:02<01:04, 3.79s/it] 97%|█████████▋| 504/520 [45:06<01:00, 3.80s/it] {'loss': 1.2324, 'grad_norm': 0.004414855140465823, 'learning_rate': 0.000496922463459859, 'epoch': 0.97} + 97%|█████████▋| 504/520 [45:06<01:00, 3.80s/it] 97%|█████████▋| 505/520 [45:09<00:57, 3.81s/it] {'loss': 1.2752, 'grad_norm': 0.0036376861869329194, 'learning_rate': 0.0004367921058866187, 'epoch': 0.97} + 97%|█████████▋| 505/520 [45:09<00:57, 3.81s/it] 97%|█████████▋| 506/520 [45:13<00:53, 3.79s/it] {'loss': 1.185, 'grad_norm': 0.0043079349287065884, 'learning_rate': 0.0003805301908254455, 'epoch': 0.97} + 97%|█████████▋| 506/520 [45:13<00:53, 3.79s/it] 98%|█████████▊| 507/520 [45:17<00:48, 3.75s/it] {'loss': 1.4419, 'grad_norm': 0.003544698374681255, 'learning_rate': 0.0003281389042844918, 'epoch': 0.97} + 98%|█████████▊| 507/520 [45:17<00:48, 3.75s/it] 98%|█████████▊| 508/520 [45:20<00:44, 3.73s/it] {'loss': 1.3119, 'grad_norm': 0.0037681601074036303, 'learning_rate': 0.00027962028188198705, 'epoch': 0.98} + 98%|█████████▊| 508/520 [45:20<00:44, 3.73s/it] 98%|█████████▊| 509/520 [45:24<00:40, 3.71s/it] {'loss': 1.2692, 'grad_norm': 0.0036264811985476588, 'learning_rate': 0.00023497620876711256, 'epoch': 0.98} + 98%|█████████▊| 509/520 [45:24<00:40, 3.71s/it] 98%|█████████▊| 510/520 [45:28<00:36, 3.70s/it] {'loss': 1.2313, 'grad_norm': 0.0035547719993240284, 'learning_rate': 0.00019420841954681523, 'epoch': 0.98} + 98%|█████████▊| 510/520 [45:28<00:36, 3.70s/it] 98%|█████████▊| 511/520 [45:31<00:33, 3.69s/it] {'loss': 1.2011, 'grad_norm': 0.003557020827185209, 'learning_rate': 0.00015731849821833956, 'epoch': 0.98} + 98%|█████████▊| 511/520 [45:31<00:33, 3.69s/it] 98%|█████████▊| 512/520 [45:35<00:29, 3.68s/it] {'loss': 1.0814, 'grad_norm': 0.0035491992636818913, 'learning_rate': 0.00012430787810776555, 'epoch': 0.98} + 98%|█████████▊| 512/520 [45:35<00:29, 3.68s/it] 99%|█████████▊| 513/520 [45:39<00:25, 3.67s/it] {'loss': 1.2835, 'grad_norm': 0.004012950528007395, 'learning_rate': 9.517784181422018e-05, 'epoch': 0.99} + 99%|█████████▊| 513/520 [45:39<00:25, 3.67s/it] 99%|█████████▉| 514/520 [45:42<00:22, 3.67s/it] {'loss': 1.2616, 'grad_norm': 0.0034084048170446523, 'learning_rate': 6.992952116013917e-05, 'epoch': 0.99} + 99%|█████████▉| 514/520 [45:42<00:22, 3.67s/it] 99%|█████████▉| 515/520 [45:46<00:18, 3.68s/it] {'loss': 1.3196, 'grad_norm': 0.00424377469956192, 'learning_rate': 4.856389714723575e-05, 'epoch': 0.99} + 99%|█████████▉| 515/520 [45:46<00:18, 3.68s/it] 99%|█████████▉| 516/520 [45:50<00:14, 3.68s/it] {'loss': 1.1913, 'grad_norm': 0.003553419302882841, 'learning_rate': 3.108179991837545e-05, 'epoch': 0.99} + 99%|█████████▉| 516/520 [45:50<00:14, 3.68s/it] 99%|█████████▉| 517/520 [45:53<00:10, 3.66s/it] {'loss': 1.3531, 'grad_norm': 0.003655067562759453, 'learning_rate': 1.7483908725357544e-05, 'epoch': 0.99} + 99%|█████████▉| 517/520 [45:53<00:10, 3.66s/it] 100%|█████████▉| 518/520 [45:57<00:07, 3.65s/it] {'loss': 1.2212, 'grad_norm': 0.0038036810845718683, 'learning_rate': 7.770751902513862e-06, 'epoch': 1.0} + 100%|█████████▉| 518/520 [45:57<00:07, 3.65s/it] 100%|█████████▉| 519/520 [46:01<00:03, 3.64s/it] {'loss': 1.2975, 'grad_norm': 0.003704586791048186, 'learning_rate': 1.9427068461808086e-06, 'epoch': 1.0} + 100%|█████████▉| 519/520 [46:01<00:03, 3.64s/it] 100%|██████████| 520/520 [46:05<00:00, 3.92s/it] {'loss': 1.3759, 'grad_norm': 0.0037164434169977578, 'learning_rate': 0.0, 'epoch': 1.0} + 100%|██████████| 520/520 [46:05<00:00, 3.92s/it] {'train_runtime': 2765.7319, 'train_samples_per_second': 24.055, 'train_steps_per_second': 0.188, 'train_loss': 1.547316185786174, 'epoch': 1.0} + 100%|██████████| 520/520 [46:05<00:00, 3.92s/it] 100%|██████████| 520/520 [46:05<00:00, 5.32s/it] +[2025-10-12 08:19:23,430] [INFO] [launch.py:348:main] Process 170062 exits successfully. +[2025-10-12 08:19:24,432] [INFO] [launch.py:348:main] Process 170061 exits successfully. +[2025-10-12 08:19:24,432] [INFO] [launch.py:348:main] Process 170059 exits successfully. +[2025-10-12 08:19:24,432] [INFO] [launch.py:348:main] Process 170058 exits successfully. +[2025-10-12 08:19:24,433] [INFO] [launch.py:348:main] Process 170060 exits successfully. +[2025-10-12 08:19:24,433] [INFO] [launch.py:348:main] Process 170057 exits successfully. +[2025-10-12 08:19:25,435] [INFO] [launch.py:348:main] Process 170056 exits successfully. +[2025-10-12 08:19:28,439] [INFO] [launch.py:348:main] Process 170055 exits successfully. +==== EXPERIMENT COMPLETED: qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.1_2e-1_connector-1.0_2.1_2e-1_ablation ==== +Log File: qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.1_2e-1_connector-1.0_2.1_2e-1_ablation_20251012_060227.log +Timestamp: 2025-10-12 08:19:31 +===================================== diff --git a/logs_oct11/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.3_2e-1_connector-1.0_2.3_2e-1_ablation_20251012_081931.log b/logs_oct11/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.3_2e-1_connector-1.0_2.3_2e-1_ablation_20251012_081931.log new file mode 100644 index 0000000000000000000000000000000000000000..317c7853540e951fc1feb858c31d74f1a44ada38 --- /dev/null +++ b/logs_oct11/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.3_2e-1_connector-1.0_2.3_2e-1_ablation_20251012_081931.log @@ -0,0 +1,843 @@ +==== STARTING EXPERIMENT: qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.3_2e-1_connector-1.0_2.3_2e-1_ablation ==== +Log File: qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.3_2e-1_connector-1.0_2.3_2e-1_ablation_20251012_081931.log +Timestamp: 2025-10-12 08:19:31 +===================================== +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 08:19:33,792] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 08:19:36,980] [WARNING] [runner.py:202:fetch_hostfile] Unable to find hostfile, will proceed with training with local resources only. +[2025-10-12 08:19:36,981] [INFO] [runner.py:568:main] cmd = /opt/conda/envs/tinyllava/bin/python3.10 -u -m deepspeed.launcher.launch --world_info=eyJsb2NhbGhvc3QiOiBbMCwgMSwgMiwgMywgNCwgNSwgNiwgN119 --master_addr=127.0.0.1 --master_port=29501 --enable_each_rank_log=None tinyllava/train/train.py --deepspeed ./scripts/zero3.json --data_path /s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json --image_folder /s3-code/ywang29/datasets/tinyllava --is_multimodal True --conv_version qwen2_base --model_name_or_path Qwen/Qwen2.5-0.5B --vision_tower google/siglip-so400m-patch14-384 --vision_tower2 --connector_type mlp2x_gelu --mm_vision_select_layer -2 --image_aspect_ratio square --attn_implementation flash_attention_2 --bf16 True --training_recipe common --tune_type_llm full --tune_type_vision_tower frozen --tune_vision_tower_from_layer 0 --tune_type_connector full --group_by_modality_length True --pretrained_model_path /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain --output_dir /nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.3_2e-1_connector-1.0_2.3_2e-1_ablation --num_train_epochs 1 --per_device_train_batch_size 4 --per_device_eval_batch_size 4 --gradient_accumulation_steps 4 --evaluation_strategy no --learning_rate 2e-1 --weight_decay 0. --warmup_ratio 0.03 --lr_scheduler_type cosine --logging_steps 1 --tf32 False --model_max_length 2048 --gradient_checkpointing True --dataloader_num_workers 8 --lazy_preprocess True --report_to tensorboard --tokenizer_use_fast False --run_name tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune --subnet_mode_text both --subnet_type_text None --mask_type_text soft --init_mean_text 1.0 --temperature_attn_text 2.3 --temperature_mlp_text 2.3 --backward_type_text normal --masked_layers_text all --subnet_mode_vision both --subnet_type_vision None --mask_type_vision soft --init_mean_vision 1.0 --temperature_attn_vision 2.3 --temperature_mlp_vision 2.3 --backward_type_vision normal --masked_layers_vision all --subnet_type_connector global --mask_type_connector soft --init_mean_connector 1.0 --temperature_connector 2.3 --backward_type_connector normal --mm_projector_lr 2e-1 --seed 42 --mask_model llm-connector --save_strategy steps --save_steps 50000 --save_total_limit 1 --train_data_ratio 0.1 +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 08:19:39,554] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 08:19:40,598] [INFO] [launch.py:138:main] 0 NCCL_VERSION=2.21.5 +[2025-10-12 08:19:40,598] [INFO] [launch.py:138:main] 0 NCCL_SOCKET_IFNAME=eth +[2025-10-12 08:19:40,598] [INFO] [launch.py:145:main] WORLD INFO DICT: {'localhost': [0, 1, 2, 3, 4, 5, 6, 7]} +[2025-10-12 08:19:40,598] [INFO] [launch.py:151:main] nnodes=1, num_local_procs=8, node_rank=0 +[2025-10-12 08:19:40,598] [INFO] [launch.py:162:main] global_rank_mapping=defaultdict(, {'localhost': [0, 1, 2, 3, 4, 5, 6, 7]}) +[2025-10-12 08:19:40,598] [INFO] [launch.py:163:main] dist_world_size=8 +[2025-10-12 08:19:40,598] [INFO] [launch.py:165:main] Setting CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 +[2025-10-12 08:19:40,601] [INFO] [launch.py:253:main] process 219289 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=0', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.3_2e-1_connector-1.0_2.3_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '1.0', '--temperature_attn_text', '2.3', '--temperature_mlp_text', '2.3', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '1.0', '--temperature_attn_vision', '2.3', '--temperature_mlp_vision', '2.3', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '1.0', '--temperature_connector', '2.3', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 08:19:40,603] [INFO] [launch.py:253:main] process 219290 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=1', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.3_2e-1_connector-1.0_2.3_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '1.0', '--temperature_attn_text', '2.3', '--temperature_mlp_text', '2.3', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '1.0', '--temperature_attn_vision', '2.3', '--temperature_mlp_vision', '2.3', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '1.0', '--temperature_connector', '2.3', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 08:19:40,605] [INFO] [launch.py:253:main] process 219291 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=2', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.3_2e-1_connector-1.0_2.3_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '1.0', '--temperature_attn_text', '2.3', '--temperature_mlp_text', '2.3', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '1.0', '--temperature_attn_vision', '2.3', '--temperature_mlp_vision', '2.3', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '1.0', '--temperature_connector', '2.3', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 08:19:40,607] [INFO] [launch.py:253:main] process 219292 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=3', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.3_2e-1_connector-1.0_2.3_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '1.0', '--temperature_attn_text', '2.3', '--temperature_mlp_text', '2.3', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '1.0', '--temperature_attn_vision', '2.3', '--temperature_mlp_vision', '2.3', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '1.0', '--temperature_connector', '2.3', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 08:19:40,609] [INFO] [launch.py:253:main] process 219293 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=4', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.3_2e-1_connector-1.0_2.3_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '1.0', '--temperature_attn_text', '2.3', '--temperature_mlp_text', '2.3', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '1.0', '--temperature_attn_vision', '2.3', '--temperature_mlp_vision', '2.3', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '1.0', '--temperature_connector', '2.3', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 08:19:40,612] [INFO] [launch.py:253:main] process 219294 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=5', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.3_2e-1_connector-1.0_2.3_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '1.0', '--temperature_attn_text', '2.3', '--temperature_mlp_text', '2.3', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '1.0', '--temperature_attn_vision', '2.3', '--temperature_mlp_vision', '2.3', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '1.0', '--temperature_connector', '2.3', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 08:19:40,614] [INFO] [launch.py:253:main] process 219295 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=6', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.3_2e-1_connector-1.0_2.3_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '1.0', '--temperature_attn_text', '2.3', '--temperature_mlp_text', '2.3', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '1.0', '--temperature_attn_vision', '2.3', '--temperature_mlp_vision', '2.3', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '1.0', '--temperature_connector', '2.3', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 08:19:40,616] [INFO] [launch.py:253:main] process 219296 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=7', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.3_2e-1_connector-1.0_2.3_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '1.0', '--temperature_attn_text', '2.3', '--temperature_mlp_text', '2.3', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '1.0', '--temperature_attn_vision', '2.3', '--temperature_mlp_vision', '2.3', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '1.0', '--temperature_connector', '2.3', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 08:19:47,390] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 08:19:47,390] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 08:19:47,391] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 08:19:47,391] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 08:19:47,407] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 08:19:47,412] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 08:19:47,460] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 08:19:47,460] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 08:19:47,941] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 08:19:47,941] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 08:19:47,941] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 08:19:47,941] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 08:19:47,941] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 08:19:47,941] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 08:19:47,941] [INFO] [comm.py:668:init_distributed] Initializing TorchBackend in DeepSpeed with backend nccl +[2025-10-12 08:19:47,941] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 08:19:47,941] [INFO] [comm.py:637:init_distributed] cdb=None +Apply masks for the following modules: ['llm', 'connector'] +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +{'llm': {'model_name_or_path': 'Qwen/Qwen2.5-0.5B', 'cache_dir': None, 'attn_implementation': 'flash_attention_2', 'subnet_mode': 'both', 'subnet_type': 'None', 'sparsity_attn': None, 'sparsity_mlp': None, 'threshold_attn': None, 'threshold_mlp': None, 'temperature_attn': 2.3, 'temperature_mlp': 2.3, 'masked_layers': 'all', 'mask_type': 'soft', 'backward_type': 'normal'}, 'vision_tower': {'model_name_or_path': 'google/siglip-so400m-patch14-384'}, 'connector': {'connector_type': 'mlp2x_gelu', 'subnet_type': 'global', 'threshold': None, 'sparsity': None, 'temperature': 2.3, 'mask_type': 'soft', 'backward_type': 'normal'}} +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +TinyLlavaConfig { + "backward_type_connector": "normal", + "cache_dir": null, + "connector_type": "mlp2x_gelu", + "hidden_size": 896, + "ignore_index": -100, + "image_aspect_ratio": "square", + "image_token_index": -200, + "llm_model_name_or_path": "Qwen/Qwen2.5-0.5B", + "mask_model": [ + "llm", + "connector" + ], + "mask_type_connector": "soft", + "model_type": "tinyllava", + "num_queries": 128, + "num_resampler_layers": 3, + "pad_token": null, + "resampler_hidden_size": 768, + "sparsity_connector": null, + "subnet_type_connector": "global", + "temperature_connector": 2.3, + "text_config": { + "_name_or_path": "Qwen/Qwen2.5-0.5B", + "architectures": [ + "Qwen2ForCausalLM" + ], + "backward_type": "normal", + "bos_token_id": 151643, + "eos_token_id": 151643, + "hidden_size": 896, + "intermediate_size": 4864, + "mask_type": "soft", + "masked_layers": "all", + "max_position_embeddings": 32768, + "max_window_layers": 24, + "model_type": "qwen2", + "num_attention_heads": 14, + "num_hidden_layers": 24, + "num_key_value_heads": 2, + "rope_theta": 1000000.0, + "sliding_window": 32768, + "subnet_mode": "both", + "subnet_type": "None", + "temperature_attn": 2.3, + "temperature_mlp": 2.3, + "tie_word_embeddings": true, + "torch_dtype": "bfloat16", + "use_mrope": false, + "use_sliding_window": false, + "vocab_size": 151936 + }, + "threshold_connector": null, + "tokenizer_model_max_length": 2048, + "tokenizer_name_or_path": "Qwen/Qwen2.5-0.5B", + "tokenizer_padding_side": "right", + "tokenizer_use_fast": false, + "transformers_version": "4.40.1", + "tune_type_connector": "frozen", + "tune_type_llm": "frozen", + "tune_type_vision_tower": "frozen", + "tune_vision_tower_from_layer": -1, + "use_cache": false, + "vision_config": { + "hidden_act": "gelu_pytorch_tanh", + "hidden_size": 1152, + "image_size": 384, + "intermediate_size": 4304, + "layer_norm_eps": 1e-06, + "model_name_or_path": "google/siglip-so400m-patch14-384", + "model_name_or_path2": "", + "model_type": "siglip_vision_model", + "num_attention_heads": 16, + "num_hidden_layers": 27, + "patch_size": 14 + }, + "vision_feature_layer": -2, + "vision_feature_select_strategy": "patch", + "vision_hidden_size": 1152, + "vision_model_name_or_path": "google/siglip-so400m-patch14-384", + "vision_model_name_or_path2": "", + "vocab_size": 151936 +} + +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test1-worker-0:219289:219289 [0] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:219289:219289 [0] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:219289:219289 [0] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:219289:219289 [0] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:219289:219289 [0] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test1-worker-0:219289:219289 [0] NCCL INFO cudaDriverVersion 12040 +NCCL version 2.21.5+cuda12.1 +ywang29-vrdb-test1-worker-0:219294:219294 [5] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test1-worker-0:219294:219294 [5] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:219294:219294 [5] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:219294:219294 [5] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:219294:219294 [5] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:219294:219294 [5] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test1-worker-0:219290:219290 [1] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test1-worker-0:219290:219290 [1] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:219290:219290 [1] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:219290:219290 [1] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:219290:219290 [1] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:219290:219290 [1] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test1-worker-0:219296:219296 [7] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test1-worker-0:219296:219296 [7] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:219296:219296 [7] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:219296:219296 [7] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:219296:219296 [7] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:219296:219296 [7] NCCL INFO NET/Plugin: Using internal network plugin. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test1-worker-0:219289:220883 [0] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:219289:220883 [0] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:219289:220883 [0] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:219289:220883 [0] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:219289:220883 [0] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:219289:220883 [0] NCCL INFO Using network Socket +ywang29-vrdb-test1-worker-0:219290:220885 [1] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:219290:220885 [1] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:219290:220885 [1] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:219290:220885 [1] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:219290:220885 [1] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:219290:220885 [1] NCCL INFO Using network Socket +ywang29-vrdb-test1-worker-0:219294:220884 [5] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:219294:220884 [5] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:219294:220884 [5] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:219294:220884 [5] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:219294:220884 [5] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:219294:220884 [5] NCCL INFO Using network Socket +ywang29-vrdb-test1-worker-0:219296:220886 [7] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:219296:220886 [7] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:219296:220886 [7] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:219296:220886 [7] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:219296:220886 [7] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:219296:220886 [7] NCCL INFO Using network Socket +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test1-worker-0:219292:219292 [3] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test1-worker-0:219292:219292 [3] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:219292:219292 [3] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:219292:219292 [3] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:219292:219292 [3] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:219292:219292 [3] NCCL INFO NET/Plugin: Using internal network plugin. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test1-worker-0:219291:219291 [2] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test1-worker-0:219291:219291 [2] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:219291:219291 [2] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:219291:219291 [2] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:219291:219291 [2] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:219291:219291 [2] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test1-worker-0:219293:219293 [4] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test1-worker-0:219293:219293 [4] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:219293:219293 [4] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:219293:219293 [4] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:219293:219293 [4] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:219293:219293 [4] NCCL INFO NET/Plugin: Using internal network plugin. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test1-worker-0:219295:219295 [6] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test1-worker-0:219295:219295 [6] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:219295:219295 [6] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:219295:219295 [6] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:219295:219295 [6] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:219295:219295 [6] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test1-worker-0:219292:220887 [3] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:219292:220887 [3] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:219292:220887 [3] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:219292:220887 [3] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:219292:220887 [3] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:219292:220887 [3] NCCL INFO Using network Socket +ywang29-vrdb-test1-worker-0:219291:220888 [2] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:219291:220888 [2] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:219291:220888 [2] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:219291:220888 [2] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:219291:220888 [2] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:219291:220888 [2] NCCL INFO Using network Socket +ywang29-vrdb-test1-worker-0:219293:220889 [4] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:219293:220889 [4] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:219293:220889 [4] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:219293:220889 [4] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:219293:220889 [4] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:219293:220889 [4] NCCL INFO Using network Socket +ywang29-vrdb-test1-worker-0:219295:220890 [6] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:219295:220890 [6] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:219295:220890 [6] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:219295:220890 [6] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:219295:220890 [6] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:219295:220890 [6] NCCL INFO Using network Socket +ywang29-vrdb-test1-worker-0:219291:220888 [2] NCCL INFO ncclCommInitRank comm 0x55a29aa0b960 rank 2 nranks 8 cudaDev 2 nvmlDev 2 busId 201c0 commId 0x7f78ceb54e343f2a - Init START +ywang29-vrdb-test1-worker-0:219292:220887 [3] NCCL INFO ncclCommInitRank comm 0x5589685aa560 rank 3 nranks 8 cudaDev 3 nvmlDev 3 busId 201d0 commId 0x7f78ceb54e343f2a - Init START +ywang29-vrdb-test1-worker-0:219293:220889 [4] NCCL INFO ncclCommInitRank comm 0x559a54f01bb0 rank 4 nranks 8 cudaDev 4 nvmlDev 4 busId 901c0 commId 0x7f78ceb54e343f2a - Init START +ywang29-vrdb-test1-worker-0:219294:220884 [5] NCCL INFO ncclCommInitRank comm 0x556a838fe7b0 rank 5 nranks 8 cudaDev 5 nvmlDev 5 busId 901d0 commId 0x7f78ceb54e343f2a - Init START +ywang29-vrdb-test1-worker-0:219290:220885 [1] NCCL INFO ncclCommInitRank comm 0x5562ee327a70 rank 1 nranks 8 cudaDev 1 nvmlDev 1 busId 101d0 commId 0x7f78ceb54e343f2a - Init START +ywang29-vrdb-test1-worker-0:219295:220890 [6] NCCL INFO ncclCommInitRank comm 0x55bc60b89780 rank 6 nranks 8 cudaDev 6 nvmlDev 6 busId a01c0 commId 0x7f78ceb54e343f2a - Init START +ywang29-vrdb-test1-worker-0:219289:220883 [0] NCCL INFO ncclCommInitRank comm 0x564c2935d890 rank 0 nranks 8 cudaDev 0 nvmlDev 0 busId 101c0 commId 0x7f78ceb54e343f2a - Init START +ywang29-vrdb-test1-worker-0:219296:220886 [7] NCCL INFO ncclCommInitRank comm 0x55708fef6a80 rank 7 nranks 8 cudaDev 7 nvmlDev 7 busId a01d0 commId 0x7f78ceb54e343f2a - Init START +ywang29-vrdb-test1-worker-0:219296:220886 [7] NCCL INFO Setting affinity for GPU 7 to ffffff00,0000ffff,ff000000 +ywang29-vrdb-test1-worker-0:219296:220886 [7] NCCL INFO NVLS multicast support is not available on dev 7 +ywang29-vrdb-test1-worker-0:219289:220883 [0] NCCL INFO Setting affinity for GPU 0 to ff,ffff0000,00ffffff +ywang29-vrdb-test1-worker-0:219289:220883 [0] NCCL INFO NVLS multicast support is not available on dev 0 +ywang29-vrdb-test1-worker-0:219290:220885 [1] NCCL INFO Setting affinity for GPU 1 to ff,ffff0000,00ffffff +ywang29-vrdb-test1-worker-0:219290:220885 [1] NCCL INFO NVLS multicast support is not available on dev 1 +ywang29-vrdb-test1-worker-0:219294:220884 [5] NCCL INFO Setting affinity for GPU 5 to ffffff00,0000ffff,ff000000 +ywang29-vrdb-test1-worker-0:219294:220884 [5] NCCL INFO NVLS multicast support is not available on dev 5 +ywang29-vrdb-test1-worker-0:219291:220888 [2] NCCL INFO Setting affinity for GPU 2 to ff,ffff0000,00ffffff +ywang29-vrdb-test1-worker-0:219291:220888 [2] NCCL INFO NVLS multicast support is not available on dev 2 +ywang29-vrdb-test1-worker-0:219292:220887 [3] NCCL INFO Setting affinity for GPU 3 to ff,ffff0000,00ffffff +ywang29-vrdb-test1-worker-0:219292:220887 [3] NCCL INFO NVLS multicast support is not available on dev 3 +ywang29-vrdb-test1-worker-0:219295:220890 [6] NCCL INFO Setting affinity for GPU 6 to ffffff00,0000ffff,ff000000 +ywang29-vrdb-test1-worker-0:219293:220889 [4] NCCL INFO Setting affinity for GPU 4 to ffffff00,0000ffff,ff000000 +ywang29-vrdb-test1-worker-0:219293:220889 [4] NCCL INFO NVLS multicast support is not available on dev 4 +ywang29-vrdb-test1-worker-0:219295:220890 [6] NCCL INFO NVLS multicast support is not available on dev 6 +ywang29-vrdb-test1-worker-0:219295:220890 [6] NCCL INFO comm 0x55bc60b89780 rank 6 nRanks 8 nNodes 1 localRanks 8 localRank 6 MNNVL 0 +ywang29-vrdb-test1-worker-0:219294:220884 [5] NCCL INFO comm 0x556a838fe7b0 rank 5 nRanks 8 nNodes 1 localRanks 8 localRank 5 MNNVL 0 +ywang29-vrdb-test1-worker-0:219293:220889 [4] NCCL INFO comm 0x559a54f01bb0 rank 4 nRanks 8 nNodes 1 localRanks 8 localRank 4 MNNVL 0 +ywang29-vrdb-test1-worker-0:219292:220887 [3] NCCL INFO comm 0x5589685aa560 rank 3 nRanks 8 nNodes 1 localRanks 8 localRank 3 MNNVL 0 +ywang29-vrdb-test1-worker-0:219291:220888 [2] NCCL INFO comm 0x55a29aa0b960 rank 2 nRanks 8 nNodes 1 localRanks 8 localRank 2 MNNVL 0 +ywang29-vrdb-test1-worker-0:219290:220885 [1] NCCL INFO comm 0x5562ee327a70 rank 1 nRanks 8 nNodes 1 localRanks 8 localRank 1 MNNVL 0 +ywang29-vrdb-test1-worker-0:219289:220883 [0] NCCL INFO comm 0x564c2935d890 rank 0 nRanks 8 nNodes 1 localRanks 8 localRank 0 MNNVL 0 +ywang29-vrdb-test1-worker-0:219296:220886 [7] NCCL INFO comm 0x55708fef6a80 rank 7 nRanks 8 nNodes 1 localRanks 8 localRank 7 MNNVL 0 +ywang29-vrdb-test1-worker-0:219289:220883 [0] NCCL INFO Channel 00/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:219289:220883 [0] NCCL INFO Channel 01/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:219289:220883 [0] NCCL INFO Channel 02/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:219289:220883 [0] NCCL INFO Channel 03/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:219292:220887 [3] NCCL INFO Trees [0] 4/-1/-1->3->2 [1] 4/-1/-1->3->2 [2] 4/-1/-1->3->2 [3] 4/-1/-1->3->2 [4] 4/-1/-1->3->2 [5] 4/-1/-1->3->2 [6] 4/-1/-1->3->2 [7] 4/-1/-1->3->2 [8] 4/-1/-1->3->2 [9] 4/-1/-1->3->2 [10] 4/-1/-1->3->2 [11] 4/-1/-1->3->2 [12] 4/-1/-1->3->2 [13] 4/-1/-1->3->2 [14] 4/-1/-1->3->2 [15] 4/-1/-1->3->2 [16] 4/-1/-1->3->2 [17] 4/-1/-1->3->2 [18] 4/-1/-1->3->2 [19] 4/-1/-1->3->2 [20] 4/-1/-1->3->2 [21] 4/-1/-1->3->2 [22] 4/-1/-1->3->2 [23] 4/-1/-1->3->2 +ywang29-vrdb-test1-worker-0:219289:220883 [0] NCCL INFO Channel 04/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:219291:220888 [2] NCCL INFO Trees [0] 3/-1/-1->2->1 [1] 3/-1/-1->2->1 [2] 3/-1/-1->2->1 [3] 3/-1/-1->2->1 [4] 3/-1/-1->2->1 [5] 3/-1/-1->2->1 [6] 3/-1/-1->2->1 [7] 3/-1/-1->2->1 [8] 3/-1/-1->2->1 [9] 3/-1/-1->2->1 [10] 3/-1/-1->2->1 [11] 3/-1/-1->2->1 [12] 3/-1/-1->2->1 [13] 3/-1/-1->2->1 [14] 3/-1/-1->2->1 [15] 3/-1/-1->2->1 [16] 3/-1/-1->2->1 [17] 3/-1/-1->2->1 [18] 3/-1/-1->2->1 [19] 3/-1/-1->2->1 [20] 3/-1/-1->2->1 [21] 3/-1/-1->2->1 [22] 3/-1/-1->2->1 [23] 3/-1/-1->2->1 +ywang29-vrdb-test1-worker-0:219289:220883 [0] NCCL INFO Channel 05/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:219294:220884 [5] NCCL INFO Trees [0] 6/-1/-1->5->4 [1] 6/-1/-1->5->4 [2] 6/-1/-1->5->4 [3] 6/-1/-1->5->4 [4] 6/-1/-1->5->4 [5] 6/-1/-1->5->4 [6] 6/-1/-1->5->4 [7] 6/-1/-1->5->4 [8] 6/-1/-1->5->4 [9] 6/-1/-1->5->4 [10] 6/-1/-1->5->4 [11] 6/-1/-1->5->4 [12] 6/-1/-1->5->4 [13] 6/-1/-1->5->4 [14] 6/-1/-1->5->4 [15] 6/-1/-1->5->4 [16] 6/-1/-1->5->4 [17] 6/-1/-1->5->4 [18] 6/-1/-1->5->4 [19] 6/-1/-1->5->4 [20] 6/-1/-1->5->4 [21] 6/-1/-1->5->4 [22] 6/-1/-1->5->4 [23] 6/-1/-1->5->4 +ywang29-vrdb-test1-worker-0:219292:220887 [3] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:219289:220883 [0] NCCL INFO Channel 06/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:219290:220885 [1] NCCL INFO Trees [0] 2/-1/-1->1->0 [1] 2/-1/-1->1->0 [2] 2/-1/-1->1->0 [3] 2/-1/-1->1->0 [4] 2/-1/-1->1->0 [5] 2/-1/-1->1->0 [6] 2/-1/-1->1->0 [7] 2/-1/-1->1->0 [8] 2/-1/-1->1->0 [9] 2/-1/-1->1->0 [10] 2/-1/-1->1->0 [11] 2/-1/-1->1->0 [12] 2/-1/-1->1->0 [13] 2/-1/-1->1->0 [14] 2/-1/-1->1->0 [15] 2/-1/-1->1->0 [16] 2/-1/-1->1->0 [17] 2/-1/-1->1->0 [18] 2/-1/-1->1->0 [19] 2/-1/-1->1->0 [20] 2/-1/-1->1->0 [21] 2/-1/-1->1->0 [22] 2/-1/-1->1->0 [23] 2/-1/-1->1->0 +ywang29-vrdb-test1-worker-0:219295:220890 [6] NCCL INFO Trees [0] 7/-1/-1->6->5 [1] 7/-1/-1->6->5 [2] 7/-1/-1->6->5 [3] 7/-1/-1->6->5 [4] 7/-1/-1->6->5 [5] 7/-1/-1->6->5 [6] 7/-1/-1->6->5 [7] 7/-1/-1->6->5 [8] 7/-1/-1->6->5 [9] 7/-1/-1->6->5 [10] 7/-1/-1->6->5 [11] 7/-1/-1->6->5 [12] 7/-1/-1->6->5 [13] 7/-1/-1->6->5 [14] 7/-1/-1->6->5 [15] 7/-1/-1->6->5 [16] 7/-1/-1->6->5 [17] 7/-1/-1->6->5 [18] 7/-1/-1->6->5 [19] 7/-1/-1->6->5 [20] 7/-1/-1->6->5 [21] 7/-1/-1->6->5 [22] 7/-1/-1->6->5 [23] 7/-1/-1->6->5 +ywang29-vrdb-test1-worker-0:219291:220888 [2] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:219289:220883 [0] NCCL INFO Channel 07/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:219289:220883 [0] NCCL INFO Channel 08/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:219290:220885 [1] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:219295:220890 [6] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:219289:220883 [0] NCCL INFO Channel 09/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:219293:220889 [4] NCCL INFO Trees [0] 5/-1/-1->4->3 [1] 5/-1/-1->4->3 [2] 5/-1/-1->4->3 [3] 5/-1/-1->4->3 [4] 5/-1/-1->4->3 [5] 5/-1/-1->4->3 [6] 5/-1/-1->4->3 [7] 5/-1/-1->4->3 [8] 5/-1/-1->4->3 [9] 5/-1/-1->4->3 [10] 5/-1/-1->4->3 [11] 5/-1/-1->4->3 [12] 5/-1/-1->4->3 [13] 5/-1/-1->4->3 [14] 5/-1/-1->4->3 [15] 5/-1/-1->4->3 [16] 5/-1/-1->4->3 [17] 5/-1/-1->4->3 [18] 5/-1/-1->4->3 [19] 5/-1/-1->4->3 [20] 5/-1/-1->4->3 [21] 5/-1/-1->4->3 [22] 5/-1/-1->4->3 [23] 5/-1/-1->4->3 +ywang29-vrdb-test1-worker-0:219294:220884 [5] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:219289:220883 [0] NCCL INFO Channel 10/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:219289:220883 [0] NCCL INFO Channel 11/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:219289:220883 [0] NCCL INFO Channel 12/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:219289:220883 [0] NCCL INFO Channel 13/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:219293:220889 [4] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:219289:220883 [0] NCCL INFO Channel 14/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:219289:220883 [0] NCCL INFO Channel 15/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:219296:220886 [7] NCCL INFO Trees [0] -1/-1/-1->7->6 [1] -1/-1/-1->7->6 [2] -1/-1/-1->7->6 [3] -1/-1/-1->7->6 [4] -1/-1/-1->7->6 [5] -1/-1/-1->7->6 [6] -1/-1/-1->7->6 [7] -1/-1/-1->7->6 [8] -1/-1/-1->7->6 [9] -1/-1/-1->7->6 [10] -1/-1/-1->7->6 [11] -1/-1/-1->7->6 [12] -1/-1/-1->7->6 [13] -1/-1/-1->7->6 [14] -1/-1/-1->7->6 [15] -1/-1/-1->7->6 [16] -1/-1/-1->7->6 [17] -1/-1/-1->7->6 [18] -1/-1/-1->7->6 [19] -1/-1/-1->7->6 [20] -1/-1/-1->7->6 [21] -1/-1/-1->7->6 [22] -1/-1/-1->7->6 [23] -1/-1/-1->7->6 +ywang29-vrdb-test1-worker-0:219289:220883 [0] NCCL INFO Channel 16/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:219289:220883 [0] NCCL INFO Channel 17/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:219289:220883 [0] NCCL INFO Channel 18/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:219289:220883 [0] NCCL INFO Channel 19/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:219296:220886 [7] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:219289:220883 [0] NCCL INFO Channel 20/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:219289:220883 [0] NCCL INFO Channel 21/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:219289:220883 [0] NCCL INFO Channel 22/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:219289:220883 [0] NCCL INFO Channel 23/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:219289:220883 [0] NCCL INFO Trees [0] 1/-1/-1->0->-1 [1] 1/-1/-1->0->-1 [2] 1/-1/-1->0->-1 [3] 1/-1/-1->0->-1 [4] 1/-1/-1->0->-1 [5] 1/-1/-1->0->-1 [6] 1/-1/-1->0->-1 [7] 1/-1/-1->0->-1 [8] 1/-1/-1->0->-1 [9] 1/-1/-1->0->-1 [10] 1/-1/-1->0->-1 [11] 1/-1/-1->0->-1 [12] 1/-1/-1->0->-1 [13] 1/-1/-1->0->-1 [14] 1/-1/-1->0->-1 [15] 1/-1/-1->0->-1 [16] 1/-1/-1->0->-1 [17] 1/-1/-1->0->-1 [18] 1/-1/-1->0->-1 [19] 1/-1/-1->0->-1 [20] 1/-1/-1->0->-1 [21] 1/-1/-1->0->-1 [22] 1/-1/-1->0->-1 [23] 1/-1/-1->0->-1 +ywang29-vrdb-test1-worker-0:219289:220883 [0] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:219291:220888 [2] NCCL INFO Channel 00/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219292:220887 [3] NCCL INFO Channel 00/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219293:220889 [4] NCCL INFO Channel 00/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219290:220885 [1] NCCL INFO Channel 00/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219292:220887 [3] NCCL INFO Channel 01/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219289:220883 [0] NCCL INFO Channel 00/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219291:220888 [2] NCCL INFO Channel 01/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219293:220889 [4] NCCL INFO Channel 01/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219290:220885 [1] NCCL INFO Channel 01/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219292:220887 [3] NCCL INFO Channel 02/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219289:220883 [0] NCCL INFO Channel 01/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219291:220888 [2] NCCL INFO Channel 02/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219293:220889 [4] NCCL INFO Channel 02/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219290:220885 [1] NCCL INFO Channel 02/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219292:220887 [3] NCCL INFO Channel 03/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219289:220883 [0] NCCL INFO Channel 02/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219291:220888 [2] NCCL INFO Channel 03/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219293:220889 [4] NCCL INFO Channel 03/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219294:220884 [5] NCCL INFO Channel 00/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219290:220885 [1] NCCL INFO Channel 03/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219292:220887 [3] NCCL INFO Channel 04/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219289:220883 [0] NCCL INFO Channel 03/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219296:220886 [7] NCCL INFO Channel 00/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219291:220888 [2] NCCL INFO Channel 04/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219293:220889 [4] NCCL INFO Channel 04/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219294:220884 [5] NCCL INFO Channel 01/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219290:220885 [1] NCCL INFO Channel 04/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219292:220887 [3] NCCL INFO Channel 05/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219289:220883 [0] NCCL INFO Channel 04/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219296:220886 [7] NCCL INFO Channel 01/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219291:220888 [2] NCCL INFO Channel 05/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219293:220889 [4] NCCL INFO Channel 05/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219294:220884 [5] NCCL INFO Channel 02/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219290:220885 [1] NCCL INFO Channel 05/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219292:220887 [3] NCCL INFO Channel 06/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219289:220883 [0] NCCL INFO Channel 05/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219291:220888 [2] NCCL INFO Channel 06/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219296:220886 [7] NCCL INFO Channel 02/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219293:220889 [4] NCCL INFO Channel 06/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219294:220884 [5] NCCL INFO Channel 03/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219290:220885 [1] NCCL INFO Channel 06/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219292:220887 [3] NCCL INFO Channel 07/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219289:220883 [0] NCCL INFO Channel 06/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219291:220888 [2] NCCL INFO Channel 07/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219296:220886 [7] NCCL INFO Channel 03/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219293:220889 [4] NCCL INFO Channel 07/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219290:220885 [1] NCCL INFO Channel 07/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219294:220884 [5] NCCL INFO Channel 04/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219292:220887 [3] NCCL INFO Channel 08/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219289:220883 [0] NCCL INFO Channel 07/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219293:220889 [4] NCCL INFO Channel 08/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219291:220888 [2] NCCL INFO Channel 08/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219296:220886 [7] NCCL INFO Channel 04/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219290:220885 [1] NCCL INFO Channel 08/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219294:220884 [5] NCCL INFO Channel 05/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219292:220887 [3] NCCL INFO Channel 09/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219289:220883 [0] NCCL INFO Channel 08/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219293:220889 [4] NCCL INFO Channel 09/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219296:220886 [7] NCCL INFO Channel 05/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219291:220888 [2] NCCL INFO Channel 09/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219290:220885 [1] NCCL INFO Channel 09/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219294:220884 [5] NCCL INFO Channel 06/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219292:220887 [3] NCCL INFO Channel 10/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219289:220883 [0] NCCL INFO Channel 09/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219293:220889 [4] NCCL INFO Channel 10/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219296:220886 [7] NCCL INFO Channel 06/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219291:220888 [2] NCCL INFO Channel 10/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219290:220885 [1] NCCL INFO Channel 10/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219294:220884 [5] NCCL INFO Channel 07/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219292:220887 [3] NCCL INFO Channel 11/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219289:220883 [0] NCCL INFO Channel 10/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219293:220889 [4] NCCL INFO Channel 11/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219291:220888 [2] NCCL INFO Channel 11/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219296:220886 [7] NCCL INFO Channel 07/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219290:220885 [1] NCCL INFO Channel 11/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219292:220887 [3] NCCL INFO Channel 12/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219294:220884 [5] NCCL INFO Channel 08/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219289:220883 [0] NCCL INFO Channel 11/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219293:220889 [4] NCCL INFO Channel 12/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219290:220885 [1] NCCL INFO Channel 12/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219291:220888 [2] NCCL INFO Channel 12/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219296:220886 [7] NCCL INFO Channel 08/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219295:220890 [6] NCCL INFO Channel 00/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219292:220887 [3] NCCL INFO Channel 13/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219294:220884 [5] NCCL INFO Channel 09/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219289:220883 [0] NCCL INFO Channel 12/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219293:220889 [4] NCCL INFO Channel 13/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219290:220885 [1] NCCL INFO Channel 13/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219291:220888 [2] NCCL INFO Channel 13/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219296:220886 [7] NCCL INFO Channel 09/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219295:220890 [6] NCCL INFO Channel 01/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219292:220887 [3] NCCL INFO Channel 14/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219289:220883 [0] NCCL INFO Channel 13/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219294:220884 [5] NCCL INFO Channel 10/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219293:220889 [4] NCCL INFO Channel 14/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219290:220885 [1] NCCL INFO Channel 14/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219291:220888 [2] NCCL INFO Channel 14/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219296:220886 [7] NCCL INFO Channel 10/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219295:220890 [6] NCCL INFO Channel 02/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219292:220887 [3] NCCL INFO Channel 15/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219289:220883 [0] NCCL INFO Channel 14/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219290:220885 [1] NCCL INFO Channel 15/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219294:220884 [5] NCCL INFO Channel 11/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219293:220889 [4] NCCL INFO Channel 15/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219291:220888 [2] NCCL INFO Channel 15/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219296:220886 [7] NCCL INFO Channel 11/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219295:220890 [6] NCCL INFO Channel 03/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219292:220887 [3] NCCL INFO Channel 16/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219289:220883 [0] NCCL INFO Channel 15/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219290:220885 [1] NCCL INFO Channel 16/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219294:220884 [5] NCCL INFO Channel 12/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219293:220889 [4] NCCL INFO Channel 16/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219291:220888 [2] NCCL INFO Channel 16/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219296:220886 [7] NCCL INFO Channel 12/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219295:220890 [6] NCCL INFO Channel 04/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219292:220887 [3] NCCL INFO Channel 17/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219289:220883 [0] NCCL INFO Channel 16/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219290:220885 [1] NCCL INFO Channel 17/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219294:220884 [5] NCCL INFO Channel 13/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219293:220889 [4] NCCL INFO Channel 17/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219291:220888 [2] NCCL INFO Channel 17/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219296:220886 [7] NCCL INFO Channel 13/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219295:220890 [6] NCCL INFO Channel 05/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219292:220887 [3] NCCL INFO Channel 18/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219289:220883 [0] NCCL INFO Channel 17/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219290:220885 [1] NCCL INFO Channel 18/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219294:220884 [5] NCCL INFO Channel 14/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219291:220888 [2] NCCL INFO Channel 18/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219293:220889 [4] NCCL INFO Channel 18/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219296:220886 [7] NCCL INFO Channel 14/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219295:220890 [6] NCCL INFO Channel 06/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219292:220887 [3] NCCL INFO Channel 19/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219289:220883 [0] NCCL INFO Channel 18/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219290:220885 [1] NCCL INFO Channel 19/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219294:220884 [5] NCCL INFO Channel 15/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219291:220888 [2] NCCL INFO Channel 19/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219293:220889 [4] NCCL INFO Channel 19/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219295:220890 [6] NCCL INFO Channel 07/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219296:220886 [7] NCCL INFO Channel 15/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219292:220887 [3] NCCL INFO Channel 20/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219289:220883 [0] NCCL INFO Channel 19/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219290:220885 [1] NCCL INFO Channel 20/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219294:220884 [5] NCCL INFO Channel 16/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219291:220888 [2] NCCL INFO Channel 20/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219293:220889 [4] NCCL INFO Channel 20/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219296:220886 [7] NCCL INFO Channel 16/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219295:220890 [6] NCCL INFO Channel 08/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219292:220887 [3] NCCL INFO Channel 21/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219289:220883 [0] NCCL INFO Channel 20/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219290:220885 [1] NCCL INFO Channel 21/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219294:220884 [5] NCCL INFO Channel 17/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219291:220888 [2] NCCL INFO Channel 21/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219293:220889 [4] NCCL INFO Channel 21/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219296:220886 [7] NCCL INFO Channel 17/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219295:220890 [6] NCCL INFO Channel 09/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219292:220887 [3] NCCL INFO Channel 22/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219289:220883 [0] NCCL INFO Channel 21/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219290:220885 [1] NCCL INFO Channel 22/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219291:220888 [2] NCCL INFO Channel 22/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219294:220884 [5] NCCL INFO Channel 18/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219293:220889 [4] NCCL INFO Channel 22/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219296:220886 [7] NCCL INFO Channel 18/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219295:220890 [6] NCCL INFO Channel 10/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219292:220887 [3] NCCL INFO Channel 23/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219289:220883 [0] NCCL INFO Channel 22/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219290:220885 [1] NCCL INFO Channel 23/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219293:220889 [4] NCCL INFO Channel 23/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219291:220888 [2] NCCL INFO Channel 23/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219294:220884 [5] NCCL INFO Channel 19/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219296:220886 [7] NCCL INFO Channel 19/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219295:220890 [6] NCCL INFO Channel 11/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219289:220883 [0] NCCL INFO Channel 23/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219294:220884 [5] NCCL INFO Channel 20/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219295:220890 [6] NCCL INFO Channel 12/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219296:220886 [7] NCCL INFO Channel 20/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219294:220884 [5] NCCL INFO Channel 21/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219296:220886 [7] NCCL INFO Channel 21/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219295:220890 [6] NCCL INFO Channel 13/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219294:220884 [5] NCCL INFO Channel 22/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219295:220890 [6] NCCL INFO Channel 14/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219294:220884 [5] NCCL INFO Channel 23/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219295:220890 [6] NCCL INFO Channel 15/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219296:220886 [7] NCCL INFO Channel 22/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219295:220890 [6] NCCL INFO Channel 16/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219296:220886 [7] NCCL INFO Channel 23/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219295:220890 [6] NCCL INFO Channel 17/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219295:220890 [6] NCCL INFO Channel 18/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219295:220890 [6] NCCL INFO Channel 19/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219295:220890 [6] NCCL INFO Channel 20/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219295:220890 [6] NCCL INFO Channel 21/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219295:220890 [6] NCCL INFO Channel 22/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219295:220890 [6] NCCL INFO Channel 23/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219291:220888 [2] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:219292:220887 [3] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:219293:220889 [4] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:219290:220885 [1] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:219289:220883 [0] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:219291:220888 [2] NCCL INFO Channel 00/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219291:220888 [2] NCCL INFO Channel 01/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219296:220886 [7] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:219296:220886 [7] NCCL INFO Channel 00/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219291:220888 [2] NCCL INFO Channel 02/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219294:220884 [5] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:219295:220890 [6] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:219296:220886 [7] NCCL INFO Channel 01/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219291:220888 [2] NCCL INFO Channel 03/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219296:220886 [7] NCCL INFO Channel 02/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219291:220888 [2] NCCL INFO Channel 04/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219292:220887 [3] NCCL INFO Channel 00/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219296:220886 [7] NCCL INFO Channel 03/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219291:220888 [2] NCCL INFO Channel 05/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219292:220887 [3] NCCL INFO Channel 01/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219296:220886 [7] NCCL INFO Channel 04/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219291:220888 [2] NCCL INFO Channel 06/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219292:220887 [3] NCCL INFO Channel 02/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219296:220886 [7] NCCL INFO Channel 05/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219291:220888 [2] NCCL INFO Channel 07/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219292:220887 [3] NCCL INFO Channel 03/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219296:220886 [7] NCCL INFO Channel 06/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219291:220888 [2] NCCL INFO Channel 08/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219292:220887 [3] NCCL INFO Channel 04/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219296:220886 [7] NCCL INFO Channel 07/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219291:220888 [2] NCCL INFO Channel 09/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219292:220887 [3] NCCL INFO Channel 05/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219296:220886 [7] NCCL INFO Channel 08/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219291:220888 [2] NCCL INFO Channel 10/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219292:220887 [3] NCCL INFO Channel 06/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219296:220886 [7] NCCL INFO Channel 09/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219291:220888 [2] NCCL INFO Channel 11/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219292:220887 [3] NCCL INFO Channel 07/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219296:220886 [7] NCCL INFO Channel 10/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219291:220888 [2] NCCL INFO Channel 12/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219292:220887 [3] NCCL INFO Channel 08/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219296:220886 [7] NCCL INFO Channel 11/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219291:220888 [2] NCCL INFO Channel 13/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219292:220887 [3] NCCL INFO Channel 09/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219296:220886 [7] NCCL INFO Channel 12/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219291:220888 [2] NCCL INFO Channel 14/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219292:220887 [3] NCCL INFO Channel 10/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219296:220886 [7] NCCL INFO Channel 13/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219291:220888 [2] NCCL INFO Channel 15/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219292:220887 [3] NCCL INFO Channel 11/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219296:220886 [7] NCCL INFO Channel 14/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219291:220888 [2] NCCL INFO Channel 16/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219292:220887 [3] NCCL INFO Channel 12/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219291:220888 [2] NCCL INFO Channel 17/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219296:220886 [7] NCCL INFO Channel 15/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219292:220887 [3] NCCL INFO Channel 13/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219291:220888 [2] NCCL INFO Channel 18/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219296:220886 [7] NCCL INFO Channel 16/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219293:220889 [4] NCCL INFO Channel 00/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219292:220887 [3] NCCL INFO Channel 14/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219291:220888 [2] NCCL INFO Channel 19/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219296:220886 [7] NCCL INFO Channel 17/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219293:220889 [4] NCCL INFO Channel 01/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219292:220887 [3] NCCL INFO Channel 15/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219291:220888 [2] NCCL INFO Channel 20/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219296:220886 [7] NCCL INFO Channel 18/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219293:220889 [4] NCCL INFO Channel 02/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219292:220887 [3] NCCL INFO Channel 16/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219291:220888 [2] NCCL INFO Channel 21/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219296:220886 [7] NCCL INFO Channel 19/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219293:220889 [4] NCCL INFO Channel 03/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219292:220887 [3] NCCL INFO Channel 17/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219291:220888 [2] NCCL INFO Channel 22/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219290:220885 [1] NCCL INFO Channel 00/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219296:220886 [7] NCCL INFO Channel 20/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219293:220889 [4] NCCL INFO Channel 04/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219292:220887 [3] NCCL INFO Channel 18/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219291:220888 [2] NCCL INFO Channel 23/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219290:220885 [1] NCCL INFO Channel 01/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219296:220886 [7] NCCL INFO Channel 21/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219293:220889 [4] NCCL INFO Channel 05/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219292:220887 [3] NCCL INFO Channel 19/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219290:220885 [1] NCCL INFO Channel 02/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219296:220886 [7] NCCL INFO Channel 22/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219293:220889 [4] NCCL INFO Channel 06/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219292:220887 [3] NCCL INFO Channel 20/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219290:220885 [1] NCCL INFO Channel 03/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219296:220886 [7] NCCL INFO Channel 23/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219293:220889 [4] NCCL INFO Channel 07/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219292:220887 [3] NCCL INFO Channel 21/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219290:220885 [1] NCCL INFO Channel 04/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219293:220889 [4] NCCL INFO Channel 08/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219292:220887 [3] NCCL INFO Channel 22/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219295:220890 [6] NCCL INFO Channel 00/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219290:220885 [1] NCCL INFO Channel 05/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219293:220889 [4] NCCL INFO Channel 09/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219292:220887 [3] NCCL INFO Channel 23/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219295:220890 [6] NCCL INFO Channel 01/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219294:220884 [5] NCCL INFO Channel 00/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219290:220885 [1] NCCL INFO Channel 06/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219293:220889 [4] NCCL INFO Channel 10/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219295:220890 [6] NCCL INFO Channel 02/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219294:220884 [5] NCCL INFO Channel 01/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219290:220885 [1] NCCL INFO Channel 07/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219293:220889 [4] NCCL INFO Channel 11/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219295:220890 [6] NCCL INFO Channel 03/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219294:220884 [5] NCCL INFO Channel 02/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219290:220885 [1] NCCL INFO Channel 08/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219293:220889 [4] NCCL INFO Channel 12/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219295:220890 [6] NCCL INFO Channel 04/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219294:220884 [5] NCCL INFO Channel 03/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219290:220885 [1] NCCL INFO Channel 09/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219293:220889 [4] NCCL INFO Channel 13/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219295:220890 [6] NCCL INFO Channel 05/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219290:220885 [1] NCCL INFO Channel 10/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219293:220889 [4] NCCL INFO Channel 14/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219295:220890 [6] NCCL INFO Channel 06/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219290:220885 [1] NCCL INFO Channel 11/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219293:220889 [4] NCCL INFO Channel 15/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219295:220890 [6] NCCL INFO Channel 07/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219290:220885 [1] NCCL INFO Channel 12/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219293:220889 [4] NCCL INFO Channel 16/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219295:220890 [6] NCCL INFO Channel 08/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219290:220885 [1] NCCL INFO Channel 13/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219293:220889 [4] NCCL INFO Channel 17/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219295:220890 [6] NCCL INFO Channel 09/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219290:220885 [1] NCCL INFO Channel 14/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219293:220889 [4] NCCL INFO Channel 18/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219295:220890 [6] NCCL INFO Channel 10/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219290:220885 [1] NCCL INFO Channel 15/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219293:220889 [4] NCCL INFO Channel 19/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219295:220890 [6] NCCL INFO Channel 11/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219290:220885 [1] NCCL INFO Channel 16/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219293:220889 [4] NCCL INFO Channel 20/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219295:220890 [6] NCCL INFO Channel 12/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219290:220885 [1] NCCL INFO Channel 17/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219293:220889 [4] NCCL INFO Channel 21/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219295:220890 [6] NCCL INFO Channel 13/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219290:220885 [1] NCCL INFO Channel 18/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219293:220889 [4] NCCL INFO Channel 22/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219295:220890 [6] NCCL INFO Channel 14/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219290:220885 [1] NCCL INFO Channel 19/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219293:220889 [4] NCCL INFO Channel 23/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219295:220890 [6] NCCL INFO Channel 15/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219290:220885 [1] NCCL INFO Channel 20/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219295:220890 [6] NCCL INFO Channel 16/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219290:220885 [1] NCCL INFO Channel 21/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219295:220890 [6] NCCL INFO Channel 17/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219290:220885 [1] NCCL INFO Channel 22/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219295:220890 [6] NCCL INFO Channel 18/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219290:220885 [1] NCCL INFO Channel 23/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219294:220884 [5] NCCL INFO Channel 04/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219295:220890 [6] NCCL INFO Channel 19/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219294:220884 [5] NCCL INFO Channel 05/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219295:220890 [6] NCCL INFO Channel 20/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219294:220884 [5] NCCL INFO Channel 06/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219295:220890 [6] NCCL INFO Channel 21/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219294:220884 [5] NCCL INFO Channel 07/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219295:220890 [6] NCCL INFO Channel 22/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219294:220884 [5] NCCL INFO Channel 08/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219295:220890 [6] NCCL INFO Channel 23/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219294:220884 [5] NCCL INFO Channel 09/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219294:220884 [5] NCCL INFO Channel 10/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219294:220884 [5] NCCL INFO Channel 11/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219294:220884 [5] NCCL INFO Channel 12/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219294:220884 [5] NCCL INFO Channel 13/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219294:220884 [5] NCCL INFO Channel 14/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219294:220884 [5] NCCL INFO Channel 15/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219294:220884 [5] NCCL INFO Channel 16/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219294:220884 [5] NCCL INFO Channel 17/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219294:220884 [5] NCCL INFO Channel 18/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219294:220884 [5] NCCL INFO Channel 19/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219294:220884 [5] NCCL INFO Channel 20/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219294:220884 [5] NCCL INFO Channel 21/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219294:220884 [5] NCCL INFO Channel 22/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219294:220884 [5] NCCL INFO Channel 23/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:219289:220883 [0] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:219289:220883 [0] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:219289:220883 [0] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:219290:220885 [1] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:219290:220885 [1] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:219290:220885 [1] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:219291:220888 [2] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:219291:220888 [2] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:219291:220888 [2] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:219292:220887 [3] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:219292:220887 [3] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:219292:220887 [3] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:219293:220889 [4] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:219293:220889 [4] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:219293:220889 [4] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:219294:220884 [5] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:219294:220884 [5] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:219294:220884 [5] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:219296:220886 [7] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:219296:220886 [7] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:219296:220886 [7] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:219295:220890 [6] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:219295:220890 [6] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:219295:220890 [6] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:219293:220889 [4] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:219293:220889 [4] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:219293:220889 [4] NCCL INFO ncclCommInitRank comm 0x559a54f01bb0 rank 4 nranks 8 cudaDev 4 nvmlDev 4 busId 901c0 commId 0x7f78ceb54e343f2a - Init COMPLETE +ywang29-vrdb-test1-worker-0:219296:220886 [7] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:219289:220883 [0] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:219295:220890 [6] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:219296:220886 [7] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:219289:220883 [0] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:219292:220887 [3] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:219295:220890 [6] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:219290:220885 [1] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:219292:220887 [3] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:219291:220888 [2] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:219289:220883 [0] NCCL INFO ncclCommInitRank comm 0x564c2935d890 rank 0 nranks 8 cudaDev 0 nvmlDev 0 busId 101c0 commId 0x7f78ceb54e343f2a - Init COMPLETE +ywang29-vrdb-test1-worker-0:219296:220886 [7] NCCL INFO ncclCommInitRank comm 0x55708fef6a80 rank 7 nranks 8 cudaDev 7 nvmlDev 7 busId a01d0 commId 0x7f78ceb54e343f2a - Init COMPLETE +ywang29-vrdb-test1-worker-0:219290:220885 [1] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:219292:220887 [3] NCCL INFO ncclCommInitRank comm 0x5589685aa560 rank 3 nranks 8 cudaDev 3 nvmlDev 3 busId 201d0 commId 0x7f78ceb54e343f2a - Init COMPLETE +ywang29-vrdb-test1-worker-0:219295:220890 [6] NCCL INFO ncclCommInitRank comm 0x55bc60b89780 rank 6 nranks 8 cudaDev 6 nvmlDev 6 busId a01c0 commId 0x7f78ceb54e343f2a - Init COMPLETE +ywang29-vrdb-test1-worker-0:219291:220888 [2] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:219294:220884 [5] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:219290:220885 [1] NCCL INFO ncclCommInitRank comm 0x5562ee327a70 rank 1 nranks 8 cudaDev 1 nvmlDev 1 busId 101d0 commId 0x7f78ceb54e343f2a - Init COMPLETE +ywang29-vrdb-test1-worker-0:219291:220888 [2] NCCL INFO ncclCommInitRank comm 0x55a29aa0b960 rank 2 nranks 8 cudaDev 2 nvmlDev 2 busId 201c0 commId 0x7f78ceb54e343f2a - Init COMPLETE +ywang29-vrdb-test1-worker-0:219294:220884 [5] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:219294:220884 [5] NCCL INFO ncclCommInitRank comm 0x556a838fe7b0 rank 5 nranks 8 cudaDev 5 nvmlDev 5 busId 901d0 commId 0x7f78ceb54e343f2a - Init COMPLETE +[2025-10-12 08:20:31,004] [INFO] [partition_parameters.py:343:__exit__] finished initializing model - num_params = 459, num_elems = 0.99B +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +[E ProcessGroupNCCL.cpp:474] [Rank 3] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=1858, OpType=BROADCAST, NumelIn=677376, NumelOut=677376, Timeout(ms)=1800000) ran for 1800584 milliseconds before timing out. +[E ProcessGroupNCCL.cpp:474] [Rank 7] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=1858, OpType=BROADCAST, NumelIn=677376, NumelOut=677376, Timeout(ms)=1800000) ran for 1800621 milliseconds before timing out. +ywang29-vrdb-test1-worker-0:219296:220903 [7] NCCL INFO [Service thread] Connection closed by localRank 7 +ywang29-vrdb-test1-worker-0:219296:220394 [7] NCCL INFO comm 0x55708fef6a80 rank 7 nranks 8 cudaDev 7 busId a01d0 - Abort COMPLETE +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +[E ProcessGroupNCCL.cpp:488] Some NCCL operations have failed or timed out. Due to the asynchronous nature of CUDA kernels, subsequent GPU operations might run on corrupted/incomplete data. +Traceback (most recent call last): +[E ProcessGroupNCCL.cpp:494] To avoid data inconsistency, we are taking the entire process down. +ywang29-vrdb-test1-worker-0:219292:220891 [3] NCCL INFO [Service thread] Connection closed by localRank 3 + File "/nfs/ywang29/TinyLLaVA/tinyllava/train/train.py", line 193, in +[E ProcessGroupNCCL.cpp:915] [Rank 7] NCCL watchdog thread terminated with exception: [Rank 7] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=1858, OpType=BROADCAST, NumelIn=677376, NumelOut=677376, Timeout(ms)=1800000) ran for 1800621 milliseconds before timing out. +terminate called after throwing an instance of 'std::runtime_error' + what(): [Rank 7] NCCL watchdog thread terminated with exception: [Rank 7] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=1858, OpType=BROADCAST, NumelIn=677376, NumelOut=677376, Timeout(ms)=1800000) ran for 1800621 milliseconds before timing out. +ywang29-vrdb-test1-worker-0:219292:220425 [3] NCCL INFO comm 0x5589685aa560 rank 3 nranks 8 cudaDev 3 busId 201d0 - Abort COMPLETE +[E ProcessGroupNCCL.cpp:488] Some NCCL operations have failed or timed out. Due to the asynchronous nature of CUDA kernels, subsequent GPU operations might run on corrupted/incomplete data. +[E ProcessGroupNCCL.cpp:494] To avoid data inconsistency, we are taking the entire process down. +[E ProcessGroupNCCL.cpp:915] [Rank 3] NCCL watchdog thread terminated with exception: [Rank 3] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=1858, OpType=BROADCAST, NumelIn=677376, NumelOut=677376, Timeout(ms)=1800000) ran for 1800584 milliseconds before timing out. +terminate called after throwing an instance of 'std::runtime_error' + what(): [Rank 3] NCCL watchdog thread terminated with exception: [Rank 3] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=1858, OpType=BROADCAST, NumelIn=677376, NumelOut=677376, Timeout(ms)=1800000) ran for 1800584 milliseconds before timing out. +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +[2025-10-12 09:02:31,676] [INFO] [partition_parameters.py:343:__exit__] finished initializing model - num_params = 907, num_elems = 1.42B +[2025-10-12 09:02:42,351] [INFO] [launch.py:316:sigkill_handler] Killing subprocess 219289 +[2025-10-12 09:02:42,688] [INFO] [launch.py:316:sigkill_handler] Killing subprocess 219290 +[2025-10-12 09:02:43,025] [INFO] [launch.py:316:sigkill_handler] Killing subprocess 219291 +[2025-10-12 09:02:43,402] [INFO] [launch.py:316:sigkill_handler] Killing subprocess 219292 +[2025-10-12 09:02:43,404] [INFO] [launch.py:316:sigkill_handler] Killing subprocess 219293 +[2025-10-12 09:02:43,820] [INFO] [launch.py:316:sigkill_handler] Killing subprocess 219294 +[2025-10-12 09:02:44,197] [INFO] [launch.py:316:sigkill_handler] Killing subprocess 219295 +[2025-10-12 09:02:44,534] [INFO] [launch.py:316:sigkill_handler] Killing subprocess 219296 +[2025-10-12 09:02:44,534] [ERROR] [launch.py:322:sigkill_handler] ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=7', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.3_2e-1_connector-1.0_2.3_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '1.0', '--temperature_attn_text', '2.3', '--temperature_mlp_text', '2.3', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '1.0', '--temperature_attn_vision', '2.3', '--temperature_mlp_vision', '2.3', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '1.0', '--temperature_connector', '2.3', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] exits with return code = -6 +==== EXPERIMENT COMPLETED: qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.3_2e-1_connector-1.0_2.3_2e-1_ablation ==== +Log File: qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.3_2e-1_connector-1.0_2.3_2e-1_ablation_20251012_081931.log +Timestamp: 2025-10-12 09:02:45 +===================================== diff --git a/logs_oct11/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.5_2e-1_connector-1.0_2.5_2e-1_ablation_20251012_090245.log b/logs_oct11/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.5_2e-1_connector-1.0_2.5_2e-1_ablation_20251012_090245.log new file mode 100644 index 0000000000000000000000000000000000000000..d326b5ae01ca4259b8c91569b847b6fc3a2baf9d --- /dev/null +++ b/logs_oct11/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.5_2e-1_connector-1.0_2.5_2e-1_ablation_20251012_090245.log @@ -0,0 +1,2312 @@ +==== STARTING EXPERIMENT: qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.5_2e-1_connector-1.0_2.5_2e-1_ablation ==== +Log File: qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.5_2e-1_connector-1.0_2.5_2e-1_ablation_20251012_090245.log +Timestamp: 2025-10-12 09:02:45 +===================================== +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 09:02:48,481] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 09:02:51,154] [WARNING] [runner.py:202:fetch_hostfile] Unable to find hostfile, will proceed with training with local resources only. +[2025-10-12 09:02:51,156] [INFO] [runner.py:568:main] cmd = /opt/conda/envs/tinyllava/bin/python3.10 -u -m deepspeed.launcher.launch --world_info=eyJsb2NhbGhvc3QiOiBbMCwgMSwgMiwgMywgNCwgNSwgNiwgN119 --master_addr=127.0.0.1 --master_port=29501 --enable_each_rank_log=None tinyllava/train/train.py --deepspeed ./scripts/zero3.json --data_path /s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json --image_folder /s3-code/ywang29/datasets/tinyllava --is_multimodal True --conv_version qwen2_base --model_name_or_path Qwen/Qwen2.5-0.5B --vision_tower google/siglip-so400m-patch14-384 --vision_tower2 --connector_type mlp2x_gelu --mm_vision_select_layer -2 --image_aspect_ratio square --attn_implementation flash_attention_2 --bf16 True --training_recipe common --tune_type_llm full --tune_type_vision_tower frozen --tune_vision_tower_from_layer 0 --tune_type_connector full --group_by_modality_length True --pretrained_model_path /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain --output_dir /nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.5_2e-1_connector-1.0_2.5_2e-1_ablation --num_train_epochs 1 --per_device_train_batch_size 4 --per_device_eval_batch_size 4 --gradient_accumulation_steps 4 --evaluation_strategy no --learning_rate 2e-1 --weight_decay 0. --warmup_ratio 0.03 --lr_scheduler_type cosine --logging_steps 1 --tf32 False --model_max_length 2048 --gradient_checkpointing True --dataloader_num_workers 8 --lazy_preprocess True --report_to tensorboard --tokenizer_use_fast False --run_name tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune --subnet_mode_text both --subnet_type_text None --mask_type_text soft --init_mean_text 1.0 --temperature_attn_text 2.5 --temperature_mlp_text 2.5 --backward_type_text normal --masked_layers_text all --subnet_mode_vision both --subnet_type_vision None --mask_type_vision soft --init_mean_vision 1.0 --temperature_attn_vision 2.5 --temperature_mlp_vision 2.5 --backward_type_vision normal --masked_layers_vision all --subnet_type_connector global --mask_type_connector soft --init_mean_connector 1.0 --temperature_connector 2.5 --backward_type_connector normal --mm_projector_lr 2e-1 --seed 42 --mask_model llm-connector --save_strategy steps --save_steps 50000 --save_total_limit 1 --train_data_ratio 0.1 +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 09:02:53,780] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 09:02:54,790] [INFO] [launch.py:138:main] 0 NCCL_VERSION=2.21.5 +[2025-10-12 09:02:54,790] [INFO] [launch.py:138:main] 0 NCCL_SOCKET_IFNAME=eth +[2025-10-12 09:02:54,790] [INFO] [launch.py:145:main] WORLD INFO DICT: {'localhost': [0, 1, 2, 3, 4, 5, 6, 7]} +[2025-10-12 09:02:54,790] [INFO] [launch.py:151:main] nnodes=1, num_local_procs=8, node_rank=0 +[2025-10-12 09:02:54,790] [INFO] [launch.py:162:main] global_rank_mapping=defaultdict(, {'localhost': [0, 1, 2, 3, 4, 5, 6, 7]}) +[2025-10-12 09:02:54,790] [INFO] [launch.py:163:main] dist_world_size=8 +[2025-10-12 09:02:54,790] [INFO] [launch.py:165:main] Setting CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 +[2025-10-12 09:02:54,793] [INFO] [launch.py:253:main] process 221839 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=0', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.5_2e-1_connector-1.0_2.5_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '1.0', '--temperature_attn_text', '2.5', '--temperature_mlp_text', '2.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '1.0', '--temperature_attn_vision', '2.5', '--temperature_mlp_vision', '2.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '1.0', '--temperature_connector', '2.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 09:02:54,795] [INFO] [launch.py:253:main] process 221840 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=1', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.5_2e-1_connector-1.0_2.5_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '1.0', '--temperature_attn_text', '2.5', '--temperature_mlp_text', '2.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '1.0', '--temperature_attn_vision', '2.5', '--temperature_mlp_vision', '2.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '1.0', '--temperature_connector', '2.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 09:02:54,797] [INFO] [launch.py:253:main] process 221841 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=2', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.5_2e-1_connector-1.0_2.5_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '1.0', '--temperature_attn_text', '2.5', '--temperature_mlp_text', '2.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '1.0', '--temperature_attn_vision', '2.5', '--temperature_mlp_vision', '2.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '1.0', '--temperature_connector', '2.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 09:02:54,799] [INFO] [launch.py:253:main] process 221842 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=3', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.5_2e-1_connector-1.0_2.5_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '1.0', '--temperature_attn_text', '2.5', '--temperature_mlp_text', '2.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '1.0', '--temperature_attn_vision', '2.5', '--temperature_mlp_vision', '2.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '1.0', '--temperature_connector', '2.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 09:02:54,802] [INFO] [launch.py:253:main] process 221843 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=4', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.5_2e-1_connector-1.0_2.5_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '1.0', '--temperature_attn_text', '2.5', '--temperature_mlp_text', '2.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '1.0', '--temperature_attn_vision', '2.5', '--temperature_mlp_vision', '2.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '1.0', '--temperature_connector', '2.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 09:02:54,804] [INFO] [launch.py:253:main] process 221844 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=5', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.5_2e-1_connector-1.0_2.5_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '1.0', '--temperature_attn_text', '2.5', '--temperature_mlp_text', '2.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '1.0', '--temperature_attn_vision', '2.5', '--temperature_mlp_vision', '2.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '1.0', '--temperature_connector', '2.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 09:02:54,806] [INFO] [launch.py:253:main] process 221845 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=6', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.5_2e-1_connector-1.0_2.5_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '1.0', '--temperature_attn_text', '2.5', '--temperature_mlp_text', '2.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '1.0', '--temperature_attn_vision', '2.5', '--temperature_mlp_vision', '2.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '1.0', '--temperature_connector', '2.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 09:02:54,808] [INFO] [launch.py:253:main] process 221846 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=7', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.5_2e-1_connector-1.0_2.5_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '1.0', '--temperature_attn_text', '2.5', '--temperature_mlp_text', '2.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '1.0', '--temperature_attn_vision', '2.5', '--temperature_mlp_vision', '2.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '1.0', '--temperature_connector', '2.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 09:03:01,286] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 09:03:01,547] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 09:03:01,547] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 09:03:01,591] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 09:03:01,597] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 09:03:01,599] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 09:03:01,607] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 09:03:01,607] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 09:03:01,693] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 09:03:01,949] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 09:03:01,949] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 09:03:01,994] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 09:03:01,996] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 09:03:01,996] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 09:03:01,997] [INFO] [comm.py:668:init_distributed] Initializing TorchBackend in DeepSpeed with backend nccl +[2025-10-12 09:03:02,008] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 09:03:02,008] [INFO] [comm.py:637:init_distributed] cdb=None +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +{'llm': {'model_name_or_path': 'Qwen/Qwen2.5-0.5B', 'cache_dir': None, 'attn_implementation': 'flash_attention_2', 'subnet_mode': 'both', 'subnet_type': 'None', 'sparsity_attn': None, 'sparsity_mlp': None, 'threshold_attn': None, 'threshold_mlp': None, 'temperature_attn': 2.5, 'temperature_mlp': 2.5, 'masked_layers': 'all', 'mask_type': 'soft', 'backward_type': 'normal'}, 'vision_tower': {'model_name_or_path': 'google/siglip-so400m-patch14-384'}, 'connector': {'connector_type': 'mlp2x_gelu', 'subnet_type': 'global', 'threshold': None, 'sparsity': None, 'temperature': 2.5, 'mask_type': 'soft', 'backward_type': 'normal'}} +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +Apply masks for the following modules: ['llm', 'connector'] +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +TinyLlavaConfig { + "backward_type_connector": "normal", + "cache_dir": null, + "connector_type": "mlp2x_gelu", + "hidden_size": 896, + "ignore_index": -100, + "image_aspect_ratio": "square", + "image_token_index": -200, + "llm_model_name_or_path": "Qwen/Qwen2.5-0.5B", + "mask_model": [ + "llm", + "connector" + ], + "mask_type_connector": "soft", + "model_type": "tinyllava", + "num_queries": 128, + "num_resampler_layers": 3, + "pad_token": null, + "resampler_hidden_size": 768, + "sparsity_connector": null, + "subnet_type_connector": "global", + "temperature_connector": 2.5, + "text_config": { + "_name_or_path": "Qwen/Qwen2.5-0.5B", + "architectures": [ + "Qwen2ForCausalLM" + ], + "backward_type": "normal", + "bos_token_id": 151643, + "eos_token_id": 151643, + "hidden_size": 896, + "intermediate_size": 4864, + "mask_type": "soft", + "masked_layers": "all", + "max_position_embeddings": 32768, + "max_window_layers": 24, + "model_type": "qwen2", + "num_attention_heads": 14, + "num_hidden_layers": 24, + "num_key_value_heads": 2, + "rope_theta": 1000000.0, + "sliding_window": 32768, + "subnet_mode": "both", + "subnet_type": "None", + "temperature_attn": 2.5, + "temperature_mlp": 2.5, + "tie_word_embeddings": true, + "torch_dtype": "bfloat16", + "use_mrope": false, + "use_sliding_window": false, + "vocab_size": 151936 + }, + "threshold_connector": null, + "tokenizer_model_max_length": 2048, + "tokenizer_name_or_path": "Qwen/Qwen2.5-0.5B", + "tokenizer_padding_side": "right", + "tokenizer_use_fast": false, + "transformers_version": "4.40.1", + "tune_type_connector": "frozen", + "tune_type_llm": "frozen", + "tune_type_vision_tower": "frozen", + "tune_vision_tower_from_layer": -1, + "use_cache": false, + "vision_config": { + "hidden_act": "gelu_pytorch_tanh", + "hidden_size": 1152, + "image_size": 384, + "intermediate_size": 4304, + "layer_norm_eps": 1e-06, + "model_name_or_path": "google/siglip-so400m-patch14-384", + "model_name_or_path2": "", + "model_type": "siglip_vision_model", + "num_attention_heads": 16, + "num_hidden_layers": 27, + "patch_size": 14 + }, + "vision_feature_layer": -2, + "vision_feature_select_strategy": "patch", + "vision_hidden_size": 1152, + "vision_model_name_or_path": "google/siglip-so400m-patch14-384", + "vision_model_name_or_path2": "", + "vocab_size": 151936 +} + +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test1-worker-0:221839:221839 [0] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:221839:221839 [0] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:221839:221839 [0] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:221839:221839 [0] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:221839:221839 [0] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test1-worker-0:221839:221839 [0] NCCL INFO cudaDriverVersion 12040 +NCCL version 2.21.5+cuda12.1 +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +ywang29-vrdb-test1-worker-0:221846:221846 [7] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test1-worker-0:221846:221846 [7] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:221846:221846 [7] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:221841:221841 [2] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test1-worker-0:221841:221841 [2] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:221841:221841 [2] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:221846:221846 [7] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:221846:221846 [7] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:221846:221846 [7] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test1-worker-0:221841:221841 [2] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:221841:221841 [2] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:221841:221841 [2] NCCL INFO NET/Plugin: Using internal network plugin. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +ywang29-vrdb-test1-worker-0:221839:223458 [0] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:221839:223458 [0] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:221839:223458 [0] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:221839:223458 [0] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:221839:223458 [0] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:221839:223458 [0] NCCL INFO Using network Socket +ywang29-vrdb-test1-worker-0:221841:223460 [2] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:221841:223460 [2] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:221841:223460 [2] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:221841:223460 [2] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:221841:223460 [2] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:221841:223460 [2] NCCL INFO Using network Socket +ywang29-vrdb-test1-worker-0:221846:223459 [7] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:221846:223459 [7] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:221846:223459 [7] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:221846:223459 [7] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:221846:223459 [7] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:221846:223459 [7] NCCL INFO Using network Socket +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test1-worker-0:221843:221843 [4] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test1-worker-0:221843:221843 [4] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:221843:221843 [4] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:221843:221843 [4] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:221843:221843 [4] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:221843:221843 [4] NCCL INFO NET/Plugin: Using internal network plugin. +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +ywang29-vrdb-test1-worker-0:221843:223461 [4] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:221843:223461 [4] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:221843:223461 [4] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:221843:223461 [4] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:221843:223461 [4] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:221843:223461 [4] NCCL INFO Using network Socket +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test1-worker-0:221844:221844 [5] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test1-worker-0:221844:221844 [5] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:221844:221844 [5] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:221842:221842 [3] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test1-worker-0:221842:221842 [3] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:221842:221842 [3] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:221844:221844 [5] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:221844:221844 [5] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:221844:221844 [5] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test1-worker-0:221842:221842 [3] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:221842:221842 [3] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:221842:221842 [3] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test1-worker-0:221842:223463 [3] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:221842:223463 [3] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:221842:223463 [3] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:221842:223463 [3] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:221842:223463 [3] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:221842:223463 [3] NCCL INFO Using network Socket +ywang29-vrdb-test1-worker-0:221844:223462 [5] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:221844:223462 [5] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:221844:223462 [5] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:221844:223462 [5] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:221844:223462 [5] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:221844:223462 [5] NCCL INFO Using network Socket +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test1-worker-0:221840:221840 [1] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test1-worker-0:221840:221840 [1] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:221840:221840 [1] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:221840:221840 [1] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:221840:221840 [1] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:221840:221840 [1] NCCL INFO NET/Plugin: Using internal network plugin. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test1-worker-0:221845:221845 [6] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test1-worker-0:221845:221845 [6] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:221845:221845 [6] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:221845:221845 [6] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:221845:221845 [6] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:221845:221845 [6] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test1-worker-0:221840:223464 [1] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:221840:223464 [1] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:221840:223464 [1] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:221840:223464 [1] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:221840:223464 [1] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:221840:223464 [1] NCCL INFO Using network Socket +ywang29-vrdb-test1-worker-0:221845:223465 [6] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:221845:223465 [6] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:221845:223465 [6] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:221845:223465 [6] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:221845:223465 [6] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:221845:223465 [6] NCCL INFO Using network Socket +ywang29-vrdb-test1-worker-0:221845:223465 [6] NCCL INFO ncclCommInitRank comm 0x562b6bb81240 rank 6 nranks 8 cudaDev 6 nvmlDev 6 busId a01c0 commId 0xd63d2a52f14618ac - Init START +ywang29-vrdb-test1-worker-0:221842:223463 [3] NCCL INFO ncclCommInitRank comm 0x559fa9cacd80 rank 3 nranks 8 cudaDev 3 nvmlDev 3 busId 201d0 commId 0xd63d2a52f14618ac - Init START +ywang29-vrdb-test1-worker-0:221843:223461 [4] NCCL INFO ncclCommInitRank comm 0x55f355a97140 rank 4 nranks 8 cudaDev 4 nvmlDev 4 busId 901c0 commId 0xd63d2a52f14618ac - Init START +ywang29-vrdb-test1-worker-0:221840:223464 [1] NCCL INFO ncclCommInitRank comm 0x55e2fbfac080 rank 1 nranks 8 cudaDev 1 nvmlDev 1 busId 101d0 commId 0xd63d2a52f14618ac - Init START +ywang29-vrdb-test1-worker-0:221841:223460 [2] NCCL INFO ncclCommInitRank comm 0x55f2449d8200 rank 2 nranks 8 cudaDev 2 nvmlDev 2 busId 201c0 commId 0xd63d2a52f14618ac - Init START +ywang29-vrdb-test1-worker-0:221846:223459 [7] NCCL INFO ncclCommInitRank comm 0x5581de1ed740 rank 7 nranks 8 cudaDev 7 nvmlDev 7 busId a01d0 commId 0xd63d2a52f14618ac - Init START +ywang29-vrdb-test1-worker-0:221839:223458 [0] NCCL INFO ncclCommInitRank comm 0x55589fe9d150 rank 0 nranks 8 cudaDev 0 nvmlDev 0 busId 101c0 commId 0xd63d2a52f14618ac - Init START +ywang29-vrdb-test1-worker-0:221844:223462 [5] NCCL INFO ncclCommInitRank comm 0x56140b4605f0 rank 5 nranks 8 cudaDev 5 nvmlDev 5 busId 901d0 commId 0xd63d2a52f14618ac - Init START +ywang29-vrdb-test1-worker-0:221839:223458 [0] NCCL INFO Setting affinity for GPU 0 to ff,ffff0000,00ffffff +ywang29-vrdb-test1-worker-0:221839:223458 [0] NCCL INFO NVLS multicast support is not available on dev 0 +ywang29-vrdb-test1-worker-0:221843:223461 [4] NCCL INFO Setting affinity for GPU 4 to ffffff00,0000ffff,ff000000 +ywang29-vrdb-test1-worker-0:221843:223461 [4] NCCL INFO NVLS multicast support is not available on dev 4 +ywang29-vrdb-test1-worker-0:221846:223459 [7] NCCL INFO Setting affinity for GPU 7 to ffffff00,0000ffff,ff000000 +ywang29-vrdb-test1-worker-0:221846:223459 [7] NCCL INFO NVLS multicast support is not available on dev 7 +ywang29-vrdb-test1-worker-0:221840:223464 [1] NCCL INFO Setting affinity for GPU 1 to ff,ffff0000,00ffffff +ywang29-vrdb-test1-worker-0:221840:223464 [1] NCCL INFO NVLS multicast support is not available on dev 1 +ywang29-vrdb-test1-worker-0:221845:223465 [6] NCCL INFO Setting affinity for GPU 6 to ffffff00,0000ffff,ff000000 +ywang29-vrdb-test1-worker-0:221844:223462 [5] NCCL INFO Setting affinity for GPU 5 to ffffff00,0000ffff,ff000000 +ywang29-vrdb-test1-worker-0:221841:223460 [2] NCCL INFO Setting affinity for GPU 2 to ff,ffff0000,00ffffff +ywang29-vrdb-test1-worker-0:221841:223460 [2] NCCL INFO NVLS multicast support is not available on dev 2 +ywang29-vrdb-test1-worker-0:221844:223462 [5] NCCL INFO NVLS multicast support is not available on dev 5 +ywang29-vrdb-test1-worker-0:221845:223465 [6] NCCL INFO NVLS multicast support is not available on dev 6 +ywang29-vrdb-test1-worker-0:221842:223463 [3] NCCL INFO Setting affinity for GPU 3 to ff,ffff0000,00ffffff +ywang29-vrdb-test1-worker-0:221842:223463 [3] NCCL INFO NVLS multicast support is not available on dev 3 +ywang29-vrdb-test1-worker-0:221841:223460 [2] NCCL INFO comm 0x55f2449d8200 rank 2 nRanks 8 nNodes 1 localRanks 8 localRank 2 MNNVL 0 +ywang29-vrdb-test1-worker-0:221839:223458 [0] NCCL INFO comm 0x55589fe9d150 rank 0 nRanks 8 nNodes 1 localRanks 8 localRank 0 MNNVL 0 +ywang29-vrdb-test1-worker-0:221840:223464 [1] NCCL INFO comm 0x55e2fbfac080 rank 1 nRanks 8 nNodes 1 localRanks 8 localRank 1 MNNVL 0 +ywang29-vrdb-test1-worker-0:221846:223459 [7] NCCL INFO comm 0x5581de1ed740 rank 7 nRanks 8 nNodes 1 localRanks 8 localRank 7 MNNVL 0 +ywang29-vrdb-test1-worker-0:221845:223465 [6] NCCL INFO comm 0x562b6bb81240 rank 6 nRanks 8 nNodes 1 localRanks 8 localRank 6 MNNVL 0 +ywang29-vrdb-test1-worker-0:221839:223458 [0] NCCL INFO Channel 00/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:221839:223458 [0] NCCL INFO Channel 01/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:221844:223462 [5] NCCL INFO comm 0x56140b4605f0 rank 5 nRanks 8 nNodes 1 localRanks 8 localRank 5 MNNVL 0 +ywang29-vrdb-test1-worker-0:221842:223463 [3] NCCL INFO comm 0x559fa9cacd80 rank 3 nRanks 8 nNodes 1 localRanks 8 localRank 3 MNNVL 0 +ywang29-vrdb-test1-worker-0:221839:223458 [0] NCCL INFO Channel 02/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:221839:223458 [0] NCCL INFO Channel 03/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:221843:223461 [4] NCCL INFO comm 0x55f355a97140 rank 4 nRanks 8 nNodes 1 localRanks 8 localRank 4 MNNVL 0 +ywang29-vrdb-test1-worker-0:221839:223458 [0] NCCL INFO Channel 04/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:221839:223458 [0] NCCL INFO Channel 05/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:221841:223460 [2] NCCL INFO Trees [0] 3/-1/-1->2->1 [1] 3/-1/-1->2->1 [2] 3/-1/-1->2->1 [3] 3/-1/-1->2->1 [4] 3/-1/-1->2->1 [5] 3/-1/-1->2->1 [6] 3/-1/-1->2->1 [7] 3/-1/-1->2->1 [8] 3/-1/-1->2->1 [9] 3/-1/-1->2->1 [10] 3/-1/-1->2->1 [11] 3/-1/-1->2->1 [12] 3/-1/-1->2->1 [13] 3/-1/-1->2->1 [14] 3/-1/-1->2->1 [15] 3/-1/-1->2->1 [16] 3/-1/-1->2->1 [17] 3/-1/-1->2->1 [18] 3/-1/-1->2->1 [19] 3/-1/-1->2->1 [20] 3/-1/-1->2->1 [21] 3/-1/-1->2->1 [22] 3/-1/-1->2->1 [23] 3/-1/-1->2->1 +ywang29-vrdb-test1-worker-0:221839:223458 [0] NCCL INFO Channel 06/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:221839:223458 [0] NCCL INFO Channel 07/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:221841:223460 [2] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:221839:223458 [0] NCCL INFO Channel 08/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:221839:223458 [0] NCCL INFO Channel 09/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:221846:223459 [7] NCCL INFO Trees [0] -1/-1/-1->7->6 [1] -1/-1/-1->7->6 [2] -1/-1/-1->7->6 [3] -1/-1/-1->7->6 [4] -1/-1/-1->7->6 [5] -1/-1/-1->7->6 [6] -1/-1/-1->7->6 [7] -1/-1/-1->7->6 [8] -1/-1/-1->7->6 [9] -1/-1/-1->7->6 [10] -1/-1/-1->7->6 [11] -1/-1/-1->7->6 [12] -1/-1/-1->7->6 [13] -1/-1/-1->7->6 [14] -1/-1/-1->7->6 [15] -1/-1/-1->7->6 [16] -1/-1/-1->7->6 [17] -1/-1/-1->7->6 [18] -1/-1/-1->7->6 [19] -1/-1/-1->7->6 [20] -1/-1/-1->7->6 [21] -1/-1/-1->7->6 [22] -1/-1/-1->7->6 [23] -1/-1/-1->7->6 +ywang29-vrdb-test1-worker-0:221839:223458 [0] NCCL INFO Channel 10/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:221839:223458 [0] NCCL INFO Channel 11/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:221839:223458 [0] NCCL INFO Channel 12/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:221846:223459 [7] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:221839:223458 [0] NCCL INFO Channel 13/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:221839:223458 [0] NCCL INFO Channel 14/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:221839:223458 [0] NCCL INFO Channel 15/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:221839:223458 [0] NCCL INFO Channel 16/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:221840:223464 [1] NCCL INFO Trees [0] 2/-1/-1->1->0 [1] 2/-1/-1->1->0 [2] 2/-1/-1->1->0 [3] 2/-1/-1->1->0 [4] 2/-1/-1->1->0 [5] 2/-1/-1->1->0 [6] 2/-1/-1->1->0 [7] 2/-1/-1->1->0 [8] 2/-1/-1->1->0 [9] 2/-1/-1->1->0 [10] 2/-1/-1->1->0 [11] 2/-1/-1->1->0 [12] 2/-1/-1->1->0 [13] 2/-1/-1->1->0 [14] 2/-1/-1->1->0 [15] 2/-1/-1->1->0 [16] 2/-1/-1->1->0 [17] 2/-1/-1->1->0 [18] 2/-1/-1->1->0 [19] 2/-1/-1->1->0 [20] 2/-1/-1->1->0 [21] 2/-1/-1->1->0 [22] 2/-1/-1->1->0 [23] 2/-1/-1->1->0 +ywang29-vrdb-test1-worker-0:221839:223458 [0] NCCL INFO Channel 17/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:221839:223458 [0] NCCL INFO Channel 18/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:221845:223465 [6] NCCL INFO Trees [0] 7/-1/-1->6->5 [1] 7/-1/-1->6->5 [2] 7/-1/-1->6->5 [3] 7/-1/-1->6->5 [4] 7/-1/-1->6->5 [5] 7/-1/-1->6->5 [6] 7/-1/-1->6->5 [7] 7/-1/-1->6->5 [8] 7/-1/-1->6->5 [9] 7/-1/-1->6->5 [10] 7/-1/-1->6->5 [11] 7/-1/-1->6->5 [12] 7/-1/-1->6->5 [13] 7/-1/-1->6->5 [14] 7/-1/-1->6->5 [15] 7/-1/-1->6->5 [16] 7/-1/-1->6->5 [17] 7/-1/-1->6->5 [18] 7/-1/-1->6->5 [19] 7/-1/-1->6->5 [20] 7/-1/-1->6->5 [21] 7/-1/-1->6->5 [22] 7/-1/-1->6->5 [23] 7/-1/-1->6->5 +ywang29-vrdb-test1-worker-0:221839:223458 [0] NCCL INFO Channel 19/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:221840:223464 [1] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:221839:223458 [0] NCCL INFO Channel 20/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:221839:223458 [0] NCCL INFO Channel 21/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:221843:223461 [4] NCCL INFO Trees [0] 5/-1/-1->4->3 [1] 5/-1/-1->4->3 [2] 5/-1/-1->4->3 [3] 5/-1/-1->4->3 [4] 5/-1/-1->4->3 [5] 5/-1/-1->4->3 [6] 5/-1/-1->4->3 [7] 5/-1/-1->4->3 [8] 5/-1/-1->4->3 [9] 5/-1/-1->4->3 [10] 5/-1/-1->4->3 [11] 5/-1/-1->4->3 [12] 5/-1/-1->4->3 [13] 5/-1/-1->4->3 [14] 5/-1/-1->4->3 [15] 5/-1/-1->4->3 [16] 5/-1/-1->4->3 [17] 5/-1/-1->4->3 [18] 5/-1/-1->4->3 [19] 5/-1/-1->4->3 [20] 5/-1/-1->4->3 [21] 5/-1/-1->4->3 [22] 5/-1/-1->4->3 [23] 5/-1/-1->4->3 +ywang29-vrdb-test1-worker-0:221844:223462 [5] NCCL INFO Trees [0] 6/-1/-1->5->4 [1] 6/-1/-1->5->4 [2] 6/-1/-1->5->4 [3] 6/-1/-1->5->4 [4] 6/-1/-1->5->4 [5] 6/-1/-1->5->4 [6] 6/-1/-1->5->4 [7] 6/-1/-1->5->4 [8] 6/-1/-1->5->4 [9] 6/-1/-1->5->4 [10] 6/-1/-1->5->4 [11] 6/-1/-1->5->4 [12] 6/-1/-1->5->4 [13] 6/-1/-1->5->4 [14] 6/-1/-1->5->4 [15] 6/-1/-1->5->4 [16] 6/-1/-1->5->4 [17] 6/-1/-1->5->4 [18] 6/-1/-1->5->4 [19] 6/-1/-1->5->4 [20] 6/-1/-1->5->4 [21] 6/-1/-1->5->4 [22] 6/-1/-1->5->4 [23] 6/-1/-1->5->4 +ywang29-vrdb-test1-worker-0:221842:223463 [3] NCCL INFO Trees [0] 4/-1/-1->3->2 [1] 4/-1/-1->3->2 [2] 4/-1/-1->3->2 [3] 4/-1/-1->3->2 [4] 4/-1/-1->3->2 [5] 4/-1/-1->3->2 [6] 4/-1/-1->3->2 [7] 4/-1/-1->3->2 [8] 4/-1/-1->3->2 [9] 4/-1/-1->3->2 [10] 4/-1/-1->3->2 [11] 4/-1/-1->3->2 [12] 4/-1/-1->3->2 [13] 4/-1/-1->3->2 [14] 4/-1/-1->3->2 [15] 4/-1/-1->3->2 [16] 4/-1/-1->3->2 [17] 4/-1/-1->3->2 [18] 4/-1/-1->3->2 [19] 4/-1/-1->3->2 [20] 4/-1/-1->3->2 [21] 4/-1/-1->3->2 [22] 4/-1/-1->3->2 [23] 4/-1/-1->3->2 +ywang29-vrdb-test1-worker-0:221845:223465 [6] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:221839:223458 [0] NCCL INFO Channel 22/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:221839:223458 [0] NCCL INFO Channel 23/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:221843:223461 [4] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:221844:223462 [5] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:221842:223463 [3] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:221839:223458 [0] NCCL INFO Trees [0] 1/-1/-1->0->-1 [1] 1/-1/-1->0->-1 [2] 1/-1/-1->0->-1 [3] 1/-1/-1->0->-1 [4] 1/-1/-1->0->-1 [5] 1/-1/-1->0->-1 [6] 1/-1/-1->0->-1 [7] 1/-1/-1->0->-1 [8] 1/-1/-1->0->-1 [9] 1/-1/-1->0->-1 [10] 1/-1/-1->0->-1 [11] 1/-1/-1->0->-1 [12] 1/-1/-1->0->-1 [13] 1/-1/-1->0->-1 [14] 1/-1/-1->0->-1 [15] 1/-1/-1->0->-1 [16] 1/-1/-1->0->-1 [17] 1/-1/-1->0->-1 [18] 1/-1/-1->0->-1 [19] 1/-1/-1->0->-1 [20] 1/-1/-1->0->-1 [21] 1/-1/-1->0->-1 [22] 1/-1/-1->0->-1 [23] 1/-1/-1->0->-1 +ywang29-vrdb-test1-worker-0:221839:223458 [0] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:221841:223460 [2] NCCL INFO Channel 00/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:223461 [4] NCCL INFO Channel 00/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:223460 [2] NCCL INFO Channel 01/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:223461 [4] NCCL INFO Channel 01/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:223460 [2] NCCL INFO Channel 02/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:223461 [4] NCCL INFO Channel 02/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:223462 [5] NCCL INFO Channel 00/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:223465 [6] NCCL INFO Channel 00/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:223460 [2] NCCL INFO Channel 03/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:223461 [4] NCCL INFO Channel 03/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:223462 [5] NCCL INFO Channel 01/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:223465 [6] NCCL INFO Channel 01/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:223460 [2] NCCL INFO Channel 04/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:223461 [4] NCCL INFO Channel 04/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:223462 [5] NCCL INFO Channel 02/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:223465 [6] NCCL INFO Channel 02/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:223460 [2] NCCL INFO Channel 05/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:223461 [4] NCCL INFO Channel 05/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:223462 [5] NCCL INFO Channel 03/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:223465 [6] NCCL INFO Channel 03/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:223460 [2] NCCL INFO Channel 06/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:223461 [4] NCCL INFO Channel 06/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:223462 [5] NCCL INFO Channel 04/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:223465 [6] NCCL INFO Channel 04/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:223460 [2] NCCL INFO Channel 07/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:223461 [4] NCCL INFO Channel 07/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:223462 [5] NCCL INFO Channel 05/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:223465 [6] NCCL INFO Channel 05/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:223460 [2] NCCL INFO Channel 08/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:223461 [4] NCCL INFO Channel 08/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:223462 [5] NCCL INFO Channel 06/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:223465 [6] NCCL INFO Channel 06/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:223460 [2] NCCL INFO Channel 09/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:223461 [4] NCCL INFO Channel 09/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:223462 [5] NCCL INFO Channel 07/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:223465 [6] NCCL INFO Channel 07/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:223461 [4] NCCL INFO Channel 10/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:223460 [2] NCCL INFO Channel 10/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:223462 [5] NCCL INFO Channel 08/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:223465 [6] NCCL INFO Channel 08/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:223461 [4] NCCL INFO Channel 11/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:223460 [2] NCCL INFO Channel 11/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:223462 [5] NCCL INFO Channel 09/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:223465 [6] NCCL INFO Channel 09/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:223461 [4] NCCL INFO Channel 12/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:223462 [5] NCCL INFO Channel 10/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:223460 [2] NCCL INFO Channel 12/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:223465 [6] NCCL INFO Channel 10/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:223461 [4] NCCL INFO Channel 13/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:223462 [5] NCCL INFO Channel 11/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:223460 [2] NCCL INFO Channel 13/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:223465 [6] NCCL INFO Channel 11/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:223461 [4] NCCL INFO Channel 14/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:223462 [5] NCCL INFO Channel 12/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:223460 [2] NCCL INFO Channel 14/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:223465 [6] NCCL INFO Channel 12/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:223461 [4] NCCL INFO Channel 15/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:223462 [5] NCCL INFO Channel 13/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:223460 [2] NCCL INFO Channel 15/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:223465 [6] NCCL INFO Channel 13/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:223461 [4] NCCL INFO Channel 16/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:223462 [5] NCCL INFO Channel 14/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:223459 [7] NCCL INFO Channel 00/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:223460 [2] NCCL INFO Channel 16/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:223465 [6] NCCL INFO Channel 14/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:223461 [4] NCCL INFO Channel 17/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:223462 [5] NCCL INFO Channel 15/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:223459 [7] NCCL INFO Channel 01/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:223460 [2] NCCL INFO Channel 17/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:223465 [6] NCCL INFO Channel 15/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:223461 [4] NCCL INFO Channel 18/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:223462 [5] NCCL INFO Channel 16/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:223459 [7] NCCL INFO Channel 02/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221839:223458 [0] NCCL INFO Channel 00/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:223460 [2] NCCL INFO Channel 18/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:223465 [6] NCCL INFO Channel 16/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:223461 [4] NCCL INFO Channel 19/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:223462 [5] NCCL INFO Channel 17/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:223459 [7] NCCL INFO Channel 03/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221839:223458 [0] NCCL INFO Channel 01/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:223460 [2] NCCL INFO Channel 19/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:223465 [6] NCCL INFO Channel 17/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:223461 [4] NCCL INFO Channel 20/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:223462 [5] NCCL INFO Channel 18/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:223459 [7] NCCL INFO Channel 04/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221839:223458 [0] NCCL INFO Channel 02/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:223460 [2] NCCL INFO Channel 20/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:223465 [6] NCCL INFO Channel 18/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:223461 [4] NCCL INFO Channel 21/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:223463 [3] NCCL INFO Channel 00/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:223462 [5] NCCL INFO Channel 19/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:223459 [7] NCCL INFO Channel 05/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221839:223458 [0] NCCL INFO Channel 03/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:223460 [2] NCCL INFO Channel 21/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:223465 [6] NCCL INFO Channel 19/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:223461 [4] NCCL INFO Channel 22/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:223464 [1] NCCL INFO Channel 00/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:223463 [3] NCCL INFO Channel 01/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:223462 [5] NCCL INFO Channel 20/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:223459 [7] NCCL INFO Channel 06/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221839:223458 [0] NCCL INFO Channel 04/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:223460 [2] NCCL INFO Channel 22/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:223465 [6] NCCL INFO Channel 20/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:223461 [4] NCCL INFO Channel 23/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:223464 [1] NCCL INFO Channel 01/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:223463 [3] NCCL INFO Channel 02/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:223462 [5] NCCL INFO Channel 21/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:223459 [7] NCCL INFO Channel 07/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221839:223458 [0] NCCL INFO Channel 05/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:223460 [2] NCCL INFO Channel 23/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:223465 [6] NCCL INFO Channel 21/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:223464 [1] NCCL INFO Channel 02/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:223463 [3] NCCL INFO Channel 03/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:223462 [5] NCCL INFO Channel 22/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:223459 [7] NCCL INFO Channel 08/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221839:223458 [0] NCCL INFO Channel 06/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:223465 [6] NCCL INFO Channel 22/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:223464 [1] NCCL INFO Channel 03/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:223463 [3] NCCL INFO Channel 04/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:223462 [5] NCCL INFO Channel 23/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:223459 [7] NCCL INFO Channel 09/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221839:223458 [0] NCCL INFO Channel 07/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:223465 [6] NCCL INFO Channel 23/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:223464 [1] NCCL INFO Channel 04/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:223463 [3] NCCL INFO Channel 05/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:223459 [7] NCCL INFO Channel 10/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221839:223458 [0] NCCL INFO Channel 08/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:223464 [1] NCCL INFO Channel 05/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:223459 [7] NCCL INFO Channel 11/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221839:223458 [0] NCCL INFO Channel 09/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:223463 [3] NCCL INFO Channel 06/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:223464 [1] NCCL INFO Channel 06/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:223459 [7] NCCL INFO Channel 12/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221839:223458 [0] NCCL INFO Channel 10/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:223463 [3] NCCL INFO Channel 07/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:223464 [1] NCCL INFO Channel 07/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:223459 [7] NCCL INFO Channel 13/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221839:223458 [0] NCCL INFO Channel 11/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:223464 [1] NCCL INFO Channel 08/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:223459 [7] NCCL INFO Channel 14/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221839:223458 [0] NCCL INFO Channel 12/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:223463 [3] NCCL INFO Channel 08/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:223464 [1] NCCL INFO Channel 09/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:223459 [7] NCCL INFO Channel 15/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:223459 [7] NCCL INFO Channel 16/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:223464 [1] NCCL INFO Channel 10/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221839:223458 [0] NCCL INFO Channel 13/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:223463 [3] NCCL INFO Channel 09/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:223459 [7] NCCL INFO Channel 17/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:223464 [1] NCCL INFO Channel 11/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221839:223458 [0] NCCL INFO Channel 14/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:223463 [3] NCCL INFO Channel 10/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:223459 [7] NCCL INFO Channel 18/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:223464 [1] NCCL INFO Channel 12/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221839:223458 [0] NCCL INFO Channel 15/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:223459 [7] NCCL INFO Channel 19/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:223463 [3] NCCL INFO Channel 11/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221839:223458 [0] NCCL INFO Channel 16/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:223459 [7] NCCL INFO Channel 20/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221839:223458 [0] NCCL INFO Channel 17/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:223464 [1] NCCL INFO Channel 13/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:223463 [3] NCCL INFO Channel 12/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:223459 [7] NCCL INFO Channel 21/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221839:223458 [0] NCCL INFO Channel 18/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:223464 [1] NCCL INFO Channel 14/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:223463 [3] NCCL INFO Channel 13/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:223459 [7] NCCL INFO Channel 22/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221839:223458 [0] NCCL INFO Channel 19/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:223464 [1] NCCL INFO Channel 15/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:223463 [3] NCCL INFO Channel 14/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:223459 [7] NCCL INFO Channel 23/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221839:223458 [0] NCCL INFO Channel 20/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:223464 [1] NCCL INFO Channel 16/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:223463 [3] NCCL INFO Channel 15/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:223464 [1] NCCL INFO Channel 17/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221839:223458 [0] NCCL INFO Channel 21/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:223463 [3] NCCL INFO Channel 16/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:223464 [1] NCCL INFO Channel 18/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221839:223458 [0] NCCL INFO Channel 22/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:223463 [3] NCCL INFO Channel 17/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:223464 [1] NCCL INFO Channel 19/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:223463 [3] NCCL INFO Channel 18/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:223464 [1] NCCL INFO Channel 20/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:223463 [3] NCCL INFO Channel 19/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221839:223458 [0] NCCL INFO Channel 23/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:223464 [1] NCCL INFO Channel 21/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:223463 [3] NCCL INFO Channel 20/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:223464 [1] NCCL INFO Channel 22/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:223463 [3] NCCL INFO Channel 21/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:223464 [1] NCCL INFO Channel 23/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:223463 [3] NCCL INFO Channel 22/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:223463 [3] NCCL INFO Channel 23/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:223460 [2] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:221845:223465 [6] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:221840:223464 [1] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:221846:223459 [7] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:221839:223458 [0] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:221846:223459 [7] NCCL INFO Channel 00/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:223462 [5] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:221842:223463 [3] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:221843:223461 [4] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:221841:223460 [2] NCCL INFO Channel 00/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:223460 [2] NCCL INFO Channel 01/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:223460 [2] NCCL INFO Channel 02/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:223459 [7] NCCL INFO Channel 01/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:223460 [2] NCCL INFO Channel 03/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:223459 [7] NCCL INFO Channel 02/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:223460 [2] NCCL INFO Channel 04/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:223459 [7] NCCL INFO Channel 03/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:223460 [2] NCCL INFO Channel 05/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:223459 [7] NCCL INFO Channel 04/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:223460 [2] NCCL INFO Channel 06/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:223459 [7] NCCL INFO Channel 05/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:223460 [2] NCCL INFO Channel 07/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:223459 [7] NCCL INFO Channel 06/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:223460 [2] NCCL INFO Channel 08/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:223459 [7] NCCL INFO Channel 07/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:223460 [2] NCCL INFO Channel 09/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:223460 [2] NCCL INFO Channel 10/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:223459 [7] NCCL INFO Channel 08/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:223465 [6] NCCL INFO Channel 00/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:223460 [2] NCCL INFO Channel 11/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:223459 [7] NCCL INFO Channel 09/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:223465 [6] NCCL INFO Channel 01/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:223460 [2] NCCL INFO Channel 12/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:223459 [7] NCCL INFO Channel 10/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:223465 [6] NCCL INFO Channel 02/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:223460 [2] NCCL INFO Channel 13/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:223459 [7] NCCL INFO Channel 11/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:223465 [6] NCCL INFO Channel 03/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:223460 [2] NCCL INFO Channel 14/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:223459 [7] NCCL INFO Channel 12/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:223465 [6] NCCL INFO Channel 04/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:223464 [1] NCCL INFO Channel 00/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:223460 [2] NCCL INFO Channel 15/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:223465 [6] NCCL INFO Channel 05/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:223459 [7] NCCL INFO Channel 13/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:223464 [1] NCCL INFO Channel 01/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:223460 [2] NCCL INFO Channel 16/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:223465 [6] NCCL INFO Channel 06/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:223464 [1] NCCL INFO Channel 02/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:223459 [7] NCCL INFO Channel 14/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:223461 [4] NCCL INFO Channel 00/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:223460 [2] NCCL INFO Channel 17/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:223465 [6] NCCL INFO Channel 07/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:223464 [1] NCCL INFO Channel 03/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:223459 [7] NCCL INFO Channel 15/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:223461 [4] NCCL INFO Channel 01/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:223460 [2] NCCL INFO Channel 18/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:223465 [6] NCCL INFO Channel 08/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:223464 [1] NCCL INFO Channel 04/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:223459 [7] NCCL INFO Channel 16/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:223460 [2] NCCL INFO Channel 19/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:223461 [4] NCCL INFO Channel 02/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:223465 [6] NCCL INFO Channel 09/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:223464 [1] NCCL INFO Channel 05/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:223459 [7] NCCL INFO Channel 17/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:223460 [2] NCCL INFO Channel 20/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:223461 [4] NCCL INFO Channel 03/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:223465 [6] NCCL INFO Channel 10/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:223464 [1] NCCL INFO Channel 06/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:223460 [2] NCCL INFO Channel 21/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:223459 [7] NCCL INFO Channel 18/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:223461 [4] NCCL INFO Channel 04/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:223465 [6] NCCL INFO Channel 11/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:223464 [1] NCCL INFO Channel 07/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:223460 [2] NCCL INFO Channel 22/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:223459 [7] NCCL INFO Channel 19/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:223461 [4] NCCL INFO Channel 05/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:223465 [6] NCCL INFO Channel 12/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:223464 [1] NCCL INFO Channel 08/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:223460 [2] NCCL INFO Channel 23/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:223459 [7] NCCL INFO Channel 20/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:223461 [4] NCCL INFO Channel 06/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:223465 [6] NCCL INFO Channel 13/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:223464 [1] NCCL INFO Channel 09/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:223459 [7] NCCL INFO Channel 21/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:223465 [6] NCCL INFO Channel 14/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:223461 [4] NCCL INFO Channel 07/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:223464 [1] NCCL INFO Channel 10/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:223465 [6] NCCL INFO Channel 15/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:223459 [7] NCCL INFO Channel 22/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:223464 [1] NCCL INFO Channel 11/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:223461 [4] NCCL INFO Channel 08/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:223463 [3] NCCL INFO Channel 00/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:223465 [6] NCCL INFO Channel 16/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:223464 [1] NCCL INFO Channel 12/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:223459 [7] NCCL INFO Channel 23/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:223461 [4] NCCL INFO Channel 09/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:223463 [3] NCCL INFO Channel 01/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:223465 [6] NCCL INFO Channel 17/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:223464 [1] NCCL INFO Channel 13/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:223463 [3] NCCL INFO Channel 02/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:223461 [4] NCCL INFO Channel 10/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:223462 [5] NCCL INFO Channel 00/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:223465 [6] NCCL INFO Channel 18/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:223464 [1] NCCL INFO Channel 14/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:223463 [3] NCCL INFO Channel 03/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:223461 [4] NCCL INFO Channel 11/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:223462 [5] NCCL INFO Channel 01/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:223465 [6] NCCL INFO Channel 19/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:223464 [1] NCCL INFO Channel 15/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:223463 [3] NCCL INFO Channel 04/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:223461 [4] NCCL INFO Channel 12/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:223462 [5] NCCL INFO Channel 02/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:223465 [6] NCCL INFO Channel 20/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:223464 [1] NCCL INFO Channel 16/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:223463 [3] NCCL INFO Channel 05/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:223461 [4] NCCL INFO Channel 13/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:223462 [5] NCCL INFO Channel 03/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:223465 [6] NCCL INFO Channel 21/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:223464 [1] NCCL INFO Channel 17/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:223463 [3] NCCL INFO Channel 06/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:223461 [4] NCCL INFO Channel 14/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:223462 [5] NCCL INFO Channel 04/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:223465 [6] NCCL INFO Channel 22/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:223464 [1] NCCL INFO Channel 18/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:223463 [3] NCCL INFO Channel 07/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:223461 [4] NCCL INFO Channel 15/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:223462 [5] NCCL INFO Channel 05/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:223465 [6] NCCL INFO Channel 23/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:223464 [1] NCCL INFO Channel 19/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:223463 [3] NCCL INFO Channel 08/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:223461 [4] NCCL INFO Channel 16/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:223462 [5] NCCL INFO Channel 06/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:223464 [1] NCCL INFO Channel 20/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:223461 [4] NCCL INFO Channel 17/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:223461 [4] NCCL INFO Channel 18/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:223461 [4] NCCL INFO Channel 19/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:223461 [4] NCCL INFO Channel 20/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:223461 [4] NCCL INFO Channel 21/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:223461 [4] NCCL INFO Channel 22/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:223461 [4] NCCL INFO Channel 23/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:223462 [5] NCCL INFO Channel 07/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:223463 [3] NCCL INFO Channel 09/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:223462 [5] NCCL INFO Channel 08/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:223463 [3] NCCL INFO Channel 10/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:223462 [5] NCCL INFO Channel 09/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:223463 [3] NCCL INFO Channel 11/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:223462 [5] NCCL INFO Channel 10/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:223463 [3] NCCL INFO Channel 12/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:223462 [5] NCCL INFO Channel 11/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:223463 [3] NCCL INFO Channel 13/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:223462 [5] NCCL INFO Channel 12/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:223463 [3] NCCL INFO Channel 14/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:223462 [5] NCCL INFO Channel 13/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:223464 [1] NCCL INFO Channel 21/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:223463 [3] NCCL INFO Channel 15/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:223462 [5] NCCL INFO Channel 14/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:223464 [1] NCCL INFO Channel 22/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:223463 [3] NCCL INFO Channel 16/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:223464 [1] NCCL INFO Channel 23/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:223463 [3] NCCL INFO Channel 17/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:223462 [5] NCCL INFO Channel 15/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:223463 [3] NCCL INFO Channel 18/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:223462 [5] NCCL INFO Channel 16/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:223463 [3] NCCL INFO Channel 19/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:223462 [5] NCCL INFO Channel 17/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:223463 [3] NCCL INFO Channel 20/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:223462 [5] NCCL INFO Channel 18/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:223463 [3] NCCL INFO Channel 21/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:223462 [5] NCCL INFO Channel 19/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:223463 [3] NCCL INFO Channel 22/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:223462 [5] NCCL INFO Channel 20/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:223462 [5] NCCL INFO Channel 21/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:223463 [3] NCCL INFO Channel 23/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:223462 [5] NCCL INFO Channel 22/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:223462 [5] NCCL INFO Channel 23/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221839:223458 [0] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:221839:223458 [0] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:221839:223458 [0] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:221840:223464 [1] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:221840:223464 [1] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:221840:223464 [1] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:221841:223460 [2] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:221841:223460 [2] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:221841:223460 [2] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:221843:223461 [4] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:221842:223463 [3] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:221843:223461 [4] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:221843:223461 [4] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:221842:223463 [3] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:221842:223463 [3] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:221844:223462 [5] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:221844:223462 [5] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:221844:223462 [5] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:221845:223465 [6] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:221845:223465 [6] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:221845:223465 [6] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:221846:223459 [7] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:221846:223459 [7] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:221846:223459 [7] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:221846:223459 [7] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:221846:223459 [7] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:221846:223459 [7] NCCL INFO ncclCommInitRank comm 0x5581de1ed740 rank 7 nranks 8 cudaDev 7 nvmlDev 7 busId a01d0 commId 0xd63d2a52f14618ac - Init COMPLETE +ywang29-vrdb-test1-worker-0:221842:223463 [3] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:221842:223463 [3] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:221842:223463 [3] NCCL INFO ncclCommInitRank comm 0x559fa9cacd80 rank 3 nranks 8 cudaDev 3 nvmlDev 3 busId 201d0 commId 0xd63d2a52f14618ac - Init COMPLETE +ywang29-vrdb-test1-worker-0:221841:223460 [2] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:221841:223460 [2] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:221845:223465 [6] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:221841:223460 [2] NCCL INFO ncclCommInitRank comm 0x55f2449d8200 rank 2 nranks 8 cudaDev 2 nvmlDev 2 busId 201c0 commId 0xd63d2a52f14618ac - Init COMPLETE +ywang29-vrdb-test1-worker-0:221845:223465 [6] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:221839:223458 [0] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:221845:223465 [6] NCCL INFO ncclCommInitRank comm 0x562b6bb81240 rank 6 nranks 8 cudaDev 6 nvmlDev 6 busId a01c0 commId 0xd63d2a52f14618ac - Init COMPLETE +ywang29-vrdb-test1-worker-0:221839:223458 [0] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:221839:223458 [0] NCCL INFO ncclCommInitRank comm 0x55589fe9d150 rank 0 nranks 8 cudaDev 0 nvmlDev 0 busId 101c0 commId 0xd63d2a52f14618ac - Init COMPLETE +ywang29-vrdb-test1-worker-0:221843:223461 [4] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:221843:223461 [4] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:221843:223461 [4] NCCL INFO ncclCommInitRank comm 0x55f355a97140 rank 4 nranks 8 cudaDev 4 nvmlDev 4 busId 901c0 commId 0xd63d2a52f14618ac - Init COMPLETE +ywang29-vrdb-test1-worker-0:221844:223462 [5] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:221844:223462 [5] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:221844:223462 [5] NCCL INFO ncclCommInitRank comm 0x56140b4605f0 rank 5 nranks 8 cudaDev 5 nvmlDev 5 busId 901d0 commId 0xd63d2a52f14618ac - Init COMPLETE +ywang29-vrdb-test1-worker-0:221840:223464 [1] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:221840:223464 [1] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:221840:223464 [1] NCCL INFO ncclCommInitRank comm 0x55e2fbfac080 rank 1 nranks 8 cudaDev 1 nvmlDev 1 busId 101d0 commId 0xd63d2a52f14618ac - Init COMPLETE +[2025-10-12 09:03:45,550] [INFO] [partition_parameters.py:343:__exit__] finished initializing model - num_params = 459, num_elems = 0.99B +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +[2025-10-12 09:03:47,235] [INFO] [partition_parameters.py:343:__exit__] finished initializing model - num_params = 907, num_elems = 1.42B +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +TinyLlavaForConditionalGeneration( + (language_model): Qwen2ForCausalLM( + (model): Qwen2Model( + (embed_tokens): Embedding(151936, 896) + (layers): ModuleList( + (0-23): 24 x Qwen2DecoderLayer( + (self_attn): Qwen2FlashAttention2( + (q_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=True) + (k_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=128, bias=True) + (v_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=128, bias=True) + (o_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=False) + (rotary_emb): Qwen2RotaryEmbedding() + ) + (mlp): Qwen2MLP( + (gate_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=4864, bias=False) + (up_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=4864, bias=False) + (down_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=4864, out_features=896, bias=False) + (act_fn): SiLU() + ) + (input_layernorm): Qwen2RMSNorm() + (post_attention_layernorm): Qwen2RMSNorm() + ) + ) + (norm): Qwen2RMSNorm() + ) + (lm_head): Linear(in_features=896, out_features=151936, bias=False) + ) + (vision_tower): SIGLIPVisionTower( + (_vision_tower): SiglipVisionModel( + (vision_model): SiglipVisionTransformer( + (embeddings): SiglipVisionEmbeddings( + (patch_embedding): Conv2d(3, 1152, kernel_size=(14, 14), stride=(14, 14), padding=valid) + (position_embedding): Embedding(729, 1152) + ) + (encoder): SiglipEncoder( + (layers): ModuleList( + (0-26): 27 x SiglipEncoderLayer( + (self_attn): SiglipAttention( + (k_proj): Linear(in_features=1152, out_features=1152, bias=True) + (v_proj): Linear(in_features=1152, out_features=1152, bias=True) + (q_proj): Linear(in_features=1152, out_features=1152, bias=True) + (out_proj): Linear(in_features=1152, out_features=1152, bias=True) + ) + (layer_norm1): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + (layer_norm2): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + ) + ) + ) + (post_layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (head): SiglipMultiheadAttentionPoolingHead( + (attention): MultiheadAttention( + (out_proj): NonDynamicallyQuantizableLinear(in_features=1152, out_features=1152, bias=True) + ) + (layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + ) + ) + ) + ) + (connector): MLPConnector( + (_connector): Sequential( + (0): SupermaskLinearSparsity_SoftForward_Normal(in_features=1152, out_features=896, bias=True) + (1): GELU(approximate='none') + (2): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=True) + ) + ) +) +Pre-training init language_model.model.layers.0.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.0.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.0.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.0.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.0.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.0.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.0.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.1.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.1.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.1.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.1.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.1.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.1.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.1.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.2.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.2.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.2.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.2.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.2.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.2.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.2.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.3.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.3.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.3.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.3.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.3.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.3.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.3.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.4.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.4.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.4.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.4.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.4.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.4.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.4.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.5.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.5.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.5.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.5.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.5.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.5.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.5.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.6.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.6.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.6.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.6.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.6.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.6.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.6.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.7.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.7.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.7.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.7.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.7.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.7.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.7.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.8.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.8.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.8.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.8.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.8.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.8.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.8.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.9.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.9.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.9.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.9.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.9.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.9.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.9.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.10.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.10.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.10.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.10.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.10.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.10.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.10.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.11.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.11.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.11.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.11.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.11.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.11.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.11.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.12.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.12.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.12.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.12.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.12.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.12.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.12.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.13.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.13.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.13.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.13.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.13.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.13.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.13.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.14.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.14.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.14.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.14.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.14.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.14.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.14.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.15.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.15.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.15.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.15.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.15.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.15.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.15.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.16.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.16.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.16.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.16.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.16.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.16.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.16.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.17.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.17.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.17.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.17.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.17.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.17.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.17.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.18.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.18.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.18.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.18.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.18.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.18.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.18.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.19.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.19.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.19.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.19.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.19.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.19.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.19.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.20.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.20.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.20.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.20.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.20.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.20.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.20.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.21.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.21.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.21.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.21.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.21.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.21.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.21.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.22.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.22.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.22.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.22.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.22.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.22.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.22.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.23.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.23.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.23.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.23.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.23.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.23.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.23.mlp.down_proj.scores: Mean=1.000000 +Pre-training init connector._connector.0.scores: Mean=1.000005 +Pre-training init connector._connector.2.scores: Mean=0.999970 +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +2025-10-12 09:04:05,213 | INFO: Total Parameters: 1283756736, Total Trainable Parameters: 359661568 +2025-10-12 09:04:05,218 | INFO: Trainable Parameters: +language_model.model.layers.0.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.0.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.0.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.0.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.0.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.0.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.0.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.1.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.1.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.1.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.1.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.1.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.1.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.1.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.2.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.2.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.2.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.2.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.2.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.2.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.2.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.3.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.3.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.3.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.3.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.3.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.3.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.3.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.4.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.4.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.4.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.4.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.4.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.4.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.4.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.5.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.5.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.5.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.5.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.5.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.5.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.5.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.6.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.6.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.6.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.6.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.6.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.6.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.6.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.7.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.7.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.7.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.7.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.7.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.7.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.7.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.8.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.8.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.8.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.8.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.8.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.8.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.8.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.9.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.9.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.9.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.9.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.9.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.9.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.9.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.10.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.10.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.10.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.10.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.10.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.10.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.10.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.11.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.11.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.11.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.11.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.11.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.11.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.11.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.12.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.12.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.12.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.12.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.12.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.12.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.12.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.13.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.13.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.13.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.13.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.13.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.13.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.13.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.14.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.14.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.14.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.14.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.14.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.14.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.14.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.15.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.15.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.15.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.15.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.15.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.15.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.15.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.16.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.16.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.16.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.16.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.16.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.16.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.16.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.17.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.17.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.17.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.17.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.17.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.17.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.17.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.18.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.18.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.18.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.18.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.18.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.18.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.18.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.19.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.19.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.19.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.19.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.19.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.19.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.19.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.20.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.20.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.20.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.20.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.20.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.20.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.20.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.21.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.21.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.21.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.21.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.21.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.21.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.21.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.22.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.22.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.22.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.22.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.22.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.22.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.22.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.23.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.23.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.23.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.23.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.23.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.23.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.23.mlp.down_proj.scores: 4358144 parameters +connector._connector.0.scores: 1032192 parameters +connector._connector.2.scores: 802816 parameters +Parameter Offload: Total persistent parameters: 486464 in 403 params + 0%| | 0/520 [00:003->2 [1] 4/-1/-1->3->2 [2] 4/-1/-1->3->2 [3] 4/-1/-1->3->2 [4] 4/-1/-1->3->2 [5] 4/-1/-1->3->2 [6] 4/-1/-1->3->2 [7] 4/-1/-1->3->2 [8] 4/-1/-1->3->2 [9] 4/-1/-1->3->2 [10] 4/-1/-1->3->2 [11] 4/-1/-1->3->2 [12] 4/-1/-1->3->2 [13] 4/-1/-1->3->2 [14] 4/-1/-1->3->2 [15] 4/-1/-1->3->2 [16] 4/-1/-1->3->2 [17] 4/-1/-1->3->2 [18] 4/-1/-1->3->2 [19] 4/-1/-1->3->2 [20] 4/-1/-1->3->2 [21] 4/-1/-1->3->2 [22] 4/-1/-1->3->2 [23] 4/-1/-1->3->2 +ywang29-vrdb-test1-worker-0:221846:228362 [7] NCCL INFO comm 0x7efbb006a970 rank 7 nRanks 8 nNodes 1 localRanks 8 localRank 7 MNNVL 0 +ywang29-vrdb-test1-worker-0:221842:228361 [3] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:221839:228360 [0] NCCL INFO Channel 00/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:221844:228364 [5] NCCL INFO Trees [0] 6/-1/-1->5->4 [1] 6/-1/-1->5->4 [2] 6/-1/-1->5->4 [3] 6/-1/-1->5->4 [4] 6/-1/-1->5->4 [5] 6/-1/-1->5->4 [6] 6/-1/-1->5->4 [7] 6/-1/-1->5->4 [8] 6/-1/-1->5->4 [9] 6/-1/-1->5->4 [10] 6/-1/-1->5->4 [11] 6/-1/-1->5->4 [12] 6/-1/-1->5->4 [13] 6/-1/-1->5->4 [14] 6/-1/-1->5->4 [15] 6/-1/-1->5->4 [16] 6/-1/-1->5->4 [17] 6/-1/-1->5->4 [18] 6/-1/-1->5->4 [19] 6/-1/-1->5->4 [20] 6/-1/-1->5->4 [21] 6/-1/-1->5->4 [22] 6/-1/-1->5->4 [23] 6/-1/-1->5->4 +ywang29-vrdb-test1-worker-0:221845:228363 [6] NCCL INFO Trees [0] 7/-1/-1->6->5 [1] 7/-1/-1->6->5 [2] 7/-1/-1->6->5 [3] 7/-1/-1->6->5 [4] 7/-1/-1->6->5 [5] 7/-1/-1->6->5 [6] 7/-1/-1->6->5 [7] 7/-1/-1->6->5 [8] 7/-1/-1->6->5 [9] 7/-1/-1->6->5 [10] 7/-1/-1->6->5 [11] 7/-1/-1->6->5 [12] 7/-1/-1->6->5 [13] 7/-1/-1->6->5 [14] 7/-1/-1->6->5 [15] 7/-1/-1->6->5 [16] 7/-1/-1->6->5 [17] 7/-1/-1->6->5 [18] 7/-1/-1->6->5 [19] 7/-1/-1->6->5 [20] 7/-1/-1->6->5 [21] 7/-1/-1->6->5 [22] 7/-1/-1->6->5 [23] 7/-1/-1->6->5 +ywang29-vrdb-test1-worker-0:221839:228360 [0] NCCL INFO Channel 01/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:221844:228364 [5] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:221845:228363 [6] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:221839:228360 [0] NCCL INFO Channel 02/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:221839:228360 [0] NCCL INFO Channel 03/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:221839:228360 [0] NCCL INFO Channel 04/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:221841:228366 [2] NCCL INFO Trees [0] 3/-1/-1->2->1 [1] 3/-1/-1->2->1 [2] 3/-1/-1->2->1 [3] 3/-1/-1->2->1 [4] 3/-1/-1->2->1 [5] 3/-1/-1->2->1 [6] 3/-1/-1->2->1 [7] 3/-1/-1->2->1 [8] 3/-1/-1->2->1 [9] 3/-1/-1->2->1 [10] 3/-1/-1->2->1 [11] 3/-1/-1->2->1 [12] 3/-1/-1->2->1 [13] 3/-1/-1->2->1 [14] 3/-1/-1->2->1 [15] 3/-1/-1->2->1 [16] 3/-1/-1->2->1 [17] 3/-1/-1->2->1 [18] 3/-1/-1->2->1 [19] 3/-1/-1->2->1 [20] 3/-1/-1->2->1 [21] 3/-1/-1->2->1 [22] 3/-1/-1->2->1 [23] 3/-1/-1->2->1 +ywang29-vrdb-test1-worker-0:221839:228360 [0] NCCL INFO Channel 05/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:221841:228366 [2] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:221839:228360 [0] NCCL INFO Channel 06/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:221839:228360 [0] NCCL INFO Channel 07/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:221840:228365 [1] NCCL INFO Trees [0] 2/-1/-1->1->0 [1] 2/-1/-1->1->0 [2] 2/-1/-1->1->0 [3] 2/-1/-1->1->0 [4] 2/-1/-1->1->0 [5] 2/-1/-1->1->0 [6] 2/-1/-1->1->0 [7] 2/-1/-1->1->0 [8] 2/-1/-1->1->0 [9] 2/-1/-1->1->0 [10] 2/-1/-1->1->0 [11] 2/-1/-1->1->0 [12] 2/-1/-1->1->0 [13] 2/-1/-1->1->0 [14] 2/-1/-1->1->0 [15] 2/-1/-1->1->0 [16] 2/-1/-1->1->0 [17] 2/-1/-1->1->0 [18] 2/-1/-1->1->0 [19] 2/-1/-1->1->0 [20] 2/-1/-1->1->0 [21] 2/-1/-1->1->0 [22] 2/-1/-1->1->0 [23] 2/-1/-1->1->0 +ywang29-vrdb-test1-worker-0:221839:228360 [0] NCCL INFO Channel 08/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:221840:228365 [1] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:221839:228360 [0] NCCL INFO Channel 09/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:221839:228360 [0] NCCL INFO Channel 10/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:221839:228360 [0] NCCL INFO Channel 11/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:221839:228360 [0] NCCL INFO Channel 12/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:221839:228360 [0] NCCL INFO Channel 13/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:221839:228360 [0] NCCL INFO Channel 14/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:221843:228367 [4] NCCL INFO Trees [0] 5/-1/-1->4->3 [1] 5/-1/-1->4->3 [2] 5/-1/-1->4->3 [3] 5/-1/-1->4->3 [4] 5/-1/-1->4->3 [5] 5/-1/-1->4->3 [6] 5/-1/-1->4->3 [7] 5/-1/-1->4->3 [8] 5/-1/-1->4->3 [9] 5/-1/-1->4->3 [10] 5/-1/-1->4->3 [11] 5/-1/-1->4->3 [12] 5/-1/-1->4->3 [13] 5/-1/-1->4->3 [14] 5/-1/-1->4->3 [15] 5/-1/-1->4->3 [16] 5/-1/-1->4->3 [17] 5/-1/-1->4->3 [18] 5/-1/-1->4->3 [19] 5/-1/-1->4->3 [20] 5/-1/-1->4->3 [21] 5/-1/-1->4->3 [22] 5/-1/-1->4->3 [23] 5/-1/-1->4->3 +ywang29-vrdb-test1-worker-0:221839:228360 [0] NCCL INFO Channel 15/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:221839:228360 [0] NCCL INFO Channel 16/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:221843:228367 [4] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:221839:228360 [0] NCCL INFO Channel 17/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:221839:228360 [0] NCCL INFO Channel 18/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:221846:228362 [7] NCCL INFO Trees [0] -1/-1/-1->7->6 [1] -1/-1/-1->7->6 [2] -1/-1/-1->7->6 [3] -1/-1/-1->7->6 [4] -1/-1/-1->7->6 [5] -1/-1/-1->7->6 [6] -1/-1/-1->7->6 [7] -1/-1/-1->7->6 [8] -1/-1/-1->7->6 [9] -1/-1/-1->7->6 [10] -1/-1/-1->7->6 [11] -1/-1/-1->7->6 [12] -1/-1/-1->7->6 [13] -1/-1/-1->7->6 [14] -1/-1/-1->7->6 [15] -1/-1/-1->7->6 [16] -1/-1/-1->7->6 [17] -1/-1/-1->7->6 [18] -1/-1/-1->7->6 [19] -1/-1/-1->7->6 [20] -1/-1/-1->7->6 [21] -1/-1/-1->7->6 [22] -1/-1/-1->7->6 [23] -1/-1/-1->7->6 +ywang29-vrdb-test1-worker-0:221839:228360 [0] NCCL INFO Channel 19/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:221846:228362 [7] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:221839:228360 [0] NCCL INFO Channel 20/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:221839:228360 [0] NCCL INFO Channel 21/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:221839:228360 [0] NCCL INFO Channel 22/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:221839:228360 [0] NCCL INFO Channel 23/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:221839:228360 [0] NCCL INFO Trees [0] 1/-1/-1->0->-1 [1] 1/-1/-1->0->-1 [2] 1/-1/-1->0->-1 [3] 1/-1/-1->0->-1 [4] 1/-1/-1->0->-1 [5] 1/-1/-1->0->-1 [6] 1/-1/-1->0->-1 [7] 1/-1/-1->0->-1 [8] 1/-1/-1->0->-1 [9] 1/-1/-1->0->-1 [10] 1/-1/-1->0->-1 [11] 1/-1/-1->0->-1 [12] 1/-1/-1->0->-1 [13] 1/-1/-1->0->-1 [14] 1/-1/-1->0->-1 [15] 1/-1/-1->0->-1 [16] 1/-1/-1->0->-1 [17] 1/-1/-1->0->-1 [18] 1/-1/-1->0->-1 [19] 1/-1/-1->0->-1 [20] 1/-1/-1->0->-1 [21] 1/-1/-1->0->-1 [22] 1/-1/-1->0->-1 [23] 1/-1/-1->0->-1 +ywang29-vrdb-test1-worker-0:221839:228360 [0] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:221842:228361 [3] NCCL INFO Channel 00/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:228364 [5] NCCL INFO Channel 00/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:228363 [6] NCCL INFO Channel 00/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:228365 [1] NCCL INFO Channel 00/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:228361 [3] NCCL INFO Channel 01/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:228366 [2] NCCL INFO Channel 00/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:228364 [5] NCCL INFO Channel 01/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221839:228360 [0] NCCL INFO Channel 00/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:228363 [6] NCCL INFO Channel 01/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:228365 [1] NCCL INFO Channel 01/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:228361 [3] NCCL INFO Channel 02/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:228366 [2] NCCL INFO Channel 01/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:228364 [5] NCCL INFO Channel 02/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221839:228360 [0] NCCL INFO Channel 01/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:228363 [6] NCCL INFO Channel 02/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:228365 [1] NCCL INFO Channel 02/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:228361 [3] NCCL INFO Channel 03/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:228366 [2] NCCL INFO Channel 02/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:228364 [5] NCCL INFO Channel 03/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:228363 [6] NCCL INFO Channel 03/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221839:228360 [0] NCCL INFO Channel 02/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:228367 [4] NCCL INFO Channel 00/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:228365 [1] NCCL INFO Channel 03/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:228361 [3] NCCL INFO Channel 04/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:228366 [2] NCCL INFO Channel 03/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:228364 [5] NCCL INFO Channel 04/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:228363 [6] NCCL INFO Channel 04/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:228362 [7] NCCL INFO Channel 00/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221839:228360 [0] NCCL INFO Channel 03/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:228367 [4] NCCL INFO Channel 01/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:228365 [1] NCCL INFO Channel 04/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:228361 [3] NCCL INFO Channel 05/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:228366 [2] NCCL INFO Channel 04/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:228364 [5] NCCL INFO Channel 05/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:228363 [6] NCCL INFO Channel 05/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221839:228360 [0] NCCL INFO Channel 04/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:228362 [7] NCCL INFO Channel 01/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:228367 [4] NCCL INFO Channel 02/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:228365 [1] NCCL INFO Channel 05/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:228366 [2] NCCL INFO Channel 05/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:228361 [3] NCCL INFO Channel 06/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:228364 [5] NCCL INFO Channel 06/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:228363 [6] NCCL INFO Channel 06/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221839:228360 [0] NCCL INFO Channel 05/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:228367 [4] NCCL INFO Channel 03/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:228362 [7] NCCL INFO Channel 02/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:228365 [1] NCCL INFO Channel 06/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:228366 [2] NCCL INFO Channel 06/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:228361 [3] NCCL INFO Channel 07/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:228364 [5] NCCL INFO Channel 07/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:228363 [6] NCCL INFO Channel 07/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221839:228360 [0] NCCL INFO Channel 06/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:228367 [4] NCCL INFO Channel 04/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:228362 [7] NCCL INFO Channel 03/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:228365 [1] NCCL INFO Channel 07/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:228366 [2] NCCL INFO Channel 07/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:228361 [3] NCCL INFO Channel 08/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:228364 [5] NCCL INFO Channel 08/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:228363 [6] NCCL INFO Channel 08/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221839:228360 [0] NCCL INFO Channel 07/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:228367 [4] NCCL INFO Channel 05/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:228362 [7] NCCL INFO Channel 04/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:228365 [1] NCCL INFO Channel 08/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:228366 [2] NCCL INFO Channel 08/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:228361 [3] NCCL INFO Channel 09/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:228364 [5] NCCL INFO Channel 09/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:228363 [6] NCCL INFO Channel 09/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221839:228360 [0] NCCL INFO Channel 08/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:228367 [4] NCCL INFO Channel 06/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:228362 [7] NCCL INFO Channel 05/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:228365 [1] NCCL INFO Channel 09/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:228366 [2] NCCL INFO Channel 09/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:228361 [3] NCCL INFO Channel 10/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:228364 [5] NCCL INFO Channel 10/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:228363 [6] NCCL INFO Channel 10/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221839:228360 [0] NCCL INFO Channel 09/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:228367 [4] NCCL INFO Channel 07/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:228362 [7] NCCL INFO Channel 06/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:228365 [1] NCCL INFO Channel 10/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:228366 [2] NCCL INFO Channel 10/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:228361 [3] NCCL INFO Channel 11/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:228364 [5] NCCL INFO Channel 11/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:228363 [6] NCCL INFO Channel 11/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221839:228360 [0] NCCL INFO Channel 10/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:228367 [4] NCCL INFO Channel 08/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:228362 [7] NCCL INFO Channel 07/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:228365 [1] NCCL INFO Channel 11/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:228366 [2] NCCL INFO Channel 11/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:228364 [5] NCCL INFO Channel 12/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:228361 [3] NCCL INFO Channel 12/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:228363 [6] NCCL INFO Channel 12/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:228367 [4] NCCL INFO Channel 09/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221839:228360 [0] NCCL INFO Channel 11/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:228362 [7] NCCL INFO Channel 08/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:228365 [1] NCCL INFO Channel 12/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:228366 [2] NCCL INFO Channel 12/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:228364 [5] NCCL INFO Channel 13/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:228361 [3] NCCL INFO Channel 13/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:228363 [6] NCCL INFO Channel 13/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:228367 [4] NCCL INFO Channel 10/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221839:228360 [0] NCCL INFO Channel 12/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:228362 [7] NCCL INFO Channel 09/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:228365 [1] NCCL INFO Channel 13/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:228366 [2] NCCL INFO Channel 13/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:228364 [5] NCCL INFO Channel 14/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:228361 [3] NCCL INFO Channel 14/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:228363 [6] NCCL INFO Channel 14/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:228367 [4] NCCL INFO Channel 11/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221839:228360 [0] NCCL INFO Channel 13/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:228362 [7] NCCL INFO Channel 10/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:228365 [1] NCCL INFO Channel 14/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:228366 [2] NCCL INFO Channel 14/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:228364 [5] NCCL INFO Channel 15/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:228363 [6] NCCL INFO Channel 15/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:228361 [3] NCCL INFO Channel 15/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:228367 [4] NCCL INFO Channel 12/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221839:228360 [0] NCCL INFO Channel 14/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:228362 [7] NCCL INFO Channel 11/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:228365 [1] NCCL INFO Channel 15/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:228366 [2] NCCL INFO Channel 15/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:228364 [5] NCCL INFO Channel 16/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:228363 [6] NCCL INFO Channel 16/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:228361 [3] NCCL INFO Channel 16/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:228367 [4] NCCL INFO Channel 13/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221839:228360 [0] NCCL INFO Channel 15/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:228362 [7] NCCL INFO Channel 12/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:228365 [1] NCCL INFO Channel 16/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:228366 [2] NCCL INFO Channel 16/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:228364 [5] NCCL INFO Channel 17/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:228363 [6] NCCL INFO Channel 17/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:228361 [3] NCCL INFO Channel 17/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:228367 [4] NCCL INFO Channel 14/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221839:228360 [0] NCCL INFO Channel 16/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:228362 [7] NCCL INFO Channel 13/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:228365 [1] NCCL INFO Channel 17/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:228366 [2] NCCL INFO Channel 17/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:228364 [5] NCCL INFO Channel 18/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:228363 [6] NCCL INFO Channel 18/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:228361 [3] NCCL INFO Channel 18/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:228367 [4] NCCL INFO Channel 15/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221839:228360 [0] NCCL INFO Channel 17/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:228362 [7] NCCL INFO Channel 14/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:228365 [1] NCCL INFO Channel 18/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:228366 [2] NCCL INFO Channel 18/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:228364 [5] NCCL INFO Channel 19/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:228361 [3] NCCL INFO Channel 19/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:228363 [6] NCCL INFO Channel 19/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:228367 [4] NCCL INFO Channel 16/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221839:228360 [0] NCCL INFO Channel 18/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:228362 [7] NCCL INFO Channel 15/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:228365 [1] NCCL INFO Channel 19/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:228366 [2] NCCL INFO Channel 19/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:228364 [5] NCCL INFO Channel 20/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:228363 [6] NCCL INFO Channel 20/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:228361 [3] NCCL INFO Channel 20/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:228367 [4] NCCL INFO Channel 17/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221839:228360 [0] NCCL INFO Channel 19/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:228362 [7] NCCL INFO Channel 16/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:228365 [1] NCCL INFO Channel 20/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:228366 [2] NCCL INFO Channel 20/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:228363 [6] NCCL INFO Channel 21/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:228361 [3] NCCL INFO Channel 21/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:228367 [4] NCCL INFO Channel 18/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221839:228360 [0] NCCL INFO Channel 20/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:228364 [5] NCCL INFO Channel 21/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:228362 [7] NCCL INFO Channel 17/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:228365 [1] NCCL INFO Channel 21/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:228366 [2] NCCL INFO Channel 21/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:228361 [3] NCCL INFO Channel 22/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:228363 [6] NCCL INFO Channel 22/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221839:228360 [0] NCCL INFO Channel 21/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:228367 [4] NCCL INFO Channel 19/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:228364 [5] NCCL INFO Channel 22/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:228362 [7] NCCL INFO Channel 18/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:228366 [2] NCCL INFO Channel 22/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:228365 [1] NCCL INFO Channel 22/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:228361 [3] NCCL INFO Channel 23/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:228363 [6] NCCL INFO Channel 23/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221839:228360 [0] NCCL INFO Channel 22/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:228367 [4] NCCL INFO Channel 20/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:228364 [5] NCCL INFO Channel 23/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:228362 [7] NCCL INFO Channel 19/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:228365 [1] NCCL INFO Channel 23/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:228366 [2] NCCL INFO Channel 23/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221839:228360 [0] NCCL INFO Channel 23/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:228367 [4] NCCL INFO Channel 21/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:228362 [7] NCCL INFO Channel 20/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:228367 [4] NCCL INFO Channel 22/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:228362 [7] NCCL INFO Channel 21/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:228362 [7] NCCL INFO Channel 22/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:228367 [4] NCCL INFO Channel 23/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:228362 [7] NCCL INFO Channel 23/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:228365 [1] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:221839:228360 [0] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:221841:228366 [2] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:221842:228361 [3] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:221843:228367 [4] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:221840:228365 [1] NCCL INFO Channel 00/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:228365 [1] NCCL INFO Channel 01/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:228366 [2] NCCL INFO Channel 00/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:228365 [1] NCCL INFO Channel 02/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:228366 [2] NCCL INFO Channel 01/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:228365 [1] NCCL INFO Channel 03/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:228365 [1] NCCL INFO Channel 04/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:228366 [2] NCCL INFO Channel 02/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:228367 [4] NCCL INFO Channel 00/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:228365 [1] NCCL INFO Channel 05/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:228366 [2] NCCL INFO Channel 03/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:228361 [3] NCCL INFO Channel 00/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:228367 [4] NCCL INFO Channel 01/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:228366 [2] NCCL INFO Channel 04/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:228365 [1] NCCL INFO Channel 06/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:228361 [3] NCCL INFO Channel 01/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:228367 [4] NCCL INFO Channel 02/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:228365 [1] NCCL INFO Channel 07/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:228361 [3] NCCL INFO Channel 02/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:228366 [2] NCCL INFO Channel 05/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:228362 [7] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:221846:228362 [7] NCCL INFO Channel 00/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:228365 [1] NCCL INFO Channel 08/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:228364 [5] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:221842:228361 [3] NCCL INFO Channel 03/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:228363 [6] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:221841:228366 [2] NCCL INFO Channel 06/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:228367 [4] NCCL INFO Channel 03/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:228365 [1] NCCL INFO Channel 09/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:228362 [7] NCCL INFO Channel 01/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:228361 [3] NCCL INFO Channel 04/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:228366 [2] NCCL INFO Channel 07/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:228367 [4] NCCL INFO Channel 04/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:228365 [1] NCCL INFO Channel 10/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:228362 [7] NCCL INFO Channel 02/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:228361 [3] NCCL INFO Channel 05/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:228366 [2] NCCL INFO Channel 08/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:228367 [4] NCCL INFO Channel 05/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:228365 [1] NCCL INFO Channel 11/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:228366 [2] NCCL INFO Channel 09/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:228367 [4] NCCL INFO Channel 06/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:228365 [1] NCCL INFO Channel 12/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:228366 [2] NCCL INFO Channel 10/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:228367 [4] NCCL INFO Channel 07/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:228365 [1] NCCL INFO Channel 13/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:228366 [2] NCCL INFO Channel 11/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:228367 [4] NCCL INFO Channel 08/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:228365 [1] NCCL INFO Channel 14/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:228361 [3] NCCL INFO Channel 06/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:228366 [2] NCCL INFO Channel 12/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:228367 [4] NCCL INFO Channel 09/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:228365 [1] NCCL INFO Channel 15/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:228361 [3] NCCL INFO Channel 07/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:228366 [2] NCCL INFO Channel 13/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:228367 [4] NCCL INFO Channel 10/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:228365 [1] NCCL INFO Channel 16/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:228361 [3] NCCL INFO Channel 08/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:228367 [4] NCCL INFO Channel 11/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:228366 [2] NCCL INFO Channel 14/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:228365 [1] NCCL INFO Channel 17/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:228361 [3] NCCL INFO Channel 09/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:228367 [4] NCCL INFO Channel 12/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:228366 [2] NCCL INFO Channel 15/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:228365 [1] NCCL INFO Channel 18/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:228361 [3] NCCL INFO Channel 10/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:228367 [4] NCCL INFO Channel 13/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:228366 [2] NCCL INFO Channel 16/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:228362 [7] NCCL INFO Channel 03/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:228365 [1] NCCL INFO Channel 19/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:228361 [3] NCCL INFO Channel 11/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:228366 [2] NCCL INFO Channel 17/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:228367 [4] NCCL INFO Channel 14/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:228362 [7] NCCL INFO Channel 04/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:228365 [1] NCCL INFO Channel 20/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:228361 [3] NCCL INFO Channel 12/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:228366 [2] NCCL INFO Channel 18/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:228367 [4] NCCL INFO Channel 15/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:228362 [7] NCCL INFO Channel 05/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:228365 [1] NCCL INFO Channel 21/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:228361 [3] NCCL INFO Channel 13/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:228366 [2] NCCL INFO Channel 19/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:228367 [4] NCCL INFO Channel 16/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:228362 [7] NCCL INFO Channel 06/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:228365 [1] NCCL INFO Channel 22/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:228361 [3] NCCL INFO Channel 14/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:228366 [2] NCCL INFO Channel 20/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:228367 [4] NCCL INFO Channel 17/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:228362 [7] NCCL INFO Channel 07/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221840:228365 [1] NCCL INFO Channel 23/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:228361 [3] NCCL INFO Channel 15/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:228366 [2] NCCL INFO Channel 21/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:228367 [4] NCCL INFO Channel 18/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:228362 [7] NCCL INFO Channel 08/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:228361 [3] NCCL INFO Channel 16/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:228366 [2] NCCL INFO Channel 22/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:228367 [4] NCCL INFO Channel 19/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:228362 [7] NCCL INFO Channel 09/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:228367 [4] NCCL INFO Channel 20/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221841:228366 [2] NCCL INFO Channel 23/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:228361 [3] NCCL INFO Channel 17/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:228361 [3] NCCL INFO Channel 18/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:228367 [4] NCCL INFO Channel 21/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:228361 [3] NCCL INFO Channel 19/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:228367 [4] NCCL INFO Channel 22/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:228361 [3] NCCL INFO Channel 20/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:228361 [3] NCCL INFO Channel 21/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:228362 [7] NCCL INFO Channel 10/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:228361 [3] NCCL INFO Channel 22/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221843:228367 [4] NCCL INFO Channel 23/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221842:228361 [3] NCCL INFO Channel 23/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:228363 [6] NCCL INFO Channel 00/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:228363 [6] NCCL INFO Channel 01/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:228362 [7] NCCL INFO Channel 11/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:228363 [6] NCCL INFO Channel 02/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:228362 [7] NCCL INFO Channel 12/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:228362 [7] NCCL INFO Channel 13/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:228363 [6] NCCL INFO Channel 03/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:228362 [7] NCCL INFO Channel 14/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:228362 [7] NCCL INFO Channel 15/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:228363 [6] NCCL INFO Channel 04/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:228362 [7] NCCL INFO Channel 16/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:228363 [6] NCCL INFO Channel 05/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:228362 [7] NCCL INFO Channel 17/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:228364 [5] NCCL INFO Channel 00/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:228363 [6] NCCL INFO Channel 06/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:228363 [6] NCCL INFO Channel 07/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:228362 [7] NCCL INFO Channel 18/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:228364 [5] NCCL INFO Channel 01/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:228364 [5] NCCL INFO Channel 02/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:228364 [5] NCCL INFO Channel 03/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:228364 [5] NCCL INFO Channel 04/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:228364 [5] NCCL INFO Channel 05/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:228364 [5] NCCL INFO Channel 06/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:228364 [5] NCCL INFO Channel 07/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:228364 [5] NCCL INFO Channel 08/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:228364 [5] NCCL INFO Channel 09/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:228364 [5] NCCL INFO Channel 10/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:228364 [5] NCCL INFO Channel 11/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:228364 [5] NCCL INFO Channel 12/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:228364 [5] NCCL INFO Channel 13/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:228363 [6] NCCL INFO Channel 08/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:228362 [7] NCCL INFO Channel 19/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:228364 [5] NCCL INFO Channel 14/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:228363 [6] NCCL INFO Channel 09/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:228362 [7] NCCL INFO Channel 20/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:228363 [6] NCCL INFO Channel 10/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:228364 [5] NCCL INFO Channel 15/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:228362 [7] NCCL INFO Channel 21/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:228363 [6] NCCL INFO Channel 11/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:228363 [6] NCCL INFO Channel 12/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:228362 [7] NCCL INFO Channel 22/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:228364 [5] NCCL INFO Channel 16/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:228363 [6] NCCL INFO Channel 13/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:228364 [5] NCCL INFO Channel 17/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:228363 [6] NCCL INFO Channel 14/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:228364 [5] NCCL INFO Channel 18/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221846:228362 [7] NCCL INFO Channel 23/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:228363 [6] NCCL INFO Channel 15/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:228364 [5] NCCL INFO Channel 19/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:228364 [5] NCCL INFO Channel 20/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:228363 [6] NCCL INFO Channel 16/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:228364 [5] NCCL INFO Channel 21/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:228363 [6] NCCL INFO Channel 17/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:228364 [5] NCCL INFO Channel 22/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:228363 [6] NCCL INFO Channel 18/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221844:228364 [5] NCCL INFO Channel 23/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:228363 [6] NCCL INFO Channel 19/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:228363 [6] NCCL INFO Channel 20/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:228363 [6] NCCL INFO Channel 21/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:228363 [6] NCCL INFO Channel 22/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221845:228363 [6] NCCL INFO Channel 23/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:221839:228360 [0] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:221839:228360 [0] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:221839:228360 [0] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:221840:228365 [1] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:221840:228365 [1] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:221840:228365 [1] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:221841:228366 [2] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:221841:228366 [2] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:221841:228366 [2] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:221846:228362 [7] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:221846:228362 [7] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:221846:228362 [7] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:221842:228361 [3] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:221842:228361 [3] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:221842:228361 [3] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:221843:228367 [4] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:221843:228367 [4] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:221843:228367 [4] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:221845:228363 [6] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:221844:228364 [5] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:221845:228363 [6] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:221844:228364 [5] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:221844:228364 [5] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:221845:228363 [6] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:221842:228361 [3] NCCL INFO ncclCommInitRank comm 0x7f1a3406b5d0 rank 3 nranks 8 cudaDev 3 nvmlDev 3 busId 201d0 commId 0x526e71cd8b341e4 - Init COMPLETE +ywang29-vrdb-test1-worker-0:221845:228363 [6] NCCL INFO ncclCommInitRank comm 0x7ef6dc06b5a0 rank 6 nranks 8 cudaDev 6 nvmlDev 6 busId a01c0 commId 0x526e71cd8b341e4 - Init COMPLETE +ywang29-vrdb-test1-worker-0:221844:228364 [5] NCCL INFO ncclCommInitRank comm 0x7f419c06a250 rank 5 nranks 8 cudaDev 5 nvmlDev 5 busId 901d0 commId 0x526e71cd8b341e4 - Init COMPLETE +ywang29-vrdb-test1-worker-0:221846:228362 [7] NCCL INFO ncclCommInitRank comm 0x7efbb006a970 rank 7 nranks 8 cudaDev 7 nvmlDev 7 busId a01d0 commId 0x526e71cd8b341e4 - Init COMPLETE +ywang29-vrdb-test1-worker-0:221841:228366 [2] NCCL INFO ncclCommInitRank comm 0x7ff13806b070 rank 2 nranks 8 cudaDev 2 nvmlDev 2 busId 201c0 commId 0x526e71cd8b341e4 - Init COMPLETE +ywang29-vrdb-test1-worker-0:221840:228365 [1] NCCL INFO ncclCommInitRank comm 0x7faf4006b040 rank 1 nranks 8 cudaDev 1 nvmlDev 1 busId 101d0 commId 0x526e71cd8b341e4 - Init COMPLETE +ywang29-vrdb-test1-worker-0:221843:228367 [4] NCCL INFO ncclCommInitRank comm 0x7f235806b110 rank 4 nranks 8 cudaDev 4 nvmlDev 4 busId 901c0 commId 0x526e71cd8b341e4 - Init COMPLETE +ywang29-vrdb-test1-worker-0:221839:228360 [0] NCCL INFO ncclCommInitRank comm 0x7fabcc06b220 rank 0 nranks 8 cudaDev 0 nvmlDev 0 busId 101c0 commId 0x526e71cd8b341e4 - Init COMPLETE + 0%| | 1/520 [28:23<245:37:10, 1703.72s/it] {'loss': 8.7188, 'grad_norm': 0.3859809681693823, 'learning_rate': 0.0125, 'epoch': 0.0} + 0%| | 1/520 [28:23<245:37:10, 1703.72s/it] 0%| | 2/520 [28:27<101:14:58, 703.66s/it] {'loss': 7.9499, 'grad_norm': 0.3897968511443435, 'learning_rate': 0.025, 'epoch': 0.0} + 0%| | 2/520 [28:27<101:14:58, 703.66s/it] 1%| | 3/520 [28:30<55:09:00, 384.02s/it] {'loss': 7.0406, 'grad_norm': 0.1507521132629217, 'learning_rate': 0.037500000000000006, 'epoch': 0.01} + 1%| | 3/520 [28:31<55:09:00, 384.02s/it] 1%| | 4/520 [28:34<33:31:09, 233.86s/it] {'loss': 6.3353, 'grad_norm': 0.09772430172772587, 'learning_rate': 0.05, 'epoch': 0.01} + 1%| | 4/520 [28:34<33:31:09, 233.86s/it] 1%| | 5/520 [28:38<21:34:37, 150.83s/it] {'loss': 5.3206, 'grad_norm': 0.08094453952016659, 'learning_rate': 0.0625, 'epoch': 0.01} + 1%| | 5/520 [28:38<21:34:37, 150.83s/it] 1%| | 6/520 [28:41<14:23:23, 100.78s/it] {'loss': 9.0023, 'grad_norm': 0.5867617970990923, 'learning_rate': 0.07500000000000001, 'epoch': 0.01} + 1%| | 6/520 [28:41<14:23:23, 100.78s/it] 1%|▏ | 7/520 [28:45<9:50:13, 69.03s/it] {'loss': 5.2204, 'grad_norm': 0.0839475390920157, 'learning_rate': 0.08750000000000001, 'epoch': 0.01} + 1%|▏ | 7/520 [28:45<9:50:13, 69.03s/it] 2%|▏ | 8/520 [28:49<6:53:04, 48.41s/it] {'loss': 4.9451, 'grad_norm': 0.0443608512863493, 'learning_rate': 0.1, 'epoch': 0.02} + 2%|▏ | 8/520 [28:49<6:53:04, 48.41s/it] 2%|▏ | 9/520 [28:53<4:54:24, 34.57s/it] {'loss': 4.498, 'grad_norm': 0.0390480169612291, 'learning_rate': 0.1125, 'epoch': 0.02} + 2%|▏ | 9/520 [28:53<4:54:24, 34.57s/it] 2%|▏ | 10/520 [28:57<3:32:35, 25.01s/it] {'loss': 3.3962, 'grad_norm': 0.03508605262818374, 'learning_rate': 0.125, 'epoch': 0.02} + 2%|▏ | 10/520 [28:57<3:32:35, 25.01s/it] 2%|▏ | 11/520 [29:01<2:36:53, 18.49s/it] {'loss': 3.4329, 'grad_norm': 0.06108776475396543, 'learning_rate': 0.1375, 'epoch': 0.02} + 2%|▏ | 11/520 [29:01<2:36:53, 18.49s/it] 2%|▏ | 12/520 [29:04<1:58:28, 13.99s/it] {'loss': 3.8533, 'grad_norm': 0.09524746551976003, 'learning_rate': 0.15000000000000002, 'epoch': 0.02} + 2%|▏ | 12/520 [29:04<1:58:28, 13.99s/it][2025-10-12 09:33:19,037] [WARNING] [stage3.py:2069:step] 1 pytorch allocator cache flushes since last step. this happens when there is high memory pressure and is detrimental to performance. if this is happening frequently consider adjusting settings to reduce memory consumption. If you are unable to make the cache flushes go away consider adding get_accelerator().empty_cache() calls in your training loop to ensure that all ranks flush their caches at the same time + 2%|▎ | 13/520 [29:09<1:33:35, 11.08s/it] {'loss': 2.7593, 'grad_norm': 0.018994188772617857, 'learning_rate': 0.1625, 'epoch': 0.03} + 2%|▎ | 13/520 [29:09<1:33:35, 11.08s/it] 3%|▎ | 14/520 [29:13<1:14:45, 8.86s/it] {'loss': 2.7902, 'grad_norm': 0.024867277424468596, 'learning_rate': 0.17500000000000002, 'epoch': 0.03} + 3%|▎ | 14/520 [29:13<1:14:45, 8.86s/it] 3%|▎ | 15/520 [29:16<1:01:33, 7.31s/it] {'loss': 3.1329, 'grad_norm': 0.06677190454399579, 'learning_rate': 0.1875, 'epoch': 0.03} + 3%|▎ | 15/520 [29:16<1:01:33, 7.31s/it] 3%|▎ | 16/520 [29:20<52:25, 6.24s/it] {'loss': 2.8128, 'grad_norm': 0.02774407238975918, 'learning_rate': 0.2, 'epoch': 0.03} + 3%|▎ | 16/520 [29:20<52:25, 6.24s/it] 3%|▎ | 17/520 [29:24<46:04, 5.50s/it] {'loss': 2.5394, 'grad_norm': 0.030983325360444406, 'learning_rate': 0.1999980572931538, 'epoch': 0.03} + 3%|▎ | 17/520 [29:24<46:04, 5.50s/it] 3%|▎ | 18/520 [29:28<41:35, 4.97s/it] {'loss': 2.3238, 'grad_norm': 0.04278543342772941, 'learning_rate': 0.19999222924809748, 'epoch': 0.03} + 3%|▎ | 18/520 [29:28<41:35, 4.97s/it] 4%|▎ | 19/520 [29:31<38:23, 4.60s/it] {'loss': 2.8243, 'grad_norm': 0.02618630100663896, 'learning_rate': 0.19998251609127465, 'epoch': 0.04} + 4%|▎ | 19/520 [29:31<38:23, 4.60s/it] 4%|▍ | 20/520 [29:35<36:09, 4.34s/it] {'loss': 2.0985, 'grad_norm': 0.016353114922794627, 'learning_rate': 0.19996891820008164, 'epoch': 0.04} + 4%|▍ | 20/520 [29:35<36:09, 4.34s/it] 4%|▍ | 21/520 [29:39<34:37, 4.16s/it] {'loss': 2.6487, 'grad_norm': 0.04625653288318181, 'learning_rate': 0.19995143610285276, 'epoch': 0.04} + 4%|▍ | 21/520 [29:39<34:37, 4.16s/it] 4%|▍ | 22/520 [29:43<33:30, 4.04s/it] {'loss': 2.1768, 'grad_norm': 0.016716490334120787, 'learning_rate': 0.19993007047883987, 'epoch': 0.04} + 4%|▍ | 22/520 [29:43<33:30, 4.04s/it] 4%|▍ | 23/520 [29:46<32:46, 3.96s/it] {'loss': 2.0555, 'grad_norm': 0.012352712390782186, 'learning_rate': 0.1999048221581858, 'epoch': 0.04} + 4%|▍ | 23/520 [29:46<32:46, 3.96s/it] 5%|▍ | 24/520 [29:50<32:06, 3.88s/it] {'loss': 2.3748, 'grad_norm': 0.011805395395677399, 'learning_rate': 0.19987569212189224, 'epoch': 0.05} + 5%|▍ | 24/520 [29:50<32:06, 3.88s/it] 5%|▍ | 25/520 [29:54<31:38, 3.84s/it] {'loss': 2.0517, 'grad_norm': 0.0149132103220536, 'learning_rate': 0.19984268150178167, 'epoch': 0.05} + 5%|▍ | 25/520 [29:54<31:38, 3.84s/it] 5%|▌ | 26/520 [29:57<31:21, 3.81s/it] {'loss': 1.9631, 'grad_norm': 0.013576815752012734, 'learning_rate': 0.1998057915804532, 'epoch': 0.05} + 5%|▌ | 26/520 [29:57<31:21, 3.81s/it] 5%|▌ | 27/520 [30:01<31:06, 3.79s/it] {'loss': 1.7527, 'grad_norm': 0.007534106161443889, 'learning_rate': 0.1997650237912329, 'epoch': 0.05} + 5%|▌ | 27/520 [30:01<31:06, 3.79s/it] 5%|▌ | 28/520 [30:05<30:57, 3.78s/it] {'loss': 1.7445, 'grad_norm': 0.007403739067752161, 'learning_rate': 0.199720379718118, 'epoch': 0.05} + 5%|▌ | 28/520 [30:05<30:57, 3.78s/it] 6%|▌ | 29/520 [30:09<30:53, 3.78s/it] {'loss': 1.7549, 'grad_norm': 0.008032636339836263, 'learning_rate': 0.19967186109571552, 'epoch': 0.06} + 6%|▌ | 29/520 [30:09<30:53, 3.78s/it] 6%|▌ | 30/520 [30:12<30:44, 3.76s/it] {'loss': 2.176, 'grad_norm': 0.009808225954532141, 'learning_rate': 0.19961946980917455, 'epoch': 0.06} + 6%|▌ | 30/520 [30:12<30:44, 3.76s/it] 6%|▌ | 31/520 [30:16<30:36, 3.75s/it] {'loss': 1.7256, 'grad_norm': 0.006474603856198807, 'learning_rate': 0.1995632078941134, 'epoch': 0.06} + 6%|▌ | 31/520 [30:16<30:36, 3.75s/it] 6%|▌ | 32/520 [30:20<30:28, 3.75s/it] {'loss': 2.2747, 'grad_norm': 0.015977760888736398, 'learning_rate': 0.19950307753654017, 'epoch': 0.06} + 6%|▌ | 32/520 [30:20<30:28, 3.75s/it] 6%|▋ | 33/520 [30:24<30:22, 3.74s/it] {'loss': 1.689, 'grad_norm': 0.008436844414268466, 'learning_rate': 0.19943908107276798, 'epoch': 0.06} + 6%|▋ | 33/520 [30:24<30:22, 3.74s/it] 7%|▋ | 34/520 [30:27<30:14, 3.73s/it] {'loss': 1.6713, 'grad_norm': 0.0067146610992935285, 'learning_rate': 0.19937122098932428, 'epoch': 0.07} + 7%|▋ | 34/520 [30:27<30:14, 3.73s/it] 7%|▋ | 35/520 [30:31<30:10, 3.73s/it] {'loss': 1.6535, 'grad_norm': 0.007805141423506956, 'learning_rate': 0.19929949992285395, 'epoch': 0.07} + 7%|▋ | 35/520 [30:31<30:10, 3.73s/it] 7%|▋ | 36/520 [30:35<30:09, 3.74s/it] {'loss': 1.7764, 'grad_norm': 0.006748972495783535, 'learning_rate': 0.19922392066001723, 'epoch': 0.07} + 7%|▋ | 36/520 [30:35<30:09, 3.74s/it] 7%|▋ | 37/520 [30:39<30:03, 3.73s/it] {'loss': 2.0255, 'grad_norm': 0.010496157933151267, 'learning_rate': 0.19914448613738106, 'epoch': 0.07} + 7%|▋ | 37/520 [30:39<30:03, 3.73s/it] 7%|▋ | 38/520 [30:42<30:00, 3.74s/it] {'loss': 1.8356, 'grad_norm': 0.0072817944428879375, 'learning_rate': 0.1990611994413053, 'epoch': 0.07} + 7%|▋ | 38/520 [30:42<30:00, 3.74s/it] 8%|▊ | 39/520 [30:46<29:57, 3.74s/it] {'loss': 1.6809, 'grad_norm': 0.006967458651543056, 'learning_rate': 0.19897406380782262, 'epoch': 0.07} + 8%|▊ | 39/520 [30:46<29:57, 3.74s/it] 8%|▊ | 40/520 [30:50<29:56, 3.74s/it] {'loss': 1.6683, 'grad_norm': 0.005442844443566999, 'learning_rate': 0.19888308262251286, 'epoch': 0.08} + 8%|▊ | 40/520 [30:50<29:56, 3.74s/it] 8%|▊ | 41/520 [30:54<30:02, 3.76s/it] {'loss': 1.6401, 'grad_norm': 0.0054021544604150465, 'learning_rate': 0.19878825942037148, 'epoch': 0.08} + 8%|▊ | 41/520 [30:54<30:02, 3.76s/it] 8%|▊ | 42/520 [30:57<29:53, 3.75s/it] {'loss': 1.6727, 'grad_norm': 0.006500567085433235, 'learning_rate': 0.19868959788567211, 'epoch': 0.08} + 8%|▊ | 42/520 [30:57<29:53, 3.75s/it] 8%|▊ | 43/520 [31:01<29:54, 3.76s/it] {'loss': 1.8041, 'grad_norm': 0.00721310959116978, 'learning_rate': 0.1985871018518236, 'epoch': 0.08} + 8%|▊ | 43/520 [31:01<29:54, 3.76s/it] 8%|▊ | 44/520 [31:05<29:48, 3.76s/it] {'loss': 1.945, 'grad_norm': 0.007940187018704458, 'learning_rate': 0.19848077530122082, 'epoch': 0.08} + 8%|▊ | 44/520 [31:05<29:48, 3.76s/it] 9%|▊ | 45/520 [31:09<29:26, 3.72s/it] {'loss': 1.637, 'grad_norm': 0.00621257019375351, 'learning_rate': 0.19837062236509015, 'epoch': 0.09} + 9%|▊ | 45/520 [31:09<29:26, 3.72s/it] 9%|▉ | 46/520 [31:12<29:12, 3.70s/it] {'loss': 1.9858, 'grad_norm': 0.006228443537586687, 'learning_rate': 0.19825664732332884, 'epoch': 0.09} + 9%|▉ | 46/520 [31:12<29:12, 3.70s/it] 9%|▉ | 47/520 [31:16<28:57, 3.67s/it] {'loss': 1.6117, 'grad_norm': 0.005515185670549935, 'learning_rate': 0.19813885460433878, 'epoch': 0.09} + 9%|▉ | 47/520 [31:16<28:57, 3.67s/it] 9%|▉ | 48/520 [31:19<28:42, 3.65s/it] {'loss': 1.6047, 'grad_norm': 0.005359564752180208, 'learning_rate': 0.19801724878485438, 'epoch': 0.09} + 9%|▉ | 48/520 [31:19<28:42, 3.65s/it] 9%|▉ | 49/520 [31:23<28:36, 3.64s/it] {'loss': 1.6104, 'grad_norm': 0.004826091819411638, 'learning_rate': 0.19789183458976486, 'epoch': 0.09} + 9%|▉ | 49/520 [31:23<28:36, 3.64s/it] 10%|▉ | 50/520 [31:27<28:32, 3.64s/it] {'loss': 1.5823, 'grad_norm': 0.0045233471804394885, 'learning_rate': 0.19776261689193048, 'epoch': 0.1} + 10%|▉ | 50/520 [31:27<28:32, 3.64s/it] 10%|▉ | 51/520 [31:30<28:31, 3.65s/it] {'loss': 1.5092, 'grad_norm': 0.0049472658931127575, 'learning_rate': 0.19762960071199334, 'epoch': 0.1} + 10%|▉ | 51/520 [31:30<28:31, 3.65s/it] 10%|█ | 52/520 [31:34<28:22, 3.64s/it] {'loss': 1.6487, 'grad_norm': 0.005809506155131544, 'learning_rate': 0.19749279121818236, 'epoch': 0.1} + 10%|█ | 52/520 [31:34<28:22, 3.64s/it] 10%|█ | 53/520 [31:38<28:18, 3.64s/it] {'loss': 1.646, 'grad_norm': 0.005044076582088536, 'learning_rate': 0.19735219372611235, 'epoch': 0.1} + 10%|█ | 53/520 [31:38<28:18, 3.64s/it] 10%|█ | 54/520 [31:41<28:13, 3.63s/it] {'loss': 1.5173, 'grad_norm': 0.005635381693219882, 'learning_rate': 0.19720781369857746, 'epoch': 0.1} + 10%|█ | 54/520 [31:41<28:13, 3.63s/it] 11%|█ | 55/520 [31:45<28:20, 3.66s/it] {'loss': 1.5226, 'grad_norm': 0.005958582990394733, 'learning_rate': 0.1970596567453391, 'epoch': 0.11} + 11%|█ | 55/520 [31:45<28:20, 3.66s/it] 11%|█ | 56/520 [31:49<28:42, 3.71s/it] {'loss': 1.673, 'grad_norm': 0.005182996261115935, 'learning_rate': 0.1969077286229078, 'epoch': 0.11} + 11%|█ | 56/520 [31:49<28:42, 3.71s/it] 11%|█ | 57/520 [31:52<28:29, 3.69s/it] {'loss': 1.5174, 'grad_norm': 0.007817855331913265, 'learning_rate': 0.19675203523431964, 'epoch': 0.11} + 11%|█ | 57/520 [31:52<28:29, 3.69s/it] 11%|█ | 58/520 [31:56<28:48, 3.74s/it] {'loss': 1.6668, 'grad_norm': 0.005889332931321159, 'learning_rate': 0.19659258262890683, 'epoch': 0.11} + 11%|█ | 58/520 [31:56<28:48, 3.74s/it] 11%|█▏ | 59/520 [32:00<29:08, 3.79s/it] {'loss': 1.6554, 'grad_norm': 0.007155219876727041, 'learning_rate': 0.19642937700206278, 'epoch': 0.11} + 11%|█▏ | 59/520 [32:00<29:08, 3.79s/it] 12%|█▏ | 60/520 [32:04<29:07, 3.80s/it] {'loss': 1.5873, 'grad_norm': 0.008216530849993405, 'learning_rate': 0.19626242469500121, 'epoch': 0.12} + 12%|█▏ | 60/520 [32:04<29:07, 3.80s/it] 12%|█▏ | 61/520 [32:08<28:44, 3.76s/it] {'loss': 1.7984, 'grad_norm': 0.009889415000365841, 'learning_rate': 0.19609173219450998, 'epoch': 0.12} + 12%|█▏ | 61/520 [32:08<28:44, 3.76s/it] 12%|█▏ | 62/520 [32:11<28:19, 3.71s/it] {'loss': 1.5633, 'grad_norm': 0.005243516279610177, 'learning_rate': 0.19591730613269878, 'epoch': 0.12} + 12%|█▏ | 62/520 [32:11<28:19, 3.71s/it] 12%|█▏ | 63/520 [32:15<28:02, 3.68s/it] {'loss': 1.5527, 'grad_norm': 0.005809869726931619, 'learning_rate': 0.19573915328674182, 'epoch': 0.12} + 12%|█▏ | 63/520 [32:15<28:02, 3.68s/it] 12%|█▏ | 64/520 [32:18<27:51, 3.67s/it] {'loss': 1.578, 'grad_norm': 0.00543749844687564, 'learning_rate': 0.1955572805786141, 'epoch': 0.12} + 12%|█▏ | 64/520 [32:18<27:51, 3.67s/it] 12%|█▎ | 65/520 [32:22<27:45, 3.66s/it] {'loss': 1.5754, 'grad_norm': 0.005465695481192723, 'learning_rate': 0.1953716950748227, 'epoch': 0.12} + 12%|█▎ | 65/520 [32:22<27:45, 3.66s/it] 13%|█▎ | 66/520 [32:26<27:35, 3.65s/it] {'loss': 1.5351, 'grad_norm': 0.006051278292362363, 'learning_rate': 0.19518240398613226, 'epoch': 0.13} + 13%|█▎ | 66/520 [32:26<27:35, 3.65s/it] 13%|█▎ | 67/520 [32:30<27:48, 3.68s/it] {'loss': 1.4061, 'grad_norm': 0.0053500482555356044, 'learning_rate': 0.1949894146672846, 'epoch': 0.13} + 13%|█▎ | 67/520 [32:30<27:48, 3.68s/it] 13%|█▎ | 68/520 [32:33<28:00, 3.72s/it] {'loss': 1.4633, 'grad_norm': 0.004673480745703422, 'learning_rate': 0.1947927346167132, 'epoch': 0.13} + 13%|█▎ | 68/520 [32:33<28:00, 3.72s/it] 13%|█▎ | 69/520 [32:37<28:10, 3.75s/it] {'loss': 1.4394, 'grad_norm': 0.0064212675478031245, 'learning_rate': 0.1945923714762516, 'epoch': 0.13} + 13%|█▎ | 69/520 [32:37<28:10, 3.75s/it] 13%|█▎ | 70/520 [32:41<28:15, 3.77s/it] {'loss': 1.5151, 'grad_norm': 0.006098534243232547, 'learning_rate': 0.19438833303083677, 'epoch': 0.13} + 13%|█▎ | 70/520 [32:41<28:15, 3.77s/it] 14%|█▎ | 71/520 [32:45<28:16, 3.78s/it] {'loss': 1.4107, 'grad_norm': 0.004501684921119413, 'learning_rate': 0.19418062720820636, 'epoch': 0.14} + 14%|█▎ | 71/520 [32:45<28:16, 3.78s/it] 14%|█▍ | 72/520 [32:49<28:12, 3.78s/it] {'loss': 1.5753, 'grad_norm': 0.005192203083219257, 'learning_rate': 0.19396926207859086, 'epoch': 0.14} + 14%|█▍ | 72/520 [32:49<28:12, 3.78s/it] 14%|█▍ | 73/520 [32:52<28:21, 3.81s/it] {'loss': 1.3936, 'grad_norm': 0.004156899495525804, 'learning_rate': 0.19375424585439993, 'epoch': 0.14} + 14%|█▍ | 73/520 [32:52<28:21, 3.81s/it] 14%|█▍ | 74/520 [32:56<28:17, 3.81s/it] {'loss': 1.5133, 'grad_norm': 0.005477490621331273, 'learning_rate': 0.1935355868899034, 'epoch': 0.14} + 14%|█▍ | 74/520 [32:56<28:17, 3.81s/it] 14%|█▍ | 75/520 [33:00<28:12, 3.80s/it] {'loss': 1.4017, 'grad_norm': 0.004074280192382839, 'learning_rate': 0.19331329368090666, 'epoch': 0.14} + 14%|█▍ | 75/520 [33:00<28:12, 3.80s/it] 15%|█▍ | 76/520 [33:04<28:13, 3.81s/it] {'loss': 1.8194, 'grad_norm': 0.015809994817813925, 'learning_rate': 0.19308737486442043, 'epoch': 0.15} + 15%|█▍ | 76/520 [33:04<28:13, 3.81s/it] 15%|█▍ | 77/520 [33:08<28:10, 3.82s/it] {'loss': 1.3433, 'grad_norm': 0.005021435917378024, 'learning_rate': 0.19285783921832536, 'epoch': 0.15} + 15%|█▍ | 77/520 [33:08<28:10, 3.82s/it] 15%|█▌ | 78/520 [33:12<28:10, 3.82s/it] {'loss': 1.4694, 'grad_norm': 0.004326078591477138, 'learning_rate': 0.19262469566103088, 'epoch': 0.15} + 15%|█▌ | 78/520 [33:12<28:10, 3.82s/it] 15%|█▌ | 79/520 [33:15<27:56, 3.80s/it] {'loss': 1.448, 'grad_norm': 0.004207339698926696, 'learning_rate': 0.19238795325112867, 'epoch': 0.15} + 15%|█▌ | 79/520 [33:15<27:56, 3.80s/it] 15%|█▌ | 80/520 [33:19<27:31, 3.75s/it] {'loss': 1.8686, 'grad_norm': 0.006953138811364748, 'learning_rate': 0.19214762118704076, 'epoch': 0.15} + 15%|█▌ | 80/520 [33:19<27:31, 3.75s/it] 16%|█▌ | 81/520 [33:23<27:12, 3.72s/it] {'loss': 1.6189, 'grad_norm': 0.005474950644871538, 'learning_rate': 0.19190370880666208, 'epoch': 0.16} + 16%|█▌ | 81/520 [33:23<27:12, 3.72s/it] 16%|█▌ | 82/520 [33:26<27:01, 3.70s/it] {'loss': 1.5134, 'grad_norm': 0.004289174076829843, 'learning_rate': 0.19165622558699763, 'epoch': 0.16} + 16%|█▌ | 82/520 [33:26<27:01, 3.70s/it] 16%|█▌ | 83/520 [33:30<26:43, 3.67s/it] {'loss': 1.5637, 'grad_norm': 0.004660154555669566, 'learning_rate': 0.19140518114379435, 'epoch': 0.16} + 16%|█▌ | 83/520 [33:30<26:43, 3.67s/it] 16%|█▌ | 84/520 [33:33<26:34, 3.66s/it] {'loss': 1.5498, 'grad_norm': 0.005372570656706105, 'learning_rate': 0.19115058523116735, 'epoch': 0.16} + 16%|█▌ | 84/520 [33:33<26:34, 3.66s/it] 16%|█▋ | 85/520 [33:37<26:30, 3.66s/it] {'loss': 1.558, 'grad_norm': 0.0045887856557548625, 'learning_rate': 0.1908924477412211, 'epoch': 0.16} + 16%|█▋ | 85/520 [33:37<26:30, 3.66s/it] 17%|█▋ | 86/520 [33:41<26:21, 3.64s/it] {'loss': 1.6029, 'grad_norm': 0.005146793631463206, 'learning_rate': 0.19063077870366502, 'epoch': 0.17} + 17%|█▋ | 86/520 [33:41<26:21, 3.64s/it] 17%|█▋ | 87/520 [33:44<26:16, 3.64s/it] {'loss': 1.7834, 'grad_norm': 0.009997900849501822, 'learning_rate': 0.1903655882854237, 'epoch': 0.17} + 17%|█▋ | 87/520 [33:44<26:16, 3.64s/it] 17%|█▋ | 88/520 [33:48<26:13, 3.64s/it] {'loss': 1.8005, 'grad_norm': 0.007587084971918682, 'learning_rate': 0.19009688679024192, 'epoch': 0.17} + 17%|█▋ | 88/520 [33:48<26:13, 3.64s/it] 17%|█▋ | 89/520 [33:52<26:09, 3.64s/it] {'loss': 1.5227, 'grad_norm': 0.004323120017181859, 'learning_rate': 0.18982468465828442, 'epoch': 0.17} + 17%|█▋ | 89/520 [33:52<26:09, 3.64s/it] 17%|█▋ | 90/520 [33:55<26:06, 3.64s/it] {'loss': 1.4488, 'grad_norm': 0.004391023593828015, 'learning_rate': 0.1895489924657301, 'epoch': 0.17} + 17%|█▋ | 90/520 [33:55<26:06, 3.64s/it] 18%|█▊ | 91/520 [33:59<26:00, 3.64s/it] {'loss': 1.5361, 'grad_norm': 0.0041776331635069285, 'learning_rate': 0.18926982092436118, 'epoch': 0.17} + 18%|█▊ | 91/520 [33:59<26:00, 3.64s/it] 18%|█▊ | 92/520 [34:02<25:54, 3.63s/it] {'loss': 1.4655, 'grad_norm': 0.004453747626909981, 'learning_rate': 0.18898718088114688, 'epoch': 0.18} + 18%|█▊ | 92/520 [34:02<25:54, 3.63s/it] 18%|█▊ | 93/520 [34:06<25:54, 3.64s/it] {'loss': 1.4627, 'grad_norm': 0.004352247582734059, 'learning_rate': 0.18870108331782218, 'epoch': 0.18} + 18%|█▊ | 93/520 [34:06<25:54, 3.64s/it] 18%|█▊ | 94/520 [34:10<25:46, 3.63s/it] {'loss': 1.5762, 'grad_norm': 0.005131460267356873, 'learning_rate': 0.18841153935046098, 'epoch': 0.18} + 18%|█▊ | 94/520 [34:10<25:46, 3.63s/it] 18%|█▊ | 95/520 [34:13<25:44, 3.63s/it] {'loss': 1.4543, 'grad_norm': 0.005340419776514996, 'learning_rate': 0.18811856022904425, 'epoch': 0.18} + 18%|█▊ | 95/520 [34:13<25:44, 3.63s/it] 18%|█▊ | 96/520 [34:17<25:41, 3.64s/it] {'loss': 1.4553, 'grad_norm': 0.0037241883958914752, 'learning_rate': 0.18782215733702287, 'epoch': 0.18} + 18%|█▊ | 96/520 [34:17<25:41, 3.64s/it] 19%|█▊ | 97/520 [34:21<25:36, 3.63s/it] {'loss': 1.4206, 'grad_norm': 0.0050320501437953755, 'learning_rate': 0.18752234219087538, 'epoch': 0.19} + 19%|█▊ | 97/520 [34:21<25:36, 3.63s/it] 19%|█▉ | 98/520 [34:24<25:34, 3.64s/it] {'loss': 1.4197, 'grad_norm': 0.0037354523983947887, 'learning_rate': 0.18721912643966054, 'epoch': 0.19} + 19%|█▉ | 98/520 [34:24<25:34, 3.64s/it] 19%|█▉ | 99/520 [34:28<25:35, 3.65s/it] {'loss': 1.4449, 'grad_norm': 0.004254859059317896, 'learning_rate': 0.18691252186456464, 'epoch': 0.19} + 19%|█▉ | 99/520 [34:28<25:35, 3.65s/it] 19%|█▉ | 100/520 [34:32<25:27, 3.64s/it] {'loss': 1.612, 'grad_norm': 0.005814642827458626, 'learning_rate': 0.1866025403784439, 'epoch': 0.19} + 19%|█▉ | 100/520 [34:32<25:27, 3.64s/it] 19%|█▉ | 101/520 [34:35<25:22, 3.63s/it] {'loss': 1.4342, 'grad_norm': 0.004323870968467543, 'learning_rate': 0.18628919402536132, 'epoch': 0.19} + 19%|█▉ | 101/520 [34:35<25:22, 3.63s/it] 20%|█▉ | 102/520 [34:39<25:16, 3.63s/it] {'loss': 1.4501, 'grad_norm': 0.004547149655832278, 'learning_rate': 0.18597249498011903, 'epoch': 0.2} + 20%|█▉ | 102/520 [34:39<25:16, 3.63s/it] 20%|█▉ | 103/520 [34:42<25:12, 3.63s/it] {'loss': 1.3686, 'grad_norm': 0.00362613952282307, 'learning_rate': 0.18565245554778517, 'epoch': 0.2} + 20%|█▉ | 103/520 [34:42<25:12, 3.63s/it] 20%|██ | 104/520 [34:46<25:10, 3.63s/it] {'loss': 1.4533, 'grad_norm': 0.004545028742829835, 'learning_rate': 0.18532908816321558, 'epoch': 0.2} + 20%|██ | 104/520 [34:46<25:10, 3.63s/it] 20%|██ | 105/520 [34:50<25:05, 3.63s/it] {'loss': 1.4437, 'grad_norm': 0.003914610734405442, 'learning_rate': 0.18500240539057092, 'epoch': 0.2} + 20%|██ | 105/520 [34:50<25:05, 3.63s/it] 20%|██ | 106/520 [34:53<25:18, 3.67s/it] {'loss': 1.5905, 'grad_norm': 0.004539814696239539, 'learning_rate': 0.18467241992282843, 'epoch': 0.2} + 20%|██ | 106/520 [34:53<25:18, 3.67s/it] 21%|██ | 107/520 [34:57<25:37, 3.72s/it] {'loss': 1.598, 'grad_norm': 0.004941940583234571, 'learning_rate': 0.18433914458128858, 'epoch': 0.21} + 21%|██ | 107/520 [34:57<25:37, 3.72s/it] 21%|██ | 108/520 [35:01<25:51, 3.76s/it] {'loss': 1.4064, 'grad_norm': 0.004609390420194848, 'learning_rate': 0.18400259231507718, 'epoch': 0.21} + 21%|██ | 108/520 [35:01<25:51, 3.76s/it] 21%|██ | 109/520 [35:05<25:55, 3.79s/it] {'loss': 1.5704, 'grad_norm': 0.007293457961964297, 'learning_rate': 0.18366277620064198, 'epoch': 0.21} + 21%|██ | 109/520 [35:05<25:55, 3.79s/it] 21%|██ | 110/520 [35:09<25:58, 3.80s/it] {'loss': 1.5854, 'grad_norm': 0.004458644065710611, 'learning_rate': 0.1833197094412449, 'epoch': 0.21} + 21%|██ | 110/520 [35:09<25:58, 3.80s/it] 21%|██▏ | 111/520 [35:13<26:01, 3.82s/it] {'loss': 1.602, 'grad_norm': 0.004724924196807223, 'learning_rate': 0.18297340536644877, 'epoch': 0.21} + 21%|██▏ | 111/520 [35:13<26:01, 3.82s/it] 22%|██▏ | 112/520 [35:17<26:03, 3.83s/it] {'loss': 1.4758, 'grad_norm': 0.004114388056751404, 'learning_rate': 0.1826238774315995, 'epoch': 0.22} + 22%|██▏ | 112/520 [35:17<26:03, 3.83s/it] 22%|██▏ | 113/520 [35:20<26:00, 3.83s/it] {'loss': 1.3444, 'grad_norm': 0.0038649007119221077, 'learning_rate': 0.18227113921730334, 'epoch': 0.22} + 22%|██▏ | 113/520 [35:20<26:00, 3.83s/it] 22%|██▏ | 114/520 [35:24<25:58, 3.84s/it] {'loss': 1.4543, 'grad_norm': 0.004328350661631083, 'learning_rate': 0.1819152044288992, 'epoch': 0.22} + 22%|██▏ | 114/520 [35:24<25:58, 3.84s/it] 22%|██▏ | 115/520 [35:28<25:37, 3.80s/it] {'loss': 1.5757, 'grad_norm': 0.004143067234394189, 'learning_rate': 0.18155608689592603, 'epoch': 0.22} + 22%|██▏ | 115/520 [35:28<25:37, 3.80s/it] 22%|██▏ | 116/520 [35:32<25:13, 3.75s/it] {'loss': 1.5671, 'grad_norm': 0.003872392142894355, 'learning_rate': 0.18119380057158568, 'epoch': 0.22} + 22%|██▏ | 116/520 [35:32<25:13, 3.75s/it] 22%|██▎ | 117/520 [35:35<24:56, 3.71s/it] {'loss': 1.5552, 'grad_norm': 0.004339211808765579, 'learning_rate': 0.18082835953220056, 'epoch': 0.23} + 22%|██▎ | 117/520 [35:35<24:56, 3.71s/it] 23%|██▎ | 118/520 [35:39<24:41, 3.69s/it] {'loss': 1.4327, 'grad_norm': 0.0038004873172320435, 'learning_rate': 0.18045977797666685, 'epoch': 0.23} + 23%|██▎ | 118/520 [35:39<24:41, 3.69s/it] 23%|██▎ | 119/520 [35:43<24:32, 3.67s/it] {'loss': 1.3732, 'grad_norm': 0.0039453891253911815, 'learning_rate': 0.1800880702259028, 'epoch': 0.23} + 23%|██▎ | 119/520 [35:43<24:32, 3.67s/it] 23%|██▎ | 120/520 [35:46<24:25, 3.66s/it] {'loss': 1.4147, 'grad_norm': 0.004676735935140681, 'learning_rate': 0.17971325072229227, 'epoch': 0.23} + 23%|██▎ | 120/520 [35:46<24:25, 3.66s/it] 23%|██▎ | 121/520 [35:50<24:17, 3.65s/it] {'loss': 1.4632, 'grad_norm': 0.004692064574680109, 'learning_rate': 0.17933533402912352, 'epoch': 0.23} + 23%|██▎ | 121/520 [35:50<24:17, 3.65s/it] 23%|██▎ | 122/520 [35:53<24:16, 3.66s/it] {'loss': 1.3405, 'grad_norm': 0.0038033949553781528, 'learning_rate': 0.17895433483002354, 'epoch': 0.23} + 23%|██▎ | 122/520 [35:53<24:16, 3.66s/it] 24%|██▎ | 123/520 [35:57<24:09, 3.65s/it] {'loss': 1.6291, 'grad_norm': 0.005738748135194191, 'learning_rate': 0.17857026792838737, 'epoch': 0.24} + 24%|██▎ | 123/520 [35:57<24:09, 3.65s/it] 24%|██▍ | 124/520 [36:01<24:04, 3.65s/it] {'loss': 1.4382, 'grad_norm': 0.004561914987719681, 'learning_rate': 0.178183148246803, 'epoch': 0.24} + 24%|██▍ | 124/520 [36:01<24:04, 3.65s/it] 24%|██▍ | 125/520 [36:04<24:05, 3.66s/it] {'loss': 1.4223, 'grad_norm': 0.004220373152262201, 'learning_rate': 0.1777929908264715, 'epoch': 0.24} + 24%|██▍ | 125/520 [36:04<24:05, 3.66s/it] 24%|██▍ | 126/520 [36:09<25:18, 3.85s/it] {'loss': 1.5214, 'grad_norm': 0.003726154026520092, 'learning_rate': 0.17739981082662276, 'epoch': 0.24} + 24%|██▍ | 126/520 [36:09<25:18, 3.85s/it] 24%|██▍ | 127/520 [36:12<24:47, 3.79s/it] {'loss': 1.3989, 'grad_norm': 0.004936214846085809, 'learning_rate': 0.1770036235239263, 'epoch': 0.24} + 24%|██▍ | 127/520 [36:12<24:47, 3.79s/it] 25%|██▍ | 128/520 [36:16<24:25, 3.74s/it] {'loss': 1.4481, 'grad_norm': 0.00426717158520045, 'learning_rate': 0.1766044443118978, 'epoch': 0.25} + 25%|██▍ | 128/520 [36:16<24:25, 3.74s/it] 25%|██▍ | 129/520 [36:20<24:11, 3.71s/it] {'loss': 1.3643, 'grad_norm': 0.003711899704576205, 'learning_rate': 0.17620228870030108, 'epoch': 0.25} + 25%|██▍ | 129/520 [36:20<24:11, 3.71s/it] 25%|██▌ | 130/520 [36:23<24:01, 3.70s/it] {'loss': 1.4324, 'grad_norm': 0.0039741728606241125, 'learning_rate': 0.1757971723145453, 'epoch': 0.25} + 25%|██▌ | 130/520 [36:23<24:01, 3.70s/it] 25%|██▌ | 131/520 [36:27<23:53, 3.68s/it] {'loss': 1.494, 'grad_norm': 0.004550308079625895, 'learning_rate': 0.175389110895078, 'epoch': 0.25} + 25%|██▌ | 131/520 [36:27<23:53, 3.68s/it] 25%|██▌ | 132/520 [36:31<23:42, 3.67s/it] {'loss': 1.4781, 'grad_norm': 0.004084481782383805, 'learning_rate': 0.17497812029677343, 'epoch': 0.25} + 25%|██▌ | 132/520 [36:31<23:42, 3.67s/it] 26%|██▌ | 133/520 [36:34<23:36, 3.66s/it] {'loss': 1.3854, 'grad_norm': 0.004129854232440035, 'learning_rate': 0.17456421648831655, 'epoch': 0.26} + 26%|██▌ | 133/520 [36:34<23:36, 3.66s/it] 26%|██▌ | 134/520 [36:38<23:31, 3.66s/it] {'loss': 1.4653, 'grad_norm': 0.0042427687581316, 'learning_rate': 0.17414741555158267, 'epoch': 0.26} + 26%|██▌ | 134/520 [36:38<23:31, 3.66s/it] 26%|██▌ | 135/520 [36:41<23:23, 3.64s/it] {'loss': 1.539, 'grad_norm': 0.004015439063691244, 'learning_rate': 0.1737277336810124, 'epoch': 0.26} + 26%|██▌ | 135/520 [36:41<23:23, 3.64s/it] 26%|██▌ | 136/520 [36:45<23:15, 3.63s/it] {'loss': 1.4586, 'grad_norm': 0.004034671335344313, 'learning_rate': 0.17330518718298263, 'epoch': 0.26} + 26%|██▌ | 136/520 [36:45<23:15, 3.63s/it] 26%|██▋ | 137/520 [36:49<23:13, 3.64s/it] {'loss': 1.3866, 'grad_norm': 0.004474186859136332, 'learning_rate': 0.17287979247517285, 'epoch': 0.26} + 26%|██▋ | 137/520 [36:49<23:13, 3.64s/it] 27%|██▋ | 138/520 [36:52<23:07, 3.63s/it] {'loss': 1.3874, 'grad_norm': 0.0037354510236017233, 'learning_rate': 0.17245156608592727, 'epoch': 0.27} + 27%|██▋ | 138/520 [36:52<23:07, 3.63s/it] 27%|██▋ | 139/520 [36:56<23:08, 3.64s/it] {'loss': 1.3741, 'grad_norm': 0.003911284060417325, 'learning_rate': 0.17202052465361267, 'epoch': 0.27} + 27%|██▋ | 139/520 [36:56<23:08, 3.64s/it] 27%|██▋ | 140/520 [37:00<23:01, 3.64s/it] {'loss': 1.5564, 'grad_norm': 0.005361978978746797, 'learning_rate': 0.17158668492597184, 'epoch': 0.27} + 27%|██▋ | 140/520 [37:00<23:01, 3.64s/it] 27%|██▋ | 141/520 [37:03<22:59, 3.64s/it] {'loss': 1.5078, 'grad_norm': 0.0038325362590747505, 'learning_rate': 0.17115006375947303, 'epoch': 0.27} + 27%|██▋ | 141/520 [37:03<22:59, 3.64s/it] 27%|██▋ | 142/520 [37:07<22:59, 3.65s/it] {'loss': 1.5754, 'grad_norm': 0.004198567404826, 'learning_rate': 0.17071067811865476, 'epoch': 0.27} + 27%|██▋ | 142/520 [37:07<22:59, 3.65s/it] 28%|██▊ | 143/520 [37:11<22:54, 3.65s/it] {'loss': 1.4306, 'grad_norm': 0.004326515295842536, 'learning_rate': 0.17026854507546693, 'epoch': 0.28} + 28%|██▊ | 143/520 [37:11<22:54, 3.65s/it] 28%|██▊ | 144/520 [37:14<22:59, 3.67s/it] {'loss': 1.383, 'grad_norm': 0.00393988378564092, 'learning_rate': 0.1698236818086073, 'epoch': 0.28} + 28%|██▊ | 144/520 [37:14<22:59, 3.67s/it] 28%|██▊ | 145/520 [37:18<23:15, 3.72s/it] {'loss': 1.3004, 'grad_norm': 0.003343283669471192, 'learning_rate': 0.16937610560285418, 'epoch': 0.28} + 28%|██▊ | 145/520 [37:18<23:15, 3.72s/it] 28%|██▊ | 146/520 [37:22<23:27, 3.76s/it] {'loss': 1.6136, 'grad_norm': 0.004467271473955934, 'learning_rate': 0.1689258338483947, 'epoch': 0.28} + 28%|██▊ | 146/520 [37:22<23:27, 3.76s/it] 28%|██▊ | 147/520 [37:26<23:24, 3.76s/it] {'loss': 1.3488, 'grad_norm': 0.0040523519254499546, 'learning_rate': 0.16847288404014937, 'epoch': 0.28} + 28%|██▊ | 147/520 [37:26<23:24, 3.76s/it] 28%|██▊ | 148/520 [37:29<23:03, 3.72s/it] {'loss': 1.3925, 'grad_norm': 0.0036906001009539445, 'learning_rate': 0.16801727377709194, 'epoch': 0.28} + 28%|██▊ | 148/520 [37:29<23:03, 3.72s/it] 29%|██▊ | 149/520 [37:33<22:53, 3.70s/it] {'loss': 1.3328, 'grad_norm': 0.0036557138260473704, 'learning_rate': 0.16755902076156604, 'epoch': 0.29} + 29%|██▊ | 149/520 [37:33<22:53, 3.70s/it] 29%|██▉ | 150/520 [37:37<22:39, 3.67s/it] {'loss': 1.5651, 'grad_norm': 0.0042758094057677365, 'learning_rate': 0.16709814279859703, 'epoch': 0.29} + 29%|██▉ | 150/520 [37:37<22:39, 3.67s/it] 29%|██▉ | 151/520 [37:40<22:42, 3.69s/it] {'loss': 1.3696, 'grad_norm': 0.0037852107600247703, 'learning_rate': 0.1666346577952004, 'epoch': 0.29} + 29%|██▉ | 151/520 [37:40<22:42, 3.69s/it] 29%|██▉ | 152/520 [37:44<22:35, 3.68s/it] {'loss': 1.3483, 'grad_norm': 0.003981548957084503, 'learning_rate': 0.16616858375968596, 'epoch': 0.29} + 29%|██▉ | 152/520 [37:44<22:35, 3.68s/it] 29%|██▉ | 153/520 [37:48<22:23, 3.66s/it] {'loss': 1.3883, 'grad_norm': 0.004193547240367004, 'learning_rate': 0.16569993880095807, 'epoch': 0.29} + 29%|██▉ | 153/520 [37:48<22:23, 3.66s/it] 30%|██▉ | 154/520 [37:51<22:15, 3.65s/it] {'loss': 1.462, 'grad_norm': 0.0035857671285553566, 'learning_rate': 0.16522874112781213, 'epoch': 0.3} + 30%|██▉ | 154/520 [37:51<22:15, 3.65s/it] 30%|██▉ | 155/520 [37:55<22:11, 3.65s/it] {'loss': 1.3708, 'grad_norm': 0.0040193835577251235, 'learning_rate': 0.16475500904822704, 'epoch': 0.3} + 30%|██▉ | 155/520 [37:55<22:11, 3.65s/it] 30%|███ | 156/520 [37:59<22:11, 3.66s/it] {'loss': 1.4005, 'grad_norm': 0.004067652815510889, 'learning_rate': 0.16427876096865393, 'epoch': 0.3} + 30%|███ | 156/520 [37:59<22:11, 3.66s/it] 30%|███ | 157/520 [38:02<22:07, 3.66s/it] {'loss': 1.6429, 'grad_norm': 0.004364624575199187, 'learning_rate': 0.16380001539330089, 'epoch': 0.3} + 30%|███ | 157/520 [38:02<22:07, 3.66s/it] 30%|███ | 158/520 [38:06<22:01, 3.65s/it] {'loss': 1.379, 'grad_norm': 0.005011709827033225, 'learning_rate': 0.163318790923414, 'epoch': 0.3} + 30%|███ | 158/520 [38:06<22:01, 3.65s/it] 31%|███ | 159/520 [38:10<21:53, 3.64s/it] {'loss': 1.4084, 'grad_norm': 0.0035882057295059467, 'learning_rate': 0.16283510625655473, 'epoch': 0.31} + 31%|███ | 159/520 [38:10<21:53, 3.64s/it] 31%|███ | 160/520 [38:13<21:49, 3.64s/it] {'loss': 1.4436, 'grad_norm': 0.0039385807581538085, 'learning_rate': 0.16234898018587338, 'epoch': 0.31} + 31%|███ | 160/520 [38:13<21:49, 3.64s/it] 31%|███ | 161/520 [38:17<21:44, 3.63s/it] {'loss': 1.4193, 'grad_norm': 0.004011406719342968, 'learning_rate': 0.16186043159937882, 'epoch': 0.31} + 31%|███ | 161/520 [38:17<21:44, 3.63s/it] 31%|███ | 162/520 [38:20<21:42, 3.64s/it] {'loss': 1.5312, 'grad_norm': 0.0042701577735882994, 'learning_rate': 0.16136947947920477, 'epoch': 0.31} + 31%|███ | 162/520 [38:20<21:42, 3.64s/it] 31%|███▏ | 163/520 [38:24<21:42, 3.65s/it] {'loss': 1.2934, 'grad_norm': 0.004519990457122133, 'learning_rate': 0.16087614290087207, 'epoch': 0.31} + 31%|███▏ | 163/520 [38:24<21:42, 3.65s/it] 32%|███▏ | 164/520 [38:28<21:39, 3.65s/it] {'loss': 1.2577, 'grad_norm': 0.0035371416321396437, 'learning_rate': 0.16038044103254775, 'epoch': 0.32} + 32%|███▏ | 164/520 [38:28<21:39, 3.65s/it] 32%|███▏ | 165/520 [38:31<21:34, 3.65s/it] {'loss': 1.4122, 'grad_norm': 0.0038722786177289637, 'learning_rate': 0.15988239313430005, 'epoch': 0.32} + 32%|███▏ | 165/520 [38:31<21:34, 3.65s/it] 32%|███▏ | 166/520 [38:35<21:29, 3.64s/it] {'loss': 1.3982, 'grad_norm': 0.004309284563937565, 'learning_rate': 0.15938201855735015, 'epoch': 0.32} + 32%|███▏ | 166/520 [38:35<21:29, 3.64s/it] 32%|███▏ | 167/520 [38:39<21:26, 3.65s/it] {'loss': 1.3901, 'grad_norm': 0.004353333113814727, 'learning_rate': 0.15887933674332047, 'epoch': 0.32} + 32%|███▏ | 167/520 [38:39<21:26, 3.65s/it] 32%|███▏ | 168/520 [38:42<21:21, 3.64s/it] {'loss': 1.3288, 'grad_norm': 0.003799055776435206, 'learning_rate': 0.158374367223479, 'epoch': 0.32} + 32%|███▏ | 168/520 [38:42<21:21, 3.64s/it] 32%|███▎ | 169/520 [38:46<21:22, 3.65s/it] {'loss': 1.4035, 'grad_norm': 0.003895575322203202, 'learning_rate': 0.1578671296179806, 'epoch': 0.33} + 32%|███▎ | 169/520 [38:46<21:22, 3.65s/it] 33%|███▎ | 170/520 [38:50<21:17, 3.65s/it] {'loss': 1.4402, 'grad_norm': 0.004505407479909258, 'learning_rate': 0.15735764363510463, 'epoch': 0.33} + 33%|███▎ | 170/520 [38:50<21:17, 3.65s/it] 33%|███▎ | 171/520 [38:53<21:13, 3.65s/it] {'loss': 1.3357, 'grad_norm': 0.004126025212930989, 'learning_rate': 0.15684592907048925, 'epoch': 0.33} + 33%|███▎ | 171/520 [38:53<21:13, 3.65s/it] 33%|███▎ | 172/520 [38:57<21:12, 3.66s/it] {'loss': 1.4056, 'grad_norm': 0.004544078957934825, 'learning_rate': 0.1563320058063622, 'epoch': 0.33} + 33%|███▎ | 172/520 [38:57<21:12, 3.66s/it] 33%|███▎ | 173/520 [39:01<21:03, 3.64s/it] {'loss': 1.3389, 'grad_norm': 0.003962735344338306, 'learning_rate': 0.15581589381076844, 'epoch': 0.33} + 33%|███▎ | 173/520 [39:01<21:03, 3.64s/it] 33%|███▎ | 174/520 [39:04<21:02, 3.65s/it] {'loss': 1.4087, 'grad_norm': 0.004796440651973641, 'learning_rate': 0.15529761313679394, 'epoch': 0.33} + 33%|███▎ | 174/520 [39:04<21:02, 3.65s/it] 34%|███▎ | 175/520 [39:08<20:56, 3.64s/it] {'loss': 1.3075, 'grad_norm': 0.0036174104496714604, 'learning_rate': 0.15477718392178716, 'epoch': 0.34} + 34%|███▎ | 175/520 [39:08<20:56, 3.64s/it] 34%|███▍ | 176/520 [39:12<20:51, 3.64s/it] {'loss': 1.5413, 'grad_norm': 0.004257210946187459, 'learning_rate': 0.15425462638657594, 'epoch': 0.34} + 34%|███▍ | 176/520 [39:12<20:51, 3.64s/it] 34%|███▍ | 177/520 [39:15<20:45, 3.63s/it] {'loss': 1.3979, 'grad_norm': 0.0044645861814767915, 'learning_rate': 0.1537299608346824, 'epoch': 0.34} + 34%|███▍ | 177/520 [39:15<20:45, 3.63s/it] 34%|███▍ | 178/520 [39:19<20:38, 3.62s/it] {'loss': 1.387, 'grad_norm': 0.004308219799903836, 'learning_rate': 0.15320320765153367, 'epoch': 0.34} + 34%|███▍ | 178/520 [39:19<20:38, 3.62s/it] 34%|███▍ | 179/520 [39:22<20:36, 3.63s/it] {'loss': 1.4612, 'grad_norm': 0.0035989027427931817, 'learning_rate': 0.15267438730367008, 'epoch': 0.34} + 34%|███▍ | 179/520 [39:22<20:36, 3.63s/it] 35%|███▍ | 180/520 [39:26<20:30, 3.62s/it] {'loss': 1.3849, 'grad_norm': 0.004411894008909646, 'learning_rate': 0.1521435203379498, 'epoch': 0.35} + 35%|███▍ | 180/520 [39:26<20:30, 3.62s/it] 35%|███▍ | 181/520 [39:30<20:26, 3.62s/it] {'loss': 1.3402, 'grad_norm': 0.003818786805518146, 'learning_rate': 0.15161062738075068, 'epoch': 0.35} + 35%|███▍ | 181/520 [39:30<20:26, 3.62s/it] 35%|███▌ | 182/520 [39:33<20:22, 3.62s/it] {'loss': 1.3641, 'grad_norm': 0.0036686668172155714, 'learning_rate': 0.1510757291371686, 'epoch': 0.35} + 35%|███▌ | 182/520 [39:33<20:22, 3.62s/it] 35%|███▌ | 183/520 [39:37<20:19, 3.62s/it] {'loss': 1.3931, 'grad_norm': 0.00410702347935001, 'learning_rate': 0.1505388463902131, 'epoch': 0.35} + 35%|███▌ | 183/520 [39:37<20:19, 3.62s/it] 35%|███▌ | 184/520 [39:40<20:14, 3.61s/it] {'loss': 1.3033, 'grad_norm': 0.003872409683741245, 'learning_rate': 0.15000000000000002, 'epoch': 0.35} + 35%|███▌ | 184/520 [39:40<20:14, 3.61s/it] 36%|███▌ | 185/520 [39:44<20:11, 3.62s/it] {'loss': 1.489, 'grad_norm': 0.003718134678036352, 'learning_rate': 0.14945921090294076, 'epoch': 0.36} + 36%|███▌ | 185/520 [39:44<20:11, 3.62s/it] 36%|███▌ | 186/520 [39:48<20:06, 3.61s/it] {'loss': 1.3252, 'grad_norm': 0.003687598728595689, 'learning_rate': 0.14891650011092894, 'epoch': 0.36} + 36%|███▌ | 186/520 [39:48<20:06, 3.61s/it] 36%|███▌ | 187/520 [39:51<20:00, 3.60s/it] {'loss': 1.3336, 'grad_norm': 0.0044003913858696835, 'learning_rate': 0.14837188871052398, 'epoch': 0.36} + 36%|███▌ | 187/520 [39:51<20:00, 3.60s/it] 36%|███▌ | 188/520 [39:55<19:58, 3.61s/it] {'loss': 1.4115, 'grad_norm': 0.003919351784728416, 'learning_rate': 0.14782539786213184, 'epoch': 0.36} + 36%|███▌ | 188/520 [39:55<19:58, 3.61s/it] 36%|███▋ | 189/520 [39:58<19:56, 3.62s/it] {'loss': 1.4267, 'grad_norm': 0.0035177074412893685, 'learning_rate': 0.1472770487991827, 'epoch': 0.36} + 36%|███▋ | 189/520 [39:58<19:56, 3.62s/it] 37%|███▋ | 190/520 [40:02<19:54, 3.62s/it] {'loss': 1.3423, 'grad_norm': 0.003976510694228002, 'learning_rate': 0.1467268628273062, 'epoch': 0.37} + 37%|███▋ | 190/520 [40:02<19:54, 3.62s/it] 37%|███▋ | 191/520 [40:06<19:50, 3.62s/it] {'loss': 1.2869, 'grad_norm': 0.00351501669796275, 'learning_rate': 0.1461748613235034, 'epoch': 0.37} + 37%|███▋ | 191/520 [40:06<19:50, 3.62s/it] 37%|███▋ | 192/520 [40:09<19:47, 3.62s/it] {'loss': 1.3837, 'grad_norm': 0.0035987769274372686, 'learning_rate': 0.1456210657353163, 'epoch': 0.37} + 37%|███▋ | 192/520 [40:09<19:47, 3.62s/it] 37%|███▋ | 193/520 [40:13<19:41, 3.61s/it] {'loss': 1.4702, 'grad_norm': 0.004660369535149904, 'learning_rate': 0.14506549757999454, 'epoch': 0.37} + 37%|███▋ | 193/520 [40:13<19:41, 3.61s/it] 37%|███▋ | 194/520 [40:17<19:38, 3.61s/it] {'loss': 1.3391, 'grad_norm': 0.00393511591601138, 'learning_rate': 0.14450817844365924, 'epoch': 0.37} + 37%|███▋ | 194/520 [40:17<19:38, 3.61s/it] 38%|███▊ | 195/520 [40:20<19:33, 3.61s/it] {'loss': 1.4063, 'grad_norm': 0.0037780208725938227, 'learning_rate': 0.1439491299804645, 'epoch': 0.38} + 38%|███▊ | 195/520 [40:20<19:33, 3.61s/it] 38%|███▊ | 196/520 [40:24<19:31, 3.61s/it] {'loss': 1.3747, 'grad_norm': 0.0041945084692933746, 'learning_rate': 0.14338837391175582, 'epoch': 0.38} + 38%|███▊ | 196/520 [40:24<19:31, 3.61s/it] 38%|███▊ | 197/520 [40:27<19:23, 3.60s/it] {'loss': 1.3373, 'grad_norm': 0.0036946064180175334, 'learning_rate': 0.14282593202522628, 'epoch': 0.38} + 38%|███▊ | 197/520 [40:27<19:23, 3.60s/it] 38%|███▊ | 198/520 [40:31<19:19, 3.60s/it] {'loss': 1.4121, 'grad_norm': 0.003842905154080363, 'learning_rate': 0.14226182617406996, 'epoch': 0.38} + 38%|███▊ | 198/520 [40:31<19:19, 3.60s/it] 38%|███▊ | 199/520 [40:35<19:16, 3.60s/it] {'loss': 1.3239, 'grad_norm': 0.003568065056272864, 'learning_rate': 0.14169607827613281, 'epoch': 0.38} + 38%|███▊ | 199/520 [40:35<19:16, 3.60s/it] 38%|███▊ | 200/520 [40:38<19:12, 3.60s/it] {'loss': 1.4067, 'grad_norm': 0.004190135307768848, 'learning_rate': 0.14112871031306118, 'epoch': 0.38} + 38%|███▊ | 200/520 [40:38<19:12, 3.60s/it] 39%|███▊ | 201/520 [40:42<19:09, 3.60s/it] {'loss': 1.4089, 'grad_norm': 0.003577477523305466, 'learning_rate': 0.1405597443294475, 'epoch': 0.39} + 39%|███▊ | 201/520 [40:42<19:09, 3.60s/it] 39%|███▉ | 202/520 [40:45<19:05, 3.60s/it] {'loss': 1.3148, 'grad_norm': 0.0036571069469449385, 'learning_rate': 0.13998920243197407, 'epoch': 0.39} + 39%|███▉ | 202/520 [40:45<19:05, 3.60s/it] 39%|███▉ | 203/520 [40:49<19:02, 3.60s/it] {'loss': 1.376, 'grad_norm': 0.0037223199223374486, 'learning_rate': 0.13941710678855396, 'epoch': 0.39} + 39%|███▉ | 203/520 [40:49<19:02, 3.60s/it] 39%|███▉ | 204/520 [40:53<18:57, 3.60s/it] {'loss': 1.4068, 'grad_norm': 0.003954213746810331, 'learning_rate': 0.13884347962746948, 'epoch': 0.39} + 39%|███▉ | 204/520 [40:53<18:57, 3.60s/it] 39%|███▉ | 205/520 [40:56<18:55, 3.60s/it] {'loss': 1.4238, 'grad_norm': 0.0039685772749924264, 'learning_rate': 0.138268343236509, 'epoch': 0.39} + 39%|███▉ | 205/520 [40:56<18:55, 3.60s/it] 40%|███▉ | 206/520 [41:00<18:51, 3.60s/it] {'loss': 1.4323, 'grad_norm': 0.0037251415495615885, 'learning_rate': 0.13769171996210053, 'epoch': 0.4} + 40%|███▉ | 206/520 [41:00<18:51, 3.60s/it] 40%|███▉ | 207/520 [41:03<18:53, 3.62s/it] {'loss': 1.3934, 'grad_norm': 0.003347181754090643, 'learning_rate': 0.1371136322084438, 'epoch': 0.4} + 40%|███▉ | 207/520 [41:03<18:53, 3.62s/it] 40%|████ | 208/520 [41:07<18:47, 3.61s/it] {'loss': 1.4018, 'grad_norm': 0.004024013008406622, 'learning_rate': 0.13653410243663952, 'epoch': 0.4} + 40%|████ | 208/520 [41:07<18:47, 3.61s/it] 40%|████ | 209/520 [41:11<18:48, 3.63s/it] {'loss': 1.3349, 'grad_norm': 0.003703200161818385, 'learning_rate': 0.13595315316381676, 'epoch': 0.4} + 40%|████ | 209/520 [41:11<18:48, 3.63s/it] 40%|████ | 210/520 [41:14<18:41, 3.62s/it] {'loss': 1.408, 'grad_norm': 0.004134585197024966, 'learning_rate': 0.13537080696225814, 'epoch': 0.4} + 40%|████ | 210/520 [41:14<18:41, 3.62s/it] 41%|████ | 211/520 [41:18<18:36, 3.61s/it] {'loss': 1.4073, 'grad_norm': 0.0033345979606731416, 'learning_rate': 0.13478708645852272, 'epoch': 0.41} + 41%|████ | 211/520 [41:18<18:36, 3.61s/it] 41%|████ | 212/520 [41:21<18:31, 3.61s/it] {'loss': 1.3796, 'grad_norm': 0.0035433296780865557, 'learning_rate': 0.1342020143325669, 'epoch': 0.41} + 41%|████ | 212/520 [41:22<18:31, 3.61s/it] 41%|████ | 213/520 [41:25<18:28, 3.61s/it] {'loss': 1.3397, 'grad_norm': 0.00407230245279814, 'learning_rate': 0.13361561331686309, 'epoch': 0.41} + 41%|████ | 213/520 [41:25<18:28, 3.61s/it] 41%|████ | 214/520 [41:29<18:23, 3.61s/it] {'loss': 1.338, 'grad_norm': 0.003950563358028356, 'learning_rate': 0.13302790619551672, 'epoch': 0.41} + 41%|████ | 214/520 [41:29<18:23, 3.61s/it] 41%|████▏ | 215/520 [41:32<18:21, 3.61s/it] {'loss': 1.3299, 'grad_norm': 0.003685626607460062, 'learning_rate': 0.1324389158033807, 'epoch': 0.41} + 41%|████▏ | 215/520 [41:32<18:21, 3.61s/it] 42%|████▏ | 216/520 [41:36<18:15, 3.61s/it] {'loss': 1.2501, 'grad_norm': 0.0035072414040066093, 'learning_rate': 0.13184866502516845, 'epoch': 0.42} + 42%|████▏ | 216/520 [41:36<18:15, 3.61s/it] 42%|████▏ | 217/520 [41:40<18:14, 3.61s/it] {'loss': 1.3754, 'grad_norm': 0.0037137288376801574, 'learning_rate': 0.13125717679456447, 'epoch': 0.42} + 42%|████▏ | 217/520 [41:40<18:14, 3.61s/it] 42%|████▏ | 218/520 [41:43<18:12, 3.62s/it] {'loss': 1.3661, 'grad_norm': 0.0038228205361671972, 'learning_rate': 0.13066447409333345, 'epoch': 0.42} + 42%|████▏ | 218/520 [41:43<18:12, 3.62s/it] 42%|████▏ | 219/520 [41:47<18:06, 3.61s/it] {'loss': 1.3393, 'grad_norm': 0.0033792943100348118, 'learning_rate': 0.1300705799504273, 'epoch': 0.42} + 42%|████▏ | 219/520 [41:47<18:06, 3.61s/it] 42%|████▏ | 220/520 [41:51<18:13, 3.65s/it] {'loss': 1.3813, 'grad_norm': 0.0037013013746120035, 'learning_rate': 0.12947551744109043, 'epoch': 0.42} + 42%|████▏ | 220/520 [41:51<18:13, 3.65s/it] 42%|████▎ | 221/520 [41:54<18:32, 3.72s/it] {'loss': 1.3801, 'grad_norm': 0.0037900817836325623, 'learning_rate': 0.128879309685963, 'epoch': 0.42} + 42%|████▎ | 221/520 [41:54<18:32, 3.72s/it] 43%|████▎ | 222/520 [41:58<18:39, 3.76s/it] {'loss': 1.2819, 'grad_norm': 0.0036152703917971454, 'learning_rate': 0.12828197985018275, 'epoch': 0.43} + 43%|████▎ | 222/520 [41:58<18:39, 3.76s/it] 43%|████▎ | 223/520 [42:02<19:02, 3.85s/it] {'loss': 1.2714, 'grad_norm': 0.0035760683142389738, 'learning_rate': 0.12768355114248495, 'epoch': 0.43} + 43%|████▎ | 223/520 [42:02<19:02, 3.85s/it] 43%|████▎ | 224/520 [42:06<19:19, 3.92s/it] {'loss': 1.5727, 'grad_norm': 0.004849276810712764, 'learning_rate': 0.12708404681430052, 'epoch': 0.43} + 43%|████▎ | 224/520 [42:06<19:19, 3.92s/it] 43%|████▎ | 225/520 [42:10<19:08, 3.89s/it] {'loss': 1.3044, 'grad_norm': 0.004022258640423639, 'learning_rate': 0.1264834901588527, 'epoch': 0.43} + 43%|████▎ | 225/520 [42:10<19:08, 3.89s/it] 43%|████▎ | 226/520 [42:14<19:06, 3.90s/it] {'loss': 1.3969, 'grad_norm': 0.0035065599459515487, 'learning_rate': 0.12588190451025208, 'epoch': 0.43} + 43%|████▎ | 226/520 [42:14<19:06, 3.90s/it] 44%|████▎ | 227/520 [42:18<18:58, 3.88s/it] {'loss': 1.3787, 'grad_norm': 0.0033038428211443964, 'learning_rate': 0.12527931324258976, 'epoch': 0.44} + 44%|████▎ | 227/520 [42:18<18:58, 3.88s/it] 44%|████▍ | 228/520 [42:22<18:51, 3.87s/it] {'loss': 1.5256, 'grad_norm': 0.004005228927037191, 'learning_rate': 0.12467573976902935, 'epoch': 0.44} + 44%|████▍ | 228/520 [42:22<18:51, 3.87s/it] 44%|████▍ | 229/520 [42:25<18:25, 3.80s/it] {'loss': 1.3585, 'grad_norm': 0.0033114929044888025, 'learning_rate': 0.12407120754089732, 'epoch': 0.44} + 44%|████▍ | 229/520 [42:25<18:25, 3.80s/it] 44%|████▍ | 230/520 [42:29<18:22, 3.80s/it] {'loss': 1.2351, 'grad_norm': 0.0037197547761530446, 'learning_rate': 0.12346574004677154, 'epoch': 0.44} + 44%|████▍ | 230/520 [42:29<18:22, 3.80s/it] 44%|████▍ | 231/520 [42:33<18:17, 3.80s/it] {'loss': 1.2946, 'grad_norm': 0.0031815464840595966, 'learning_rate': 0.12285936081156897, 'epoch': 0.44} + 44%|████▍ | 231/520 [42:33<18:17, 3.80s/it] 45%|████▍ | 232/520 [42:37<18:18, 3.81s/it] {'loss': 1.5589, 'grad_norm': 0.004232948371421238, 'learning_rate': 0.12225209339563144, 'epoch': 0.45} + 45%|████▍ | 232/520 [42:37<18:18, 3.81s/it] 45%|████▍ | 233/520 [42:41<18:13, 3.81s/it] {'loss': 1.4284, 'grad_norm': 0.003941684458506941, 'learning_rate': 0.12164396139381028, 'epoch': 0.45} + 45%|████▍ | 233/520 [42:41<18:13, 3.81s/it] 45%|████▌ | 234/520 [42:45<18:11, 3.82s/it] {'loss': 1.256, 'grad_norm': 0.003800758816311683, 'learning_rate': 0.12103498843454959, 'epoch': 0.45} + 45%|████▌ | 234/520 [42:45<18:11, 3.82s/it] 45%|████▌ | 235/520 [42:48<18:06, 3.81s/it] {'loss': 1.2998, 'grad_norm': 0.003964479634537282, 'learning_rate': 0.12042519817896805, 'epoch': 0.45} + 45%|████▌ | 235/520 [42:48<18:06, 3.81s/it] 45%|████▌ | 236/520 [42:52<18:00, 3.80s/it] {'loss': 1.4201, 'grad_norm': 0.003409744485166509, 'learning_rate': 0.11981461431993977, 'epoch': 0.45} + 45%|████▌ | 236/520 [42:52<18:00, 3.80s/it] 46%|████▌ | 237/520 [42:56<17:57, 3.81s/it] {'loss': 1.3801, 'grad_norm': 0.0034894124589987786, 'learning_rate': 0.11920326058117364, 'epoch': 0.46} + 46%|████▌ | 237/520 [42:56<17:57, 3.81s/it] 46%|████▌ | 238/520 [43:00<17:54, 3.81s/it] {'loss': 1.3192, 'grad_norm': 0.003697362041574563, 'learning_rate': 0.11859116071629149, 'epoch': 0.46} + 46%|████▌ | 238/520 [43:00<17:54, 3.81s/it] 46%|████▌ | 239/520 [43:04<17:51, 3.81s/it] {'loss': 1.4179, 'grad_norm': 0.0038370425139642927, 'learning_rate': 0.11797833850790528, 'epoch': 0.46} + 46%|████▌ | 239/520 [43:04<17:51, 3.81s/it] 46%|████▌ | 240/520 [43:07<17:47, 3.81s/it] {'loss': 1.1856, 'grad_norm': 0.00357538175980193, 'learning_rate': 0.11736481776669305, 'epoch': 0.46} + 46%|████▌ | 240/520 [43:07<17:47, 3.81s/it] 46%|████▋ | 241/520 [43:11<17:39, 3.80s/it] {'loss': 1.2772, 'grad_norm': 0.003807150491785162, 'learning_rate': 0.11675062233047365, 'epoch': 0.46} + 46%|████▋ | 241/520 [43:11<17:39, 3.80s/it] 47%|████▋ | 242/520 [43:15<17:23, 3.75s/it] {'loss': 1.2968, 'grad_norm': 0.0033575849478141096, 'learning_rate': 0.11613577606328068, 'epoch': 0.47} + 47%|████▋ | 242/520 [43:15<17:23, 3.75s/it] 47%|████▋ | 243/520 [43:18<17:08, 3.71s/it] {'loss': 1.2859, 'grad_norm': 0.0035547665953754, 'learning_rate': 0.11552030285443515, 'epoch': 0.47} + 47%|████▋ | 243/520 [43:18<17:08, 3.71s/it] 47%|████▋ | 244/520 [43:22<17:00, 3.70s/it] {'loss': 1.4211, 'grad_norm': 0.003762810058029991, 'learning_rate': 0.11490422661761744, 'epoch': 0.47} + 47%|████▋ | 244/520 [43:22<17:00, 3.70s/it] 47%|████▋ | 245/520 [43:26<17:05, 3.73s/it] {'loss': 1.2705, 'grad_norm': 0.0034283364487626006, 'learning_rate': 0.11428757128993801, 'epoch': 0.47} + 47%|████▋ | 245/520 [43:26<17:05, 3.73s/it] 47%|████▋ | 246/520 [43:30<17:00, 3.72s/it] {'loss': 1.527, 'grad_norm': 0.004114473304535244, 'learning_rate': 0.11367036083100734, 'epoch': 0.47} + 47%|████▋ | 246/520 [43:30<17:00, 3.72s/it] 48%|████▊ | 247/520 [43:33<16:56, 3.72s/it] {'loss': 1.4478, 'grad_norm': 0.0036415936995512384, 'learning_rate': 0.11305261922200519, 'epoch': 0.47} + 48%|████▊ | 247/520 [43:33<16:56, 3.72s/it] 48%|████▊ | 248/520 [43:37<16:55, 3.73s/it] {'loss': 1.2756, 'grad_norm': 0.0035612182599107367, 'learning_rate': 0.11243437046474854, 'epoch': 0.48} + 48%|████▊ | 248/520 [43:37<16:55, 3.73s/it] 48%|████▊ | 249/520 [43:41<16:51, 3.73s/it] {'loss': 1.3876, 'grad_norm': 0.003804822221307175, 'learning_rate': 0.1118156385807593, 'epoch': 0.48} + 48%|████▊ | 249/520 [43:41<16:51, 3.73s/it] 48%|████▊ | 250/520 [43:45<16:48, 3.74s/it] {'loss': 1.3298, 'grad_norm': 0.0038364723147465466, 'learning_rate': 0.11119644761033079, 'epoch': 0.48} + 48%|████▊ | 250/520 [43:45<16:48, 3.74s/it] 48%|████▊ | 251/520 [43:48<16:46, 3.74s/it] {'loss': 1.3863, 'grad_norm': 0.0033011404893182775, 'learning_rate': 0.1105768216115938, 'epoch': 0.48} + 48%|████▊ | 251/520 [43:48<16:46, 3.74s/it] 48%|████▊ | 252/520 [43:52<16:41, 3.74s/it] {'loss': 1.406, 'grad_norm': 0.003616334347506286, 'learning_rate': 0.10995678465958168, 'epoch': 0.48} + 48%|████▊ | 252/520 [43:52<16:41, 3.74s/it] 49%|████▊ | 253/520 [43:56<16:40, 3.75s/it] {'loss': 1.3761, 'grad_norm': 0.003698318593815572, 'learning_rate': 0.10933636084529506, 'epoch': 0.49} + 49%|████▊ | 253/520 [43:56<16:40, 3.75s/it] 49%|████▉ | 254/520 [44:00<16:37, 3.75s/it] {'loss': 1.2942, 'grad_norm': 0.0033135784326587337, 'learning_rate': 0.10871557427476584, 'epoch': 0.49} + 49%|████▉ | 254/520 [44:00<16:37, 3.75s/it] 49%|████▉ | 255/520 [44:03<16:32, 3.74s/it] {'loss': 1.3081, 'grad_norm': 0.003608584851827315, 'learning_rate': 0.10809444906812034, 'epoch': 0.49} + 49%|████▉ | 255/520 [44:03<16:32, 3.74s/it] 49%|████▉ | 256/520 [44:07<16:27, 3.74s/it] {'loss': 1.3517, 'grad_norm': 0.003598423865952307, 'learning_rate': 0.10747300935864244, 'epoch': 0.49} + 49%|████▉ | 256/520 [44:07<16:27, 3.74s/it] 49%|████▉ | 257/520 [44:11<16:22, 3.74s/it] {'loss': 1.3479, 'grad_norm': 0.0035144288414186236, 'learning_rate': 0.10685127929183567, 'epoch': 0.49} + 49%|████▉ | 257/520 [44:11<16:22, 3.74s/it] 50%|████▉ | 258/520 [44:15<16:21, 3.75s/it] {'loss': 1.3614, 'grad_norm': 0.003051586247402253, 'learning_rate': 0.10622928302448523, 'epoch': 0.5} + 50%|████▉ | 258/520 [44:15<16:21, 3.75s/it] 50%|████▉ | 259/520 [44:18<16:16, 3.74s/it] {'loss': 1.4213, 'grad_norm': 0.0037914118853892413, 'learning_rate': 0.10560704472371918, 'epoch': 0.5} + 50%|████▉ | 259/520 [44:18<16:16, 3.74s/it] 50%|█████ | 260/520 [44:22<16:12, 3.74s/it] {'loss': 1.4834, 'grad_norm': 0.003549345286914332, 'learning_rate': 0.10498458856606972, 'epoch': 0.5} + 50%|█████ | 260/520 [44:22<16:12, 3.74s/it] 50%|█████ | 261/520 [44:26<16:07, 3.73s/it] {'loss': 1.4101, 'grad_norm': 0.003798473298104541, 'learning_rate': 0.10436193873653361, 'epoch': 0.5} + 50%|█████ | 261/520 [44:26<16:07, 3.73s/it] 50%|█████ | 262/520 [44:29<16:02, 3.73s/it] {'loss': 1.2618, 'grad_norm': 0.0035929352826392386, 'learning_rate': 0.10373911942763259, 'epoch': 0.5} + 50%|█████ | 262/520 [44:29<16:02, 3.73s/it] 51%|█████ | 263/520 [44:33<15:56, 3.72s/it] {'loss': 1.4122, 'grad_norm': 0.004054100820567659, 'learning_rate': 0.10311615483847332, 'epoch': 0.51} + 51%|█████ | 263/520 [44:33<15:56, 3.72s/it] 51%|█████ | 264/520 [44:37<15:58, 3.74s/it] {'loss': 1.3845, 'grad_norm': 0.0035355560383537746, 'learning_rate': 0.1024930691738073, 'epoch': 0.51} + 51%|█████ | 264/520 [44:37<15:58, 3.74s/it] 51%|█████ | 265/520 [44:41<15:52, 3.73s/it] {'loss': 1.2801, 'grad_norm': 0.0040393041328921275, 'learning_rate': 0.10186988664309023, 'epoch': 0.51} + 51%|█████ | 265/520 [44:41<15:52, 3.73s/it] 51%|█████ | 266/520 [44:44<15:47, 3.73s/it] {'loss': 1.1348, 'grad_norm': 0.0031434648349133094, 'learning_rate': 0.10124663145954152, 'epoch': 0.51} + 51%|█████ | 266/520 [44:44<15:47, 3.73s/it] 51%|█████▏ | 267/520 [44:48<15:44, 3.73s/it] {'loss': 1.273, 'grad_norm': 0.0033991026657662795, 'learning_rate': 0.10062332783920336, 'epoch': 0.51} + 51%|█████▏ | 267/520 [44:48<15:44, 3.73s/it] 52%|█████▏ | 268/520 [44:52<15:38, 3.72s/it] {'loss': 1.5476, 'grad_norm': 0.004023900806650206, 'learning_rate': 0.1, 'epoch': 0.52} + 52%|█████▏ | 268/520 [44:52<15:38, 3.72s/it] 52%|█████▏ | 269/520 [44:56<15:38, 3.74s/it] {'loss': 1.3995, 'grad_norm': 0.003646416441340005, 'learning_rate': 0.09937667216079665, 'epoch': 0.52} + 52%|█████▏ | 269/520 [44:56<15:38, 3.74s/it] 52%|█████▏ | 270/520 [44:59<15:35, 3.74s/it] {'loss': 1.3333, 'grad_norm': 0.003747959232438602, 'learning_rate': 0.0987533685404585, 'epoch': 0.52} + 52%|█████▏ | 270/520 [44:59<15:35, 3.74s/it] 52%|█████▏ | 271/520 [45:03<15:31, 3.74s/it] {'loss': 1.3775, 'grad_norm': 0.003476599611045054, 'learning_rate': 0.0981301133569098, 'epoch': 0.52} + 52%|█████▏ | 271/520 [45:03<15:31, 3.74s/it] 52%|█████▏ | 272/520 [45:07<15:25, 3.73s/it] {'loss': 1.3409, 'grad_norm': 0.0037002461622462597, 'learning_rate': 0.09750693082619273, 'epoch': 0.52} + 52%|█████▏ | 272/520 [45:07<15:25, 3.73s/it] 52%|█████▎ | 273/520 [45:10<15:20, 3.73s/it] {'loss': 1.4754, 'grad_norm': 0.0037102894773717353, 'learning_rate': 0.0968838451615267, 'epoch': 0.53} + 52%|█████▎ | 273/520 [45:10<15:20, 3.73s/it] 53%|█████▎ | 274/520 [45:14<15:18, 3.73s/it] {'loss': 1.3357, 'grad_norm': 0.0037013728013377176, 'learning_rate': 0.09626088057236745, 'epoch': 0.53} + 53%|█████▎ | 274/520 [45:14<15:18, 3.73s/it] 53%|█████▎ | 275/520 [45:18<15:16, 3.74s/it] {'loss': 1.2841, 'grad_norm': 0.003929726871690139, 'learning_rate': 0.09563806126346641, 'epoch': 0.53} + 53%|█████▎ | 275/520 [45:18<15:16, 3.74s/it] 53%|█████▎ | 276/520 [45:22<15:16, 3.76s/it] {'loss': 1.3642, 'grad_norm': 0.003880146215133112, 'learning_rate': 0.09501541143393027, 'epoch': 0.53} + 53%|█████▎ | 276/520 [45:22<15:16, 3.76s/it] 53%|█████▎ | 277/520 [45:26<15:14, 3.76s/it] {'loss': 1.4647, 'grad_norm': 0.004381930278153175, 'learning_rate': 0.09439295527628082, 'epoch': 0.53} + 53%|█████▎ | 277/520 [45:26<15:14, 3.76s/it] 53%|█████▎ | 278/520 [45:29<15:09, 3.76s/it] {'loss': 1.2305, 'grad_norm': 0.003306595875930885, 'learning_rate': 0.0937707169755148, 'epoch': 0.53} + 53%|█████▎ | 278/520 [45:29<15:09, 3.76s/it] 54%|█████▎ | 279/520 [45:33<15:03, 3.75s/it] {'loss': 1.3678, 'grad_norm': 0.00414906778600391, 'learning_rate': 0.09314872070816434, 'epoch': 0.54} + 54%|█████▎ | 279/520 [45:33<15:03, 3.75s/it] 54%|█████▍ | 280/520 [45:37<15:00, 3.75s/it] {'loss': 1.2845, 'grad_norm': 0.003954956452792023, 'learning_rate': 0.09252699064135758, 'epoch': 0.54} + 54%|█████▍ | 280/520 [45:37<15:00, 3.75s/it] 54%|█████▍ | 281/520 [45:41<15:00, 3.77s/it] {'loss': 1.3962, 'grad_norm': 0.003774181817967413, 'learning_rate': 0.09190555093187967, 'epoch': 0.54} + 54%|█████▍ | 281/520 [45:41<15:00, 3.77s/it] 54%|█████▍ | 282/520 [45:44<15:00, 3.78s/it] {'loss': 1.2361, 'grad_norm': 0.0032898147323206727, 'learning_rate': 0.09128442572523418, 'epoch': 0.54} + 54%|█████▍ | 282/520 [45:44<15:00, 3.78s/it] 54%|█████▍ | 283/520 [45:48<14:46, 3.74s/it] {'loss': 1.4169, 'grad_norm': 0.003969503721409043, 'learning_rate': 0.09066363915470495, 'epoch': 0.54} + 54%|█████▍ | 283/520 [45:48<14:46, 3.74s/it] 55%|█████▍ | 284/520 [45:52<14:45, 3.75s/it] {'loss': 1.3377, 'grad_norm': 0.004120966309721341, 'learning_rate': 0.09004321534041836, 'epoch': 0.55} + 55%|█████▍ | 284/520 [45:52<14:45, 3.75s/it] 55%|█████▍ | 285/520 [45:56<15:01, 3.84s/it] {'loss': 1.2669, 'grad_norm': 0.0037395983836401186, 'learning_rate': 0.08942317838840624, 'epoch': 0.55} + 55%|█████▍ | 285/520 [45:56<15:01, 3.84s/it] 55%|█████▌ | 286/520 [46:00<15:18, 3.92s/it] {'loss': 1.1352, 'grad_norm': 0.0039749124343163075, 'learning_rate': 0.08880355238966922, 'epoch': 0.55} + 55%|█████▌ | 286/520 [46:00<15:18, 3.92s/it] 55%|█████▌ | 287/520 [46:04<15:27, 3.98s/it] {'loss': 1.3917, 'grad_norm': 0.003945794846590786, 'learning_rate': 0.08818436141924073, 'epoch': 0.55} + 55%|█████▌ | 287/520 [46:04<15:27, 3.98s/it] 55%|█████▌ | 288/520 [46:08<15:33, 4.02s/it] {'loss': 1.4333, 'grad_norm': 0.0034776434962741102, 'learning_rate': 0.08756562953525152, 'epoch': 0.55} + 55%|█████▌ | 288/520 [46:08<15:33, 4.02s/it] 56%|█████▌ | 289/520 [46:12<15:36, 4.05s/it] {'loss': 1.2866, 'grad_norm': 0.0033360216129259823, 'learning_rate': 0.08694738077799487, 'epoch': 0.56} + 56%|█████▌ | 289/520 [46:12<15:36, 4.05s/it] 56%|█████▌ | 290/520 [46:16<15:37, 4.07s/it] {'loss': 1.1964, 'grad_norm': 0.0034206196168580544, 'learning_rate': 0.08632963916899268, 'epoch': 0.56} + 56%|█████▌ | 290/520 [46:16<15:37, 4.07s/it] 56%|█████▌ | 291/520 [46:21<15:34, 4.08s/it] {'loss': 1.2629, 'grad_norm': 0.003914662925832046, 'learning_rate': 0.08571242871006202, 'epoch': 0.56} + 56%|█████▌ | 291/520 [46:21<15:34, 4.08s/it] 56%|█████▌ | 292/520 [46:25<15:32, 4.09s/it] {'loss': 1.3058, 'grad_norm': 0.0034100584937384526, 'learning_rate': 0.08509577338238256, 'epoch': 0.56} + 56%|█████▌ | 292/520 [46:25<15:32, 4.09s/it] 56%|█████▋ | 293/520 [46:29<15:28, 4.09s/it] {'loss': 1.2528, 'grad_norm': 0.003746483114090666, 'learning_rate': 0.08447969714556484, 'epoch': 0.56} + 56%|█████▋ | 293/520 [46:29<15:28, 4.09s/it] 57%|█████▋ | 294/520 [46:33<15:23, 4.09s/it] {'loss': 1.2873, 'grad_norm': 0.0037334005679994756, 'learning_rate': 0.08386422393671933, 'epoch': 0.57} + 57%|█████▋ | 294/520 [46:33<15:23, 4.09s/it] 57%|█████▋ | 295/520 [46:37<14:58, 3.99s/it] {'loss': 1.4123, 'grad_norm': 0.003913863181970147, 'learning_rate': 0.08324937766952638, 'epoch': 0.57} + 57%|█████▋ | 295/520 [46:37<14:58, 3.99s/it] 57%|█████▋ | 296/520 [46:40<14:31, 3.89s/it] {'loss': 1.2212, 'grad_norm': 0.0037760025159363267, 'learning_rate': 0.08263518223330697, 'epoch': 0.57} + 57%|█████▋ | 296/520 [46:40<14:31, 3.89s/it] 57%|█████▋ | 297/520 [46:44<14:09, 3.81s/it] {'loss': 1.3646, 'grad_norm': 0.0040549754829717405, 'learning_rate': 0.08202166149209474, 'epoch': 0.57} + 57%|█████▋ | 297/520 [46:44<14:09, 3.81s/it] 57%|█████▋ | 298/520 [46:48<13:54, 3.76s/it] {'loss': 1.3215, 'grad_norm': 0.003139918916627074, 'learning_rate': 0.08140883928370855, 'epoch': 0.57} + 57%|█████▋ | 298/520 [46:48<13:54, 3.76s/it] 57%|█████▊ | 299/520 [46:51<13:41, 3.72s/it] {'loss': 1.417, 'grad_norm': 0.003566052965941467, 'learning_rate': 0.0807967394188264, 'epoch': 0.57} + 57%|█████▊ | 299/520 [46:51<13:41, 3.72s/it] 58%|█████▊ | 300/520 [46:55<13:33, 3.70s/it] {'loss': 1.3811, 'grad_norm': 0.0034787265674308125, 'learning_rate': 0.08018538568006027, 'epoch': 0.58} + 58%|█████▊ | 300/520 [46:55<13:33, 3.70s/it] 58%|█████▊ | 301/520 [46:58<13:25, 3.68s/it] {'loss': 1.3508, 'grad_norm': 0.0035572886163223304, 'learning_rate': 0.07957480182103199, 'epoch': 0.58} + 58%|█████▊ | 301/520 [46:58<13:25, 3.68s/it] 58%|█████▊ | 302/520 [47:02<13:19, 3.67s/it] {'loss': 1.435, 'grad_norm': 0.0037824171192813974, 'learning_rate': 0.07896501156545044, 'epoch': 0.58} + 58%|█████▊ | 302/520 [47:02<13:19, 3.67s/it] 58%|█████▊ | 303/520 [47:06<13:14, 3.66s/it] {'loss': 1.2852, 'grad_norm': 0.0038638758643428956, 'learning_rate': 0.07835603860618973, 'epoch': 0.58} + 58%|█████▊ | 303/520 [47:06<13:14, 3.66s/it] 58%|█████▊ | 304/520 [47:09<13:09, 3.65s/it] {'loss': 1.3213, 'grad_norm': 0.0036431324248720207, 'learning_rate': 0.07774790660436857, 'epoch': 0.58} + 58%|█████▊ | 304/520 [47:09<13:09, 3.65s/it] 59%|█████▊ | 305/520 [47:13<13:03, 3.64s/it] {'loss': 1.3975, 'grad_norm': 0.004024240067454315, 'learning_rate': 0.07714063918843106, 'epoch': 0.59} + 59%|█████▊ | 305/520 [47:13<13:03, 3.64s/it] 59%|█████▉ | 306/520 [47:17<13:03, 3.66s/it] {'loss': 1.321, 'grad_norm': 0.003396879142419131, 'learning_rate': 0.0765342599532285, 'epoch': 0.59} + 59%|█████▉ | 306/520 [47:17<13:03, 3.66s/it] 59%|█████▉ | 307/520 [47:21<13:17, 3.74s/it] {'loss': 1.2679, 'grad_norm': 0.0034491165800698237, 'learning_rate': 0.07592879245910272, 'epoch': 0.59} + 59%|█████▉ | 307/520 [47:21<13:17, 3.74s/it] 59%|█████▉ | 308/520 [47:24<13:07, 3.71s/it] {'loss': 1.3898, 'grad_norm': 0.003491913689290631, 'learning_rate': 0.07532426023097064, 'epoch': 0.59} + 59%|█████▉ | 308/520 [47:24<13:07, 3.71s/it] 59%|█████▉ | 309/520 [47:28<12:58, 3.69s/it] {'loss': 1.2619, 'grad_norm': 0.0032866549086798796, 'learning_rate': 0.07472068675741024, 'epoch': 0.59} + 59%|█████▉ | 309/520 [47:28<12:58, 3.69s/it] 60%|█████▉ | 310/520 [47:32<12:52, 3.68s/it] {'loss': 1.2399, 'grad_norm': 0.003469950588808316, 'learning_rate': 0.07411809548974792, 'epoch': 0.6} + 60%|█████▉ | 310/520 [47:32<12:52, 3.68s/it] 60%|█████▉ | 311/520 [47:35<12:47, 3.67s/it] {'loss': 1.2131, 'grad_norm': 0.0034847437273003985, 'learning_rate': 0.07351650984114727, 'epoch': 0.6} + 60%|█████▉ | 311/520 [47:35<12:47, 3.67s/it] 60%|██████ | 312/520 [47:39<12:47, 3.69s/it] {'loss': 1.2053, 'grad_norm': 0.0039734610516397725, 'learning_rate': 0.0729159531856995, 'epoch': 0.6} + 60%|██████ | 312/520 [47:39<12:47, 3.69s/it] 60%|██████ | 313/520 [47:43<12:54, 3.74s/it] {'loss': 1.1989, 'grad_norm': 0.0031783974790794006, 'learning_rate': 0.07231644885751508, 'epoch': 0.6} + 60%|██████ | 313/520 [47:43<12:54, 3.74s/it] 60%|██████ | 314/520 [47:47<13:28, 3.92s/it] {'loss': 1.2283, 'grad_norm': 0.0032340152999021433, 'learning_rate': 0.07171802014981725, 'epoch': 0.6} + 60%|██████ | 314/520 [47:47<13:28, 3.92s/it] 61%|██████ | 315/520 [47:51<13:09, 3.85s/it] {'loss': 1.4017, 'grad_norm': 0.004046578462401684, 'learning_rate': 0.07112069031403703, 'epoch': 0.61} + 61%|██████ | 315/520 [47:51<13:09, 3.85s/it] 61%|██████ | 316/520 [47:55<13:18, 3.91s/it] {'loss': 1.2045, 'grad_norm': 0.004028374378606471, 'learning_rate': 0.07052448255890957, 'epoch': 0.61} + 61%|██████ | 316/520 [47:55<13:18, 3.91s/it] 61%|██████ | 317/520 [47:59<12:55, 3.82s/it] {'loss': 1.2255, 'grad_norm': 0.003184250850669721, 'learning_rate': 0.0699294200495727, 'epoch': 0.61} + 61%|██████ | 317/520 [47:59<12:55, 3.82s/it] 61%|██████ | 318/520 [48:02<12:39, 3.76s/it] {'loss': 1.352, 'grad_norm': 0.003675451679235439, 'learning_rate': 0.06933552590666658, 'epoch': 0.61} + 61%|██████ | 318/520 [48:02<12:39, 3.76s/it] 61%|██████▏ | 319/520 [48:06<12:51, 3.84s/it] {'loss': 1.2173, 'grad_norm': 0.003540224535584222, 'learning_rate': 0.06874282320543557, 'epoch': 0.61} + 61%|██████▏ | 319/520 [48:06<12:51, 3.84s/it] 62%|██████▏ | 320/520 [48:10<12:33, 3.77s/it] {'loss': 1.1573, 'grad_norm': 0.003711145656094201, 'learning_rate': 0.06815133497483157, 'epoch': 0.62} + 62%|██████▏ | 320/520 [48:10<12:33, 3.77s/it] 62%|██████▏ | 321/520 [48:13<12:19, 3.72s/it] {'loss': 1.3651, 'grad_norm': 0.0037501182864168544, 'learning_rate': 0.06756108419661931, 'epoch': 0.62} + 62%|██████▏ | 321/520 [48:13<12:19, 3.72s/it] 62%|██████▏ | 322/520 [48:17<12:10, 3.69s/it] {'loss': 1.2523, 'grad_norm': 0.0034701512258828574, 'learning_rate': 0.06697209380448332, 'epoch': 0.62} + 62%|██████▏ | 322/520 [48:17<12:10, 3.69s/it] 62%|██████▏ | 323/520 [48:21<12:04, 3.68s/it] {'loss': 1.337, 'grad_norm': 0.004055708491955918, 'learning_rate': 0.06638438668313694, 'epoch': 0.62} + 62%|██████▏ | 323/520 [48:21<12:04, 3.68s/it] 62%|██████▏ | 324/520 [48:24<11:57, 3.66s/it] {'loss': 1.2888, 'grad_norm': 0.004212673285826627, 'learning_rate': 0.06579798566743314, 'epoch': 0.62} + 62%|██████▏ | 324/520 [48:24<11:57, 3.66s/it] 62%|██████▎ | 325/520 [48:28<11:51, 3.65s/it] {'loss': 1.305, 'grad_norm': 0.0037195179809842, 'learning_rate': 0.06521291354147728, 'epoch': 0.62} + 62%|██████▎ | 325/520 [48:28<11:51, 3.65s/it] 63%|██████▎ | 326/520 [48:31<11:45, 3.64s/it] {'loss': 1.2842, 'grad_norm': 0.0035668487849822555, 'learning_rate': 0.06462919303774187, 'epoch': 0.63} + 63%|██████▎ | 326/520 [48:31<11:45, 3.64s/it] 63%|██████▎ | 327/520 [48:35<11:45, 3.66s/it] {'loss': 1.4069, 'grad_norm': 0.004346587907569975, 'learning_rate': 0.06404684683618325, 'epoch': 0.63} + 63%|██████▎ | 327/520 [48:35<11:45, 3.66s/it] 63%|██████▎ | 328/520 [48:39<11:38, 3.64s/it] {'loss': 1.3551, 'grad_norm': 0.0037689728291500535, 'learning_rate': 0.0634658975633605, 'epoch': 0.63} + 63%|██████▎ | 328/520 [48:39<11:38, 3.64s/it] 63%|██████▎ | 329/520 [48:42<11:32, 3.63s/it] {'loss': 1.2027, 'grad_norm': 0.00310952592141935, 'learning_rate': 0.06288636779155621, 'epoch': 0.63} + 63%|██████▎ | 329/520 [48:42<11:32, 3.63s/it] 63%|██████▎ | 330/520 [48:46<11:28, 3.63s/it] {'loss': 1.2722, 'grad_norm': 0.0031468662475264936, 'learning_rate': 0.06230828003789948, 'epoch': 0.63} + 63%|██████▎ | 330/520 [48:46<11:28, 3.63s/it] 64%|██████▎ | 331/520 [48:50<11:25, 3.63s/it] {'loss': 1.2393, 'grad_norm': 0.0034068400481865346, 'learning_rate': 0.06173165676349103, 'epoch': 0.64} + 64%|██████▎ | 331/520 [48:50<11:25, 3.63s/it] 64%|██████▍ | 332/520 [48:53<11:21, 3.62s/it] {'loss': 1.4082, 'grad_norm': 0.003557597100034733, 'learning_rate': 0.06115652037253053, 'epoch': 0.64} + 64%|██████▍ | 332/520 [48:53<11:21, 3.62s/it] 64%|██████▍ | 333/520 [48:57<11:26, 3.67s/it] {'loss': 1.4046, 'grad_norm': 0.0037605672171747632, 'learning_rate': 0.06058289321144608, 'epoch': 0.64} + 64%|██████▍ | 333/520 [48:57<11:26, 3.67s/it] 64%|██████▍ | 334/520 [49:01<11:31, 3.72s/it] {'loss': 1.2932, 'grad_norm': 0.004099642055839644, 'learning_rate': 0.06001079756802592, 'epoch': 0.64} + 64%|██████▍ | 334/520 [49:01<11:31, 3.72s/it] 64%|██████▍ | 335/520 [49:05<11:33, 3.75s/it] {'loss': 1.2861, 'grad_norm': 0.0033017062457026624, 'learning_rate': 0.059440255670552514, 'epoch': 0.64} + 64%|██████▍ | 335/520 [49:05<11:33, 3.75s/it] 65%|██████▍ | 336/520 [49:08<11:33, 3.77s/it] {'loss': 1.1797, 'grad_norm': 0.0037946129308590298, 'learning_rate': 0.05887128968693887, 'epoch': 0.65} + 65%|██████▍ | 336/520 [49:08<11:33, 3.77s/it] 65%|██████▍ | 337/520 [49:12<11:31, 3.78s/it] {'loss': 1.1759, 'grad_norm': 0.0034124983662197545, 'learning_rate': 0.058303921723867225, 'epoch': 0.65} + 65%|██████▍ | 337/520 [49:12<11:31, 3.78s/it] 65%|██████▌ | 338/520 [49:16<11:31, 3.80s/it] {'loss': 1.3013, 'grad_norm': 0.0035554634489926625, 'learning_rate': 0.05773817382593008, 'epoch': 0.65} + 65%|██████▌ | 338/520 [49:16<11:31, 3.80s/it] 65%|██████▌ | 339/520 [49:20<11:19, 3.75s/it] {'loss': 1.2379, 'grad_norm': 0.003432835532957329, 'learning_rate': 0.057174067974773715, 'epoch': 0.65} + 65%|██████▌ | 339/520 [49:20<11:19, 3.75s/it] 65%|██████▌ | 340/520 [49:23<11:08, 3.72s/it] {'loss': 1.2294, 'grad_norm': 0.0034400431782692464, 'learning_rate': 0.056611626088244195, 'epoch': 0.65} + 65%|██████▌ | 340/520 [49:23<11:08, 3.72s/it] 66%|██████▌ | 341/520 [49:27<10:59, 3.69s/it] {'loss': 1.2517, 'grad_norm': 0.003795372169265797, 'learning_rate': 0.056050870019535494, 'epoch': 0.66} + 66%|██████▌ | 341/520 [49:27<10:59, 3.69s/it] 66%|██████▌ | 342/520 [49:31<10:50, 3.66s/it] {'loss': 1.3752, 'grad_norm': 0.0041489717578705775, 'learning_rate': 0.05549182155634076, 'epoch': 0.66} + 66%|██████▌ | 342/520 [49:31<10:50, 3.66s/it] 66%|██████▌ | 343/520 [49:34<10:45, 3.64s/it] {'loss': 1.3432, 'grad_norm': 0.003500629152937682, 'learning_rate': 0.054934502420005464, 'epoch': 0.66} + 66%|██████▌ | 343/520 [49:34<10:45, 3.64s/it] 66%|██████▌ | 344/520 [49:38<10:40, 3.64s/it] {'loss': 1.2022, 'grad_norm': 0.0035952964934561184, 'learning_rate': 0.0543789342646837, 'epoch': 0.66} + 66%|██████▌ | 344/520 [49:38<10:40, 3.64s/it] 66%|██████▋ | 345/520 [49:42<10:36, 3.64s/it] {'loss': 1.3217, 'grad_norm': 0.004078387742009941, 'learning_rate': 0.05382513867649663, 'epoch': 0.66} + 66%|██████▋ | 345/520 [49:42<10:36, 3.64s/it] 67%|██████▋ | 346/520 [49:45<10:31, 3.63s/it] {'loss': 1.3327, 'grad_norm': 0.0033473796593199855, 'learning_rate': 0.0532731371726938, 'epoch': 0.67} + 67%|██████▋ | 346/520 [49:45<10:31, 3.63s/it] 67%|██████▋ | 347/520 [49:49<10:27, 3.63s/it] {'loss': 1.2194, 'grad_norm': 0.003342724535516176, 'learning_rate': 0.05272295120081732, 'epoch': 0.67} + 67%|██████▋ | 347/520 [49:49<10:27, 3.63s/it]Token indices sequence length is longer than the specified maximum sequence length for this model (2778 > 2048). Running this sequence through the model will result in indexing errors + 67%|██████▋ | 348/520 [49:52<10:29, 3.66s/it] {'loss': 1.1794, 'grad_norm': 0.004175225540912436, 'learning_rate': 0.05217460213786821, 'epoch': 0.67} + 67%|██████▋ | 348/520 [49:52<10:29, 3.66s/it] 67%|██████▋ | 349/520 [49:56<10:35, 3.72s/it] {'loss': 1.2238, 'grad_norm': 0.003762201022626968, 'learning_rate': 0.051628111289476024, 'epoch': 0.67} + 67%|██████▋ | 349/520 [49:56<10:35, 3.72s/it] 67%|██████▋ | 350/520 [50:00<10:38, 3.76s/it] {'loss': 1.2609, 'grad_norm': 0.0037624166297652957, 'learning_rate': 0.051083499889071106, 'epoch': 0.67} + 67%|██████▋ | 350/520 [50:00<10:38, 3.76s/it] 68%|██████▊ | 351/520 [50:04<10:41, 3.79s/it] {'loss': 1.1622, 'grad_norm': 0.0033014991614454745, 'learning_rate': 0.05054078909705926, 'epoch': 0.68} + 68%|██████▊ | 351/520 [50:04<10:41, 3.79s/it] 68%|██████▊ | 352/520 [50:08<10:41, 3.82s/it] {'loss': 1.2902, 'grad_norm': 0.003511294442435778, 'learning_rate': 0.050000000000000024, 'epoch': 0.68} + 68%|██████▊ | 352/520 [50:08<10:41, 3.82s/it] 68%|██████▊ | 353/520 [50:12<10:44, 3.86s/it] {'loss': 1.2761, 'grad_norm': 0.002929827413260095, 'learning_rate': 0.04946115360978696, 'epoch': 0.68} + 68%|██████▊ | 353/520 [50:12<10:44, 3.86s/it] 68%|██████▊ | 354/520 [50:16<10:40, 3.86s/it] {'loss': 1.4149, 'grad_norm': 0.0034612772325375925, 'learning_rate': 0.048924270862831465, 'epoch': 0.68} + 68%|██████▊ | 354/520 [50:16<10:40, 3.86s/it] 68%|██████▊ | 355/520 [50:20<10:36, 3.86s/it] {'loss': 1.2305, 'grad_norm': 0.0035298958519495607, 'learning_rate': 0.04838937261924933, 'epoch': 0.68} + 68%|██████▊ | 355/520 [50:20<10:36, 3.86s/it] 68%|██████▊ | 356/520 [50:23<10:32, 3.86s/it] {'loss': 1.2269, 'grad_norm': 0.0035930637247615278, 'learning_rate': 0.0478564796620502, 'epoch': 0.68} + 68%|██████▊ | 356/520 [50:23<10:32, 3.86s/it] 69%|██████▊ | 357/520 [50:27<10:19, 3.80s/it] {'loss': 1.2495, 'grad_norm': 0.003196795138465662, 'learning_rate': 0.04732561269632993, 'epoch': 0.69} + 69%|██████▊ | 357/520 [50:27<10:19, 3.80s/it] 69%|██████▉ | 358/520 [50:31<10:06, 3.74s/it] {'loss': 1.1766, 'grad_norm': 0.003472022706374636, 'learning_rate': 0.04679679234846636, 'epoch': 0.69} + 69%|██████▉ | 358/520 [50:31<10:06, 3.74s/it] 69%|██████▉ | 359/520 [50:34<09:55, 3.70s/it] {'loss': 1.346, 'grad_norm': 0.00392841268192239, 'learning_rate': 0.046270039165317606, 'epoch': 0.69} + 69%|██████▉ | 359/520 [50:34<09:55, 3.70s/it] 69%|██████▉ | 360/520 [50:38<09:51, 3.70s/it] {'loss': 1.3541, 'grad_norm': 0.003712437616806545, 'learning_rate': 0.04574537361342407, 'epoch': 0.69} + 69%|██████▉ | 360/520 [50:38<09:51, 3.70s/it] 69%|██████▉ | 361/520 [50:42<09:43, 3.67s/it] {'loss': 1.3482, 'grad_norm': 0.003219780584473458, 'learning_rate': 0.04522281607821288, 'epoch': 0.69} + 69%|██████▉ | 361/520 [50:42<09:43, 3.67s/it] 70%|██████▉ | 362/520 [50:45<09:37, 3.65s/it] {'loss': 1.2413, 'grad_norm': 0.0036014707588468784, 'learning_rate': 0.04470238686320606, 'epoch': 0.7} + 70%|██████▉ | 362/520 [50:45<09:37, 3.65s/it] 70%|██████▉ | 363/520 [50:49<09:33, 3.65s/it] {'loss': 1.2722, 'grad_norm': 0.0033621301507862065, 'learning_rate': 0.044184106189231624, 'epoch': 0.7} + 70%|██████▉ | 363/520 [50:49<09:33, 3.65s/it] 70%|███████ | 364/520 [50:53<09:41, 3.73s/it] {'loss': 1.3658, 'grad_norm': 0.0034001832022133973, 'learning_rate': 0.043667994193637795, 'epoch': 0.7} + 70%|███████ | 364/520 [50:53<09:41, 3.73s/it] 70%|███████ | 365/520 [50:57<09:42, 3.76s/it] {'loss': 1.3323, 'grad_norm': 0.0035922406756768365, 'learning_rate': 0.043154070929510784, 'epoch': 0.7} + 70%|███████ | 365/520 [50:57<09:42, 3.76s/it] 70%|███████ | 366/520 [51:01<09:46, 3.81s/it] {'loss': 1.2824, 'grad_norm': 0.0031177944491531987, 'learning_rate': 0.04264235636489542, 'epoch': 0.7} + 70%|███████ | 366/520 [51:01<09:46, 3.81s/it] 71%|███████ | 367/520 [51:04<09:42, 3.81s/it] {'loss': 1.2812, 'grad_norm': 0.003306765035627669, 'learning_rate': 0.04213287038201943, 'epoch': 0.71} + 71%|███████ | 367/520 [51:04<09:42, 3.81s/it] 71%|███████ | 368/520 [51:08<09:40, 3.82s/it] {'loss': 1.1288, 'grad_norm': 0.003740571436544966, 'learning_rate': 0.04162563277652104, 'epoch': 0.71} + 71%|███████ | 368/520 [51:08<09:40, 3.82s/it] 71%|███████ | 369/520 [51:12<09:36, 3.82s/it] {'loss': 1.3223, 'grad_norm': 0.003298887467629233, 'learning_rate': 0.04112066325667954, 'epoch': 0.71} + 71%|███████ | 369/520 [51:12<09:36, 3.82s/it] 71%|███████ | 370/520 [51:16<09:33, 3.82s/it] {'loss': 1.1946, 'grad_norm': 0.00312102950500504, 'learning_rate': 0.04061798144264986, 'epoch': 0.71} + 71%|███████ | 370/520 [51:16<09:33, 3.82s/it] 71%|███████▏ | 371/520 [51:20<09:30, 3.83s/it] {'loss': 1.1882, 'grad_norm': 0.003529713925754621, 'learning_rate': 0.04011760686569998, 'epoch': 0.71} + 71%|███████▏ | 371/520 [51:20<09:30, 3.83s/it] 72%|███████▏ | 372/520 [51:24<09:26, 3.83s/it] {'loss': 1.416, 'grad_norm': 0.0032227058951242553, 'learning_rate': 0.03961955896745224, 'epoch': 0.72} + 72%|███████▏ | 372/520 [51:24<09:26, 3.83s/it] 72%|███████▏ | 373/520 [51:28<09:36, 3.92s/it] {'loss': 1.2932, 'grad_norm': 0.003658291926230795, 'learning_rate': 0.03912385709912794, 'epoch': 0.72} + 72%|███████▏ | 373/520 [51:28<09:36, 3.92s/it] 72%|███████▏ | 374/520 [51:32<09:43, 3.99s/it] {'loss': 1.2755, 'grad_norm': 0.0033287387456653315, 'learning_rate': 0.038630520520795276, 'epoch': 0.72} + 72%|███████▏ | 374/520 [51:32<09:43, 3.99s/it] 72%|███████▏ | 375/520 [51:36<09:46, 4.04s/it] {'loss': 1.1835, 'grad_norm': 0.0035579427640495555, 'learning_rate': 0.03813956840062119, 'epoch': 0.72} + 72%|███████▏ | 375/520 [51:36<09:46, 4.04s/it] 72%|███████▏ | 376/520 [51:40<09:41, 4.04s/it] {'loss': 1.3096, 'grad_norm': 0.003258397548130967, 'learning_rate': 0.037651019814126656, 'epoch': 0.72} + 72%|███████▏ | 376/520 [51:40<09:41, 4.04s/it] 72%|███████▎ | 377/520 [51:44<09:28, 3.98s/it] {'loss': 1.2432, 'grad_norm': 0.004293107513298662, 'learning_rate': 0.037164893743445275, 'epoch': 0.72} + 72%|███████▎ | 377/520 [51:44<09:28, 3.98s/it] 73%|███████▎ | 378/520 [51:48<09:18, 3.93s/it] {'loss': 1.2947, 'grad_norm': 0.003312870154483922, 'learning_rate': 0.03668120907658603, 'epoch': 0.73} + 73%|███████▎ | 378/520 [51:48<09:18, 3.93s/it] 73%|███████▎ | 379/520 [51:51<09:09, 3.90s/it] {'loss': 1.2821, 'grad_norm': 0.003200454112072779, 'learning_rate': 0.036199984606699154, 'epoch': 0.73} + 73%|███████▎ | 379/520 [51:52<09:09, 3.90s/it] 73%|███████▎ | 380/520 [51:55<09:03, 3.88s/it] {'loss': 1.4017, 'grad_norm': 0.004065664200211769, 'learning_rate': 0.035721239031346066, 'epoch': 0.73} + 73%|███████▎ | 380/520 [51:55<09:03, 3.88s/it] 73%|███████▎ | 381/520 [51:59<08:58, 3.88s/it] {'loss': 1.2707, 'grad_norm': 0.0034079450968011534, 'learning_rate': 0.03524499095177297, 'epoch': 0.73} + 73%|███████▎ | 381/520 [51:59<08:58, 3.88s/it] 73%|███████▎ | 382/520 [52:03<08:53, 3.87s/it] {'loss': 1.332, 'grad_norm': 0.0033624037390302067, 'learning_rate': 0.03477125887218792, 'epoch': 0.73} + 73%|███████▎ | 382/520 [52:03<08:53, 3.87s/it] 74%|███████▎ | 383/520 [52:07<08:48, 3.86s/it] {'loss': 1.1149, 'grad_norm': 0.0036975715117314803, 'learning_rate': 0.03430006119904196, 'epoch': 0.74} + 74%|███████▎ | 383/520 [52:07<08:48, 3.86s/it] 74%|███████▍ | 384/520 [52:11<08:37, 3.81s/it] {'loss': 1.4467, 'grad_norm': 0.0036887066112302313, 'learning_rate': 0.033831416240314084, 'epoch': 0.74} + 74%|███████▍ | 384/520 [52:11<08:37, 3.81s/it] 74%|███████▍ | 385/520 [52:14<08:26, 3.75s/it] {'loss': 1.2553, 'grad_norm': 0.0033024521647414856, 'learning_rate': 0.03336534220479961, 'epoch': 0.74} + 74%|███████▍ | 385/520 [52:14<08:26, 3.75s/it] 74%|███████▍ | 386/520 [52:18<08:18, 3.72s/it] {'loss': 1.2053, 'grad_norm': 0.0030422635736444153, 'learning_rate': 0.032901857201403005, 'epoch': 0.74} + 74%|███████▍ | 386/520 [52:18<08:18, 3.72s/it] 74%|███████▍ | 387/520 [52:21<08:11, 3.70s/it] {'loss': 1.413, 'grad_norm': 0.003447560456041051, 'learning_rate': 0.032440979238433976, 'epoch': 0.74} + 74%|███████▍ | 387/520 [52:22<08:11, 3.70s/it] 75%|███████▍ | 388/520 [52:25<08:06, 3.68s/it] {'loss': 1.1523, 'grad_norm': 0.0031604379005835065, 'learning_rate': 0.03198272622290804, 'epoch': 0.75} + 75%|███████▍ | 388/520 [52:25<08:06, 3.68s/it] 75%|███████▍ | 389/520 [52:29<08:00, 3.67s/it] {'loss': 1.217, 'grad_norm': 0.003830913606774636, 'learning_rate': 0.03152711595985065, 'epoch': 0.75} + 75%|███████▍ | 389/520 [52:29<08:00, 3.67s/it] 75%|███████▌ | 390/520 [52:32<07:56, 3.67s/it] {'loss': 1.269, 'grad_norm': 0.003204143108967365, 'learning_rate': 0.031074166151605298, 'epoch': 0.75} + 75%|███████▌ | 390/520 [52:32<07:56, 3.67s/it] 75%|███████▌ | 391/520 [52:36<07:52, 3.67s/it] {'loss': 1.3542, 'grad_norm': 0.003483128046061712, 'learning_rate': 0.030623894397145836, 'epoch': 0.75} + 75%|███████▌ | 391/520 [52:36<07:52, 3.67s/it] 75%|███████▌ | 392/520 [52:40<07:48, 3.66s/it] {'loss': 1.1682, 'grad_norm': 0.0033363172372213017, 'learning_rate': 0.03017631819139273, 'epoch': 0.75} + 75%|███████▌ | 392/520 [52:40<07:48, 3.66s/it] 76%|███████▌ | 393/520 [52:43<07:43, 3.65s/it] {'loss': 1.2162, 'grad_norm': 0.0032190020725585467, 'learning_rate': 0.029731454924533086, 'epoch': 0.76} + 76%|███████▌ | 393/520 [52:43<07:43, 3.65s/it] 76%|███████▌ | 394/520 [52:47<07:39, 3.64s/it] {'loss': 1.2312, 'grad_norm': 0.0037661258781197552, 'learning_rate': 0.029289321881345254, 'epoch': 0.76} + 76%|███████▌ | 394/520 [52:47<07:39, 3.64s/it] 76%|███████▌ | 395/520 [52:51<07:34, 3.64s/it] {'loss': 1.1894, 'grad_norm': 0.0037131642994389837, 'learning_rate': 0.028849936240527008, 'epoch': 0.76} + 76%|███████▌ | 395/520 [52:51<07:34, 3.64s/it] 76%|███████▌ | 396/520 [52:54<07:31, 3.64s/it] {'loss': 1.2744, 'grad_norm': 0.0035255116564252894, 'learning_rate': 0.028413315074028157, 'epoch': 0.76} + 76%|███████▌ | 396/520 [52:54<07:31, 3.64s/it] 76%|███████▋ | 397/520 [52:58<07:27, 3.64s/it] {'loss': 1.2611, 'grad_norm': 0.0032675428284185265, 'learning_rate': 0.027979475346387363, 'epoch': 0.76} + 76%|███████▋ | 397/520 [52:58<07:27, 3.64s/it] 77%|███████▋ | 398/520 [53:02<07:23, 3.64s/it] {'loss': 1.2453, 'grad_norm': 0.0035431249056562604, 'learning_rate': 0.027548433914072735, 'epoch': 0.77} + 77%|███████▋ | 398/520 [53:02<07:23, 3.64s/it] 77%|███████▋ | 399/520 [53:05<07:26, 3.69s/it] {'loss': 1.274, 'grad_norm': 0.0037020505133212142, 'learning_rate': 0.027120207524827168, 'epoch': 0.77} + 77%|███████▋ | 399/520 [53:05<07:26, 3.69s/it] 77%|███████▋ | 400/520 [53:09<07:34, 3.78s/it] {'loss': 1.3148, 'grad_norm': 0.003460450111701259, 'learning_rate': 0.02669481281701739, 'epoch': 0.77} + 77%|███████▋ | 400/520 [53:09<07:34, 3.78s/it] 77%|███████▋ | 401/520 [53:14<07:43, 3.90s/it] {'loss': 1.0818, 'grad_norm': 0.003571472347547154, 'learning_rate': 0.026272266318987603, 'epoch': 0.77} + 77%|███████▋ | 401/520 [53:14<07:43, 3.90s/it] 77%|███████▋ | 402/520 [53:18<07:48, 3.97s/it] {'loss': 1.1918, 'grad_norm': 0.0033534129762451785, 'learning_rate': 0.02585258444841733, 'epoch': 0.77} + 77%|███████▋ | 402/520 [53:18<07:48, 3.97s/it] 78%|███████▊ | 403/520 [53:22<07:44, 3.97s/it] {'loss': 1.226, 'grad_norm': 0.0037035354692649203, 'learning_rate': 0.025435783511683442, 'epoch': 0.78} + 78%|███████▊ | 403/520 [53:22<07:44, 3.97s/it] 78%|███████▊ | 404/520 [53:25<07:36, 3.94s/it] {'loss': 1.1421, 'grad_norm': 0.0041577136313601295, 'learning_rate': 0.02502187970322657, 'epoch': 0.78} + 78%|███████▊ | 404/520 [53:26<07:36, 3.94s/it] 78%|███████▊ | 405/520 [53:29<07:30, 3.92s/it] {'loss': 1.2664, 'grad_norm': 0.0032268985318865163, 'learning_rate': 0.02461088910492202, 'epoch': 0.78} + 78%|███████▊ | 405/520 [53:29<07:30, 3.92s/it] 78%|███████▊ | 406/520 [53:33<07:24, 3.90s/it] {'loss': 1.2112, 'grad_norm': 0.004127381341049021, 'learning_rate': 0.02420282768545469, 'epoch': 0.78} + 78%|███████▊ | 406/520 [53:33<07:24, 3.90s/it] 78%|███████▊ | 407/520 [53:37<07:18, 3.88s/it] {'loss': 1.3271, 'grad_norm': 0.0035242749344864275, 'learning_rate': 0.02379771129969892, 'epoch': 0.78} + 78%|███████▊ | 407/520 [53:37<07:18, 3.88s/it] 78%|███████▊ | 408/520 [53:41<07:06, 3.81s/it] {'loss': 1.221, 'grad_norm': 0.003655353593353105, 'learning_rate': 0.023395555688102213, 'epoch': 0.78} + 78%|███████▊ | 408/520 [53:41<07:06, 3.81s/it] 79%|███████▊ | 409/520 [53:44<06:57, 3.76s/it] {'loss': 1.3447, 'grad_norm': 0.00402499399188512, 'learning_rate': 0.02299637647607372, 'epoch': 0.79} + 79%|███████▊ | 409/520 [53:44<06:57, 3.76s/it] 79%|███████▉ | 410/520 [53:48<06:48, 3.71s/it] {'loss': 1.0571, 'grad_norm': 0.0033932407104098022, 'learning_rate': 0.022600189173377264, 'epoch': 0.79} + 79%|███████▉ | 410/520 [53:48<06:48, 3.71s/it] 79%|███████▉ | 411/520 [53:52<06:41, 3.68s/it] {'loss': 1.313, 'grad_norm': 0.003805198455258464, 'learning_rate': 0.022207009173528525, 'epoch': 0.79} + 79%|███████▉ | 411/520 [53:52<06:41, 3.68s/it] 79%|███████▉ | 412/520 [53:55<06:35, 3.66s/it] {'loss': 1.2304, 'grad_norm': 0.0034026243703978772, 'learning_rate': 0.02181685175319702, 'epoch': 0.79} + 79%|███████▉ | 412/520 [53:55<06:35, 3.66s/it] 79%|███████▉ | 413/520 [53:59<06:30, 3.65s/it] {'loss': 1.3078, 'grad_norm': 0.00340010490326893, 'learning_rate': 0.021429732071612653, 'epoch': 0.79} + 79%|███████▉ | 413/520 [53:59<06:30, 3.65s/it] 80%|███████▉ | 414/520 [54:02<06:26, 3.65s/it] {'loss': 1.096, 'grad_norm': 0.003074577504582332, 'learning_rate': 0.02104566516997647, 'epoch': 0.8} + 80%|███████▉ | 414/520 [54:02<06:26, 3.65s/it] 80%|███████▉ | 415/520 [54:06<06:22, 3.65s/it] {'loss': 1.2066, 'grad_norm': 0.0032734610740577474, 'learning_rate': 0.020664665970876496, 'epoch': 0.8} + 80%|███████▉ | 415/520 [54:06<06:22, 3.65s/it] 80%|████████ | 416/520 [54:10<06:18, 3.64s/it] {'loss': 1.1237, 'grad_norm': 0.004039260652320527, 'learning_rate': 0.020286749277707784, 'epoch': 0.8} + 80%|████████ | 416/520 [54:10<06:18, 3.64s/it] 80%|████████ | 417/520 [54:13<06:14, 3.64s/it] {'loss': 1.2972, 'grad_norm': 0.003909034595537411, 'learning_rate': 0.019911929774097215, 'epoch': 0.8} + 80%|████████ | 417/520 [54:13<06:14, 3.64s/it] 80%|████████ | 418/520 [54:17<06:11, 3.65s/it] {'loss': 1.2706, 'grad_norm': 0.003227205004481354, 'learning_rate': 0.019540222023333165, 'epoch': 0.8} + 80%|████████ | 418/520 [54:17<06:11, 3.65s/it] 81%|████████ | 419/520 [54:21<06:07, 3.64s/it] {'loss': 1.2554, 'grad_norm': 0.003608855993148585, 'learning_rate': 0.01917164046779948, 'epoch': 0.81} + 81%|████████ | 419/520 [54:21<06:07, 3.64s/it] 81%|████████ | 420/520 [54:24<06:03, 3.63s/it] {'loss': 1.1389, 'grad_norm': 0.003560983669384936, 'learning_rate': 0.018806199428414352, 'epoch': 0.81} + 81%|████████ | 420/520 [54:24<06:03, 3.63s/it] 81%|████████ | 421/520 [54:28<06:00, 3.65s/it] {'loss': 1.0746, 'grad_norm': 0.0036057014595561846, 'learning_rate': 0.018443913104073985, 'epoch': 0.81} + 81%|████████ | 421/520 [54:28<06:00, 3.65s/it] 81%|████████ | 422/520 [54:32<05:56, 3.64s/it] {'loss': 1.2013, 'grad_norm': 0.003816967184224707, 'learning_rate': 0.01808479557110081, 'epoch': 0.81} + 81%|████████ | 422/520 [54:32<05:56, 3.64s/it] 81%|████████▏ | 423/520 [54:35<05:52, 3.63s/it] {'loss': 1.1869, 'grad_norm': 0.003916884297521358, 'learning_rate': 0.017728860782696667, 'epoch': 0.81} + 81%|████████▏ | 423/520 [54:35<05:52, 3.63s/it] 82%|████████▏ | 424/520 [54:39<05:49, 3.64s/it] {'loss': 1.3841, 'grad_norm': 0.003787442008907636, 'learning_rate': 0.017376122568400532, 'epoch': 0.82} + 82%|████████▏ | 424/520 [54:39<05:49, 3.64s/it] 82%|████████▏ | 425/520 [54:42<05:45, 3.64s/it] {'loss': 1.1925, 'grad_norm': 0.003250585625077971, 'learning_rate': 0.017026594633551252, 'epoch': 0.82} + 82%|████████▏ | 425/520 [54:42<05:45, 3.64s/it] 82%|████████▏ | 426/520 [54:46<05:42, 3.64s/it] {'loss': 1.2279, 'grad_norm': 0.004452222114689869, 'learning_rate': 0.01668029055875512, 'epoch': 0.82} + 82%|████████▏ | 426/520 [54:46<05:42, 3.64s/it] 82%|████████▏ | 427/520 [54:50<05:38, 3.64s/it] {'loss': 1.1278, 'grad_norm': 0.0032498340342846203, 'learning_rate': 0.016337223799358026, 'epoch': 0.82} + 82%|████████▏ | 427/520 [54:50<05:38, 3.64s/it] 82%|████████▏ | 428/520 [54:53<05:35, 3.65s/it] {'loss': 1.101, 'grad_norm': 0.003512107702361509, 'learning_rate': 0.01599740768492286, 'epoch': 0.82} + 82%|████████▏ | 428/520 [54:53<05:35, 3.65s/it] 82%|████████▎ | 429/520 [54:57<05:32, 3.66s/it] {'loss': 1.2126, 'grad_norm': 0.0033047988762930987, 'learning_rate': 0.015660855418711452, 'epoch': 0.82} + 82%|████████▎ | 429/520 [54:57<05:32, 3.66s/it]Token indices sequence length is longer than the specified maximum sequence length for this model (2076 > 2048). Running this sequence through the model will result in indexing errors + 83%|████████▎ | 430/520 [55:01<05:27, 3.64s/it] {'loss': 1.2041, 'grad_norm': 0.003175761716822632, 'learning_rate': 0.015327580077171589, 'epoch': 0.83} + 83%|████████▎ | 430/520 [55:01<05:27, 3.64s/it] 83%|████████▎ | 431/520 [55:04<05:22, 3.63s/it] {'loss': 1.2614, 'grad_norm': 0.003854392804853736, 'learning_rate': 0.014997594609429088, 'epoch': 0.83} + 83%|████████▎ | 431/520 [55:04<05:22, 3.63s/it] 83%|████████▎ | 432/520 [55:08<05:19, 3.63s/it] {'loss': 1.1122, 'grad_norm': 0.003848102889409118, 'learning_rate': 0.01467091183678444, 'epoch': 0.83} + 83%|████████▎ | 432/520 [55:08<05:19, 3.63s/it] 83%|████████▎ | 433/520 [55:12<05:18, 3.66s/it] {'loss': 1.2507, 'grad_norm': 0.0033818846599170385, 'learning_rate': 0.014347544452214867, 'epoch': 0.83} + 83%|████████▎ | 433/520 [55:12<05:18, 3.66s/it] 83%|████████▎ | 434/520 [55:15<05:19, 3.71s/it] {'loss': 0.9898, 'grad_norm': 0.003232108118815194, 'learning_rate': 0.014027505019880971, 'epoch': 0.83} + 83%|████████▎ | 434/520 [55:15<05:19, 3.71s/it] 84%|████████▎ | 435/520 [55:19<05:19, 3.76s/it] {'loss': 1.2829, 'grad_norm': 0.0037286828239841536, 'learning_rate': 0.013710805974638696, 'epoch': 0.84} + 84%|████████▎ | 435/520 [55:19<05:19, 3.76s/it] 84%|████████▍ | 436/520 [55:23<05:17, 3.78s/it] {'loss': 1.0793, 'grad_norm': 0.0033781588316481657, 'learning_rate': 0.01339745962155613, 'epoch': 0.84} + 84%|████████▍ | 436/520 [55:23<05:17, 3.78s/it] 84%|████████▍ | 437/520 [55:27<05:11, 3.75s/it] {'loss': 1.317, 'grad_norm': 0.0034538019185198407, 'learning_rate': 0.01308747813543536, 'epoch': 0.84} + 84%|████████▍ | 437/520 [55:27<05:11, 3.75s/it] 84%|████████▍ | 438/520 [55:31<05:05, 3.72s/it] {'loss': 1.117, 'grad_norm': 0.003352154243948411, 'learning_rate': 0.012780873560339467, 'epoch': 0.84} + 84%|████████▍ | 438/520 [55:31<05:05, 3.72s/it] 84%|████████▍ | 439/520 [55:34<04:59, 3.69s/it] {'loss': 1.2352, 'grad_norm': 0.003090079018990124, 'learning_rate': 0.012477657809124632, 'epoch': 0.84} + 84%|████████▍ | 439/520 [55:34<04:59, 3.69s/it] 85%|████████▍ | 440/520 [55:38<04:54, 3.69s/it] {'loss': 1.1687, 'grad_norm': 0.0033685511286661546, 'learning_rate': 0.012177842662977134, 'epoch': 0.85} + 85%|████████▍ | 440/520 [55:38<04:54, 3.69s/it] 85%|████████▍ | 441/520 [55:41<04:49, 3.67s/it] {'loss': 1.2678, 'grad_norm': 0.0034512054546632616, 'learning_rate': 0.01188143977095576, 'epoch': 0.85} + 85%|████████▍ | 441/520 [55:41<04:49, 3.67s/it] 85%|████████▌ | 442/520 [55:45<04:45, 3.66s/it] {'loss': 1.2221, 'grad_norm': 0.003895376032642527, 'learning_rate': 0.011588460649539035, 'epoch': 0.85} + 85%|████████▌ | 442/520 [55:45<04:45, 3.66s/it] 85%|████████▌ | 443/520 [55:49<04:40, 3.65s/it] {'loss': 1.2463, 'grad_norm': 0.0035516737186484246, 'learning_rate': 0.011298916682177829, 'epoch': 0.85} + 85%|████████▌ | 443/520 [55:49<04:40, 3.65s/it] 85%|████████▌ | 444/520 [55:52<04:37, 3.66s/it] {'loss': 1.2148, 'grad_norm': 0.0031207676491493805, 'learning_rate': 0.011012819118853146, 'epoch': 0.85} + 85%|████████▌ | 444/520 [55:52<04:37, 3.66s/it] 86%|████████▌ | 445/520 [55:56<04:32, 3.64s/it] {'loss': 1.133, 'grad_norm': 0.0033871557613706503, 'learning_rate': 0.01073017907563887, 'epoch': 0.86} + 86%|████████▌ | 445/520 [55:56<04:32, 3.64s/it] 86%|████████▌ | 446/520 [56:00<04:30, 3.66s/it] {'loss': 1.3425, 'grad_norm': 0.0035539705401286727, 'learning_rate': 0.010451007534269908, 'epoch': 0.86} + 86%|████████▌ | 446/520 [56:00<04:30, 3.66s/it] 86%|████████▌ | 447/520 [56:03<04:27, 3.67s/it] {'loss': 1.2159, 'grad_norm': 0.003384642906191233, 'learning_rate': 0.010175315341715598, 'epoch': 0.86} + 86%|████████▌ | 447/520 [56:03<04:27, 3.67s/it] 86%|████████▌ | 448/520 [56:07<04:23, 3.65s/it] {'loss': 1.1911, 'grad_norm': 0.003517662839678474, 'learning_rate': 0.009903113209758098, 'epoch': 0.86} + 86%|████████▌ | 448/520 [56:07<04:23, 3.65s/it] 86%|████████▋ | 449/520 [56:11<04:18, 3.63s/it] {'loss': 1.2926, 'grad_norm': 0.0035132756486898576, 'learning_rate': 0.009634411714576352, 'epoch': 0.86} + 86%|████████▋ | 449/520 [56:11<04:18, 3.63s/it] 87%|████████▋ | 450/520 [56:14<04:14, 3.63s/it] {'loss': 1.2382, 'grad_norm': 0.003408500435363446, 'learning_rate': 0.009369221296335007, 'epoch': 0.87} + 87%|████████▋ | 450/520 [56:14<04:14, 3.63s/it] 87%|████████▋ | 451/520 [56:18<04:10, 3.62s/it] {'loss': 1.2274, 'grad_norm': 0.0033992194908634673, 'learning_rate': 0.009107552258778906, 'epoch': 0.87} + 87%|████████▋ | 451/520 [56:18<04:10, 3.62s/it] 87%|████████▋ | 452/520 [56:21<04:06, 3.62s/it] {'loss': 1.33, 'grad_norm': 0.0033291100803524054, 'learning_rate': 0.008849414768832687, 'epoch': 0.87} + 87%|████████▋ | 452/520 [56:21<04:06, 3.62s/it] 87%|████████▋ | 453/520 [56:25<04:02, 3.62s/it] {'loss': 1.3069, 'grad_norm': 0.0035821194199880894, 'learning_rate': 0.008594818856205699, 'epoch': 0.87} + 87%|████████▋ | 453/520 [56:25<04:02, 3.62s/it] 87%|████████▋ | 454/520 [56:29<03:59, 3.62s/it] {'loss': 1.1408, 'grad_norm': 0.0037731583803362716, 'learning_rate': 0.00834377441300238, 'epoch': 0.87} + 87%|████████▋ | 454/520 [56:29<03:59, 3.62s/it] 88%|████████▊ | 455/520 [56:32<03:55, 3.62s/it] {'loss': 1.2699, 'grad_norm': 0.0033350588247176693, 'learning_rate': 0.008096291193337934, 'epoch': 0.88} + 88%|████████▊ | 455/520 [56:32<03:55, 3.62s/it] 88%|████████▊ | 456/520 [56:36<03:51, 3.61s/it] {'loss': 1.194, 'grad_norm': 0.0033938722039167095, 'learning_rate': 0.007852378812959226, 'epoch': 0.88} + 88%|████████▊ | 456/520 [56:36<03:51, 3.61s/it] 88%|████████▊ | 457/520 [56:39<03:47, 3.61s/it] {'loss': 1.2798, 'grad_norm': 0.003418542029101496, 'learning_rate': 0.007612046748871327, 'epoch': 0.88} + 88%|████████▊ | 457/520 [56:39<03:47, 3.61s/it] 88%|████████▊ | 458/520 [56:43<03:44, 3.63s/it] {'loss': 1.3405, 'grad_norm': 0.0036291941326611196, 'learning_rate': 0.007375304338969136, 'epoch': 0.88} + 88%|████████▊ | 458/520 [56:43<03:44, 3.63s/it] 88%|████████▊ | 459/520 [56:47<03:40, 3.62s/it] {'loss': 1.2726, 'grad_norm': 0.0033936107494467846, 'learning_rate': 0.007142160781674645, 'epoch': 0.88} + 88%|████████▊ | 459/520 [56:47<03:40, 3.62s/it] 88%|████████▊ | 460/520 [56:50<03:37, 3.62s/it] {'loss': 1.1385, 'grad_norm': 0.0032579960376884196, 'learning_rate': 0.006912625135579587, 'epoch': 0.88} + 88%|████████▊ | 460/520 [56:50<03:37, 3.62s/it] 89%|████████▊ | 461/520 [56:54<03:34, 3.63s/it] {'loss': 1.3564, 'grad_norm': 0.0028275841260925623, 'learning_rate': 0.0066867063190933496, 'epoch': 0.89} + 89%|████████▊ | 461/520 [56:54<03:34, 3.63s/it] 89%|████████▉ | 462/520 [56:58<03:31, 3.64s/it] {'loss': 1.3798, 'grad_norm': 0.0032599632031589405, 'learning_rate': 0.006464413110096601, 'epoch': 0.89} + 89%|████████▉ | 462/520 [56:58<03:31, 3.64s/it] 89%|████████▉ | 463/520 [57:01<03:27, 3.64s/it] {'loss': 1.1036, 'grad_norm': 0.003514537214880104, 'learning_rate': 0.006245754145600091, 'epoch': 0.89} + 89%|████████▉ | 463/520 [57:01<03:27, 3.64s/it] 89%|████████▉ | 464/520 [57:05<03:23, 3.64s/it] {'loss': 1.254, 'grad_norm': 0.003549660898839814, 'learning_rate': 0.006030737921409169, 'epoch': 0.89} + 89%|████████▉ | 464/520 [57:05<03:23, 3.64s/it] 89%|████████▉ | 465/520 [57:09<03:19, 3.64s/it] {'loss': 1.3673, 'grad_norm': 0.0039328583765327095, 'learning_rate': 0.005819372791793654, 'epoch': 0.89} + 89%|████████▉ | 465/520 [57:09<03:19, 3.64s/it] 90%|████████▉ | 466/520 [57:12<03:16, 3.64s/it] {'loss': 1.2362, 'grad_norm': 0.0030754865270199087, 'learning_rate': 0.005611666969163243, 'epoch': 0.9} + 90%|████████▉ | 466/520 [57:12<03:16, 3.64s/it] 90%|████████▉ | 467/520 [57:16<03:12, 3.64s/it] {'loss': 1.2636, 'grad_norm': 0.003255908345363593, 'learning_rate': 0.005407628523748398, 'epoch': 0.9} + 90%|████████▉ | 467/520 [57:16<03:12, 3.64s/it] 90%|█████████ | 468/520 [57:20<03:09, 3.64s/it] {'loss': 1.2136, 'grad_norm': 0.003765811207200945, 'learning_rate': 0.00520726538328683, 'epoch': 0.9} + 90%|█████████ | 468/520 [57:20<03:09, 3.64s/it] 90%|█████████ | 469/520 [57:23<03:06, 3.65s/it] {'loss': 1.2726, 'grad_norm': 0.003618060679352385, 'learning_rate': 0.005010585332715401, 'epoch': 0.9} + 90%|█████████ | 469/520 [57:23<03:06, 3.65s/it] 90%|█████████ | 470/520 [57:27<03:02, 3.65s/it] {'loss': 1.1485, 'grad_norm': 0.002999090809388193, 'learning_rate': 0.004817596013867765, 'epoch': 0.9} + 90%|█████████ | 470/520 [57:27<03:02, 3.65s/it] 91%|█████████ | 471/520 [57:30<02:58, 3.64s/it] {'loss': 1.1654, 'grad_norm': 0.003415223214033775, 'learning_rate': 0.004628304925177318, 'epoch': 0.91} + 91%|█████████ | 471/520 [57:30<02:58, 3.64s/it] 91%|█████████ | 472/520 [57:34<02:54, 3.65s/it] {'loss': 1.14, 'grad_norm': 0.0034651004630823694, 'learning_rate': 0.004442719421385921, 'epoch': 0.91} + 91%|█████████ | 472/520 [57:34<02:54, 3.65s/it] 91%|█████████ | 473/520 [57:38<02:51, 3.65s/it] {'loss': 1.1955, 'grad_norm': 0.00342811182753492, 'learning_rate': 0.004260846713258193, 'epoch': 0.91} + 91%|█████████ | 473/520 [57:38<02:51, 3.65s/it] 91%|█████████ | 474/520 [57:41<02:47, 3.64s/it] {'loss': 1.2951, 'grad_norm': 0.003099887000273694, 'learning_rate': 0.004082693867301224, 'epoch': 0.91} + 91%|█████████ | 474/520 [57:41<02:47, 3.64s/it] 91%|█████████▏| 475/520 [57:45<02:44, 3.65s/it] {'loss': 1.2055, 'grad_norm': 0.0031683417010907296, 'learning_rate': 0.003908267805490051, 'epoch': 0.91} + 91%|█████████▏| 475/520 [57:45<02:44, 3.65s/it] 92%|█████████▏| 476/520 [57:49<02:40, 3.64s/it] {'loss': 1.1931, 'grad_norm': 0.003544114211331306, 'learning_rate': 0.003737575304998797, 'epoch': 0.92} + 92%|█████████▏| 476/520 [57:49<02:40, 3.64s/it] 92%|█████████▏| 477/520 [57:52<02:36, 3.63s/it] {'loss': 1.1809, 'grad_norm': 0.003899922594314614, 'learning_rate': 0.003570622997937234, 'epoch': 0.92} + 92%|█████████▏| 477/520 [57:52<02:36, 3.63s/it] 92%|█████████▏| 478/520 [57:56<02:32, 3.63s/it] {'loss': 1.1398, 'grad_norm': 0.003391093908339409, 'learning_rate': 0.00340741737109318, 'epoch': 0.92} + 92%|█████████▏| 478/520 [57:56<02:32, 3.63s/it] 92%|█████████▏| 479/520 [58:00<02:28, 3.63s/it] {'loss': 1.2674, 'grad_norm': 0.0036324598640141495, 'learning_rate': 0.003247964765680389, 'epoch': 0.92} + 92%|█████████▏| 479/520 [58:00<02:28, 3.63s/it] 92%|█████████▏| 480/520 [58:03<02:25, 3.65s/it] {'loss': 1.2898, 'grad_norm': 0.0032663069044733843, 'learning_rate': 0.0030922713770922153, 'epoch': 0.92} + 92%|█████████▏| 480/520 [58:03<02:25, 3.65s/it] 92%|█████████▎| 481/520 [58:07<02:22, 3.65s/it] {'loss': 1.301, 'grad_norm': 0.0032476158662820514, 'learning_rate': 0.0029403432546609046, 'epoch': 0.93} + 92%|█████████▎| 481/520 [58:07<02:22, 3.65s/it] 93%|█████████▎| 482/520 [58:11<02:18, 3.65s/it] {'loss': 1.3059, 'grad_norm': 0.0038010947190573836, 'learning_rate': 0.0027921863014225504, 'epoch': 0.93} + 93%|█████████▎| 482/520 [58:11<02:18, 3.65s/it] 93%|█████████▎| 483/520 [58:14<02:15, 3.65s/it] {'loss': 1.2138, 'grad_norm': 0.003614996085757096, 'learning_rate': 0.002647806273887665, 'epoch': 0.93} + 93%|█████████▎| 483/520 [58:14<02:15, 3.65s/it] 93%|█████████▎| 484/520 [58:18<02:11, 3.65s/it] {'loss': 1.2167, 'grad_norm': 0.0034160193533761154, 'learning_rate': 0.0025072087818176383, 'epoch': 0.93} + 93%|█████████▎| 484/520 [58:18<02:11, 3.65s/it] 93%|█████████▎| 485/520 [58:21<02:07, 3.64s/it] {'loss': 1.1603, 'grad_norm': 0.003249981893976588, 'learning_rate': 0.002370399288006664, 'epoch': 0.93} + 93%|█████████▎| 485/520 [58:21<02:07, 3.64s/it] 93%|█████████▎| 486/520 [58:25<02:03, 3.64s/it] {'loss': 1.2866, 'grad_norm': 0.0035145187581953337, 'learning_rate': 0.0022373831080695463, 'epoch': 0.93} + 93%|█████████▎| 486/520 [58:25<02:03, 3.64s/it] 94%|█████████▎| 487/520 [58:29<02:00, 3.64s/it] {'loss': 1.1374, 'grad_norm': 0.0032038359551554395, 'learning_rate': 0.0021081654102351635, 'epoch': 0.94} + 94%|█████████▎| 487/520 [58:29<02:00, 3.64s/it] 94%|█████████▍| 488/520 [58:32<01:57, 3.66s/it] {'loss': 1.0789, 'grad_norm': 0.003429372963666013, 'learning_rate': 0.0019827512151456175, 'epoch': 0.94} + 94%|█████████▍| 488/520 [58:32<01:57, 3.66s/it] 94%|█████████▍| 489/520 [58:36<01:53, 3.66s/it] {'loss': 1.2905, 'grad_norm': 0.0028827245036051607, 'learning_rate': 0.0018611453956612345, 'epoch': 0.94} + 94%|█████████▍| 489/520 [58:36<01:53, 3.66s/it] 94%|█████████▍| 490/520 [58:40<01:49, 3.66s/it] {'loss': 1.2033, 'grad_norm': 0.0036237013736954335, 'learning_rate': 0.0017433526766711727, 'epoch': 0.94} + 94%|█████████▍| 490/520 [58:40<01:49, 3.66s/it] 94%|█████████▍| 491/520 [58:43<01:45, 3.65s/it] {'loss': 1.1641, 'grad_norm': 0.003412441843929199, 'learning_rate': 0.0016293776349098677, 'epoch': 0.94} + 94%|█████████▍| 491/520 [58:43<01:45, 3.65s/it] 95%|█████████▍| 492/520 [58:47<01:42, 3.66s/it] {'loss': 1.2927, 'grad_norm': 0.003633045430976383, 'learning_rate': 0.0015192246987791981, 'epoch': 0.95} + 95%|█████████▍| 492/520 [58:47<01:42, 3.66s/it] 95%|█████████▍| 493/520 [58:51<01:38, 3.64s/it] {'loss': 1.3358, 'grad_norm': 0.003517870248536432, 'learning_rate': 0.0014128981481764114, 'epoch': 0.95} + 95%|█████████▍| 493/520 [58:51<01:38, 3.64s/it] 95%|█████████▌| 494/520 [58:54<01:34, 3.65s/it] {'loss': 1.2179, 'grad_norm': 0.0031011019448468477, 'learning_rate': 0.0013104021143278911, 'epoch': 0.95} + 95%|█████████▌| 494/520 [58:54<01:34, 3.65s/it] 95%|█████████▌| 495/520 [58:58<01:30, 3.64s/it] {'loss': 1.1795, 'grad_norm': 0.0033085997539149437, 'learning_rate': 0.0012117405796285285, 'epoch': 0.95} + 95%|█████████▌| 495/520 [58:58<01:30, 3.64s/it] 95%|█████████▌| 496/520 [59:02<01:27, 3.64s/it] {'loss': 1.096, 'grad_norm': 0.0034974821769102418, 'learning_rate': 0.0011169173774871477, 'epoch': 0.95} + 95%|█████████▌| 496/520 [59:02<01:27, 3.64s/it] 96%|█████████▌| 497/520 [59:05<01:24, 3.68s/it] {'loss': 1.2238, 'grad_norm': 0.0029826819454672614, 'learning_rate': 0.0010259361921774012, 'epoch': 0.96} + 96%|█████████▌| 497/520 [59:05<01:24, 3.68s/it] 96%|█████████▌| 498/520 [59:09<01:22, 3.74s/it] {'loss': 1.1833, 'grad_norm': 0.0033931719276308644, 'learning_rate': 0.000938800558694719, 'epoch': 0.96} + 96%|█████████▌| 498/520 [59:09<01:22, 3.74s/it] 96%|█████████▌| 499/520 [59:13<01:19, 3.78s/it] {'loss': 1.3618, 'grad_norm': 0.003386511042859323, 'learning_rate': 0.0008555138626189618, 'epoch': 0.96} + 96%|█████████▌| 499/520 [59:13<01:19, 3.78s/it] 96%|█████████▌| 500/520 [59:17<01:16, 3.82s/it] {'loss': 1.3023, 'grad_norm': 0.0039395824979370315, 'learning_rate': 0.0007760793399827937, 'epoch': 0.96} + 96%|█████████▌| 500/520 [59:17<01:16, 3.82s/it] 96%|█████████▋| 501/520 [59:21<01:12, 3.83s/it] {'loss': 1.2811, 'grad_norm': 0.0037847699178282083, 'learning_rate': 0.000700500077146038, 'epoch': 0.96} + 96%|█████████▋| 501/520 [59:21<01:12, 3.83s/it] 97%|█████████▋| 502/520 [59:25<01:09, 3.86s/it] {'loss': 1.2199, 'grad_norm': 0.0031964380998923798, 'learning_rate': 0.0006287790106757397, 'epoch': 0.97} + 97%|█████████▋| 502/520 [59:25<01:09, 3.86s/it] 97%|█████████▋| 503/520 [59:29<01:05, 3.86s/it] {'loss': 1.2591, 'grad_norm': 0.00345488264650552, 'learning_rate': 0.0005609189272320237, 'epoch': 0.97} + 97%|█████████▋| 503/520 [59:29<01:05, 3.86s/it] 97%|█████████▋| 504/520 [59:33<01:01, 3.86s/it] {'loss': 1.2201, 'grad_norm': 0.004101380668768032, 'learning_rate': 0.000496922463459859, 'epoch': 0.97} + 97%|█████████▋| 504/520 [59:33<01:01, 3.86s/it] 97%|█████████▋| 505/520 [59:36<00:57, 3.86s/it] {'loss': 1.2585, 'grad_norm': 0.003370954738809132, 'learning_rate': 0.0004367921058866187, 'epoch': 0.97} + 97%|█████████▋| 505/520 [59:36<00:57, 3.86s/it] 97%|█████████▋| 506/520 [59:40<00:54, 3.88s/it] {'loss': 1.1746, 'grad_norm': 0.0036604918142711953, 'learning_rate': 0.0003805301908254455, 'epoch': 0.97} + 97%|█████████▋| 506/520 [59:40<00:54, 3.88s/it] 98%|█████████▊| 507/520 [59:44<00:50, 3.87s/it] {'loss': 1.4056, 'grad_norm': 0.003158819109885269, 'learning_rate': 0.0003281389042844918, 'epoch': 0.97} + 98%|█████████▊| 507/520 [59:44<00:50, 3.87s/it] 98%|█████████▊| 508/520 [59:48<00:46, 3.87s/it] {'loss': 1.2876, 'grad_norm': 0.0033929110320658814, 'learning_rate': 0.00027962028188198705, 'epoch': 0.98} + 98%|█████████▊| 508/520 [59:48<00:46, 3.87s/it] 98%|█████████▊| 509/520 [59:52<00:42, 3.87s/it] {'loss': 1.2626, 'grad_norm': 0.0033149781952766864, 'learning_rate': 0.00023497620876711256, 'epoch': 0.98} + 98%|█████████▊| 509/520 [59:52<00:42, 3.87s/it] 98%|█████████▊| 510/520 [59:56<00:38, 3.86s/it] {'loss': 1.214, 'grad_norm': 0.003331043660990423, 'learning_rate': 0.00019420841954681523, 'epoch': 0.98} + 98%|█████████▊| 510/520 [59:56<00:38, 3.86s/it] 98%|█████████▊| 511/520 [1:00:00<00:34, 3.84s/it] {'loss': 1.1869, 'grad_norm': 0.0032345094412991604, 'learning_rate': 0.00015731849821833956, 'epoch': 0.98} + 98%|█████████▊| 511/520 [1:00:00<00:34, 3.84s/it] 98%|█████████▊| 512/520 [1:00:03<00:30, 3.83s/it] {'loss': 1.0673, 'grad_norm': 0.003256207501708509, 'learning_rate': 0.00012430787810776555, 'epoch': 0.98} + 98%|█████████▊| 512/520 [1:00:03<00:30, 3.83s/it] 99%|█████████▊| 513/520 [1:00:07<00:26, 3.83s/it] {'loss': 1.273, 'grad_norm': 0.0038419221329909453, 'learning_rate': 9.517784181422018e-05, 'epoch': 0.99} + 99%|█████████▊| 513/520 [1:00:07<00:26, 3.83s/it] 99%|█████████▉| 514/520 [1:00:11<00:22, 3.82s/it] {'loss': 1.2468, 'grad_norm': 0.00310705551609099, 'learning_rate': 6.992952116013917e-05, 'epoch': 0.99} + 99%|█████████▉| 514/520 [1:00:11<00:22, 3.82s/it] 99%|█████████▉| 515/520 [1:00:15<00:19, 3.82s/it] {'loss': 1.2922, 'grad_norm': 0.003965858924957146, 'learning_rate': 4.856389714723575e-05, 'epoch': 0.99} + 99%|█████████▉| 515/520 [1:00:15<00:19, 3.82s/it] 99%|█████████▉| 516/520 [1:00:19<00:15, 3.79s/it] {'loss': 1.1855, 'grad_norm': 0.0032694399769498915, 'learning_rate': 3.108179991837545e-05, 'epoch': 0.99} + 99%|█████████▉| 516/520 [1:00:19<00:15, 3.79s/it] 99%|█████████▉| 517/520 [1:00:22<00:11, 3.72s/it] {'loss': 1.3189, 'grad_norm': 0.0035207954188651723, 'learning_rate': 1.7483908725357544e-05, 'epoch': 0.99} + 99%|█████████▉| 517/520 [1:00:22<00:11, 3.72s/it] 100%|█████████▉| 518/520 [1:00:26<00:07, 3.68s/it] {'loss': 1.2059, 'grad_norm': 0.003519472882342733, 'learning_rate': 7.770751902513862e-06, 'epoch': 1.0} + 100%|█████████▉| 518/520 [1:00:26<00:07, 3.68s/it] 100%|█████████▉| 519/520 [1:00:29<00:03, 3.65s/it] {'loss': 1.2594, 'grad_norm': 0.0033453282595643486, 'learning_rate': 1.9427068461808086e-06, 'epoch': 1.0} + 100%|█████████▉| 519/520 [1:00:29<00:03, 3.65s/it] 100%|██████████| 520/520 [1:00:34<00:00, 3.89s/it] {'loss': 1.3305, 'grad_norm': 0.003548538536134863, 'learning_rate': 0.0, 'epoch': 1.0} + 100%|██████████| 520/520 [1:00:34<00:00, 3.89s/it] {'train_runtime': 3634.2244, 'train_samples_per_second': 18.306, 'train_steps_per_second': 0.143, 'train_loss': 1.4941325083374977, 'epoch': 1.0} + 100%|██████████| 520/520 [1:00:34<00:00, 3.89s/it] 100%|██████████| 520/520 [1:00:34<00:00, 6.99s/it] +[2025-10-12 10:04:49,634] [INFO] [launch.py:348:main] Process 221845 exits successfully. +[2025-10-12 10:04:50,636] [INFO] [launch.py:348:main] Process 221846 exits successfully. +[2025-10-12 10:04:50,636] [INFO] [launch.py:348:main] Process 221843 exits successfully. +[2025-10-12 10:04:50,637] [INFO] [launch.py:348:main] Process 221842 exits successfully. +[2025-10-12 10:04:50,637] [INFO] [launch.py:348:main] Process 221840 exits successfully. +[2025-10-12 10:04:50,637] [INFO] [launch.py:348:main] Process 221844 exits successfully. +[2025-10-12 10:04:50,638] [INFO] [launch.py:348:main] Process 221841 exits successfully. +[2025-10-12 10:04:54,642] [INFO] [launch.py:348:main] Process 221839 exits successfully. +==== EXPERIMENT COMPLETED: qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.5_2e-1_connector-1.0_2.5_2e-1_ablation ==== +Log File: qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.5_2e-1_connector-1.0_2.5_2e-1_ablation_20251012_090245.log +Timestamp: 2025-10-12 10:04:57 +===================================== diff --git a/logs_oct11/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.7_2e-1_connector-1.0_2.7_2e-1_ablation_20251012_100457.log b/logs_oct11/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.7_2e-1_connector-1.0_2.7_2e-1_ablation_20251012_100457.log new file mode 100644 index 0000000000000000000000000000000000000000..3505ea4c88edf91643c5735ebb82c0485d76b614 --- /dev/null +++ b/logs_oct11/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.7_2e-1_connector-1.0_2.7_2e-1_ablation_20251012_100457.log @@ -0,0 +1,2312 @@ +==== STARTING EXPERIMENT: qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.7_2e-1_connector-1.0_2.7_2e-1_ablation ==== +Log File: qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.7_2e-1_connector-1.0_2.7_2e-1_ablation_20251012_100457.log +Timestamp: 2025-10-12 10:04:57 +===================================== +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 10:04:59,921] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 10:05:02,830] [WARNING] [runner.py:202:fetch_hostfile] Unable to find hostfile, will proceed with training with local resources only. +[2025-10-12 10:05:02,832] [INFO] [runner.py:568:main] cmd = /opt/conda/envs/tinyllava/bin/python3.10 -u -m deepspeed.launcher.launch --world_info=eyJsb2NhbGhvc3QiOiBbMCwgMSwgMiwgMywgNCwgNSwgNiwgN119 --master_addr=127.0.0.1 --master_port=29501 --enable_each_rank_log=None tinyllava/train/train.py --deepspeed ./scripts/zero3.json --data_path /s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json --image_folder /s3-code/ywang29/datasets/tinyllava --is_multimodal True --conv_version qwen2_base --model_name_or_path Qwen/Qwen2.5-0.5B --vision_tower google/siglip-so400m-patch14-384 --vision_tower2 --connector_type mlp2x_gelu --mm_vision_select_layer -2 --image_aspect_ratio square --attn_implementation flash_attention_2 --bf16 True --training_recipe common --tune_type_llm full --tune_type_vision_tower frozen --tune_vision_tower_from_layer 0 --tune_type_connector full --group_by_modality_length True --pretrained_model_path /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain --output_dir /nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.7_2e-1_connector-1.0_2.7_2e-1_ablation --num_train_epochs 1 --per_device_train_batch_size 4 --per_device_eval_batch_size 4 --gradient_accumulation_steps 4 --evaluation_strategy no --learning_rate 2e-1 --weight_decay 0. --warmup_ratio 0.03 --lr_scheduler_type cosine --logging_steps 1 --tf32 False --model_max_length 2048 --gradient_checkpointing True --dataloader_num_workers 8 --lazy_preprocess True --report_to tensorboard --tokenizer_use_fast False --run_name tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune --subnet_mode_text both --subnet_type_text None --mask_type_text soft --init_mean_text 1.0 --temperature_attn_text 2.7 --temperature_mlp_text 2.7 --backward_type_text normal --masked_layers_text all --subnet_mode_vision both --subnet_type_vision None --mask_type_vision soft --init_mean_vision 1.0 --temperature_attn_vision 2.7 --temperature_mlp_vision 2.7 --backward_type_vision normal --masked_layers_vision all --subnet_type_connector global --mask_type_connector soft --init_mean_connector 1.0 --temperature_connector 2.7 --backward_type_connector normal --mm_projector_lr 2e-1 --seed 42 --mask_model llm-connector --save_strategy steps --save_steps 50000 --save_total_limit 1 --train_data_ratio 0.1 +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 10:05:05,400] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 10:05:06,416] [INFO] [launch.py:138:main] 0 NCCL_VERSION=2.21.5 +[2025-10-12 10:05:06,416] [INFO] [launch.py:138:main] 0 NCCL_SOCKET_IFNAME=eth +[2025-10-12 10:05:06,416] [INFO] [launch.py:145:main] WORLD INFO DICT: {'localhost': [0, 1, 2, 3, 4, 5, 6, 7]} +[2025-10-12 10:05:06,416] [INFO] [launch.py:151:main] nnodes=1, num_local_procs=8, node_rank=0 +[2025-10-12 10:05:06,416] [INFO] [launch.py:162:main] global_rank_mapping=defaultdict(, {'localhost': [0, 1, 2, 3, 4, 5, 6, 7]}) +[2025-10-12 10:05:06,416] [INFO] [launch.py:163:main] dist_world_size=8 +[2025-10-12 10:05:06,416] [INFO] [launch.py:165:main] Setting CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 +[2025-10-12 10:05:06,419] [INFO] [launch.py:253:main] process 325944 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=0', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.7_2e-1_connector-1.0_2.7_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '1.0', '--temperature_attn_text', '2.7', '--temperature_mlp_text', '2.7', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '1.0', '--temperature_attn_vision', '2.7', '--temperature_mlp_vision', '2.7', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '1.0', '--temperature_connector', '2.7', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 10:05:06,421] [INFO] [launch.py:253:main] process 325945 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=1', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.7_2e-1_connector-1.0_2.7_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '1.0', '--temperature_attn_text', '2.7', '--temperature_mlp_text', '2.7', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '1.0', '--temperature_attn_vision', '2.7', '--temperature_mlp_vision', '2.7', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '1.0', '--temperature_connector', '2.7', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 10:05:06,423] [INFO] [launch.py:253:main] process 325946 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=2', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.7_2e-1_connector-1.0_2.7_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '1.0', '--temperature_attn_text', '2.7', '--temperature_mlp_text', '2.7', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '1.0', '--temperature_attn_vision', '2.7', '--temperature_mlp_vision', '2.7', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '1.0', '--temperature_connector', '2.7', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 10:05:06,425] [INFO] [launch.py:253:main] process 325947 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=3', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.7_2e-1_connector-1.0_2.7_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '1.0', '--temperature_attn_text', '2.7', '--temperature_mlp_text', '2.7', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '1.0', '--temperature_attn_vision', '2.7', '--temperature_mlp_vision', '2.7', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '1.0', '--temperature_connector', '2.7', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 10:05:06,428] [INFO] [launch.py:253:main] process 325948 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=4', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.7_2e-1_connector-1.0_2.7_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '1.0', '--temperature_attn_text', '2.7', '--temperature_mlp_text', '2.7', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '1.0', '--temperature_attn_vision', '2.7', '--temperature_mlp_vision', '2.7', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '1.0', '--temperature_connector', '2.7', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 10:05:06,430] [INFO] [launch.py:253:main] process 325949 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=5', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.7_2e-1_connector-1.0_2.7_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '1.0', '--temperature_attn_text', '2.7', '--temperature_mlp_text', '2.7', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '1.0', '--temperature_attn_vision', '2.7', '--temperature_mlp_vision', '2.7', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '1.0', '--temperature_connector', '2.7', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 10:05:06,432] [INFO] [launch.py:253:main] process 325950 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=6', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.7_2e-1_connector-1.0_2.7_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '1.0', '--temperature_attn_text', '2.7', '--temperature_mlp_text', '2.7', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '1.0', '--temperature_attn_vision', '2.7', '--temperature_mlp_vision', '2.7', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '1.0', '--temperature_connector', '2.7', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 10:05:06,434] [INFO] [launch.py:253:main] process 325951 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=7', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.7_2e-1_connector-1.0_2.7_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '1.0', '--temperature_attn_text', '2.7', '--temperature_mlp_text', '2.7', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '1.0', '--temperature_attn_vision', '2.7', '--temperature_mlp_vision', '2.7', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '1.0', '--temperature_connector', '2.7', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 10:05:13,075] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 10:05:13,308] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 10:05:13,360] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 10:05:13,360] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 10:05:13,370] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 10:05:13,378] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 10:05:13,410] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 10:05:13,462] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 10:05:13,487] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 10:05:13,709] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 10:05:13,761] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 10:05:13,762] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 10:05:13,772] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 10:05:13,811] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 10:05:13,811] [INFO] [comm.py:668:init_distributed] Initializing TorchBackend in DeepSpeed with backend nccl +[2025-10-12 10:05:13,836] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 10:05:13,896] [INFO] [comm.py:637:init_distributed] cdb=None +Apply masks for the following modules: ['llm', 'connector'] +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +{'llm': {'model_name_or_path': 'Qwen/Qwen2.5-0.5B', 'cache_dir': None, 'attn_implementation': 'flash_attention_2', 'subnet_mode': 'both', 'subnet_type': 'None', 'sparsity_attn': None, 'sparsity_mlp': None, 'threshold_attn': None, 'threshold_mlp': None, 'temperature_attn': 2.7, 'temperature_mlp': 2.7, 'masked_layers': 'all', 'mask_type': 'soft', 'backward_type': 'normal'}, 'vision_tower': {'model_name_or_path': 'google/siglip-so400m-patch14-384'}, 'connector': {'connector_type': 'mlp2x_gelu', 'subnet_type': 'global', 'threshold': None, 'sparsity': None, 'temperature': 2.7, 'mask_type': 'soft', 'backward_type': 'normal'}} +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +TinyLlavaConfig { + "backward_type_connector": "normal", + "cache_dir": null, + "connector_type": "mlp2x_gelu", + "hidden_size": 896, + "ignore_index": -100, + "image_aspect_ratio": "square", + "image_token_index": -200, + "llm_model_name_or_path": "Qwen/Qwen2.5-0.5B", + "mask_model": [ + "llm", + "connector" + ], + "mask_type_connector": "soft", + "model_type": "tinyllava", + "num_queries": 128, + "num_resampler_layers": 3, + "pad_token": null, + "resampler_hidden_size": 768, + "sparsity_connector": null, + "subnet_type_connector": "global", + "temperature_connector": 2.7, + "text_config": { + "_name_or_path": "Qwen/Qwen2.5-0.5B", + "architectures": [ + "Qwen2ForCausalLM" + ], + "backward_type": "normal", + "bos_token_id": 151643, + "eos_token_id": 151643, + "hidden_size": 896, + "intermediate_size": 4864, + "mask_type": "soft", + "masked_layers": "all", + "max_position_embeddings": 32768, + "max_window_layers": 24, + "model_type": "qwen2", + "num_attention_heads": 14, + "num_hidden_layers": 24, + "num_key_value_heads": 2, + "rope_theta": 1000000.0, + "sliding_window": 32768, + "subnet_mode": "both", + "subnet_type": "None", + "temperature_attn": 2.7, + "temperature_mlp": 2.7, + "tie_word_embeddings": true, + "torch_dtype": "bfloat16", + "use_mrope": false, + "use_sliding_window": false, + "vocab_size": 151936 + }, + "threshold_connector": null, + "tokenizer_model_max_length": 2048, + "tokenizer_name_or_path": "Qwen/Qwen2.5-0.5B", + "tokenizer_padding_side": "right", + "tokenizer_use_fast": false, + "transformers_version": "4.40.1", + "tune_type_connector": "frozen", + "tune_type_llm": "frozen", + "tune_type_vision_tower": "frozen", + "tune_vision_tower_from_layer": -1, + "use_cache": false, + "vision_config": { + "hidden_act": "gelu_pytorch_tanh", + "hidden_size": 1152, + "image_size": 384, + "intermediate_size": 4304, + "layer_norm_eps": 1e-06, + "model_name_or_path": "google/siglip-so400m-patch14-384", + "model_name_or_path2": "", + "model_type": "siglip_vision_model", + "num_attention_heads": 16, + "num_hidden_layers": 27, + "patch_size": 14 + }, + "vision_feature_layer": -2, + "vision_feature_select_strategy": "patch", + "vision_hidden_size": 1152, + "vision_model_name_or_path": "google/siglip-so400m-patch14-384", + "vision_model_name_or_path2": "", + "vocab_size": 151936 +} + +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +ywang29-vrdb-test1-worker-0:325944:325944 [0] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:325944:325944 [0] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:325944:325944 [0] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:325944:325944 [0] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:325944:325944 [0] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test1-worker-0:325944:325944 [0] NCCL INFO cudaDriverVersion 12040 +NCCL version 2.21.5+cuda12.1 +ywang29-vrdb-test1-worker-0:325946:325946 [2] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test1-worker-0:325946:325946 [2] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:325946:325946 [2] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:325951:325951 [7] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test1-worker-0:325946:325946 [2] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:325946:325946 [2] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:325946:325946 [2] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test1-worker-0:325951:325951 [7] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:325948:325948 [4] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test1-worker-0:325951:325951 [7] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:325948:325948 [4] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:325948:325948 [4] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:325951:325951 [7] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:325951:325951 [7] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:325951:325951 [7] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test1-worker-0:325948:325948 [4] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:325948:325948 [4] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:325948:325948 [4] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test1-worker-0:325944:327541 [0] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:325944:327541 [0] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:325944:327541 [0] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:325944:327541 [0] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:325944:327541 [0] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:325944:327541 [0] NCCL INFO Using network Socket +ywang29-vrdb-test1-worker-0:325946:327542 [2] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:325948:327544 [4] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:325948:327544 [4] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:325946:327542 [2] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:325946:327542 [2] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:325948:327544 [4] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:325948:327544 [4] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:325946:327542 [2] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:325948:327544 [4] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:325946:327542 [2] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:325946:327542 [2] NCCL INFO Using network Socket +ywang29-vrdb-test1-worker-0:325948:327544 [4] NCCL INFO Using network Socket +ywang29-vrdb-test1-worker-0:325951:327543 [7] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:325951:327543 [7] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:325951:327543 [7] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:325951:327543 [7] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:325951:327543 [7] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:325951:327543 [7] NCCL INFO Using network Socket +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test1-worker-0:325950:325950 [6] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test1-worker-0:325950:325950 [6] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:325950:325950 [6] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:325950:325950 [6] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:325950:325950 [6] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:325950:325950 [6] NCCL INFO NET/Plugin: Using internal network plugin. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test1-worker-0:325945:325945 [1] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test1-worker-0:325945:325945 [1] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:325945:325945 [1] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:325945:325945 [1] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:325945:325945 [1] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:325945:325945 [1] NCCL INFO NET/Plugin: Using internal network plugin. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test1-worker-0:325949:325949 [5] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test1-worker-0:325949:325949 [5] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:325949:325949 [5] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:325947:325947 [3] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test1-worker-0:325947:325947 [3] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:325947:325947 [3] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:325949:325949 [5] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:325949:325949 [5] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:325949:325949 [5] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test1-worker-0:325947:325947 [3] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:325947:325947 [3] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:325947:325947 [3] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test1-worker-0:325950:327545 [6] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:325950:327545 [6] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:325950:327545 [6] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:325950:327545 [6] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:325950:327545 [6] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:325950:327545 [6] NCCL INFO Using network Socket +ywang29-vrdb-test1-worker-0:325945:327546 [1] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:325945:327546 [1] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:325945:327546 [1] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:325945:327546 [1] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:325945:327546 [1] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:325945:327546 [1] NCCL INFO Using network Socket +ywang29-vrdb-test1-worker-0:325949:327547 [5] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:325947:327548 [3] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:325949:327547 [5] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:325949:327547 [5] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:325949:327547 [5] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:325949:327547 [5] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:325949:327547 [5] NCCL INFO Using network Socket +ywang29-vrdb-test1-worker-0:325947:327548 [3] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:325947:327548 [3] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:325947:327548 [3] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:325947:327548 [3] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:325947:327548 [3] NCCL INFO Using network Socket +ywang29-vrdb-test1-worker-0:325949:327547 [5] NCCL INFO ncclCommInitRank comm 0x55860d263180 rank 5 nranks 8 cudaDev 5 nvmlDev 5 busId 901d0 commId 0xd5e345907503f291 - Init START +ywang29-vrdb-test1-worker-0:325947:327548 [3] NCCL INFO ncclCommInitRank comm 0x55ddf43455c0 rank 3 nranks 8 cudaDev 3 nvmlDev 3 busId 201d0 commId 0xd5e345907503f291 - Init START +ywang29-vrdb-test1-worker-0:325948:327544 [4] NCCL INFO ncclCommInitRank comm 0x5590f386aa30 rank 4 nranks 8 cudaDev 4 nvmlDev 4 busId 901c0 commId 0xd5e345907503f291 - Init START +ywang29-vrdb-test1-worker-0:325945:327546 [1] NCCL INFO ncclCommInitRank comm 0x560ffe4ab290 rank 1 nranks 8 cudaDev 1 nvmlDev 1 busId 101d0 commId 0xd5e345907503f291 - Init START +ywang29-vrdb-test1-worker-0:325944:327541 [0] NCCL INFO ncclCommInitRank comm 0x55ed2032ffb0 rank 0 nranks 8 cudaDev 0 nvmlDev 0 busId 101c0 commId 0xd5e345907503f291 - Init START +ywang29-vrdb-test1-worker-0:325946:327542 [2] NCCL INFO ncclCommInitRank comm 0x55ace1634530 rank 2 nranks 8 cudaDev 2 nvmlDev 2 busId 201c0 commId 0xd5e345907503f291 - Init START +ywang29-vrdb-test1-worker-0:325951:327543 [7] NCCL INFO ncclCommInitRank comm 0x55e050ca2cf0 rank 7 nranks 8 cudaDev 7 nvmlDev 7 busId a01d0 commId 0xd5e345907503f291 - Init START +ywang29-vrdb-test1-worker-0:325950:327545 [6] NCCL INFO ncclCommInitRank comm 0x55f9a3232760 rank 6 nranks 8 cudaDev 6 nvmlDev 6 busId a01c0 commId 0xd5e345907503f291 - Init START +ywang29-vrdb-test1-worker-0:325948:327544 [4] NCCL INFO Setting affinity for GPU 4 to ffffff00,0000ffff,ff000000 +ywang29-vrdb-test1-worker-0:325948:327544 [4] NCCL INFO NVLS multicast support is not available on dev 4 +ywang29-vrdb-test1-worker-0:325945:327546 [1] NCCL INFO Setting affinity for GPU 1 to ff,ffff0000,00ffffff +ywang29-vrdb-test1-worker-0:325945:327546 [1] NCCL INFO NVLS multicast support is not available on dev 1 +ywang29-vrdb-test1-worker-0:325951:327543 [7] NCCL INFO Setting affinity for GPU 7 to ffffff00,0000ffff,ff000000 +ywang29-vrdb-test1-worker-0:325951:327543 [7] NCCL INFO NVLS multicast support is not available on dev 7 +ywang29-vrdb-test1-worker-0:325949:327547 [5] NCCL INFO Setting affinity for GPU 5 to ffffff00,0000ffff,ff000000 +ywang29-vrdb-test1-worker-0:325944:327541 [0] NCCL INFO Setting affinity for GPU 0 to ff,ffff0000,00ffffff +ywang29-vrdb-test1-worker-0:325949:327547 [5] NCCL INFO NVLS multicast support is not available on dev 5 +ywang29-vrdb-test1-worker-0:325944:327541 [0] NCCL INFO NVLS multicast support is not available on dev 0 +ywang29-vrdb-test1-worker-0:325947:327548 [3] NCCL INFO Setting affinity for GPU 3 to ff,ffff0000,00ffffff +ywang29-vrdb-test1-worker-0:325947:327548 [3] NCCL INFO NVLS multicast support is not available on dev 3 +ywang29-vrdb-test1-worker-0:325946:327542 [2] NCCL INFO Setting affinity for GPU 2 to ff,ffff0000,00ffffff +ywang29-vrdb-test1-worker-0:325946:327542 [2] NCCL INFO NVLS multicast support is not available on dev 2 +ywang29-vrdb-test1-worker-0:325950:327545 [6] NCCL INFO Setting affinity for GPU 6 to ffffff00,0000ffff,ff000000 +ywang29-vrdb-test1-worker-0:325950:327545 [6] NCCL INFO NVLS multicast support is not available on dev 6 +ywang29-vrdb-test1-worker-0:325951:327543 [7] NCCL INFO comm 0x55e050ca2cf0 rank 7 nRanks 8 nNodes 1 localRanks 8 localRank 7 MNNVL 0 +ywang29-vrdb-test1-worker-0:325950:327545 [6] NCCL INFO comm 0x55f9a3232760 rank 6 nRanks 8 nNodes 1 localRanks 8 localRank 6 MNNVL 0 +ywang29-vrdb-test1-worker-0:325949:327547 [5] NCCL INFO comm 0x55860d263180 rank 5 nRanks 8 nNodes 1 localRanks 8 localRank 5 MNNVL 0 +ywang29-vrdb-test1-worker-0:325947:327548 [3] NCCL INFO comm 0x55ddf43455c0 rank 3 nRanks 8 nNodes 1 localRanks 8 localRank 3 MNNVL 0 +ywang29-vrdb-test1-worker-0:325946:327542 [2] NCCL INFO comm 0x55ace1634530 rank 2 nRanks 8 nNodes 1 localRanks 8 localRank 2 MNNVL 0 +ywang29-vrdb-test1-worker-0:325948:327544 [4] NCCL INFO comm 0x5590f386aa30 rank 4 nRanks 8 nNodes 1 localRanks 8 localRank 4 MNNVL 0 +ywang29-vrdb-test1-worker-0:325944:327541 [0] NCCL INFO comm 0x55ed2032ffb0 rank 0 nRanks 8 nNodes 1 localRanks 8 localRank 0 MNNVL 0 +ywang29-vrdb-test1-worker-0:325945:327546 [1] NCCL INFO comm 0x560ffe4ab290 rank 1 nRanks 8 nNodes 1 localRanks 8 localRank 1 MNNVL 0 +ywang29-vrdb-test1-worker-0:325950:327545 [6] NCCL INFO Trees [0] 7/-1/-1->6->5 [1] 7/-1/-1->6->5 [2] 7/-1/-1->6->5 [3] 7/-1/-1->6->5 [4] 7/-1/-1->6->5 [5] 7/-1/-1->6->5 [6] 7/-1/-1->6->5 [7] 7/-1/-1->6->5 [8] 7/-1/-1->6->5 [9] 7/-1/-1->6->5 [10] 7/-1/-1->6->5 [11] 7/-1/-1->6->5 [12] 7/-1/-1->6->5 [13] 7/-1/-1->6->5 [14] 7/-1/-1->6->5 [15] 7/-1/-1->6->5 [16] 7/-1/-1->6->5 [17] 7/-1/-1->6->5 [18] 7/-1/-1->6->5 [19] 7/-1/-1->6->5 [20] 7/-1/-1->6->5 [21] 7/-1/-1->6->5 [22] 7/-1/-1->6->5 [23] 7/-1/-1->6->5 +ywang29-vrdb-test1-worker-0:325944:327541 [0] NCCL INFO Channel 00/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:325951:327543 [7] NCCL INFO Trees [0] -1/-1/-1->7->6 [1] -1/-1/-1->7->6 [2] -1/-1/-1->7->6 [3] -1/-1/-1->7->6 [4] -1/-1/-1->7->6 [5] -1/-1/-1->7->6 [6] -1/-1/-1->7->6 [7] -1/-1/-1->7->6 [8] -1/-1/-1->7->6 [9] -1/-1/-1->7->6 [10] -1/-1/-1->7->6 [11] -1/-1/-1->7->6 [12] -1/-1/-1->7->6 [13] -1/-1/-1->7->6 [14] -1/-1/-1->7->6 [15] -1/-1/-1->7->6 [16] -1/-1/-1->7->6 [17] -1/-1/-1->7->6 [18] -1/-1/-1->7->6 [19] -1/-1/-1->7->6 [20] -1/-1/-1->7->6 [21] -1/-1/-1->7->6 [22] -1/-1/-1->7->6 [23] -1/-1/-1->7->6 +ywang29-vrdb-test1-worker-0:325949:327547 [5] NCCL INFO Trees [0] 6/-1/-1->5->4 [1] 6/-1/-1->5->4 [2] 6/-1/-1->5->4 [3] 6/-1/-1->5->4 [4] 6/-1/-1->5->4 [5] 6/-1/-1->5->4 [6] 6/-1/-1->5->4 [7] 6/-1/-1->5->4 [8] 6/-1/-1->5->4 [9] 6/-1/-1->5->4 [10] 6/-1/-1->5->4 [11] 6/-1/-1->5->4 [12] 6/-1/-1->5->4 [13] 6/-1/-1->5->4 [14] 6/-1/-1->5->4 [15] 6/-1/-1->5->4 [16] 6/-1/-1->5->4 [17] 6/-1/-1->5->4 [18] 6/-1/-1->5->4 [19] 6/-1/-1->5->4 [20] 6/-1/-1->5->4 [21] 6/-1/-1->5->4 [22] 6/-1/-1->5->4 [23] 6/-1/-1->5->4 +ywang29-vrdb-test1-worker-0:325944:327541 [0] NCCL INFO Channel 01/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:325950:327545 [6] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:325947:327548 [3] NCCL INFO Trees [0] 4/-1/-1->3->2 [1] 4/-1/-1->3->2 [2] 4/-1/-1->3->2 [3] 4/-1/-1->3->2 [4] 4/-1/-1->3->2 [5] 4/-1/-1->3->2 [6] 4/-1/-1->3->2 [7] 4/-1/-1->3->2 [8] 4/-1/-1->3->2 [9] 4/-1/-1->3->2 [10] 4/-1/-1->3->2 [11] 4/-1/-1->3->2 [12] 4/-1/-1->3->2 [13] 4/-1/-1->3->2 [14] 4/-1/-1->3->2 [15] 4/-1/-1->3->2 [16] 4/-1/-1->3->2 [17] 4/-1/-1->3->2 [18] 4/-1/-1->3->2 [19] 4/-1/-1->3->2 [20] 4/-1/-1->3->2 [21] 4/-1/-1->3->2 [22] 4/-1/-1->3->2 [23] 4/-1/-1->3->2 +ywang29-vrdb-test1-worker-0:325951:327543 [7] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:325946:327542 [2] NCCL INFO Trees [0] 3/-1/-1->2->1 [1] 3/-1/-1->2->1 [2] 3/-1/-1->2->1 [3] 3/-1/-1->2->1 [4] 3/-1/-1->2->1 [5] 3/-1/-1->2->1 [6] 3/-1/-1->2->1 [7] 3/-1/-1->2->1 [8] 3/-1/-1->2->1 [9] 3/-1/-1->2->1 [10] 3/-1/-1->2->1 [11] 3/-1/-1->2->1 [12] 3/-1/-1->2->1 [13] 3/-1/-1->2->1 [14] 3/-1/-1->2->1 [15] 3/-1/-1->2->1 [16] 3/-1/-1->2->1 [17] 3/-1/-1->2->1 [18] 3/-1/-1->2->1 [19] 3/-1/-1->2->1 [20] 3/-1/-1->2->1 [21] 3/-1/-1->2->1 [22] 3/-1/-1->2->1 [23] 3/-1/-1->2->1 +ywang29-vrdb-test1-worker-0:325949:327547 [5] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:325944:327541 [0] NCCL INFO Channel 02/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:325947:327548 [3] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:325946:327542 [2] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:325944:327541 [0] NCCL INFO Channel 03/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:325948:327544 [4] NCCL INFO Trees [0] 5/-1/-1->4->3 [1] 5/-1/-1->4->3 [2] 5/-1/-1->4->3 [3] 5/-1/-1->4->3 [4] 5/-1/-1->4->3 [5] 5/-1/-1->4->3 [6] 5/-1/-1->4->3 [7] 5/-1/-1->4->3 [8] 5/-1/-1->4->3 [9] 5/-1/-1->4->3 [10] 5/-1/-1->4->3 [11] 5/-1/-1->4->3 [12] 5/-1/-1->4->3 [13] 5/-1/-1->4->3 [14] 5/-1/-1->4->3 [15] 5/-1/-1->4->3 [16] 5/-1/-1->4->3 [17] 5/-1/-1->4->3 [18] 5/-1/-1->4->3 [19] 5/-1/-1->4->3 [20] 5/-1/-1->4->3 [21] 5/-1/-1->4->3 [22] 5/-1/-1->4->3 [23] 5/-1/-1->4->3 +ywang29-vrdb-test1-worker-0:325944:327541 [0] NCCL INFO Channel 04/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:325944:327541 [0] NCCL INFO Channel 05/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:325948:327544 [4] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:325944:327541 [0] NCCL INFO Channel 06/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:325944:327541 [0] NCCL INFO Channel 07/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:325944:327541 [0] NCCL INFO Channel 08/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:325944:327541 [0] NCCL INFO Channel 09/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:325945:327546 [1] NCCL INFO Trees [0] 2/-1/-1->1->0 [1] 2/-1/-1->1->0 [2] 2/-1/-1->1->0 [3] 2/-1/-1->1->0 [4] 2/-1/-1->1->0 [5] 2/-1/-1->1->0 [6] 2/-1/-1->1->0 [7] 2/-1/-1->1->0 [8] 2/-1/-1->1->0 [9] 2/-1/-1->1->0 [10] 2/-1/-1->1->0 [11] 2/-1/-1->1->0 [12] 2/-1/-1->1->0 [13] 2/-1/-1->1->0 [14] 2/-1/-1->1->0 [15] 2/-1/-1->1->0 [16] 2/-1/-1->1->0 [17] 2/-1/-1->1->0 [18] 2/-1/-1->1->0 [19] 2/-1/-1->1->0 [20] 2/-1/-1->1->0 [21] 2/-1/-1->1->0 [22] 2/-1/-1->1->0 [23] 2/-1/-1->1->0 +ywang29-vrdb-test1-worker-0:325944:327541 [0] NCCL INFO Channel 10/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:325944:327541 [0] NCCL INFO Channel 11/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:325945:327546 [1] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:325944:327541 [0] NCCL INFO Channel 12/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:325944:327541 [0] NCCL INFO Channel 13/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:325944:327541 [0] NCCL INFO Channel 14/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:325944:327541 [0] NCCL INFO Channel 15/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:325944:327541 [0] NCCL INFO Channel 16/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:325944:327541 [0] NCCL INFO Channel 17/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:325944:327541 [0] NCCL INFO Channel 18/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:325944:327541 [0] NCCL INFO Channel 19/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:325944:327541 [0] NCCL INFO Channel 20/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:325944:327541 [0] NCCL INFO Channel 21/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:325944:327541 [0] NCCL INFO Channel 22/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:325944:327541 [0] NCCL INFO Channel 23/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:325944:327541 [0] NCCL INFO Trees [0] 1/-1/-1->0->-1 [1] 1/-1/-1->0->-1 [2] 1/-1/-1->0->-1 [3] 1/-1/-1->0->-1 [4] 1/-1/-1->0->-1 [5] 1/-1/-1->0->-1 [6] 1/-1/-1->0->-1 [7] 1/-1/-1->0->-1 [8] 1/-1/-1->0->-1 [9] 1/-1/-1->0->-1 [10] 1/-1/-1->0->-1 [11] 1/-1/-1->0->-1 [12] 1/-1/-1->0->-1 [13] 1/-1/-1->0->-1 [14] 1/-1/-1->0->-1 [15] 1/-1/-1->0->-1 [16] 1/-1/-1->0->-1 [17] 1/-1/-1->0->-1 [18] 1/-1/-1->0->-1 [19] 1/-1/-1->0->-1 [20] 1/-1/-1->0->-1 [21] 1/-1/-1->0->-1 [22] 1/-1/-1->0->-1 [23] 1/-1/-1->0->-1 +ywang29-vrdb-test1-worker-0:325944:327541 [0] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:325951:327543 [7] NCCL INFO Channel 00/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:327545 [6] NCCL INFO Channel 00/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:327547 [5] NCCL INFO Channel 00/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:327548 [3] NCCL INFO Channel 00/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:327542 [2] NCCL INFO Channel 00/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:327543 [7] NCCL INFO Channel 01/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:327545 [6] NCCL INFO Channel 01/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:327544 [4] NCCL INFO Channel 00/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:327547 [5] NCCL INFO Channel 01/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:327548 [3] NCCL INFO Channel 01/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:327542 [2] NCCL INFO Channel 01/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:327545 [6] NCCL INFO Channel 02/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:327543 [7] NCCL INFO Channel 02/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:327544 [4] NCCL INFO Channel 01/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:327547 [5] NCCL INFO Channel 02/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:327546 [1] NCCL INFO Channel 00/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:327548 [3] NCCL INFO Channel 02/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:327542 [2] NCCL INFO Channel 02/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325944:327541 [0] NCCL INFO Channel 00/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:327545 [6] NCCL INFO Channel 03/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:327543 [7] NCCL INFO Channel 03/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:327544 [4] NCCL INFO Channel 02/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:327547 [5] NCCL INFO Channel 03/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:327546 [1] NCCL INFO Channel 01/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:327548 [3] NCCL INFO Channel 03/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:327542 [2] NCCL INFO Channel 03/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325944:327541 [0] NCCL INFO Channel 01/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:327545 [6] NCCL INFO Channel 04/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:327543 [7] NCCL INFO Channel 04/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:327544 [4] NCCL INFO Channel 03/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:327547 [5] NCCL INFO Channel 04/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:327548 [3] NCCL INFO Channel 04/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:327546 [1] NCCL INFO Channel 02/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:327542 [2] NCCL INFO Channel 04/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325944:327541 [0] NCCL INFO Channel 02/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:327545 [6] NCCL INFO Channel 05/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:327544 [4] NCCL INFO Channel 04/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:327543 [7] NCCL INFO Channel 05/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:327547 [5] NCCL INFO Channel 05/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:327548 [3] NCCL INFO Channel 05/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:327546 [1] NCCL INFO Channel 03/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:327542 [2] NCCL INFO Channel 05/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325944:327541 [0] NCCL INFO Channel 03/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:327545 [6] NCCL INFO Channel 06/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:327543 [7] NCCL INFO Channel 06/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:327544 [4] NCCL INFO Channel 05/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:327547 [5] NCCL INFO Channel 06/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:327548 [3] NCCL INFO Channel 06/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:327542 [2] NCCL INFO Channel 06/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:327546 [1] NCCL INFO Channel 04/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325944:327541 [0] NCCL INFO Channel 04/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:327545 [6] NCCL INFO Channel 07/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:327543 [7] NCCL INFO Channel 07/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:327544 [4] NCCL INFO Channel 06/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:327547 [5] NCCL INFO Channel 07/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:327548 [3] NCCL INFO Channel 07/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:327542 [2] NCCL INFO Channel 07/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:327546 [1] NCCL INFO Channel 05/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:327545 [6] NCCL INFO Channel 08/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:327543 [7] NCCL INFO Channel 08/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325944:327541 [0] NCCL INFO Channel 05/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:327544 [4] NCCL INFO Channel 07/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:327547 [5] NCCL INFO Channel 08/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:327548 [3] NCCL INFO Channel 08/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:327542 [2] NCCL INFO Channel 08/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:327546 [1] NCCL INFO Channel 06/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:327545 [6] NCCL INFO Channel 09/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:327543 [7] NCCL INFO Channel 09/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325944:327541 [0] NCCL INFO Channel 06/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:327544 [4] NCCL INFO Channel 08/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:327547 [5] NCCL INFO Channel 09/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:327548 [3] NCCL INFO Channel 09/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:327542 [2] NCCL INFO Channel 09/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:327546 [1] NCCL INFO Channel 07/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:327544 [4] NCCL INFO Channel 09/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:327545 [6] NCCL INFO Channel 10/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:327543 [7] NCCL INFO Channel 10/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325944:327541 [0] NCCL INFO Channel 07/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:327547 [5] NCCL INFO Channel 10/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:327548 [3] NCCL INFO Channel 10/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:327542 [2] NCCL INFO Channel 10/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:327546 [1] NCCL INFO Channel 08/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:327544 [4] NCCL INFO Channel 10/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:327545 [6] NCCL INFO Channel 11/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:327543 [7] NCCL INFO Channel 11/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325944:327541 [0] NCCL INFO Channel 08/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:327547 [5] NCCL INFO Channel 11/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:327548 [3] NCCL INFO Channel 11/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:327542 [2] NCCL INFO Channel 11/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:327546 [1] NCCL INFO Channel 09/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:327544 [4] NCCL INFO Channel 11/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:327545 [6] NCCL INFO Channel 12/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:327543 [7] NCCL INFO Channel 12/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:327547 [5] NCCL INFO Channel 12/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325944:327541 [0] NCCL INFO Channel 09/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:327548 [3] NCCL INFO Channel 12/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:327542 [2] NCCL INFO Channel 12/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:327546 [1] NCCL INFO Channel 10/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:327544 [4] NCCL INFO Channel 12/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:327545 [6] NCCL INFO Channel 13/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:327543 [7] NCCL INFO Channel 13/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:327547 [5] NCCL INFO Channel 13/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325944:327541 [0] NCCL INFO Channel 10/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:327548 [3] NCCL INFO Channel 13/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:327542 [2] NCCL INFO Channel 13/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:327546 [1] NCCL INFO Channel 11/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:327544 [4] NCCL INFO Channel 13/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:327545 [6] NCCL INFO Channel 14/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:327543 [7] NCCL INFO Channel 14/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:327547 [5] NCCL INFO Channel 14/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325944:327541 [0] NCCL INFO Channel 11/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:327548 [3] NCCL INFO Channel 14/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:327542 [2] NCCL INFO Channel 14/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:327544 [4] NCCL INFO Channel 14/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:327546 [1] NCCL INFO Channel 12/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:327545 [6] NCCL INFO Channel 15/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:327543 [7] NCCL INFO Channel 15/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:327547 [5] NCCL INFO Channel 15/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325944:327541 [0] NCCL INFO Channel 12/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:327548 [3] NCCL INFO Channel 15/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:327542 [2] NCCL INFO Channel 15/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:327544 [4] NCCL INFO Channel 15/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:327546 [1] NCCL INFO Channel 13/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:327545 [6] NCCL INFO Channel 16/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:327543 [7] NCCL INFO Channel 16/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:327547 [5] NCCL INFO Channel 16/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:327548 [3] NCCL INFO Channel 16/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325944:327541 [0] NCCL INFO Channel 13/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:327542 [2] NCCL INFO Channel 16/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:327544 [4] NCCL INFO Channel 16/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:327545 [6] NCCL INFO Channel 17/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:327546 [1] NCCL INFO Channel 14/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:327543 [7] NCCL INFO Channel 17/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:327547 [5] NCCL INFO Channel 17/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:327548 [3] NCCL INFO Channel 17/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325944:327541 [0] NCCL INFO Channel 14/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:327542 [2] NCCL INFO Channel 17/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:327544 [4] NCCL INFO Channel 17/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:327545 [6] NCCL INFO Channel 18/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:327543 [7] NCCL INFO Channel 18/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:327546 [1] NCCL INFO Channel 15/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:327547 [5] NCCL INFO Channel 18/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:327548 [3] NCCL INFO Channel 18/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325944:327541 [0] NCCL INFO Channel 15/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:327542 [2] NCCL INFO Channel 18/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:327544 [4] NCCL INFO Channel 18/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:327545 [6] NCCL INFO Channel 19/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:327543 [7] NCCL INFO Channel 19/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:327547 [5] NCCL INFO Channel 19/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:327546 [1] NCCL INFO Channel 16/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:327548 [3] NCCL INFO Channel 19/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325944:327541 [0] NCCL INFO Channel 16/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:327542 [2] NCCL INFO Channel 19/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:327544 [4] NCCL INFO Channel 19/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:327545 [6] NCCL INFO Channel 20/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:327543 [7] NCCL INFO Channel 20/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:327547 [5] NCCL INFO Channel 20/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:327548 [3] NCCL INFO Channel 20/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:327546 [1] NCCL INFO Channel 17/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325944:327541 [0] NCCL INFO Channel 17/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:327542 [2] NCCL INFO Channel 20/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:327544 [4] NCCL INFO Channel 20/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:327545 [6] NCCL INFO Channel 21/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:327543 [7] NCCL INFO Channel 21/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:327547 [5] NCCL INFO Channel 21/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:327548 [3] NCCL INFO Channel 21/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:327546 [1] NCCL INFO Channel 18/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325944:327541 [0] NCCL INFO Channel 18/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:327542 [2] NCCL INFO Channel 21/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:327544 [4] NCCL INFO Channel 21/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:327545 [6] NCCL INFO Channel 22/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:327543 [7] NCCL INFO Channel 22/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:327547 [5] NCCL INFO Channel 22/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:327548 [3] NCCL INFO Channel 22/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:327546 [1] NCCL INFO Channel 19/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:327542 [2] NCCL INFO Channel 22/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325944:327541 [0] NCCL INFO Channel 19/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:327544 [4] NCCL INFO Channel 22/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:327545 [6] NCCL INFO Channel 23/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:327543 [7] NCCL INFO Channel 23/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:327547 [5] NCCL INFO Channel 23/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:327548 [3] NCCL INFO Channel 23/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:327546 [1] NCCL INFO Channel 20/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:327542 [2] NCCL INFO Channel 23/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325944:327541 [0] NCCL INFO Channel 20/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:327544 [4] NCCL INFO Channel 23/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:327546 [1] NCCL INFO Channel 21/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325944:327541 [0] NCCL INFO Channel 21/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:327546 [1] NCCL INFO Channel 22/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325944:327541 [0] NCCL INFO Channel 22/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325944:327541 [0] NCCL INFO Channel 23/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:327546 [1] NCCL INFO Channel 23/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:327542 [2] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:325945:327546 [1] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:325944:327541 [0] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:325948:327544 [4] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:325947:327548 [3] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:325946:327542 [2] NCCL INFO Channel 00/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:327542 [2] NCCL INFO Channel 01/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:327542 [2] NCCL INFO Channel 02/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:327542 [2] NCCL INFO Channel 03/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:327542 [2] NCCL INFO Channel 04/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:327547 [5] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:325946:327542 [2] NCCL INFO Channel 05/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:327545 [6] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:325946:327542 [2] NCCL INFO Channel 06/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:327542 [2] NCCL INFO Channel 07/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:327542 [2] NCCL INFO Channel 08/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:327542 [2] NCCL INFO Channel 09/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:327543 [7] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:325951:327543 [7] NCCL INFO Channel 00/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:327543 [7] NCCL INFO Channel 01/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:327542 [2] NCCL INFO Channel 10/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:327544 [4] NCCL INFO Channel 00/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:327546 [1] NCCL INFO Channel 00/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:327543 [7] NCCL INFO Channel 02/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:327542 [2] NCCL INFO Channel 11/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:327544 [4] NCCL INFO Channel 01/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:327546 [1] NCCL INFO Channel 01/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:327542 [2] NCCL INFO Channel 12/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:327543 [7] NCCL INFO Channel 03/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:327544 [4] NCCL INFO Channel 02/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:327546 [1] NCCL INFO Channel 02/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:327542 [2] NCCL INFO Channel 13/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:327543 [7] NCCL INFO Channel 04/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:327544 [4] NCCL INFO Channel 03/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:327546 [1] NCCL INFO Channel 03/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:327542 [2] NCCL INFO Channel 14/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:327543 [7] NCCL INFO Channel 05/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:327544 [4] NCCL INFO Channel 04/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:327546 [1] NCCL INFO Channel 04/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:327542 [2] NCCL INFO Channel 15/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:327543 [7] NCCL INFO Channel 06/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:327544 [4] NCCL INFO Channel 05/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:327546 [1] NCCL INFO Channel 05/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:327542 [2] NCCL INFO Channel 16/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:327543 [7] NCCL INFO Channel 07/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:327544 [4] NCCL INFO Channel 06/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:327546 [1] NCCL INFO Channel 06/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:327542 [2] NCCL INFO Channel 17/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:327543 [7] NCCL INFO Channel 08/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:327544 [4] NCCL INFO Channel 07/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:327546 [1] NCCL INFO Channel 07/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:327542 [2] NCCL INFO Channel 18/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:327543 [7] NCCL INFO Channel 09/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:327544 [4] NCCL INFO Channel 08/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:327546 [1] NCCL INFO Channel 08/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:327542 [2] NCCL INFO Channel 19/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:327543 [7] NCCL INFO Channel 10/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:327544 [4] NCCL INFO Channel 09/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:327546 [1] NCCL INFO Channel 09/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:327542 [2] NCCL INFO Channel 20/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:327543 [7] NCCL INFO Channel 11/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:327544 [4] NCCL INFO Channel 10/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:327546 [1] NCCL INFO Channel 10/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:327542 [2] NCCL INFO Channel 21/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:327543 [7] NCCL INFO Channel 12/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:327544 [4] NCCL INFO Channel 11/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:327546 [1] NCCL INFO Channel 11/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:327542 [2] NCCL INFO Channel 22/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:327543 [7] NCCL INFO Channel 13/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:327544 [4] NCCL INFO Channel 12/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:327546 [1] NCCL INFO Channel 12/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:327542 [2] NCCL INFO Channel 23/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:327543 [7] NCCL INFO Channel 14/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:327548 [3] NCCL INFO Channel 00/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:327544 [4] NCCL INFO Channel 13/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:327546 [1] NCCL INFO Channel 13/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:327543 [7] NCCL INFO Channel 15/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:327548 [3] NCCL INFO Channel 01/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:327544 [4] NCCL INFO Channel 14/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:327546 [1] NCCL INFO Channel 14/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:327543 [7] NCCL INFO Channel 16/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:327548 [3] NCCL INFO Channel 02/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:327544 [4] NCCL INFO Channel 15/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:327546 [1] NCCL INFO Channel 15/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:327543 [7] NCCL INFO Channel 17/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:327548 [3] NCCL INFO Channel 03/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:327544 [4] NCCL INFO Channel 16/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:327546 [1] NCCL INFO Channel 16/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:327543 [7] NCCL INFO Channel 18/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:327548 [3] NCCL INFO Channel 04/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:327544 [4] NCCL INFO Channel 17/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:327546 [1] NCCL INFO Channel 17/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:327543 [7] NCCL INFO Channel 19/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:327548 [3] NCCL INFO Channel 05/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:327544 [4] NCCL INFO Channel 18/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:327546 [1] NCCL INFO Channel 18/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:327545 [6] NCCL INFO Channel 00/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:327543 [7] NCCL INFO Channel 20/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:327548 [3] NCCL INFO Channel 06/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:327547 [5] NCCL INFO Channel 00/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:327544 [4] NCCL INFO Channel 19/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:327546 [1] NCCL INFO Channel 19/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:327545 [6] NCCL INFO Channel 01/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:327543 [7] NCCL INFO Channel 21/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:327548 [3] NCCL INFO Channel 07/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:327547 [5] NCCL INFO Channel 01/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:327544 [4] NCCL INFO Channel 20/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:327546 [1] NCCL INFO Channel 20/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:327545 [6] NCCL INFO Channel 02/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:327543 [7] NCCL INFO Channel 22/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:327548 [3] NCCL INFO Channel 08/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:327547 [5] NCCL INFO Channel 02/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:327544 [4] NCCL INFO Channel 21/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:327546 [1] NCCL INFO Channel 21/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:327545 [6] NCCL INFO Channel 03/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:327543 [7] NCCL INFO Channel 23/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:327548 [3] NCCL INFO Channel 09/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:327544 [4] NCCL INFO Channel 22/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:327546 [1] NCCL INFO Channel 22/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:327545 [6] NCCL INFO Channel 04/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:327548 [3] NCCL INFO Channel 10/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:327544 [4] NCCL INFO Channel 23/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:327546 [1] NCCL INFO Channel 23/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:327545 [6] NCCL INFO Channel 05/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:327548 [3] NCCL INFO Channel 11/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:327545 [6] NCCL INFO Channel 06/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:327548 [3] NCCL INFO Channel 12/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:327545 [6] NCCL INFO Channel 07/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:327548 [3] NCCL INFO Channel 13/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:327545 [6] NCCL INFO Channel 08/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:327548 [3] NCCL INFO Channel 14/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:327547 [5] NCCL INFO Channel 03/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:327545 [6] NCCL INFO Channel 09/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:327548 [3] NCCL INFO Channel 15/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:327547 [5] NCCL INFO Channel 04/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:327548 [3] NCCL INFO Channel 16/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:327545 [6] NCCL INFO Channel 10/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:327548 [3] NCCL INFO Channel 17/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:327547 [5] NCCL INFO Channel 05/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:327548 [3] NCCL INFO Channel 18/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:327547 [5] NCCL INFO Channel 06/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:327545 [6] NCCL INFO Channel 11/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:327548 [3] NCCL INFO Channel 19/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:327548 [3] NCCL INFO Channel 20/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:327545 [6] NCCL INFO Channel 12/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:327548 [3] NCCL INFO Channel 21/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:327545 [6] NCCL INFO Channel 13/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:327548 [3] NCCL INFO Channel 22/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:327545 [6] NCCL INFO Channel 14/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:327548 [3] NCCL INFO Channel 23/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:327545 [6] NCCL INFO Channel 15/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:327547 [5] NCCL INFO Channel 07/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:327545 [6] NCCL INFO Channel 16/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:327547 [5] NCCL INFO Channel 08/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:327545 [6] NCCL INFO Channel 17/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:327547 [5] NCCL INFO Channel 09/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:327547 [5] NCCL INFO Channel 10/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:327545 [6] NCCL INFO Channel 18/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:327547 [5] NCCL INFO Channel 11/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:327545 [6] NCCL INFO Channel 19/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:327547 [5] NCCL INFO Channel 12/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:327547 [5] NCCL INFO Channel 13/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:327547 [5] NCCL INFO Channel 14/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:327547 [5] NCCL INFO Channel 15/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:327545 [6] NCCL INFO Channel 20/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:327547 [5] NCCL INFO Channel 16/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:327545 [6] NCCL INFO Channel 21/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:327547 [5] NCCL INFO Channel 17/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:327545 [6] NCCL INFO Channel 22/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:327547 [5] NCCL INFO Channel 18/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:327545 [6] NCCL INFO Channel 23/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:327547 [5] NCCL INFO Channel 19/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:327547 [5] NCCL INFO Channel 20/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:327547 [5] NCCL INFO Channel 21/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:327547 [5] NCCL INFO Channel 22/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:327547 [5] NCCL INFO Channel 23/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325944:327541 [0] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:325944:327541 [0] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:325944:327541 [0] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:325945:327546 [1] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:325945:327546 [1] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:325945:327546 [1] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:325946:327542 [2] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:325946:327542 [2] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:325946:327542 [2] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:325947:327548 [3] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:325947:327548 [3] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:325947:327548 [3] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:325948:327544 [4] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:325948:327544 [4] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:325948:327544 [4] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:325949:327547 [5] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:325949:327547 [5] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:325949:327547 [5] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:325951:327543 [7] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:325950:327545 [6] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:325951:327543 [7] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:325951:327543 [7] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:325950:327545 [6] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:325950:327545 [6] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:325950:327545 [6] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:325951:327543 [7] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:325951:327543 [7] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:325950:327545 [6] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:325951:327543 [7] NCCL INFO ncclCommInitRank comm 0x55e050ca2cf0 rank 7 nranks 8 cudaDev 7 nvmlDev 7 busId a01d0 commId 0xd5e345907503f291 - Init COMPLETE +ywang29-vrdb-test1-worker-0:325950:327545 [6] NCCL INFO ncclCommInitRank comm 0x55f9a3232760 rank 6 nranks 8 cudaDev 6 nvmlDev 6 busId a01c0 commId 0xd5e345907503f291 - Init COMPLETE +ywang29-vrdb-test1-worker-0:325947:327548 [3] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:325947:327548 [3] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:325947:327548 [3] NCCL INFO ncclCommInitRank comm 0x55ddf43455c0 rank 3 nranks 8 cudaDev 3 nvmlDev 3 busId 201d0 commId 0xd5e345907503f291 - Init COMPLETE +ywang29-vrdb-test1-worker-0:325948:327544 [4] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:325948:327544 [4] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:325948:327544 [4] NCCL INFO ncclCommInitRank comm 0x5590f386aa30 rank 4 nranks 8 cudaDev 4 nvmlDev 4 busId 901c0 commId 0xd5e345907503f291 - Init COMPLETE +ywang29-vrdb-test1-worker-0:325944:327541 [0] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:325944:327541 [0] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:325944:327541 [0] NCCL INFO ncclCommInitRank comm 0x55ed2032ffb0 rank 0 nranks 8 cudaDev 0 nvmlDev 0 busId 101c0 commId 0xd5e345907503f291 - Init COMPLETE +ywang29-vrdb-test1-worker-0:325946:327542 [2] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:325946:327542 [2] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:325946:327542 [2] NCCL INFO ncclCommInitRank comm 0x55ace1634530 rank 2 nranks 8 cudaDev 2 nvmlDev 2 busId 201c0 commId 0xd5e345907503f291 - Init COMPLETE +ywang29-vrdb-test1-worker-0:325945:327546 [1] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:325945:327546 [1] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:325945:327546 [1] NCCL INFO ncclCommInitRank comm 0x560ffe4ab290 rank 1 nranks 8 cudaDev 1 nvmlDev 1 busId 101d0 commId 0xd5e345907503f291 - Init COMPLETE +ywang29-vrdb-test1-worker-0:325949:327547 [5] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:325949:327547 [5] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:325949:327547 [5] NCCL INFO ncclCommInitRank comm 0x55860d263180 rank 5 nranks 8 cudaDev 5 nvmlDev 5 busId 901d0 commId 0xd5e345907503f291 - Init COMPLETE +[2025-10-12 10:05:57,904] [INFO] [partition_parameters.py:343:__exit__] finished initializing model - num_params = 459, num_elems = 0.99B +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.laSome weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +loading language model from loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model + /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from yers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. + /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +[2025-10-12 10:33:32,418] [INFO] [partition_parameters.py:343:__exit__] finished initializing model - num_params = 907, num_elems = 1.42B +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin...Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... + +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +TinyLlavaForConditionalGeneration( + (language_model): Qwen2ForCausalLM( + (model): Qwen2Model( + (embed_tokens): Embedding(151936, 896) + (layers): ModuleList( + (0-23): 24 x Qwen2DecoderLayer( + (self_attn): Qwen2FlashAttention2( + (q_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=True) + (k_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=128, bias=True) + (v_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=128, bias=True) + (o_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=False) + (rotary_emb): Qwen2RotaryEmbedding() + ) + (mlp): Qwen2MLP( + (gate_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=4864, bias=False) + (up_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=4864, bias=False) + (down_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=4864, out_features=896, bias=False) + (act_fn): SiLU() + ) + (input_layernorm): Qwen2RMSNorm() + (post_attention_layernorm): Qwen2RMSNorm() + ) + ) + (norm): Qwen2RMSNorm() + ) + (lm_head): Linear(in_features=896, out_features=151936, bias=False) + ) + (vision_tower): SIGLIPVisionTower( + (_vision_tower): SiglipVisionModel( + (vision_model): SiglipVisionTransformer( + (embeddings): SiglipVisionEmbeddings( + (patch_embedding): Conv2d(3, 1152, kernel_size=(14, 14), stride=(14, 14), padding=valid) + (position_embedding): Embedding(729, 1152) + ) + (encoder): SiglipEncoder( + (layers): ModuleList( + (0-26): 27 x SiglipEncoderLayer( + (self_attn): SiglipAttention( + (k_proj): Linear(in_features=1152, out_features=1152, bias=True) + (v_proj): Linear(in_features=1152, out_features=1152, bias=True) + (q_proj): Linear(in_features=1152, out_features=1152, bias=True) + (out_proj): Linear(in_features=1152, out_features=1152, bias=True) + ) + (layer_norm1): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + (layer_norm2): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + ) + ) + ) + (post_layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (head): SiglipMultiheadAttentionPoolingHead( + (attention): MultiheadAttention( + (out_proj): NonDynamicallyQuantizableLinear(in_features=1152, out_features=1152, bias=True) + ) + (layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + ) + ) + ) + ) + (connector): MLPConnector( + (_connector): Sequential( + (0): SupermaskLinearSparsity_SoftForward_Normal(in_features=1152, out_features=896, bias=True) + (1): GELU(approximate='none') + (2): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=True) + ) + ) +) +Pre-training init language_model.model.layers.0.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.0.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.0.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.0.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.0.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.0.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.0.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.1.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.1.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.1.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.1.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.1.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.1.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.1.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.2.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.2.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.2.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.2.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.2.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.2.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.2.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.3.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.3.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.3.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.3.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.3.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.3.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.3.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.4.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.4.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.4.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.4.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.4.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.4.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.4.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.5.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.5.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.5.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.5.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.5.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.5.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.5.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.6.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.6.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.6.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.6.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.6.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.6.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.6.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.7.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.7.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.7.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.7.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.7.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.7.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.7.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.8.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.8.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.8.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.8.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.8.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.8.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.8.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.9.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.9.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.9.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.9.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.9.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.9.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.9.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.10.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.10.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.10.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.10.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.10.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.10.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.10.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.11.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.11.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.11.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.11.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.11.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.11.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.11.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.12.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.12.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.12.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.12.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.12.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.12.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.12.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.13.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.13.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.13.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.13.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.13.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.13.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.13.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.14.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.14.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.14.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.14.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.14.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.14.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.14.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.15.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.15.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.15.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.15.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.15.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.15.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.15.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.16.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.16.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.16.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.16.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.16.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.16.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.16.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.17.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.17.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.17.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.17.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.17.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.17.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.17.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.18.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.18.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.18.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.18.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.18.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.18.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.18.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.19.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.19.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.19.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.19.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.19.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.19.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.19.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.20.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.20.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.20.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.20.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.20.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.20.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.20.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.21.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.21.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.21.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.21.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.21.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.21.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.21.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.22.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.22.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.22.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.22.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.22.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.22.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.22.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.23.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.23.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.23.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.23.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.23.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.23.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.23.mlp.down_proj.scores: Mean=1.000000 +Pre-training init connector._connector.0.scores: Mean=1.000005 +Pre-training init connector._connector.2.scores: Mean=0.999970 +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +2025-10-12 10:33:49,461 | INFO: Total Parameters: 1283756736, Total Trainable Parameters: 359661568 +2025-10-12 10:33:49,466 | INFO: Trainable Parameters: +language_model.model.layers.0.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.0.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.0.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.0.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.0.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.0.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.0.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.1.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.1.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.1.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.1.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.1.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.1.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.1.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.2.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.2.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.2.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.2.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.2.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.2.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.2.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.3.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.3.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.3.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.3.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.3.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.3.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.3.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.4.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.4.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.4.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.4.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.4.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.4.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.4.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.5.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.5.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.5.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.5.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.5.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.5.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.5.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.6.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.6.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.6.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.6.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.6.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.6.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.6.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.7.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.7.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.7.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.7.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.7.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.7.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.7.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.8.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.8.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.8.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.8.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.8.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.8.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.8.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.9.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.9.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.9.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.9.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.9.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.9.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.9.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.10.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.10.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.10.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.10.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.10.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.10.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.10.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.11.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.11.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.11.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.11.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.11.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.11.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.11.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.12.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.12.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.12.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.12.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.12.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.12.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.12.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.13.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.13.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.13.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.13.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.13.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.13.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.13.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.14.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.14.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.14.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.14.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.14.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.14.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.14.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.15.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.15.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.15.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.15.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.15.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.15.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.15.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.16.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.16.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.16.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.16.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.16.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.16.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.16.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.17.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.17.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.17.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.17.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.17.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.17.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.17.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.18.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.18.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.18.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.18.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.18.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.18.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.18.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.19.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.19.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.19.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.19.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.19.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.19.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.19.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.20.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.20.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.20.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.20.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.20.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.20.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.20.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.21.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.21.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.21.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.21.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.21.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.21.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.21.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.22.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.22.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.22.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.22.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.22.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.22.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.22.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.23.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.23.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.23.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.23.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.23.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.23.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.23.mlp.down_proj.scores: 4358144 parameters +connector._connector.0.scores: 1032192 parameters +connector._connector.2.scores: 802816 parameters +Parameter Offload: Total persistent parameters: 486464 in 403 params + 0%| | 0/520 [00:004->3 [1] 5/-1/-1->4->3 [2] 5/-1/-1->4->3 [3] 5/-1/-1->4->3 [4] 5/-1/-1->4->3 [5] 5/-1/-1->4->3 [6] 5/-1/-1->4->3 [7] 5/-1/-1->4->3 [8] 5/-1/-1->4->3 [9] 5/-1/-1->4->3 [10] 5/-1/-1->4->3 [11] 5/-1/-1->4->3 [12] 5/-1/-1->4->3 [13] 5/-1/-1->4->3 [14] 5/-1/-1->4->3 [15] 5/-1/-1->4->3 [16] 5/-1/-1->4->3 [17] 5/-1/-1->4->3 [18] 5/-1/-1->4->3 [19] 5/-1/-1->4->3 [20] 5/-1/-1->4->3 [21] 5/-1/-1->4->3 [22] 5/-1/-1->4->3 [23] 5/-1/-1->4->3 +ywang29-vrdb-test1-worker-0:325944:333045 [0] NCCL INFO Channel 06/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:325948:333046 [4] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:325947:333050 [3] NCCL INFO Trees [0] 4/-1/-1->3->2 [1] 4/-1/-1->3->2 [2] 4/-1/-1->3->2 [3] 4/-1/-1->3->2 [4] 4/-1/-1->3->2 [5] 4/-1/-1->3->2 [6] 4/-1/-1->3->2 [7] 4/-1/-1->3->2 [8] 4/-1/-1->3->2 [9] 4/-1/-1->3->2 [10] 4/-1/-1->3->2 [11] 4/-1/-1->3->2 [12] 4/-1/-1->3->2 [13] 4/-1/-1->3->2 [14] 4/-1/-1->3->2 [15] 4/-1/-1->3->2 [16] 4/-1/-1->3->2 [17] 4/-1/-1->3->2 [18] 4/-1/-1->3->2 [19] 4/-1/-1->3->2 [20] 4/-1/-1->3->2 [21] 4/-1/-1->3->2 [22] 4/-1/-1->3->2 [23] 4/-1/-1->3->2 +ywang29-vrdb-test1-worker-0:325945:333048 [1] NCCL INFO Trees [0] 2/-1/-1->1->0 [1] 2/-1/-1->1->0 [2] 2/-1/-1->1->0 [3] 2/-1/-1->1->0 [4] 2/-1/-1->1->0 [5] 2/-1/-1->1->0 [6] 2/-1/-1->1->0 [7] 2/-1/-1->1->0 [8] 2/-1/-1->1->0 [9] 2/-1/-1->1->0 [10] 2/-1/-1->1->0 [11] 2/-1/-1->1->0 [12] 2/-1/-1->1->0 [13] 2/-1/-1->1->0 [14] 2/-1/-1->1->0 [15] 2/-1/-1->1->0 [16] 2/-1/-1->1->0 [17] 2/-1/-1->1->0 [18] 2/-1/-1->1->0 [19] 2/-1/-1->1->0 [20] 2/-1/-1->1->0 [21] 2/-1/-1->1->0 [22] 2/-1/-1->1->0 [23] 2/-1/-1->1->0 +ywang29-vrdb-test1-worker-0:325944:333045 [0] NCCL INFO Channel 07/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:325947:333050 [3] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:325945:333048 [1] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:325951:333047 [7] NCCL INFO Trees [0] -1/-1/-1->7->6 [1] -1/-1/-1->7->6 [2] -1/-1/-1->7->6 [3] -1/-1/-1->7->6 [4] -1/-1/-1->7->6 [5] -1/-1/-1->7->6 [6] -1/-1/-1->7->6 [7] -1/-1/-1->7->6 [8] -1/-1/-1->7->6 [9] -1/-1/-1->7->6 [10] -1/-1/-1->7->6 [11] -1/-1/-1->7->6 [12] -1/-1/-1->7->6 [13] -1/-1/-1->7->6 [14] -1/-1/-1->7->6 [15] -1/-1/-1->7->6 [16] -1/-1/-1->7->6 [17] -1/-1/-1->7->6 [18] -1/-1/-1->7->6 [19] -1/-1/-1->7->6 [20] -1/-1/-1->7->6 [21] -1/-1/-1->7->6 [22] -1/-1/-1->7->6 [23] -1/-1/-1->7->6 +ywang29-vrdb-test1-worker-0:325944:333045 [0] NCCL INFO Channel 08/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:325946:333049 [2] NCCL INFO Trees [0] 3/-1/-1->2->1 [1] 3/-1/-1->2->1 [2] 3/-1/-1->2->1 [3] 3/-1/-1->2->1 [4] 3/-1/-1->2->1 [5] 3/-1/-1->2->1 [6] 3/-1/-1->2->1 [7] 3/-1/-1->2->1 [8] 3/-1/-1->2->1 [9] 3/-1/-1->2->1 [10] 3/-1/-1->2->1 [11] 3/-1/-1->2->1 [12] 3/-1/-1->2->1 [13] 3/-1/-1->2->1 [14] 3/-1/-1->2->1 [15] 3/-1/-1->2->1 [16] 3/-1/-1->2->1 [17] 3/-1/-1->2->1 [18] 3/-1/-1->2->1 [19] 3/-1/-1->2->1 [20] 3/-1/-1->2->1 [21] 3/-1/-1->2->1 [22] 3/-1/-1->2->1 [23] 3/-1/-1->2->1 +ywang29-vrdb-test1-worker-0:325951:333047 [7] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:325946:333049 [2] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:325944:333045 [0] NCCL INFO Channel 09/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:325944:333045 [0] NCCL INFO Channel 10/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:325944:333045 [0] NCCL INFO Channel 11/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:325944:333045 [0] NCCL INFO Channel 12/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:325944:333045 [0] NCCL INFO Channel 13/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:325944:333045 [0] NCCL INFO Channel 14/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:325944:333045 [0] NCCL INFO Channel 15/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:325944:333045 [0] NCCL INFO Channel 16/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:325944:333045 [0] NCCL INFO Channel 17/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:325950:333051 [6] NCCL INFO Trees [0] 7/-1/-1->6->5 [1] 7/-1/-1->6->5 [2] 7/-1/-1->6->5 [3] 7/-1/-1->6->5 [4] 7/-1/-1->6->5 [5] 7/-1/-1->6->5 [6] 7/-1/-1->6->5 [7] 7/-1/-1->6->5 [8] 7/-1/-1->6->5 [9] 7/-1/-1->6->5 [10] 7/-1/-1->6->5 [11] 7/-1/-1->6->5 [12] 7/-1/-1->6->5 [13] 7/-1/-1->6->5 [14] 7/-1/-1->6->5 [15] 7/-1/-1->6->5 [16] 7/-1/-1->6->5 [17] 7/-1/-1->6->5 [18] 7/-1/-1->6->5 [19] 7/-1/-1->6->5 [20] 7/-1/-1->6->5 [21] 7/-1/-1->6->5 [22] 7/-1/-1->6->5 [23] 7/-1/-1->6->5 +ywang29-vrdb-test1-worker-0:325949:333052 [5] NCCL INFO Trees [0] 6/-1/-1->5->4 [1] 6/-1/-1->5->4 [2] 6/-1/-1->5->4 [3] 6/-1/-1->5->4 [4] 6/-1/-1->5->4 [5] 6/-1/-1->5->4 [6] 6/-1/-1->5->4 [7] 6/-1/-1->5->4 [8] 6/-1/-1->5->4 [9] 6/-1/-1->5->4 [10] 6/-1/-1->5->4 [11] 6/-1/-1->5->4 [12] 6/-1/-1->5->4 [13] 6/-1/-1->5->4 [14] 6/-1/-1->5->4 [15] 6/-1/-1->5->4 [16] 6/-1/-1->5->4 [17] 6/-1/-1->5->4 [18] 6/-1/-1->5->4 [19] 6/-1/-1->5->4 [20] 6/-1/-1->5->4 [21] 6/-1/-1->5->4 [22] 6/-1/-1->5->4 [23] 6/-1/-1->5->4 +ywang29-vrdb-test1-worker-0:325944:333045 [0] NCCL INFO Channel 18/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:325944:333045 [0] NCCL INFO Channel 19/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:325949:333052 [5] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:325950:333051 [6] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:325944:333045 [0] NCCL INFO Channel 20/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:325944:333045 [0] NCCL INFO Channel 21/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:325944:333045 [0] NCCL INFO Channel 22/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:325944:333045 [0] NCCL INFO Channel 23/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:325944:333045 [0] NCCL INFO Trees [0] 1/-1/-1->0->-1 [1] 1/-1/-1->0->-1 [2] 1/-1/-1->0->-1 [3] 1/-1/-1->0->-1 [4] 1/-1/-1->0->-1 [5] 1/-1/-1->0->-1 [6] 1/-1/-1->0->-1 [7] 1/-1/-1->0->-1 [8] 1/-1/-1->0->-1 [9] 1/-1/-1->0->-1 [10] 1/-1/-1->0->-1 [11] 1/-1/-1->0->-1 [12] 1/-1/-1->0->-1 [13] 1/-1/-1->0->-1 [14] 1/-1/-1->0->-1 [15] 1/-1/-1->0->-1 [16] 1/-1/-1->0->-1 [17] 1/-1/-1->0->-1 [18] 1/-1/-1->0->-1 [19] 1/-1/-1->0->-1 [20] 1/-1/-1->0->-1 [21] 1/-1/-1->0->-1 [22] 1/-1/-1->0->-1 [23] 1/-1/-1->0->-1 +ywang29-vrdb-test1-worker-0:325944:333045 [0] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:325945:333048 [1] NCCL INFO Channel 00/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:333046 [4] NCCL INFO Channel 00/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:333050 [3] NCCL INFO Channel 00/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:333049 [2] NCCL INFO Channel 00/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:333048 [1] NCCL INFO Channel 01/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:333046 [4] NCCL INFO Channel 01/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:333050 [3] NCCL INFO Channel 01/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325944:333045 [0] NCCL INFO Channel 00/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:333049 [2] NCCL INFO Channel 01/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:333048 [1] NCCL INFO Channel 02/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:333046 [4] NCCL INFO Channel 02/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:333050 [3] NCCL INFO Channel 02/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325944:333045 [0] NCCL INFO Channel 01/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:333049 [2] NCCL INFO Channel 02/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:333048 [1] NCCL INFO Channel 03/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:333046 [4] NCCL INFO Channel 03/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:333050 [3] NCCL INFO Channel 03/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325944:333045 [0] NCCL INFO Channel 02/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:333048 [1] NCCL INFO Channel 04/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:333049 [2] NCCL INFO Channel 03/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:333046 [4] NCCL INFO Channel 04/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:333050 [3] NCCL INFO Channel 04/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325944:333045 [0] NCCL INFO Channel 03/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:333052 [5] NCCL INFO Channel 00/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:333049 [2] NCCL INFO Channel 04/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:333048 [1] NCCL INFO Channel 05/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:333051 [6] NCCL INFO Channel 00/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:333046 [4] NCCL INFO Channel 05/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:333050 [3] NCCL INFO Channel 05/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325944:333045 [0] NCCL INFO Channel 04/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:333052 [5] NCCL INFO Channel 01/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:333049 [2] NCCL INFO Channel 05/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:333048 [1] NCCL INFO Channel 06/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:333051 [6] NCCL INFO Channel 01/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:333046 [4] NCCL INFO Channel 06/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:333050 [3] NCCL INFO Channel 06/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325944:333045 [0] NCCL INFO Channel 05/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:333052 [5] NCCL INFO Channel 02/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:333049 [2] NCCL INFO Channel 06/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:333048 [1] NCCL INFO Channel 07/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:333046 [4] NCCL INFO Channel 07/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:333050 [3] NCCL INFO Channel 07/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:333051 [6] NCCL INFO Channel 02/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325944:333045 [0] NCCL INFO Channel 06/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:333052 [5] NCCL INFO Channel 03/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:333049 [2] NCCL INFO Channel 07/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:333048 [1] NCCL INFO Channel 08/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:333046 [4] NCCL INFO Channel 08/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:333050 [3] NCCL INFO Channel 08/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:333051 [6] NCCL INFO Channel 03/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325944:333045 [0] NCCL INFO Channel 07/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:333049 [2] NCCL INFO Channel 08/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:333052 [5] NCCL INFO Channel 04/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:333048 [1] NCCL INFO Channel 09/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:333046 [4] NCCL INFO Channel 09/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:333050 [3] NCCL INFO Channel 09/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:333051 [6] NCCL INFO Channel 04/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325944:333045 [0] NCCL INFO Channel 08/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:333049 [2] NCCL INFO Channel 09/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:333052 [5] NCCL INFO Channel 05/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:333046 [4] NCCL INFO Channel 10/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:333048 [1] NCCL INFO Channel 10/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:333050 [3] NCCL INFO Channel 10/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:333051 [6] NCCL INFO Channel 05/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325944:333045 [0] NCCL INFO Channel 09/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:333049 [2] NCCL INFO Channel 10/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:333046 [4] NCCL INFO Channel 11/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:333052 [5] NCCL INFO Channel 06/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:333048 [1] NCCL INFO Channel 11/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:333050 [3] NCCL INFO Channel 11/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:333047 [7] NCCL INFO Channel 00/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:333051 [6] NCCL INFO Channel 06/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325944:333045 [0] NCCL INFO Channel 10/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:333046 [4] NCCL INFO Channel 12/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:333049 [2] NCCL INFO Channel 11/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:333052 [5] NCCL INFO Channel 07/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:333048 [1] NCCL INFO Channel 12/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:333050 [3] NCCL INFO Channel 12/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:333047 [7] NCCL INFO Channel 01/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:333051 [6] NCCL INFO Channel 07/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325944:333045 [0] NCCL INFO Channel 11/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:333046 [4] NCCL INFO Channel 13/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:333049 [2] NCCL INFO Channel 12/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:333052 [5] NCCL INFO Channel 08/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:333048 [1] NCCL INFO Channel 13/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:333050 [3] NCCL INFO Channel 13/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:333047 [7] NCCL INFO Channel 02/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325944:333045 [0] NCCL INFO Channel 12/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:333051 [6] NCCL INFO Channel 08/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:333046 [4] NCCL INFO Channel 14/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:333049 [2] NCCL INFO Channel 13/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:333048 [1] NCCL INFO Channel 14/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:333052 [5] NCCL INFO Channel 09/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:333050 [3] NCCL INFO Channel 14/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:333047 [7] NCCL INFO Channel 03/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325944:333045 [0] NCCL INFO Channel 13/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:333046 [4] NCCL INFO Channel 15/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:333051 [6] NCCL INFO Channel 09/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:333048 [1] NCCL INFO Channel 15/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:333049 [2] NCCL INFO Channel 14/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:333052 [5] NCCL INFO Channel 10/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:333050 [3] NCCL INFO Channel 15/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:333047 [7] NCCL INFO Channel 04/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325944:333045 [0] NCCL INFO Channel 14/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:333046 [4] NCCL INFO Channel 16/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:333051 [6] NCCL INFO Channel 10/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:333048 [1] NCCL INFO Channel 16/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:333049 [2] NCCL INFO Channel 15/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:333052 [5] NCCL INFO Channel 11/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:333050 [3] NCCL INFO Channel 16/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:333047 [7] NCCL INFO Channel 05/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325944:333045 [0] NCCL INFO Channel 15/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:333046 [4] NCCL INFO Channel 17/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:333051 [6] NCCL INFO Channel 11/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:333048 [1] NCCL INFO Channel 17/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:333049 [2] NCCL INFO Channel 16/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:333052 [5] NCCL INFO Channel 12/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:333050 [3] NCCL INFO Channel 17/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:333047 [7] NCCL INFO Channel 06/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325944:333045 [0] NCCL INFO Channel 16/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:333046 [4] NCCL INFO Channel 18/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:333051 [6] NCCL INFO Channel 12/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:333048 [1] NCCL INFO Channel 18/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:333049 [2] NCCL INFO Channel 17/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:333050 [3] NCCL INFO Channel 18/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:333052 [5] NCCL INFO Channel 13/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:333047 [7] NCCL INFO Channel 07/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325944:333045 [0] NCCL INFO Channel 17/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:333046 [4] NCCL INFO Channel 19/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:333048 [1] NCCL INFO Channel 19/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:333049 [2] NCCL INFO Channel 18/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:333051 [6] NCCL INFO Channel 13/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:333052 [5] NCCL INFO Channel 14/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:333050 [3] NCCL INFO Channel 19/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:333047 [7] NCCL INFO Channel 08/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325944:333045 [0] NCCL INFO Channel 18/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:333046 [4] NCCL INFO Channel 20/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:333048 [1] NCCL INFO Channel 20/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:333051 [6] NCCL INFO Channel 14/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:333049 [2] NCCL INFO Channel 19/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:333052 [5] NCCL INFO Channel 15/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:333050 [3] NCCL INFO Channel 20/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:333047 [7] NCCL INFO Channel 09/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325944:333045 [0] NCCL INFO Channel 19/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:333046 [4] NCCL INFO Channel 21/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:333048 [1] NCCL INFO Channel 21/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:333051 [6] NCCL INFO Channel 15/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:333049 [2] NCCL INFO Channel 20/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:333050 [3] NCCL INFO Channel 21/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:333052 [5] NCCL INFO Channel 16/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:333047 [7] NCCL INFO Channel 10/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325944:333045 [0] NCCL INFO Channel 20/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:333046 [4] NCCL INFO Channel 22/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:333048 [1] NCCL INFO Channel 22/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:333049 [2] NCCL INFO Channel 21/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:333051 [6] NCCL INFO Channel 16/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:333050 [3] NCCL INFO Channel 22/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:333052 [5] NCCL INFO Channel 17/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:333047 [7] NCCL INFO Channel 11/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325944:333045 [0] NCCL INFO Channel 21/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:333046 [4] NCCL INFO Channel 23/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:333048 [1] NCCL INFO Channel 23/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:333049 [2] NCCL INFO Channel 22/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:333050 [3] NCCL INFO Channel 23/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:333051 [6] NCCL INFO Channel 17/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:333047 [7] NCCL INFO Channel 12/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325944:333045 [0] NCCL INFO Channel 22/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:333052 [5] NCCL INFO Channel 18/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:333049 [2] NCCL INFO Channel 23/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:333051 [6] NCCL INFO Channel 18/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:333047 [7] NCCL INFO Channel 13/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325944:333045 [0] NCCL INFO Channel 23/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:333052 [5] NCCL INFO Channel 19/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:333047 [7] NCCL INFO Channel 14/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:333051 [6] NCCL INFO Channel 19/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:333052 [5] NCCL INFO Channel 20/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:333047 [7] NCCL INFO Channel 15/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:333047 [7] NCCL INFO Channel 16/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:333052 [5] NCCL INFO Channel 21/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:333051 [6] NCCL INFO Channel 20/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:333047 [7] NCCL INFO Channel 17/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:333052 [5] NCCL INFO Channel 22/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:333047 [7] NCCL INFO Channel 18/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:333051 [6] NCCL INFO Channel 21/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:333052 [5] NCCL INFO Channel 23/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:333051 [6] NCCL INFO Channel 22/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:333047 [7] NCCL INFO Channel 19/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:333051 [6] NCCL INFO Channel 23/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:333047 [7] NCCL INFO Channel 20/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:333047 [7] NCCL INFO Channel 21/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:333047 [7] NCCL INFO Channel 22/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:333047 [7] NCCL INFO Channel 23/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:333049 [2] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:325947:333050 [3] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:325945:333048 [1] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:325944:333045 [0] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:325946:333049 [2] NCCL INFO Channel 00/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:333049 [2] NCCL INFO Channel 01/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:333049 [2] NCCL INFO Channel 02/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:333049 [2] NCCL INFO Channel 03/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:333049 [2] NCCL INFO Channel 04/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:333049 [2] NCCL INFO Channel 05/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:333049 [2] NCCL INFO Channel 06/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:333049 [2] NCCL INFO Channel 07/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:333049 [2] NCCL INFO Channel 08/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:333049 [2] NCCL INFO Channel 09/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:333046 [4] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:325946:333049 [2] NCCL INFO Channel 10/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:333049 [2] NCCL INFO Channel 11/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:333049 [2] NCCL INFO Channel 12/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:333047 [7] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:325951:333047 [7] NCCL INFO Channel 00/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:333052 [5] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:325950:333051 [6] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:325946:333049 [2] NCCL INFO Channel 13/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:333047 [7] NCCL INFO Channel 01/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:333049 [2] NCCL INFO Channel 14/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:333047 [7] NCCL INFO Channel 02/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:333049 [2] NCCL INFO Channel 15/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:333047 [7] NCCL INFO Channel 03/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:333049 [2] NCCL INFO Channel 16/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:333047 [7] NCCL INFO Channel 04/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:333049 [2] NCCL INFO Channel 17/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:333050 [3] NCCL INFO Channel 00/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:333047 [7] NCCL INFO Channel 05/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:333049 [2] NCCL INFO Channel 18/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:333050 [3] NCCL INFO Channel 01/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:333049 [2] NCCL INFO Channel 19/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:333050 [3] NCCL INFO Channel 02/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:333049 [2] NCCL INFO Channel 20/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:333050 [3] NCCL INFO Channel 03/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:333049 [2] NCCL INFO Channel 21/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:333050 [3] NCCL INFO Channel 04/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:333049 [2] NCCL INFO Channel 22/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:333050 [3] NCCL INFO Channel 05/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:333047 [7] NCCL INFO Channel 06/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325946:333049 [2] NCCL INFO Channel 23/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:333050 [3] NCCL INFO Channel 06/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:333047 [7] NCCL INFO Channel 07/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:333050 [3] NCCL INFO Channel 07/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:333047 [7] NCCL INFO Channel 08/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:333050 [3] NCCL INFO Channel 08/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:333047 [7] NCCL INFO Channel 09/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:333050 [3] NCCL INFO Channel 09/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:333047 [7] NCCL INFO Channel 10/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:333050 [3] NCCL INFO Channel 10/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:333047 [7] NCCL INFO Channel 11/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:333050 [3] NCCL INFO Channel 11/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:333047 [7] NCCL INFO Channel 12/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:333050 [3] NCCL INFO Channel 12/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:333047 [7] NCCL INFO Channel 13/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:333050 [3] NCCL INFO Channel 13/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:333047 [7] NCCL INFO Channel 14/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:333050 [3] NCCL INFO Channel 14/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:333047 [7] NCCL INFO Channel 15/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:333050 [3] NCCL INFO Channel 15/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:333047 [7] NCCL INFO Channel 16/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:333050 [3] NCCL INFO Channel 16/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:333047 [7] NCCL INFO Channel 17/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:333050 [3] NCCL INFO Channel 17/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:333050 [3] NCCL INFO Channel 18/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:333052 [5] NCCL INFO Channel 00/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:333050 [3] NCCL INFO Channel 19/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:333048 [1] NCCL INFO Channel 00/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:333046 [4] NCCL INFO Channel 00/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:333051 [6] NCCL INFO Channel 00/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:333052 [5] NCCL INFO Channel 01/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:333050 [3] NCCL INFO Channel 20/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:333048 [1] NCCL INFO Channel 01/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:333046 [4] NCCL INFO Channel 01/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:333051 [6] NCCL INFO Channel 01/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:333052 [5] NCCL INFO Channel 02/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:333050 [3] NCCL INFO Channel 21/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:333048 [1] NCCL INFO Channel 02/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:333046 [4] NCCL INFO Channel 02/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:333051 [6] NCCL INFO Channel 02/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:333052 [5] NCCL INFO Channel 03/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:333050 [3] NCCL INFO Channel 22/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:333048 [1] NCCL INFO Channel 03/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:333046 [4] NCCL INFO Channel 03/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:333051 [6] NCCL INFO Channel 03/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:333052 [5] NCCL INFO Channel 04/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325947:333050 [3] NCCL INFO Channel 23/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:333048 [1] NCCL INFO Channel 04/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:333046 [4] NCCL INFO Channel 04/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:333051 [6] NCCL INFO Channel 04/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:333052 [5] NCCL INFO Channel 05/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:333048 [1] NCCL INFO Channel 05/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:333047 [7] NCCL INFO Channel 18/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:333046 [4] NCCL INFO Channel 05/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:333051 [6] NCCL INFO Channel 05/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:333052 [5] NCCL INFO Channel 06/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:333048 [1] NCCL INFO Channel 06/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:333047 [7] NCCL INFO Channel 19/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:333046 [4] NCCL INFO Channel 06/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:333051 [6] NCCL INFO Channel 06/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:333052 [5] NCCL INFO Channel 07/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:333048 [1] NCCL INFO Channel 07/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:333047 [7] NCCL INFO Channel 20/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:333051 [6] NCCL INFO Channel 07/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:333052 [5] NCCL INFO Channel 08/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:333048 [1] NCCL INFO Channel 08/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:333047 [7] NCCL INFO Channel 21/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:333051 [6] NCCL INFO Channel 08/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:333052 [5] NCCL INFO Channel 09/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:333048 [1] NCCL INFO Channel 09/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:333047 [7] NCCL INFO Channel 22/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:333052 [5] NCCL INFO Channel 10/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:333051 [6] NCCL INFO Channel 09/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:333048 [1] NCCL INFO Channel 10/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325951:333047 [7] NCCL INFO Channel 23/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:333052 [5] NCCL INFO Channel 11/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:333051 [6] NCCL INFO Channel 10/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:333048 [1] NCCL INFO Channel 11/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:333052 [5] NCCL INFO Channel 12/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:333051 [6] NCCL INFO Channel 11/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:333048 [1] NCCL INFO Channel 12/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:333052 [5] NCCL INFO Channel 13/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:333051 [6] NCCL INFO Channel 12/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:333048 [1] NCCL INFO Channel 13/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:333052 [5] NCCL INFO Channel 14/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:333051 [6] NCCL INFO Channel 13/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:333048 [1] NCCL INFO Channel 14/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:333046 [4] NCCL INFO Channel 07/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:333052 [5] NCCL INFO Channel 15/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:333051 [6] NCCL INFO Channel 14/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:333048 [1] NCCL INFO Channel 15/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:333046 [4] NCCL INFO Channel 08/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:333052 [5] NCCL INFO Channel 16/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:333051 [6] NCCL INFO Channel 15/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:333048 [1] NCCL INFO Channel 16/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:333046 [4] NCCL INFO Channel 09/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:333052 [5] NCCL INFO Channel 17/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:333051 [6] NCCL INFO Channel 16/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:333048 [1] NCCL INFO Channel 17/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:333046 [4] NCCL INFO Channel 10/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:333052 [5] NCCL INFO Channel 18/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:333051 [6] NCCL INFO Channel 17/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:333048 [1] NCCL INFO Channel 18/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:333046 [4] NCCL INFO Channel 11/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:333052 [5] NCCL INFO Channel 19/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:333051 [6] NCCL INFO Channel 18/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:333048 [1] NCCL INFO Channel 19/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:333046 [4] NCCL INFO Channel 12/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:333052 [5] NCCL INFO Channel 20/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:333051 [6] NCCL INFO Channel 19/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:333048 [1] NCCL INFO Channel 20/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:333046 [4] NCCL INFO Channel 13/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:333052 [5] NCCL INFO Channel 21/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:333051 [6] NCCL INFO Channel 20/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:333048 [1] NCCL INFO Channel 21/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:333046 [4] NCCL INFO Channel 14/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:333052 [5] NCCL INFO Channel 22/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:333051 [6] NCCL INFO Channel 21/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:333048 [1] NCCL INFO Channel 22/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:333046 [4] NCCL INFO Channel 15/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325949:333052 [5] NCCL INFO Channel 23/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:333051 [6] NCCL INFO Channel 22/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325945:333048 [1] NCCL INFO Channel 23/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:333046 [4] NCCL INFO Channel 16/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325950:333051 [6] NCCL INFO Channel 23/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:333046 [4] NCCL INFO Channel 17/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:333046 [4] NCCL INFO Channel 18/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:333046 [4] NCCL INFO Channel 19/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:333046 [4] NCCL INFO Channel 20/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:333046 [4] NCCL INFO Channel 21/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:333046 [4] NCCL INFO Channel 22/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325948:333046 [4] NCCL INFO Channel 23/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:325944:333045 [0] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:325944:333045 [0] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:325944:333045 [0] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:325945:333048 [1] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:325945:333048 [1] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:325945:333048 [1] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:325946:333049 [2] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:325946:333049 [2] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:325946:333049 [2] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:325947:333050 [3] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:325947:333050 [3] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:325947:333050 [3] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:325948:333046 [4] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:325948:333046 [4] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:325948:333046 [4] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:325949:333052 [5] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:325949:333052 [5] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:325949:333052 [5] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:325951:333047 [7] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:325950:333051 [6] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:325951:333047 [7] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:325951:333047 [7] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:325950:333051 [6] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:325950:333051 [6] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:325944:333045 [0] NCCL INFO ncclCommInitRank comm 0x7fb3c406b210 rank 0 nranks 8 cudaDev 0 nvmlDev 0 busId 101c0 commId 0xb91eb6918d943c50 - Init COMPLETE +ywang29-vrdb-test1-worker-0:325948:333046 [4] NCCL INFO ncclCommInitRank comm 0x7fbb7406aff0 rank 4 nranks 8 cudaDev 4 nvmlDev 4 busId 901c0 commId 0xb91eb6918d943c50 - Init COMPLETE +ywang29-vrdb-test1-worker-0:325950:333051 [6] NCCL INFO ncclCommInitRank comm 0x7fd6bc06b130 rank 6 nranks 8 cudaDev 6 nvmlDev 6 busId a01c0 commId 0xb91eb6918d943c50 - Init COMPLETE +ywang29-vrdb-test1-worker-0:325946:333049 [2] NCCL INFO ncclCommInitRank comm 0x7f7a2006ac40 rank 2 nranks 8 cudaDev 2 nvmlDev 2 busId 201c0 commId 0xb91eb6918d943c50 - Init COMPLETE +ywang29-vrdb-test1-worker-0:325947:333050 [3] NCCL INFO ncclCommInitRank comm 0x7fb01c06b0a0 rank 3 nranks 8 cudaDev 3 nvmlDev 3 busId 201d0 commId 0xb91eb6918d943c50 - Init COMPLETE +ywang29-vrdb-test1-worker-0:325951:333047 [7] NCCL INFO ncclCommInitRank comm 0x7f099c06a9f0 rank 7 nranks 8 cudaDev 7 nvmlDev 7 busId a01d0 commId 0xb91eb6918d943c50 - Init COMPLETE +ywang29-vrdb-test1-worker-0:325949:333052 [5] NCCL INFO ncclCommInitRank comm 0x7f563c06ab80 rank 5 nranks 8 cudaDev 5 nvmlDev 5 busId 901d0 commId 0xb91eb6918d943c50 - Init COMPLETE +ywang29-vrdb-test1-worker-0:325945:333048 [1] NCCL INFO ncclCommInitRank comm 0x7f2e8c06a870 rank 1 nranks 8 cudaDev 1 nvmlDev 1 busId 101d0 commId 0xb91eb6918d943c50 - Init COMPLETE + 0%| | 1/520 [00:13<1:59:03, 13.76s/it] {'loss': 8.8834, 'grad_norm': 0.4118065844018823, 'learning_rate': 0.0125, 'epoch': 0.0} + 0%| | 1/520 [00:13<1:59:03, 13.76s/it] 0%| | 2/520 [00:17<1:09:13, 8.02s/it] {'loss': 8.0887, 'grad_norm': 0.38124599467066633, 'learning_rate': 0.025, 'epoch': 0.0} + 0%| | 2/520 [00:17<1:09:13, 8.02s/it] 1%| | 3/520 [00:21<53:41, 6.23s/it] {'loss': 7.2136, 'grad_norm': 0.13980115991186218, 'learning_rate': 0.037500000000000006, 'epoch': 0.01} + 1%| | 3/520 [00:21<53:41, 6.23s/it] 1%| | 4/520 [00:25<46:21, 5.39s/it] {'loss': 6.5104, 'grad_norm': 0.08992470751655822, 'learning_rate': 0.05, 'epoch': 0.01} + 1%| | 4/520 [00:25<46:21, 5.39s/it] 1%| | 5/520 [00:29<41:46, 4.87s/it] {'loss': 5.8973, 'grad_norm': 0.09236914591424679, 'learning_rate': 0.0625, 'epoch': 0.01} + 1%| | 5/520 [00:29<41:46, 4.87s/it] 1%| | 6/520 [00:33<38:46, 4.53s/it] {'loss': 6.8181, 'grad_norm': 0.2651972547111439, 'learning_rate': 0.07500000000000001, 'epoch': 0.01} + 1%| | 6/520 [00:33<38:46, 4.53s/it] 1%|▏ | 7/520 [00:37<36:50, 4.31s/it] {'loss': 5.0461, 'grad_norm': 0.08826342818104276, 'learning_rate': 0.08750000000000001, 'epoch': 0.01} + 1%|▏ | 7/520 [00:37<36:50, 4.31s/it] 2%|▏ | 8/520 [00:42<37:15, 4.37s/it] {'loss': 4.6557, 'grad_norm': 0.04406826320024679, 'learning_rate': 0.1, 'epoch': 0.02} + 2%|▏ | 8/520 [00:42<37:15, 4.37s/it] 2%|▏ | 9/520 [00:46<37:23, 4.39s/it] {'loss': 4.1473, 'grad_norm': 0.03236520576249872, 'learning_rate': 0.1125, 'epoch': 0.02} + 2%|▏ | 9/520 [00:46<37:23, 4.39s/it] 2%|▏ | 10/520 [00:50<35:51, 4.22s/it] {'loss': 3.2344, 'grad_norm': 0.03073383701203129, 'learning_rate': 0.125, 'epoch': 0.02} + 2%|▏ | 10/520 [00:50<35:51, 4.22s/it] 2%|▏ | 11/520 [00:54<35:16, 4.16s/it] {'loss': 3.4579, 'grad_norm': 0.047100919042872585, 'learning_rate': 0.1375, 'epoch': 0.02} + 2%|▏ | 11/520 [00:54<35:16, 4.16s/it] 2%|▏ | 12/520 [00:58<34:25, 4.07s/it] {'loss': 4.1973, 'grad_norm': 0.11165857033232954, 'learning_rate': 0.15000000000000002, 'epoch': 0.02} + 2%|▏ | 12/520 [00:58<34:25, 4.07s/it][2025-10-12 10:34:57,982] [WARNING] [stage3.py:2069:step] 1 pytorch allocator cache flushes since last step. this happens when there is high memory pressure and is detrimental to performance. if this is happening frequently consider adjusting settings to reduce memory consumption. If you are unable to make the cache flushes go away consider adding get_accelerator().empty_cache() calls in your training loop to ensure that all ranks flush their caches at the same time + 2%|▎ | 13/520 [01:02<35:35, 4.21s/it] {'loss': 2.9447, 'grad_norm': 0.023561223486599402, 'learning_rate': 0.1625, 'epoch': 0.03} + 2%|▎ | 13/520 [01:02<35:35, 4.21s/it] 3%|▎ | 14/520 [01:06<34:34, 4.10s/it] {'loss': 2.8861, 'grad_norm': 0.023053248874496612, 'learning_rate': 0.17500000000000002, 'epoch': 0.03} + 3%|▎ | 14/520 [01:06<34:34, 4.10s/it] 3%|▎ | 15/520 [01:10<33:25, 3.97s/it] {'loss': 3.3424, 'grad_norm': 0.03640231743213069, 'learning_rate': 0.1875, 'epoch': 0.03} + 3%|▎ | 15/520 [01:10<33:25, 3.97s/it] 3%|▎ | 16/520 [01:13<32:31, 3.87s/it] {'loss': 3.1519, 'grad_norm': 0.04719960781710062, 'learning_rate': 0.2, 'epoch': 0.03} + 3%|▎ | 16/520 [01:13<32:31, 3.87s/it] 3%|▎ | 17/520 [01:17<32:01, 3.82s/it] {'loss': 2.4976, 'grad_norm': 0.010786900922529236, 'learning_rate': 0.1999980572931538, 'epoch': 0.03} + 3%|▎ | 17/520 [01:17<32:01, 3.82s/it] 3%|▎ | 18/520 [01:21<31:33, 3.77s/it] {'loss': 2.214, 'grad_norm': 0.01347522002176218, 'learning_rate': 0.19999222924809748, 'epoch': 0.03} + 3%|▎ | 18/520 [01:21<31:33, 3.77s/it] 4%|▎ | 19/520 [01:25<31:27, 3.77s/it] {'loss': 2.7374, 'grad_norm': 0.0461521809509178, 'learning_rate': 0.19998251609127465, 'epoch': 0.04} + 4%|▎ | 19/520 [01:25<31:27, 3.77s/it] 4%|▍ | 20/520 [01:28<31:07, 3.74s/it] {'loss': 2.0506, 'grad_norm': 0.01937266257554415, 'learning_rate': 0.19996891820008164, 'epoch': 0.04} + 4%|▍ | 20/520 [01:28<31:07, 3.74s/it] 4%|▍ | 21/520 [01:32<30:58, 3.72s/it] {'loss': 2.6227, 'grad_norm': 0.03810348713787424, 'learning_rate': 0.19995143610285276, 'epoch': 0.04} + 4%|▍ | 21/520 [01:32<30:58, 3.72s/it] 4%|▍ | 22/520 [01:36<30:41, 3.70s/it] {'loss': 2.1961, 'grad_norm': 0.01656173769587545, 'learning_rate': 0.19993007047883987, 'epoch': 0.04} + 4%|▍ | 22/520 [01:36<30:41, 3.70s/it] 4%|▍ | 23/520 [01:39<30:30, 3.68s/it] {'loss': 2.0974, 'grad_norm': 0.014412293508508747, 'learning_rate': 0.1999048221581858, 'epoch': 0.04} + 4%|▍ | 23/520 [01:39<30:30, 3.68s/it] 5%|▍ | 24/520 [01:43<30:21, 3.67s/it] {'loss': 2.2724, 'grad_norm': 0.012141096867323444, 'learning_rate': 0.19987569212189224, 'epoch': 0.05} + 5%|▍ | 24/520 [01:43<30:21, 3.67s/it] 5%|▍ | 25/520 [01:47<30:18, 3.67s/it] {'loss': 2.0477, 'grad_norm': 0.01592679611852341, 'learning_rate': 0.19984268150178167, 'epoch': 0.05} + 5%|▍ | 25/520 [01:47<30:18, 3.67s/it] 5%|▌ | 26/520 [01:50<30:14, 3.67s/it] {'loss': 1.9488, 'grad_norm': 0.010392997031941806, 'learning_rate': 0.1998057915804532, 'epoch': 0.05} + 5%|▌ | 26/520 [01:50<30:14, 3.67s/it] 5%|▌ | 27/520 [01:54<30:04, 3.66s/it] {'loss': 1.7772, 'grad_norm': 0.010153107847599669, 'learning_rate': 0.1997650237912329, 'epoch': 0.05} + 5%|▌ | 27/520 [01:54<30:04, 3.66s/it] 5%|▌ | 28/520 [01:58<30:02, 3.66s/it] {'loss': 1.7817, 'grad_norm': 0.008048473967462739, 'learning_rate': 0.199720379718118, 'epoch': 0.05} + 5%|▌ | 28/520 [01:58<30:02, 3.66s/it] 6%|▌ | 29/520 [02:01<30:14, 3.70s/it] {'loss': 1.7678, 'grad_norm': 0.008811417087269525, 'learning_rate': 0.19967186109571552, 'epoch': 0.06} + 6%|▌ | 29/520 [02:01<30:14, 3.70s/it] 6%|▌ | 30/520 [02:05<30:19, 3.71s/it] {'loss': 2.1788, 'grad_norm': 0.01420267317288418, 'learning_rate': 0.19961946980917455, 'epoch': 0.06} + 6%|▌ | 30/520 [02:05<30:19, 3.71s/it] 6%|▌ | 31/520 [02:09<30:28, 3.74s/it] {'loss': 1.7421, 'grad_norm': 0.008516722383158603, 'learning_rate': 0.1995632078941134, 'epoch': 0.06} + 6%|▌ | 31/520 [02:09<30:28, 3.74s/it] 6%|▌ | 32/520 [02:13<30:26, 3.74s/it] {'loss': 2.2366, 'grad_norm': 0.010720858117102028, 'learning_rate': 0.19950307753654017, 'epoch': 0.06} + 6%|▌ | 32/520 [02:13<30:26, 3.74s/it] 6%|▋ | 33/520 [02:16<30:24, 3.75s/it] {'loss': 1.691, 'grad_norm': 0.007151878367120141, 'learning_rate': 0.19943908107276798, 'epoch': 0.06} + 6%|▋ | 33/520 [02:16<30:24, 3.75s/it] 7%|▋ | 34/520 [02:20<30:34, 3.77s/it] {'loss': 1.6754, 'grad_norm': 0.006983034752483857, 'learning_rate': 0.19937122098932428, 'epoch': 0.07} + 7%|▋ | 34/520 [02:20<30:34, 3.77s/it] 7%|▋ | 35/520 [02:24<30:30, 3.77s/it] {'loss': 1.6577, 'grad_norm': 0.006358575371332926, 'learning_rate': 0.19929949992285395, 'epoch': 0.07} + 7%|▋ | 35/520 [02:24<30:30, 3.77s/it] 7%|▋ | 36/520 [02:28<30:25, 3.77s/it] {'loss': 1.7849, 'grad_norm': 0.007989273780218173, 'learning_rate': 0.19922392066001723, 'epoch': 0.07} + 7%|▋ | 36/520 [02:28<30:25, 3.77s/it] 7%|▋ | 37/520 [02:31<29:59, 3.73s/it] {'loss': 2.0009, 'grad_norm': 0.010864355439015502, 'learning_rate': 0.19914448613738106, 'epoch': 0.07} + 7%|▋ | 37/520 [02:31<29:59, 3.73s/it] 7%|▋ | 38/520 [02:35<29:41, 3.70s/it] {'loss': 1.8463, 'grad_norm': 0.008815629634958886, 'learning_rate': 0.1990611994413053, 'epoch': 0.07} + 7%|▋ | 38/520 [02:35<29:41, 3.70s/it] 8%|▊ | 39/520 [02:39<29:26, 3.67s/it] {'loss': 1.6765, 'grad_norm': 0.007490686677393382, 'learning_rate': 0.19897406380782262, 'epoch': 0.07} + 8%|▊ | 39/520 [02:39<29:26, 3.67s/it] 8%|▊ | 40/520 [02:42<29:18, 3.66s/it] {'loss': 1.6637, 'grad_norm': 0.006843425627268484, 'learning_rate': 0.19888308262251286, 'epoch': 0.08} + 8%|▊ | 40/520 [02:42<29:18, 3.66s/it] 8%|▊ | 41/520 [02:46<29:11, 3.66s/it] {'loss': 1.6519, 'grad_norm': 0.006204833478355677, 'learning_rate': 0.19878825942037148, 'epoch': 0.08} + 8%|▊ | 41/520 [02:46<29:11, 3.66s/it] 8%|▊ | 42/520 [02:50<29:02, 3.64s/it] {'loss': 1.6789, 'grad_norm': 0.006415877067612326, 'learning_rate': 0.19868959788567211, 'epoch': 0.08} + 8%|▊ | 42/520 [02:50<29:02, 3.64s/it] 8%|▊ | 43/520 [02:53<29:03, 3.65s/it] {'loss': 1.8514, 'grad_norm': 0.008441921156015046, 'learning_rate': 0.1985871018518236, 'epoch': 0.08} + 8%|▊ | 43/520 [02:53<29:03, 3.65s/it] 8%|▊ | 44/520 [02:57<28:56, 3.65s/it] {'loss': 1.9762, 'grad_norm': 0.012045029072344372, 'learning_rate': 0.19848077530122082, 'epoch': 0.08} + 8%|▊ | 44/520 [02:57<28:56, 3.65s/it] 9%|▊ | 45/520 [03:00<28:51, 3.65s/it] {'loss': 1.6406, 'grad_norm': 0.006204597324200579, 'learning_rate': 0.19837062236509015, 'epoch': 0.09} + 9%|▊ | 45/520 [03:00<28:51, 3.65s/it] 9%|▉ | 46/520 [03:04<28:49, 3.65s/it] {'loss': 2.0215, 'grad_norm': 0.008821802068666267, 'learning_rate': 0.19825664732332884, 'epoch': 0.09} + 9%|▉ | 46/520 [03:04<28:49, 3.65s/it] 9%|▉ | 47/520 [03:08<28:50, 3.66s/it] {'loss': 1.6122, 'grad_norm': 0.0053976162590375345, 'learning_rate': 0.19813885460433878, 'epoch': 0.09} + 9%|▉ | 47/520 [03:08<28:50, 3.66s/it] 9%|▉ | 48/520 [03:11<28:40, 3.65s/it] {'loss': 1.6109, 'grad_norm': 0.0049990368938004235, 'learning_rate': 0.19801724878485438, 'epoch': 0.09} + 9%|▉ | 48/520 [03:11<28:40, 3.65s/it] 9%|▉ | 49/520 [03:15<28:39, 3.65s/it] {'loss': 1.6029, 'grad_norm': 0.004524038741682533, 'learning_rate': 0.19789183458976486, 'epoch': 0.09} + 9%|▉ | 49/520 [03:15<28:39, 3.65s/it] 10%|▉ | 50/520 [03:19<28:36, 3.65s/it] {'loss': 1.5912, 'grad_norm': 0.00438996028438541, 'learning_rate': 0.19776261689193048, 'epoch': 0.1} + 10%|▉ | 50/520 [03:19<28:36, 3.65s/it] 10%|▉ | 51/520 [03:22<28:29, 3.65s/it] {'loss': 1.5071, 'grad_norm': 0.004613422977564665, 'learning_rate': 0.19762960071199334, 'epoch': 0.1} + 10%|▉ | 51/520 [03:22<28:29, 3.65s/it] 10%|█ | 52/520 [03:26<28:25, 3.64s/it] {'loss': 1.6475, 'grad_norm': 0.005253301174817517, 'learning_rate': 0.19749279121818236, 'epoch': 0.1} + 10%|█ | 52/520 [03:26<28:25, 3.64s/it] 10%|█ | 53/520 [03:30<28:24, 3.65s/it] {'loss': 1.6654, 'grad_norm': 0.0051478837848790435, 'learning_rate': 0.19735219372611235, 'epoch': 0.1} + 10%|█ | 53/520 [03:30<28:24, 3.65s/it] 10%|█ | 54/520 [03:33<28:22, 3.65s/it] {'loss': 1.5186, 'grad_norm': 0.004475876633793394, 'learning_rate': 0.19720781369857746, 'epoch': 0.1} + 10%|█ | 54/520 [03:33<28:22, 3.65s/it] 11%|█ | 55/520 [03:37<28:16, 3.65s/it] {'loss': 1.5191, 'grad_norm': 0.004947668221928469, 'learning_rate': 0.1970596567453391, 'epoch': 0.11} + 11%|█ | 55/520 [03:37<28:16, 3.65s/it] 11%|█ | 56/520 [03:41<28:10, 3.64s/it] {'loss': 1.6605, 'grad_norm': 0.0046701617621382014, 'learning_rate': 0.1969077286229078, 'epoch': 0.11} + 11%|█ | 56/520 [03:41<28:10, 3.64s/it] 11%|█ | 57/520 [03:44<28:02, 3.63s/it] {'loss': 1.5089, 'grad_norm': 0.0059331007279949475, 'learning_rate': 0.19675203523431964, 'epoch': 0.11} + 11%|█ | 57/520 [03:44<28:02, 3.63s/it] 11%|█ | 58/520 [03:48<28:00, 3.64s/it] {'loss': 1.6705, 'grad_norm': 0.004492459539083241, 'learning_rate': 0.19659258262890683, 'epoch': 0.11} + 11%|█ | 58/520 [03:48<28:00, 3.64s/it] 11%|█▏ | 59/520 [03:52<28:02, 3.65s/it] {'loss': 1.7013, 'grad_norm': 0.009685570130509307, 'learning_rate': 0.19642937700206278, 'epoch': 0.11} + 11%|█▏ | 59/520 [03:52<28:02, 3.65s/it] 12%|█▏ | 60/520 [03:55<28:20, 3.70s/it] {'loss': 1.5875, 'grad_norm': 0.007174055301322121, 'learning_rate': 0.19626242469500121, 'epoch': 0.12} + 12%|█▏ | 60/520 [03:55<28:20, 3.70s/it] 12%|█▏ | 61/520 [03:59<28:37, 3.74s/it] {'loss': 1.8221, 'grad_norm': 0.006769036662289895, 'learning_rate': 0.19609173219450998, 'epoch': 0.12} + 12%|█▏ | 61/520 [03:59<28:37, 3.74s/it] 12%|█▏ | 62/520 [04:03<28:49, 3.78s/it] {'loss': 1.5525, 'grad_norm': 0.004957684567873684, 'learning_rate': 0.19591730613269878, 'epoch': 0.12} + 12%|█▏ | 62/520 [04:03<28:49, 3.78s/it] 12%|█▏ | 63/520 [04:07<28:53, 3.79s/it] {'loss': 1.5567, 'grad_norm': 0.006270246744767329, 'learning_rate': 0.19573915328674182, 'epoch': 0.12} + 12%|█▏ | 63/520 [04:07<28:53, 3.79s/it] 12%|█▏ | 64/520 [04:11<28:56, 3.81s/it] {'loss': 1.5782, 'grad_norm': 0.004587645820456628, 'learning_rate': 0.1955572805786141, 'epoch': 0.12} + 12%|█▏ | 64/520 [04:11<28:56, 3.81s/it] 12%|█▎ | 65/520 [04:15<28:54, 3.81s/it] {'loss': 1.5851, 'grad_norm': 0.004792081224946108, 'learning_rate': 0.1953716950748227, 'epoch': 0.12} + 12%|█▎ | 65/520 [04:15<28:54, 3.81s/it] 13%|█▎ | 66/520 [04:18<28:59, 3.83s/it] {'loss': 1.5478, 'grad_norm': 0.006942611804324067, 'learning_rate': 0.19518240398613226, 'epoch': 0.13} + 13%|█▎ | 66/520 [04:18<28:59, 3.83s/it] 13%|█▎ | 67/520 [04:22<28:54, 3.83s/it] {'loss': 1.3998, 'grad_norm': 0.004717295641019544, 'learning_rate': 0.1949894146672846, 'epoch': 0.13} + 13%|█▎ | 67/520 [04:22<28:54, 3.83s/it] 13%|█▎ | 68/520 [04:26<28:51, 3.83s/it] {'loss': 1.4547, 'grad_norm': 0.0038678072607630995, 'learning_rate': 0.1947927346167132, 'epoch': 0.13} + 13%|█▎ | 68/520 [04:26<28:51, 3.83s/it] 13%|█▎ | 69/520 [04:30<28:42, 3.82s/it] {'loss': 1.4478, 'grad_norm': 0.007650744265677009, 'learning_rate': 0.1945923714762516, 'epoch': 0.13} + 13%|█▎ | 69/520 [04:30<28:42, 3.82s/it] 13%|█▎ | 70/520 [04:34<28:24, 3.79s/it] {'loss': 1.512, 'grad_norm': 0.0057582705883452295, 'learning_rate': 0.19438833303083677, 'epoch': 0.13} + 13%|█▎ | 70/520 [04:34<28:24, 3.79s/it] 14%|█▎ | 71/520 [04:37<28:05, 3.75s/it] {'loss': 1.4165, 'grad_norm': 0.004629528301869666, 'learning_rate': 0.19418062720820636, 'epoch': 0.14} + 14%|█▎ | 71/520 [04:37<28:05, 3.75s/it] 14%|█▍ | 72/520 [04:41<27:42, 3.71s/it] {'loss': 1.5765, 'grad_norm': 0.00612410464075126, 'learning_rate': 0.19396926207859086, 'epoch': 0.14} + 14%|█▍ | 72/520 [04:41<27:42, 3.71s/it] 14%|█▍ | 73/520 [04:45<27:44, 3.72s/it] {'loss': 1.3988, 'grad_norm': 0.004487485598990654, 'learning_rate': 0.19375424585439993, 'epoch': 0.14} + 14%|█▍ | 73/520 [04:45<27:44, 3.72s/it] 14%|█▍ | 74/520 [04:48<27:52, 3.75s/it] {'loss': 1.5304, 'grad_norm': 0.004683762916396288, 'learning_rate': 0.1935355868899034, 'epoch': 0.14} + 14%|█▍ | 74/520 [04:48<27:52, 3.75s/it] 14%|█▍ | 75/520 [04:52<27:57, 3.77s/it] {'loss': 1.4111, 'grad_norm': 0.004575553098375336, 'learning_rate': 0.19331329368090666, 'epoch': 0.14} + 14%|█▍ | 75/520 [04:52<27:57, 3.77s/it] 15%|█▍ | 76/520 [04:56<27:57, 3.78s/it] {'loss': 1.8666, 'grad_norm': 0.021482530606007907, 'learning_rate': 0.19308737486442043, 'epoch': 0.15} + 15%|█▍ | 76/520 [04:56<27:57, 3.78s/it] 15%|█▍ | 77/520 [05:00<27:54, 3.78s/it] {'loss': 1.3461, 'grad_norm': 0.004382219033988883, 'learning_rate': 0.19285783921832536, 'epoch': 0.15} + 15%|█▍ | 77/520 [05:00<27:54, 3.78s/it] 15%|█▌ | 78/520 [05:04<27:53, 3.79s/it] {'loss': 1.4849, 'grad_norm': 0.004603079250733303, 'learning_rate': 0.19262469566103088, 'epoch': 0.15} + 15%|█▌ | 78/520 [05:04<27:53, 3.79s/it] 15%|█▌ | 79/520 [05:07<27:40, 3.77s/it] {'loss': 1.4712, 'grad_norm': 0.004203046065209704, 'learning_rate': 0.19238795325112867, 'epoch': 0.15} + 15%|█▌ | 79/520 [05:07<27:40, 3.77s/it] 15%|█▌ | 80/520 [05:11<27:20, 3.73s/it] {'loss': 2.0665, 'grad_norm': 0.01166918915760408, 'learning_rate': 0.19214762118704076, 'epoch': 0.15} + 15%|█▌ | 80/520 [05:11<27:20, 3.73s/it] 16%|█▌ | 81/520 [05:15<27:03, 3.70s/it] {'loss': 1.6438, 'grad_norm': 0.006285911425354084, 'learning_rate': 0.19190370880666208, 'epoch': 0.16} + 16%|█▌ | 81/520 [05:15<27:03, 3.70s/it] 16%|█▌ | 82/520 [05:18<26:48, 3.67s/it] {'loss': 1.5413, 'grad_norm': 0.005162531016312118, 'learning_rate': 0.19165622558699763, 'epoch': 0.16} + 16%|█▌ | 82/520 [05:18<26:48, 3.67s/it] 16%|█▌ | 83/520 [05:22<26:40, 3.66s/it] {'loss': 1.5891, 'grad_norm': 0.005142032064884331, 'learning_rate': 0.19140518114379435, 'epoch': 0.16} + 16%|█▌ | 83/520 [05:22<26:40, 3.66s/it] 16%|█▌ | 84/520 [05:25<26:32, 3.65s/it] {'loss': 1.5635, 'grad_norm': 0.005808456942284469, 'learning_rate': 0.19115058523116735, 'epoch': 0.16} + 16%|█▌ | 84/520 [05:25<26:32, 3.65s/it] 16%|█▋ | 85/520 [05:29<26:24, 3.64s/it] {'loss': 1.576, 'grad_norm': 0.004279767827102761, 'learning_rate': 0.1908924477412211, 'epoch': 0.16} + 16%|█▋ | 85/520 [05:29<26:24, 3.64s/it] 17%|█▋ | 86/520 [05:33<26:18, 3.64s/it] {'loss': 1.6231, 'grad_norm': 0.005490877243914987, 'learning_rate': 0.19063077870366502, 'epoch': 0.17} + 17%|█▋ | 86/520 [05:33<26:18, 3.64s/it] 17%|█▋ | 87/520 [05:36<26:14, 3.64s/it] {'loss': 1.9017, 'grad_norm': 0.009686747781818397, 'learning_rate': 0.1903655882854237, 'epoch': 0.17} + 17%|█▋ | 87/520 [05:36<26:14, 3.64s/it] 17%|█▋ | 88/520 [05:40<26:17, 3.65s/it] {'loss': 2.0024, 'grad_norm': 0.00921672303489649, 'learning_rate': 0.19009688679024192, 'epoch': 0.17} + 17%|█▋ | 88/520 [05:40<26:17, 3.65s/it] 17%|█▋ | 89/520 [05:44<26:09, 3.64s/it] {'loss': 1.5441, 'grad_norm': 0.0049461959791441655, 'learning_rate': 0.18982468465828442, 'epoch': 0.17} + 17%|█▋ | 89/520 [05:44<26:09, 3.64s/it] 17%|█▋ | 90/520 [05:47<26:04, 3.64s/it] {'loss': 1.4652, 'grad_norm': 0.0043827439552198465, 'learning_rate': 0.1895489924657301, 'epoch': 0.17} + 17%|█▋ | 90/520 [05:47<26:04, 3.64s/it] 18%|█▊ | 91/520 [05:51<26:03, 3.65s/it] {'loss': 1.5567, 'grad_norm': 0.004329541977456745, 'learning_rate': 0.18926982092436118, 'epoch': 0.17} + 18%|█▊ | 91/520 [05:51<26:03, 3.65s/it] 18%|█▊ | 92/520 [05:55<25:57, 3.64s/it] {'loss': 1.484, 'grad_norm': 0.0042871477602864135, 'learning_rate': 0.18898718088114688, 'epoch': 0.18} + 18%|█▊ | 92/520 [05:55<25:57, 3.64s/it] 18%|█▊ | 93/520 [05:58<25:54, 3.64s/it] {'loss': 1.4812, 'grad_norm': 0.0042383874264007425, 'learning_rate': 0.18870108331782218, 'epoch': 0.18} + 18%|█▊ | 93/520 [05:58<25:54, 3.64s/it] 18%|█▊ | 94/520 [06:02<25:48, 3.64s/it] {'loss': 1.5941, 'grad_norm': 0.00496889591591579, 'learning_rate': 0.18841153935046098, 'epoch': 0.18} + 18%|█▊ | 94/520 [06:02<25:48, 3.64s/it] 18%|█▊ | 95/520 [06:05<25:41, 3.63s/it] {'loss': 1.4675, 'grad_norm': 0.005087399737832676, 'learning_rate': 0.18811856022904425, 'epoch': 0.18} + 18%|█▊ | 95/520 [06:05<25:41, 3.63s/it] 18%|█▊ | 96/520 [06:09<25:37, 3.63s/it] {'loss': 1.469, 'grad_norm': 0.004242239984471143, 'learning_rate': 0.18782215733702287, 'epoch': 0.18} + 18%|█▊ | 96/520 [06:09<25:37, 3.63s/it] 19%|█▊ | 97/520 [06:13<25:35, 3.63s/it] {'loss': 1.4413, 'grad_norm': 0.004936166881358321, 'learning_rate': 0.18752234219087538, 'epoch': 0.19} + 19%|█▊ | 97/520 [06:13<25:35, 3.63s/it] 19%|█▉ | 98/520 [06:16<25:28, 3.62s/it] {'loss': 1.4324, 'grad_norm': 0.003302759908709003, 'learning_rate': 0.18721912643966054, 'epoch': 0.19} + 19%|█▉ | 98/520 [06:16<25:28, 3.62s/it] 19%|█▉ | 99/520 [06:20<25:25, 3.62s/it] {'loss': 1.4713, 'grad_norm': 0.00425306679025449, 'learning_rate': 0.18691252186456464, 'epoch': 0.19} + 19%|█▉ | 99/520 [06:20<25:25, 3.62s/it] 19%|█▉ | 100/520 [06:24<25:19, 3.62s/it] {'loss': 1.6766, 'grad_norm': 0.006822641872584159, 'learning_rate': 0.1866025403784439, 'epoch': 0.19} + 19%|█▉ | 100/520 [06:24<25:19, 3.62s/it] 19%|█▉ | 101/520 [06:27<25:14, 3.61s/it] {'loss': 1.4428, 'grad_norm': 0.003724560634985825, 'learning_rate': 0.18628919402536132, 'epoch': 0.19} + 19%|█▉ | 101/520 [06:27<25:14, 3.61s/it] 20%|█▉ | 102/520 [06:31<25:10, 3.61s/it] {'loss': 1.4593, 'grad_norm': 0.0039898631878613035, 'learning_rate': 0.18597249498011903, 'epoch': 0.2} + 20%|█▉ | 102/520 [06:31<25:10, 3.61s/it] 20%|█▉ | 103/520 [06:34<25:05, 3.61s/it] {'loss': 1.3806, 'grad_norm': 0.0034652200471099034, 'learning_rate': 0.18565245554778517, 'epoch': 0.2} + 20%|█▉ | 103/520 [06:34<25:05, 3.61s/it] 20%|██ | 104/520 [06:38<25:03, 3.61s/it] {'loss': 1.479, 'grad_norm': 0.004825080011545796, 'learning_rate': 0.18532908816321558, 'epoch': 0.2} + 20%|██ | 104/520 [06:38<25:03, 3.61s/it] 20%|██ | 105/520 [06:42<25:05, 3.63s/it] {'loss': 1.4642, 'grad_norm': 0.0037767073822964647, 'learning_rate': 0.18500240539057092, 'epoch': 0.2} + 20%|██ | 105/520 [06:42<25:05, 3.63s/it] 20%|██ | 106/520 [06:45<24:59, 3.62s/it] {'loss': 1.6349, 'grad_norm': 0.004502954292529433, 'learning_rate': 0.18467241992282843, 'epoch': 0.2} + 20%|██ | 106/520 [06:45<24:59, 3.62s/it] 21%|██ | 107/520 [06:49<24:57, 3.62s/it] {'loss': 1.6385, 'grad_norm': 0.004508400267813085, 'learning_rate': 0.18433914458128858, 'epoch': 0.21} + 21%|██ | 107/520 [06:49<24:57, 3.62s/it] 21%|██ | 108/520 [06:53<24:56, 3.63s/it] {'loss': 1.4272, 'grad_norm': 0.004833823514675, 'learning_rate': 0.18400259231507718, 'epoch': 0.21} + 21%|██ | 108/520 [06:53<24:56, 3.63s/it] 21%|██ | 109/520 [06:56<24:56, 3.64s/it] {'loss': 1.6088, 'grad_norm': 0.004323120995450582, 'learning_rate': 0.18366277620064198, 'epoch': 0.21} + 21%|██ | 109/520 [06:56<24:56, 3.64s/it] 21%|██ | 110/520 [07:00<24:52, 3.64s/it] {'loss': 1.6038, 'grad_norm': 0.004354547883789243, 'learning_rate': 0.1833197094412449, 'epoch': 0.21} + 21%|██ | 110/520 [07:00<24:52, 3.64s/it] 21%|██▏ | 111/520 [07:03<24:47, 3.64s/it] {'loss': 1.6167, 'grad_norm': 0.004573494679117319, 'learning_rate': 0.18297340536644877, 'epoch': 0.21} + 21%|██▏ | 111/520 [07:03<24:47, 3.64s/it] 22%|██▏ | 112/520 [07:07<24:43, 3.64s/it] {'loss': 1.4792, 'grad_norm': 0.004073544685832165, 'learning_rate': 0.1826238774315995, 'epoch': 0.22} + 22%|██▏ | 112/520 [07:07<24:43, 3.64s/it] 22%|██▏ | 113/520 [07:11<24:37, 3.63s/it] {'loss': 1.3536, 'grad_norm': 0.0035366531803645547, 'learning_rate': 0.18227113921730334, 'epoch': 0.22} + 22%|██▏ | 113/520 [07:11<24:37, 3.63s/it] 22%|██▏ | 114/520 [07:14<24:33, 3.63s/it] {'loss': 1.4681, 'grad_norm': 0.004071892400606206, 'learning_rate': 0.1819152044288992, 'epoch': 0.22} + 22%|██▏ | 114/520 [07:14<24:33, 3.63s/it] 22%|██▏ | 115/520 [07:18<24:47, 3.67s/it] {'loss': 1.5877, 'grad_norm': 0.003964021985990735, 'learning_rate': 0.18155608689592603, 'epoch': 0.22} + 22%|██▏ | 115/520 [07:18<24:47, 3.67s/it] 22%|██▏ | 116/520 [07:22<25:02, 3.72s/it] {'loss': 1.5761, 'grad_norm': 0.003571623182460237, 'learning_rate': 0.18119380057158568, 'epoch': 0.22} + 22%|██▏ | 116/520 [07:22<25:02, 3.72s/it] 22%|██▎ | 117/520 [07:26<25:21, 3.78s/it] {'loss': 1.5685, 'grad_norm': 0.004240875656476483, 'learning_rate': 0.18082835953220056, 'epoch': 0.23} + 22%|██▎ | 117/520 [07:26<25:21, 3.78s/it] 23%|██▎ | 118/520 [07:30<25:24, 3.79s/it] {'loss': 1.4378, 'grad_norm': 0.0033774504456581356, 'learning_rate': 0.18045977797666685, 'epoch': 0.23} + 23%|██▎ | 118/520 [07:30<25:24, 3.79s/it] 23%|██▎ | 119/520 [07:34<25:36, 3.83s/it] {'loss': 1.3767, 'grad_norm': 0.0036057300744187343, 'learning_rate': 0.1800880702259028, 'epoch': 0.23} + 23%|██▎ | 119/520 [07:34<25:36, 3.83s/it] 23%|██▎ | 120/520 [07:37<25:37, 3.84s/it] {'loss': 1.4169, 'grad_norm': 0.004061876607035732, 'learning_rate': 0.17971325072229227, 'epoch': 0.23} + 23%|██▎ | 120/520 [07:38<25:37, 3.84s/it] 23%|██▎ | 121/520 [07:41<25:37, 3.85s/it] {'loss': 1.4723, 'grad_norm': 0.004447485184996896, 'learning_rate': 0.17933533402912352, 'epoch': 0.23} + 23%|██▎ | 121/520 [07:41<25:37, 3.85s/it] 23%|██▎ | 122/520 [07:45<25:31, 3.85s/it] {'loss': 1.3446, 'grad_norm': 0.0034485677995340592, 'learning_rate': 0.17895433483002354, 'epoch': 0.23} + 23%|██▎ | 122/520 [07:45<25:31, 3.85s/it] 24%|██▎ | 123/520 [07:49<25:27, 3.85s/it] {'loss': 1.6674, 'grad_norm': 0.005041702912428502, 'learning_rate': 0.17857026792838737, 'epoch': 0.24} + 24%|██▎ | 123/520 [07:49<25:27, 3.85s/it] 24%|██▍ | 124/520 [07:53<25:20, 3.84s/it] {'loss': 1.4441, 'grad_norm': 0.004151081876927427, 'learning_rate': 0.178183148246803, 'epoch': 0.24} + 24%|██▍ | 124/520 [07:53<25:20, 3.84s/it] 24%|██▍ | 125/520 [07:57<25:12, 3.83s/it] {'loss': 1.4381, 'grad_norm': 0.0038084439703842715, 'learning_rate': 0.1777929908264715, 'epoch': 0.24} + 24%|██▍ | 125/520 [07:57<25:12, 3.83s/it] 24%|██▍ | 126/520 [08:01<26:23, 4.02s/it] {'loss': 1.5451, 'grad_norm': 0.003526196161028334, 'learning_rate': 0.17739981082662276, 'epoch': 0.24} + 24%|██▍ | 126/520 [08:01<26:23, 4.02s/it] 24%|██▍ | 127/520 [08:05<25:53, 3.95s/it] {'loss': 1.4065, 'grad_norm': 0.004588099523508859, 'learning_rate': 0.1770036235239263, 'epoch': 0.24} + 24%|██▍ | 127/520 [08:05<25:53, 3.95s/it] 25%|██▍ | 128/520 [08:09<25:36, 3.92s/it] {'loss': 1.4583, 'grad_norm': 0.0037246546940093534, 'learning_rate': 0.1766044443118978, 'epoch': 0.25} + 25%|██▍ | 128/520 [08:09<25:36, 3.92s/it] 25%|██▍ | 129/520 [08:13<25:19, 3.89s/it] {'loss': 1.3688, 'grad_norm': 0.003158017667364362, 'learning_rate': 0.17620228870030108, 'epoch': 0.25} + 25%|██▍ | 129/520 [08:13<25:19, 3.89s/it] 25%|██▌ | 130/520 [08:16<25:08, 3.87s/it] {'loss': 1.4352, 'grad_norm': 0.0037592721537140768, 'learning_rate': 0.1757971723145453, 'epoch': 0.25} + 25%|██▌ | 130/520 [08:16<25:08, 3.87s/it] 25%|██▌ | 131/520 [08:20<24:58, 3.85s/it] {'loss': 1.5162, 'grad_norm': 0.00451388404001547, 'learning_rate': 0.175389110895078, 'epoch': 0.25} + 25%|██▌ | 131/520 [08:20<24:58, 3.85s/it] 25%|██▌ | 132/520 [08:24<24:48, 3.84s/it] {'loss': 1.4859, 'grad_norm': 0.003797881204487727, 'learning_rate': 0.17497812029677343, 'epoch': 0.25} + 25%|██▌ | 132/520 [08:24<24:48, 3.84s/it] 26%|██▌ | 133/520 [08:28<24:33, 3.81s/it] {'loss': 1.3813, 'grad_norm': 0.0037217442404023546, 'learning_rate': 0.17456421648831655, 'epoch': 0.26} + 26%|██▌ | 133/520 [08:28<24:33, 3.81s/it] 26%|██▌ | 134/520 [08:31<24:12, 3.76s/it] {'loss': 1.4665, 'grad_norm': 0.003897819682741381, 'learning_rate': 0.17414741555158267, 'epoch': 0.26} + 26%|██▌ | 134/520 [08:31<24:12, 3.76s/it] 26%|██▌ | 135/520 [08:35<23:52, 3.72s/it] {'loss': 1.5473, 'grad_norm': 0.0035691419030025226, 'learning_rate': 0.1737277336810124, 'epoch': 0.26} + 26%|██▌ | 135/520 [08:35<23:52, 3.72s/it] 26%|██▌ | 136/520 [08:39<23:38, 3.69s/it] {'loss': 1.4668, 'grad_norm': 0.0038409723361059757, 'learning_rate': 0.17330518718298263, 'epoch': 0.26} + 26%|██▌ | 136/520 [08:39<23:38, 3.69s/it] 26%|██▋ | 137/520 [08:42<23:29, 3.68s/it] {'loss': 1.3874, 'grad_norm': 0.004363549516970194, 'learning_rate': 0.17287979247517285, 'epoch': 0.26} + 26%|██▋ | 137/520 [08:42<23:29, 3.68s/it] 27%|██▋ | 138/520 [08:46<23:20, 3.67s/it] {'loss': 1.3875, 'grad_norm': 0.0034644400640070935, 'learning_rate': 0.17245156608592727, 'epoch': 0.27} + 27%|██▋ | 138/520 [08:46<23:20, 3.67s/it] 27%|██▋ | 139/520 [08:50<23:14, 3.66s/it] {'loss': 1.4043, 'grad_norm': 0.003810263083273613, 'learning_rate': 0.17202052465361267, 'epoch': 0.27} + 27%|██▋ | 139/520 [08:50<23:14, 3.66s/it] 27%|██▋ | 140/520 [08:53<23:06, 3.65s/it] {'loss': 1.581, 'grad_norm': 0.003898438066212237, 'learning_rate': 0.17158668492597184, 'epoch': 0.27} + 27%|██▋ | 140/520 [08:53<23:06, 3.65s/it] 27%|██▋ | 141/520 [08:57<23:00, 3.64s/it] {'loss': 1.5159, 'grad_norm': 0.003743976623699979, 'learning_rate': 0.17115006375947303, 'epoch': 0.27} + 27%|██▋ | 141/520 [08:57<23:00, 3.64s/it] 27%|██▋ | 142/520 [09:01<22:56, 3.64s/it] {'loss': 1.6173, 'grad_norm': 0.004044573161394752, 'learning_rate': 0.17071067811865476, 'epoch': 0.27} + 27%|██▋ | 142/520 [09:01<22:56, 3.64s/it] 28%|██▊ | 143/520 [09:04<22:51, 3.64s/it] {'loss': 1.4344, 'grad_norm': 0.003924578828421638, 'learning_rate': 0.17026854507546693, 'epoch': 0.28} + 28%|██▊ | 143/520 [09:04<22:51, 3.64s/it] 28%|██▊ | 144/520 [09:08<22:46, 3.63s/it] {'loss': 1.3837, 'grad_norm': 0.0037226493447790323, 'learning_rate': 0.1698236818086073, 'epoch': 0.28} + 28%|██▊ | 144/520 [09:08<22:46, 3.63s/it] 28%|██▊ | 145/520 [09:11<22:46, 3.64s/it] {'loss': 1.3099, 'grad_norm': 0.00322951672734, 'learning_rate': 0.16937610560285418, 'epoch': 0.28} + 28%|██▊ | 145/520 [09:11<22:46, 3.64s/it] 28%|██▊ | 146/520 [09:15<22:46, 3.65s/it] {'loss': 1.6301, 'grad_norm': 0.003704520735283498, 'learning_rate': 0.1689258338483947, 'epoch': 0.28} + 28%|██▊ | 146/520 [09:15<22:46, 3.65s/it] 28%|██▊ | 147/520 [09:19<22:49, 3.67s/it] {'loss': 1.3555, 'grad_norm': 0.0037528657858750085, 'learning_rate': 0.16847288404014937, 'epoch': 0.28} + 28%|██▊ | 147/520 [09:19<22:49, 3.67s/it] 28%|██▊ | 148/520 [09:22<22:46, 3.67s/it] {'loss': 1.3919, 'grad_norm': 0.0034202421637208385, 'learning_rate': 0.16801727377709194, 'epoch': 0.28} + 28%|██▊ | 148/520 [09:22<22:46, 3.67s/it] 29%|██▊ | 149/520 [09:26<22:39, 3.66s/it] {'loss': 1.3421, 'grad_norm': 0.004014723829041871, 'learning_rate': 0.16755902076156604, 'epoch': 0.29} + 29%|██▊ | 149/520 [09:26<22:39, 3.66s/it] 29%|██▉ | 150/520 [09:30<22:37, 3.67s/it] {'loss': 1.5718, 'grad_norm': 0.0043204353954563735, 'learning_rate': 0.16709814279859703, 'epoch': 0.29} + 29%|██▉ | 150/520 [09:30<22:37, 3.67s/it] 29%|██▉ | 151/520 [09:33<22:34, 3.67s/it] {'loss': 1.3749, 'grad_norm': 0.0034795279514702212, 'learning_rate': 0.1666346577952004, 'epoch': 0.29} + 29%|██▉ | 151/520 [09:33<22:34, 3.67s/it] 29%|██▉ | 152/520 [09:37<22:30, 3.67s/it] {'loss': 1.3414, 'grad_norm': 0.003759178727859615, 'learning_rate': 0.16616858375968596, 'epoch': 0.29} + 29%|██▉ | 152/520 [09:37<22:30, 3.67s/it] 29%|██▉ | 153/520 [09:41<22:25, 3.67s/it] {'loss': 1.3915, 'grad_norm': 0.004270174174441347, 'learning_rate': 0.16569993880095807, 'epoch': 0.29} + 29%|██▉ | 153/520 [09:41<22:25, 3.67s/it] 30%|██▉ | 154/520 [09:44<22:20, 3.66s/it] {'loss': 1.4689, 'grad_norm': 0.003450690569881415, 'learning_rate': 0.16522874112781213, 'epoch': 0.3} + 30%|██▉ | 154/520 [09:44<22:20, 3.66s/it] 30%|██▉ | 155/520 [09:48<22:33, 3.71s/it] {'loss': 1.3677, 'grad_norm': 0.003805563292856129, 'learning_rate': 0.16475500904822704, 'epoch': 0.3} + 30%|██▉ | 155/520 [09:48<22:33, 3.71s/it] 30%|███ | 156/520 [09:52<22:41, 3.74s/it] {'loss': 1.4165, 'grad_norm': 0.0042824572112847336, 'learning_rate': 0.16427876096865393, 'epoch': 0.3} + 30%|███ | 156/520 [09:52<22:41, 3.74s/it] 30%|███ | 157/520 [09:56<22:44, 3.76s/it] {'loss': 1.6521, 'grad_norm': 0.004549272395591708, 'learning_rate': 0.16380001539330089, 'epoch': 0.3} + 30%|███ | 157/520 [09:56<22:44, 3.76s/it] 30%|███ | 158/520 [10:00<22:43, 3.77s/it] {'loss': 1.3874, 'grad_norm': 0.004323814210243875, 'learning_rate': 0.163318790923414, 'epoch': 0.3} + 30%|███ | 158/520 [10:00<22:43, 3.77s/it] 31%|███ | 159/520 [10:03<22:42, 3.77s/it] {'loss': 1.4145, 'grad_norm': 0.003354040174234224, 'learning_rate': 0.16283510625655473, 'epoch': 0.31} + 31%|███ | 159/520 [10:03<22:42, 3.77s/it] 31%|███ | 160/520 [10:07<22:45, 3.79s/it] {'loss': 1.4509, 'grad_norm': 0.0038435278738981255, 'learning_rate': 0.16234898018587338, 'epoch': 0.31} + 31%|███ | 160/520 [10:07<22:45, 3.79s/it] 31%|███ | 161/520 [10:11<22:45, 3.80s/it] {'loss': 1.4254, 'grad_norm': 0.0037504054448192683, 'learning_rate': 0.16186043159937882, 'epoch': 0.31} + 31%|███ | 161/520 [10:11<22:45, 3.80s/it] 31%|███ | 162/520 [10:15<22:41, 3.80s/it] {'loss': 1.5552, 'grad_norm': 0.0036468628513667972, 'learning_rate': 0.16136947947920477, 'epoch': 0.31} + 31%|███ | 162/520 [10:15<22:41, 3.80s/it] 31%|███▏ | 163/520 [10:19<22:24, 3.77s/it] {'loss': 1.2948, 'grad_norm': 0.004258704848314915, 'learning_rate': 0.16087614290087207, 'epoch': 0.31} + 31%|███▏ | 163/520 [10:19<22:24, 3.77s/it] 32%|███▏ | 164/520 [10:22<22:16, 3.75s/it] {'loss': 1.261, 'grad_norm': 0.0033158108893890518, 'learning_rate': 0.16038044103254775, 'epoch': 0.32} + 32%|███▏ | 164/520 [10:22<22:16, 3.75s/it] 32%|███▏ | 165/520 [10:26<22:07, 3.74s/it] {'loss': 1.4153, 'grad_norm': 0.003514275043592999, 'learning_rate': 0.15988239313430005, 'epoch': 0.32} + 32%|███▏ | 165/520 [10:26<22:07, 3.74s/it] 32%|███▏ | 166/520 [10:30<21:54, 3.71s/it] {'loss': 1.4043, 'grad_norm': 0.0040814555107816595, 'learning_rate': 0.15938201855735015, 'epoch': 0.32} + 32%|███▏ | 166/520 [10:30<21:54, 3.71s/it] 32%|███▏ | 167/520 [10:33<21:46, 3.70s/it] {'loss': 1.3936, 'grad_norm': 0.004044332867877829, 'learning_rate': 0.15887933674332047, 'epoch': 0.32} + 32%|███▏ | 167/520 [10:33<21:46, 3.70s/it] 32%|███▏ | 168/520 [10:37<21:40, 3.69s/it] {'loss': 1.3322, 'grad_norm': 0.003618905322294904, 'learning_rate': 0.158374367223479, 'epoch': 0.32} + 32%|███▏ | 168/520 [10:37<21:40, 3.69s/it] 32%|███▎ | 169/520 [10:41<21:33, 3.68s/it] {'loss': 1.4187, 'grad_norm': 0.0036349638890471637, 'learning_rate': 0.1578671296179806, 'epoch': 0.33} + 32%|███▎ | 169/520 [10:41<21:33, 3.68s/it] 33%|███▎ | 170/520 [10:44<21:32, 3.69s/it] {'loss': 1.4587, 'grad_norm': 0.004033013319098783, 'learning_rate': 0.15735764363510463, 'epoch': 0.33} + 33%|███▎ | 170/520 [10:44<21:32, 3.69s/it] 33%|███▎ | 171/520 [10:48<21:27, 3.69s/it] {'loss': 1.3363, 'grad_norm': 0.003821147320441752, 'learning_rate': 0.15684592907048925, 'epoch': 0.33} + 33%|███▎ | 171/520 [10:48<21:27, 3.69s/it] 33%|███▎ | 172/520 [10:52<21:19, 3.68s/it] {'loss': 1.4102, 'grad_norm': 0.0034244646013515106, 'learning_rate': 0.1563320058063622, 'epoch': 0.33} + 33%|███▎ | 172/520 [10:52<21:19, 3.68s/it] 33%|███▎ | 173/520 [10:55<21:15, 3.67s/it] {'loss': 1.3407, 'grad_norm': 0.003714104279998068, 'learning_rate': 0.15581589381076844, 'epoch': 0.33} + 33%|███▎ | 173/520 [10:55<21:15, 3.67s/it] 33%|███▎ | 174/520 [10:59<21:15, 3.69s/it] {'loss': 1.4276, 'grad_norm': 0.004579953414247004, 'learning_rate': 0.15529761313679394, 'epoch': 0.33} + 33%|███▎ | 174/520 [10:59<21:15, 3.69s/it] 34%|███▎ | 175/520 [11:03<21:13, 3.69s/it] {'loss': 1.3292, 'grad_norm': 0.003615760530846836, 'learning_rate': 0.15477718392178716, 'epoch': 0.34} + 34%|███▎ | 175/520 [11:03<21:13, 3.69s/it] 34%|███▍ | 176/520 [11:07<21:07, 3.68s/it] {'loss': 1.5582, 'grad_norm': 0.003797165089003795, 'learning_rate': 0.15425462638657594, 'epoch': 0.34} + 34%|███▍ | 176/520 [11:07<21:07, 3.68s/it] 34%|███▍ | 177/520 [11:10<21:14, 3.72s/it] {'loss': 1.413, 'grad_norm': 0.003948214620022251, 'learning_rate': 0.1537299608346824, 'epoch': 0.34} + 34%|███▍ | 177/520 [11:10<21:14, 3.72s/it] 34%|███▍ | 178/520 [11:14<21:35, 3.79s/it] {'loss': 1.3921, 'grad_norm': 0.0038067635048640016, 'learning_rate': 0.15320320765153367, 'epoch': 0.34} + 34%|███▍ | 178/520 [11:14<21:35, 3.79s/it] 34%|███▍ | 179/520 [11:18<21:46, 3.83s/it] {'loss': 1.467, 'grad_norm': 0.003347249987115756, 'learning_rate': 0.15267438730367008, 'epoch': 0.34} + 34%|███▍ | 179/520 [11:18<21:46, 3.83s/it] 35%|███▍ | 180/520 [11:22<21:54, 3.87s/it] {'loss': 1.3893, 'grad_norm': 0.003867051939532117, 'learning_rate': 0.1521435203379498, 'epoch': 0.35} + 35%|███▍ | 180/520 [11:22<21:54, 3.87s/it] 35%|███▍ | 181/520 [11:26<21:54, 3.88s/it] {'loss': 1.3503, 'grad_norm': 0.0031364239757447837, 'learning_rate': 0.15161062738075068, 'epoch': 0.35} + 35%|███▍ | 181/520 [11:26<21:54, 3.88s/it] 35%|███▌ | 182/520 [11:30<21:40, 3.85s/it] {'loss': 1.3688, 'grad_norm': 0.003499369407504016, 'learning_rate': 0.1510757291371686, 'epoch': 0.35} + 35%|███▌ | 182/520 [11:30<21:40, 3.85s/it] 35%|███▌ | 183/520 [11:34<21:24, 3.81s/it] {'loss': 1.3997, 'grad_norm': 0.0036371258730058557, 'learning_rate': 0.1505388463902131, 'epoch': 0.35} + 35%|███▌ | 183/520 [11:34<21:24, 3.81s/it] 35%|███▌ | 184/520 [11:37<21:15, 3.80s/it] {'loss': 1.3068, 'grad_norm': 0.0035530080491531562, 'learning_rate': 0.15000000000000002, 'epoch': 0.35} + 35%|███▌ | 184/520 [11:37<21:15, 3.80s/it] 36%|███▌ | 185/520 [11:41<21:05, 3.78s/it] {'loss': 1.4909, 'grad_norm': 0.0035518482590040806, 'learning_rate': 0.14945921090294076, 'epoch': 0.36} + 36%|███▌ | 185/520 [11:41<21:05, 3.78s/it] 36%|███▌ | 186/520 [11:45<20:50, 3.74s/it] {'loss': 1.3251, 'grad_norm': 0.00332777643581187, 'learning_rate': 0.14891650011092894, 'epoch': 0.36} + 36%|███▌ | 186/520 [11:45<20:50, 3.74s/it] 36%|███▌ | 187/520 [11:49<20:59, 3.78s/it] {'loss': 1.344, 'grad_norm': 0.004283124317254829, 'learning_rate': 0.14837188871052398, 'epoch': 0.36} + 36%|███▌ | 187/520 [11:49<20:59, 3.78s/it] 36%|███▌ | 188/520 [11:53<21:08, 3.82s/it] {'loss': 1.4135, 'grad_norm': 0.0034477138226933165, 'learning_rate': 0.14782539786213184, 'epoch': 0.36} + 36%|███▌ | 188/520 [11:53<21:08, 3.82s/it] 36%|███▋ | 189/520 [11:56<21:13, 3.85s/it] {'loss': 1.4317, 'grad_norm': 0.0031634159392163468, 'learning_rate': 0.1472770487991827, 'epoch': 0.36} + 36%|███▋ | 189/520 [11:56<21:13, 3.85s/it] 37%|███▋ | 190/520 [12:00<21:20, 3.88s/it] {'loss': 1.3466, 'grad_norm': 0.003704500199372305, 'learning_rate': 0.1467268628273062, 'epoch': 0.37} + 37%|███▋ | 190/520 [12:00<21:20, 3.88s/it] 37%|███▋ | 191/520 [12:04<21:22, 3.90s/it] {'loss': 1.2985, 'grad_norm': 0.0033907810985945096, 'learning_rate': 0.1461748613235034, 'epoch': 0.37} + 37%|███▋ | 191/520 [12:04<21:22, 3.90s/it] 37%|███▋ | 192/520 [12:08<21:17, 3.89s/it] {'loss': 1.392, 'grad_norm': 0.0034574786959642874, 'learning_rate': 0.1456210657353163, 'epoch': 0.37} + 37%|███▋ | 192/520 [12:08<21:17, 3.89s/it] 37%|███▋ | 193/520 [12:12<20:51, 3.83s/it] {'loss': 1.5088, 'grad_norm': 0.0043608612194922835, 'learning_rate': 0.14506549757999454, 'epoch': 0.37} + 37%|███▋ | 193/520 [12:12<20:51, 3.83s/it] 37%|███▋ | 194/520 [12:16<20:32, 3.78s/it] {'loss': 1.3513, 'grad_norm': 0.003358073992032349, 'learning_rate': 0.14450817844365924, 'epoch': 0.37} + 37%|███▋ | 194/520 [12:16<20:32, 3.78s/it] 38%|███▊ | 195/520 [12:19<20:40, 3.82s/it] {'loss': 1.4146, 'grad_norm': 0.0037063831529636844, 'learning_rate': 0.1439491299804645, 'epoch': 0.38} + 38%|███▊ | 195/520 [12:19<20:40, 3.82s/it] 38%|███▊ | 196/520 [12:23<20:39, 3.83s/it] {'loss': 1.3819, 'grad_norm': 0.003913007318484279, 'learning_rate': 0.14338837391175582, 'epoch': 0.38} + 38%|███▊ | 196/520 [12:23<20:39, 3.83s/it] 38%|███▊ | 197/520 [12:27<20:20, 3.78s/it] {'loss': 1.3369, 'grad_norm': 0.0033445345395324414, 'learning_rate': 0.14282593202522628, 'epoch': 0.38} + 38%|███▊ | 197/520 [12:27<20:20, 3.78s/it] 38%|███▊ | 198/520 [12:31<20:09, 3.75s/it] {'loss': 1.4146, 'grad_norm': 0.0035841307315705377, 'learning_rate': 0.14226182617406996, 'epoch': 0.38} + 38%|███▊ | 198/520 [12:31<20:09, 3.75s/it] 38%|███▊ | 199/520 [12:34<20:02, 3.74s/it] {'loss': 1.3319, 'grad_norm': 0.0034324265730466437, 'learning_rate': 0.14169607827613281, 'epoch': 0.38} + 38%|███▊ | 199/520 [12:34<20:02, 3.74s/it] 38%|███▊ | 200/520 [12:38<19:54, 3.73s/it] {'loss': 1.442, 'grad_norm': 0.00417297539471079, 'learning_rate': 0.14112871031306118, 'epoch': 0.38} + 38%|███▊ | 200/520 [12:38<19:54, 3.73s/it] 39%|███▊ | 201/520 [12:42<19:47, 3.72s/it] {'loss': 1.4282, 'grad_norm': 0.003228172195580287, 'learning_rate': 0.1405597443294475, 'epoch': 0.39} + 39%|███▊ | 201/520 [12:42<19:47, 3.72s/it] 39%|███▉ | 202/520 [12:45<19:40, 3.71s/it] {'loss': 1.3153, 'grad_norm': 0.0033874062064824678, 'learning_rate': 0.13998920243197407, 'epoch': 0.39} + 39%|███▉ | 202/520 [12:45<19:40, 3.71s/it] 39%|███▉ | 203/520 [12:49<19:33, 3.70s/it] {'loss': 1.3776, 'grad_norm': 0.0035179356889349592, 'learning_rate': 0.13941710678855396, 'epoch': 0.39} + 39%|███▉ | 203/520 [12:49<19:33, 3.70s/it] 39%|███▉ | 204/520 [12:53<19:26, 3.69s/it] {'loss': 1.4115, 'grad_norm': 0.003800627777671298, 'learning_rate': 0.13884347962746948, 'epoch': 0.39} + 39%|███▉ | 204/520 [12:53<19:26, 3.69s/it] 39%|███▉ | 205/520 [12:56<19:21, 3.69s/it] {'loss': 1.4459, 'grad_norm': 0.0038546447781140475, 'learning_rate': 0.138268343236509, 'epoch': 0.39} + 39%|███▉ | 205/520 [12:56<19:21, 3.69s/it] 40%|███▉ | 206/520 [13:00<19:28, 3.72s/it] {'loss': 1.4412, 'grad_norm': 0.0037788969857651847, 'learning_rate': 0.13769171996210053, 'epoch': 0.4} + 40%|███▉ | 206/520 [13:00<19:28, 3.72s/it] 40%|███▉ | 207/520 [13:04<19:40, 3.77s/it] {'loss': 1.4066, 'grad_norm': 0.003113608003005422, 'learning_rate': 0.1371136322084438, 'epoch': 0.4} + 40%|███▉ | 207/520 [13:04<19:40, 3.77s/it] 40%|████ | 208/520 [13:08<19:42, 3.79s/it] {'loss': 1.4089, 'grad_norm': 0.003689936623729333, 'learning_rate': 0.13653410243663952, 'epoch': 0.4} + 40%|████ | 208/520 [13:08<19:42, 3.79s/it] 40%|████ | 209/520 [13:12<19:46, 3.81s/it] {'loss': 1.3366, 'grad_norm': 0.0035011803093770067, 'learning_rate': 0.13595315316381676, 'epoch': 0.4} + 40%|████ | 209/520 [13:12<19:46, 3.81s/it] 40%|████ | 210/520 [13:16<19:42, 3.82s/it] {'loss': 1.4173, 'grad_norm': 0.0040775093029619, 'learning_rate': 0.13537080696225814, 'epoch': 0.4} + 40%|████ | 210/520 [13:16<19:42, 3.82s/it] 41%|████ | 211/520 [13:20<19:45, 3.84s/it] {'loss': 1.411, 'grad_norm': 0.003132908595017193, 'learning_rate': 0.13478708645852272, 'epoch': 0.41} + 41%|████ | 211/520 [13:20<19:45, 3.84s/it] 41%|████ | 212/520 [13:23<19:43, 3.84s/it] {'loss': 1.3768, 'grad_norm': 0.0032515967924476932, 'learning_rate': 0.1342020143325669, 'epoch': 0.41} + 41%|████ | 212/520 [13:23<19:43, 3.84s/it] 41%|████ | 213/520 [13:27<19:40, 3.85s/it] {'loss': 1.3504, 'grad_norm': 0.003804274681582023, 'learning_rate': 0.13361561331686309, 'epoch': 0.41} + 41%|████ | 213/520 [13:27<19:40, 3.85s/it] 41%|████ | 214/520 [13:31<19:27, 3.82s/it] {'loss': 1.3426, 'grad_norm': 0.0036817397979113445, 'learning_rate': 0.13302790619551672, 'epoch': 0.41} + 41%|████ | 214/520 [13:31<19:27, 3.82s/it] 41%|████▏ | 215/520 [13:35<19:12, 3.78s/it] {'loss': 1.3393, 'grad_norm': 0.0033632033164437496, 'learning_rate': 0.1324389158033807, 'epoch': 0.41} + 41%|████▏ | 215/520 [13:35<19:12, 3.78s/it] 42%|████▏ | 216/520 [13:38<19:06, 3.77s/it] {'loss': 1.2516, 'grad_norm': 0.0032223825522334436, 'learning_rate': 0.13184866502516845, 'epoch': 0.42} + 42%|████▏ | 216/520 [13:38<19:06, 3.77s/it] 42%|████▏ | 217/520 [13:42<18:57, 3.75s/it] {'loss': 1.3778, 'grad_norm': 0.0034236319844349566, 'learning_rate': 0.13125717679456447, 'epoch': 0.42} + 42%|████▏ | 217/520 [13:42<18:57, 3.75s/it] 42%|████▏ | 218/520 [13:46<19:00, 3.78s/it] {'loss': 1.3726, 'grad_norm': 0.003513059030075508, 'learning_rate': 0.13066447409333345, 'epoch': 0.42} + 42%|████▏ | 218/520 [13:46<19:00, 3.78s/it] 42%|████▏ | 219/520 [13:50<19:03, 3.80s/it] {'loss': 1.3417, 'grad_norm': 0.0030658222573038706, 'learning_rate': 0.1300705799504273, 'epoch': 0.42} + 42%|████▏ | 219/520 [13:50<19:03, 3.80s/it] 42%|████▏ | 220/520 [13:54<19:02, 3.81s/it] {'loss': 1.4038, 'grad_norm': 0.0033818950691668847, 'learning_rate': 0.12947551744109043, 'epoch': 0.42} + 42%|████▏ | 220/520 [13:54<19:02, 3.81s/it] 42%|████▎ | 221/520 [13:58<19:00, 3.81s/it] {'loss': 1.3775, 'grad_norm': 0.0033287347023246497, 'learning_rate': 0.128879309685963, 'epoch': 0.42} + 42%|████▎ | 221/520 [13:58<19:00, 3.81s/it] 43%|████▎ | 222/520 [14:01<18:47, 3.78s/it] {'loss': 1.2799, 'grad_norm': 0.0031580533068553433, 'learning_rate': 0.12828197985018275, 'epoch': 0.43} + 43%|████▎ | 222/520 [14:01<18:47, 3.78s/it] 43%|████▎ | 223/520 [14:05<18:34, 3.75s/it] {'loss': 1.2799, 'grad_norm': 0.0033164481526756365, 'learning_rate': 0.12768355114248495, 'epoch': 0.43} + 43%|████▎ | 223/520 [14:05<18:34, 3.75s/it] 43%|████▎ | 224/520 [14:09<18:39, 3.78s/it] {'loss': 1.6127, 'grad_norm': 0.004993683247318963, 'learning_rate': 0.12708404681430052, 'epoch': 0.43} + 43%|████▎ | 224/520 [14:09<18:39, 3.78s/it] 43%|████▎ | 225/520 [14:13<18:41, 3.80s/it] {'loss': 1.3028, 'grad_norm': 0.003652053718053434, 'learning_rate': 0.1264834901588527, 'epoch': 0.43} + 43%|████▎ | 225/520 [14:13<18:41, 3.80s/it] 43%|████▎ | 226/520 [14:16<18:41, 3.81s/it] {'loss': 1.4074, 'grad_norm': 0.0033826099165458782, 'learning_rate': 0.12588190451025208, 'epoch': 0.43} + 43%|████▎ | 226/520 [14:16<18:41, 3.81s/it] 44%|████▎ | 227/520 [14:20<18:41, 3.83s/it] {'loss': 1.3831, 'grad_norm': 0.0030897170268345854, 'learning_rate': 0.12527931324258976, 'epoch': 0.44} + 44%|████▎ | 227/520 [14:20<18:41, 3.83s/it] 44%|████▍ | 228/520 [14:24<18:40, 3.84s/it] {'loss': 1.5423, 'grad_norm': 0.003692441141891582, 'learning_rate': 0.12467573976902935, 'epoch': 0.44} + 44%|████▍ | 228/520 [14:24<18:40, 3.84s/it] 44%|████▍ | 229/520 [14:28<18:37, 3.84s/it] {'loss': 1.3658, 'grad_norm': 0.003091714992632229, 'learning_rate': 0.12407120754089732, 'epoch': 0.44} + 44%|████▍ | 229/520 [14:28<18:37, 3.84s/it] 44%|████▍ | 230/520 [14:32<18:34, 3.84s/it] {'loss': 1.2341, 'grad_norm': 0.003430525444605673, 'learning_rate': 0.12346574004677154, 'epoch': 0.44} + 44%|████▍ | 230/520 [14:32<18:34, 3.84s/it] 44%|████▍ | 231/520 [14:36<18:29, 3.84s/it] {'loss': 1.2966, 'grad_norm': 0.0029721738695981915, 'learning_rate': 0.12285936081156897, 'epoch': 0.44} + 44%|████▍ | 231/520 [14:36<18:29, 3.84s/it] 45%|████▍ | 232/520 [14:40<18:27, 3.84s/it] {'loss': 1.5773, 'grad_norm': 0.00397550749046059, 'learning_rate': 0.12225209339563144, 'epoch': 0.45} + 45%|████▍ | 232/520 [14:40<18:27, 3.84s/it] 45%|████▍ | 233/520 [14:43<18:22, 3.84s/it] {'loss': 1.439, 'grad_norm': 0.0035159093989312, 'learning_rate': 0.12164396139381028, 'epoch': 0.45} + 45%|████▍ | 233/520 [14:43<18:22, 3.84s/it] 45%|████▌ | 234/520 [14:47<18:17, 3.84s/it] {'loss': 1.2629, 'grad_norm': 0.0036538940835133663, 'learning_rate': 0.12103498843454959, 'epoch': 0.45} + 45%|████▌ | 234/520 [14:47<18:17, 3.84s/it] 45%|████▌ | 235/520 [14:51<18:13, 3.84s/it] {'loss': 1.3081, 'grad_norm': 0.0034856612601161874, 'learning_rate': 0.12042519817896805, 'epoch': 0.45} + 45%|████▌ | 235/520 [14:51<18:13, 3.84s/it] 45%|████▌ | 236/520 [14:55<18:09, 3.84s/it] {'loss': 1.4277, 'grad_norm': 0.0031296368446798828, 'learning_rate': 0.11981461431993977, 'epoch': 0.45} + 45%|████▌ | 236/520 [14:55<18:09, 3.84s/it] 46%|████▌ | 237/520 [14:59<18:04, 3.83s/it] {'loss': 1.3789, 'grad_norm': 0.003124855111061319, 'learning_rate': 0.11920326058117364, 'epoch': 0.46} + 46%|████▌ | 237/520 [14:59<18:04, 3.83s/it] 46%|████▌ | 238/520 [15:03<17:59, 3.83s/it] {'loss': 1.3265, 'grad_norm': 0.0032584256213719154, 'learning_rate': 0.11859116071629149, 'epoch': 0.46} + 46%|████▌ | 238/520 [15:03<17:59, 3.83s/it] 46%|████▌ | 239/520 [15:06<17:55, 3.83s/it] {'loss': 1.4274, 'grad_norm': 0.003385200761969433, 'learning_rate': 0.11797833850790528, 'epoch': 0.46} + 46%|████▌ | 239/520 [15:06<17:55, 3.83s/it] 46%|████▌ | 240/520 [15:10<17:54, 3.84s/it] {'loss': 1.1917, 'grad_norm': 0.0031066016508554954, 'learning_rate': 0.11736481776669305, 'epoch': 0.46} + 46%|████▌ | 240/520 [15:10<17:54, 3.84s/it] 46%|████▋ | 241/520 [15:14<17:50, 3.84s/it] {'loss': 1.2774, 'grad_norm': 0.003425687010174329, 'learning_rate': 0.11675062233047365, 'epoch': 0.46} + 46%|████▋ | 241/520 [15:14<17:50, 3.84s/it] 47%|████▋ | 242/520 [15:18<17:44, 3.83s/it] {'loss': 1.2986, 'grad_norm': 0.003077195331454676, 'learning_rate': 0.11613577606328068, 'epoch': 0.47} + 47%|████▋ | 242/520 [15:18<17:44, 3.83s/it] 47%|████▋ | 243/520 [15:22<17:44, 3.84s/it] {'loss': 1.2845, 'grad_norm': 0.0031985568054581893, 'learning_rate': 0.11552030285443515, 'epoch': 0.47} + 47%|████▋ | 243/520 [15:22<17:44, 3.84s/it] 47%|████▋ | 244/520 [15:26<17:38, 3.84s/it] {'loss': 1.4282, 'grad_norm': 0.0036631045241334406, 'learning_rate': 0.11490422661761744, 'epoch': 0.47} + 47%|████▋ | 244/520 [15:26<17:38, 3.84s/it] 47%|████▋ | 245/520 [15:29<17:34, 3.83s/it] {'loss': 1.2759, 'grad_norm': 0.003348498633858399, 'learning_rate': 0.11428757128993801, 'epoch': 0.47} + 47%|████▋ | 245/520 [15:29<17:34, 3.83s/it] 47%|████▋ | 246/520 [15:33<17:30, 3.83s/it] {'loss': 1.5535, 'grad_norm': 0.0040264320607290354, 'learning_rate': 0.11367036083100734, 'epoch': 0.47} + 47%|████▋ | 246/520 [15:33<17:30, 3.83s/it] 48%|████▊ | 247/520 [15:37<17:25, 3.83s/it] {'loss': 1.4514, 'grad_norm': 0.0034525221869708, 'learning_rate': 0.11305261922200519, 'epoch': 0.47} + 48%|████▊ | 247/520 [15:37<17:25, 3.83s/it] 48%|████▊ | 248/520 [15:41<17:22, 3.83s/it] {'loss': 1.2705, 'grad_norm': 0.003286895118851348, 'learning_rate': 0.11243437046474854, 'epoch': 0.48} + 48%|████▊ | 248/520 [15:41<17:22, 3.83s/it] 48%|████▊ | 249/520 [15:45<17:18, 3.83s/it] {'loss': 1.3956, 'grad_norm': 0.003531798794117999, 'learning_rate': 0.1118156385807593, 'epoch': 0.48} + 48%|████▊ | 249/520 [15:45<17:18, 3.83s/it] 48%|████▊ | 250/520 [15:49<17:15, 3.83s/it] {'loss': 1.332, 'grad_norm': 0.003733282785580016, 'learning_rate': 0.11119644761033079, 'epoch': 0.48} + 48%|████▊ | 250/520 [15:49<17:15, 3.83s/it] 48%|████▊ | 251/520 [15:52<17:09, 3.83s/it] {'loss': 1.3838, 'grad_norm': 0.0030280873352345784, 'learning_rate': 0.1105768216115938, 'epoch': 0.48} + 48%|████▊ | 251/520 [15:52<17:09, 3.83s/it] 48%|████▊ | 252/520 [15:56<17:06, 3.83s/it] {'loss': 1.4146, 'grad_norm': 0.0031792945812745264, 'learning_rate': 0.10995678465958168, 'epoch': 0.48} + 48%|████▊ | 252/520 [15:56<17:06, 3.83s/it] 49%|████▊ | 253/520 [16:00<17:03, 3.83s/it] {'loss': 1.3813, 'grad_norm': 0.0035321239363212856, 'learning_rate': 0.10933636084529506, 'epoch': 0.49} + 49%|████▊ | 253/520 [16:00<17:03, 3.83s/it] 49%|████▉ | 254/520 [16:04<17:00, 3.84s/it] {'loss': 1.2917, 'grad_norm': 0.0029869069285684644, 'learning_rate': 0.10871557427476584, 'epoch': 0.49} + 49%|████▉ | 254/520 [16:04<17:00, 3.84s/it] 49%|████▉ | 255/520 [16:08<16:56, 3.83s/it] {'loss': 1.3134, 'grad_norm': 0.003351661068098975, 'learning_rate': 0.10809444906812034, 'epoch': 0.49} + 49%|████▉ | 255/520 [16:08<16:56, 3.83s/it] 49%|████▉ | 256/520 [16:11<16:44, 3.81s/it] {'loss': 1.3553, 'grad_norm': 0.003441683137191686, 'learning_rate': 0.10747300935864244, 'epoch': 0.49} + 49%|████▉ | 256/520 [16:11<16:44, 3.81s/it] 49%|████▉ | 257/520 [16:15<16:33, 3.78s/it] {'loss': 1.3476, 'grad_norm': 0.003385557725354405, 'learning_rate': 0.10685127929183567, 'epoch': 0.49} + 49%|████▉ | 257/520 [16:15<16:33, 3.78s/it] 50%|████▉ | 258/520 [16:19<16:23, 3.75s/it] {'loss': 1.3662, 'grad_norm': 0.0028295166208093824, 'learning_rate': 0.10622928302448523, 'epoch': 0.5} + 50%|████▉ | 258/520 [16:19<16:23, 3.75s/it] 50%|████▉ | 259/520 [16:23<16:12, 3.73s/it] {'loss': 1.4263, 'grad_norm': 0.0036198338821136103, 'learning_rate': 0.10560704472371918, 'epoch': 0.5} + 50%|████▉ | 259/520 [16:23<16:12, 3.73s/it] 50%|█████ | 260/520 [16:26<16:05, 3.72s/it] {'loss': 1.5107, 'grad_norm': 0.00329319785010895, 'learning_rate': 0.10498458856606972, 'epoch': 0.5} + 50%|█████ | 260/520 [16:26<16:05, 3.72s/it] 50%|█████ | 261/520 [16:30<16:00, 3.71s/it] {'loss': 1.4258, 'grad_norm': 0.0033850327955352024, 'learning_rate': 0.10436193873653361, 'epoch': 0.5} + 50%|█████ | 261/520 [16:30<16:00, 3.71s/it] 50%|█████ | 262/520 [16:34<15:52, 3.69s/it] {'loss': 1.2674, 'grad_norm': 0.0033148601745230707, 'learning_rate': 0.10373911942763259, 'epoch': 0.5} + 50%|█████ | 262/520 [16:34<15:52, 3.69s/it] 51%|█████ | 263/520 [16:37<15:44, 3.67s/it] {'loss': 1.44, 'grad_norm': 0.0034972581536710586, 'learning_rate': 0.10311615483847332, 'epoch': 0.51} + 51%|█████ | 263/520 [16:37<15:44, 3.67s/it] 51%|█████ | 264/520 [16:41<15:38, 3.67s/it] {'loss': 1.3876, 'grad_norm': 0.00330126657899592, 'learning_rate': 0.1024930691738073, 'epoch': 0.51} + 51%|█████ | 264/520 [16:41<15:38, 3.67s/it] 51%|█████ | 265/520 [16:45<15:36, 3.67s/it] {'loss': 1.2842, 'grad_norm': 0.00375813406774125, 'learning_rate': 0.10186988664309023, 'epoch': 0.51} + 51%|█████ | 265/520 [16:45<15:36, 3.67s/it] 51%|█████ | 266/520 [16:48<15:31, 3.67s/it] {'loss': 1.1326, 'grad_norm': 0.002929042623743462, 'learning_rate': 0.10124663145954152, 'epoch': 0.51} + 51%|█████ | 266/520 [16:48<15:31, 3.67s/it] 51%|█████▏ | 267/520 [16:52<15:30, 3.68s/it] {'loss': 1.2826, 'grad_norm': 0.0031420719563478427, 'learning_rate': 0.10062332783920336, 'epoch': 0.51} + 51%|█████▏ | 267/520 [16:52<15:30, 3.68s/it] 52%|█████▏ | 268/520 [16:56<15:26, 3.68s/it] {'loss': 1.5657, 'grad_norm': 0.0037909361678661346, 'learning_rate': 0.1, 'epoch': 0.52} + 52%|█████▏ | 268/520 [16:56<15:26, 3.68s/it] 52%|█████▏ | 269/520 [16:59<15:20, 3.67s/it] {'loss': 1.3993, 'grad_norm': 0.003354774464146591, 'learning_rate': 0.09937667216079665, 'epoch': 0.52} + 52%|█████▏ | 269/520 [16:59<15:20, 3.67s/it] 52%|█████▏ | 270/520 [17:03<15:17, 3.67s/it] {'loss': 1.3531, 'grad_norm': 0.0032239567337742294, 'learning_rate': 0.0987533685404585, 'epoch': 0.52} + 52%|█████▏ | 270/520 [17:03<15:17, 3.67s/it] 52%|█████▏ | 271/520 [17:07<15:26, 3.72s/it] {'loss': 1.38, 'grad_norm': 0.0032437442998318137, 'learning_rate': 0.0981301133569098, 'epoch': 0.52} + 52%|█████▏ | 271/520 [17:07<15:26, 3.72s/it] 52%|█████▏ | 272/520 [17:11<15:41, 3.80s/it] {'loss': 1.3521, 'grad_norm': 0.0034519430259002787, 'learning_rate': 0.09750693082619273, 'epoch': 0.52} + 52%|█████▏ | 272/520 [17:11<15:41, 3.80s/it] 52%|█████▎ | 273/520 [17:15<15:47, 3.84s/it] {'loss': 1.4876, 'grad_norm': 0.0035354005752913135, 'learning_rate': 0.0968838451615267, 'epoch': 0.53} + 52%|█████▎ | 273/520 [17:15<15:47, 3.84s/it] 53%|█████▎ | 274/520 [17:19<15:50, 3.87s/it] {'loss': 1.3349, 'grad_norm': 0.0035054775121587025, 'learning_rate': 0.09626088057236745, 'epoch': 0.53} + 53%|█████▎ | 274/520 [17:19<15:50, 3.87s/it] 53%|█████▎ | 275/520 [17:22<15:49, 3.88s/it] {'loss': 1.2795, 'grad_norm': 0.0036639966147471443, 'learning_rate': 0.09563806126346641, 'epoch': 0.53} + 53%|█████▎ | 275/520 [17:22<15:49, 3.88s/it] 53%|█████▎ | 276/520 [17:26<15:47, 3.88s/it] {'loss': 1.3625, 'grad_norm': 0.0034736592560321087, 'learning_rate': 0.09501541143393027, 'epoch': 0.53} + 53%|█████▎ | 276/520 [17:26<15:47, 3.88s/it] 53%|█████▎ | 277/520 [17:30<15:48, 3.90s/it] {'loss': 1.4778, 'grad_norm': 0.003781436278865195, 'learning_rate': 0.09439295527628082, 'epoch': 0.53} + 53%|█████▎ | 277/520 [17:30<15:48, 3.90s/it] 53%|█████▎ | 278/520 [17:34<15:43, 3.90s/it] {'loss': 1.2347, 'grad_norm': 0.0030904950311060005, 'learning_rate': 0.0937707169755148, 'epoch': 0.53} + 53%|█████▎ | 278/520 [17:34<15:43, 3.90s/it] 54%|█████▎ | 279/520 [17:38<15:41, 3.90s/it] {'loss': 1.3855, 'grad_norm': 0.004550042312953474, 'learning_rate': 0.09314872070816434, 'epoch': 0.54} + 54%|█████▎ | 279/520 [17:38<15:41, 3.90s/it] 54%|█████▍ | 280/520 [17:42<15:36, 3.90s/it] {'loss': 1.2901, 'grad_norm': 0.003758445736838111, 'learning_rate': 0.09252699064135758, 'epoch': 0.54} + 54%|█████▍ | 280/520 [17:42<15:36, 3.90s/it] 54%|█████▍ | 281/520 [17:46<15:31, 3.90s/it] {'loss': 1.4002, 'grad_norm': 0.003464687832693915, 'learning_rate': 0.09190555093187967, 'epoch': 0.54} + 54%|█████▍ | 281/520 [17:46<15:31, 3.90s/it] 54%|█████▍ | 282/520 [17:50<15:31, 3.91s/it] {'loss': 1.2349, 'grad_norm': 0.0034563466031420744, 'learning_rate': 0.09128442572523418, 'epoch': 0.54} + 54%|█████▍ | 282/520 [17:50<15:31, 3.91s/it] 54%|█████▍ | 283/520 [17:54<15:29, 3.92s/it] {'loss': 1.4193, 'grad_norm': 0.0037452193155578075, 'learning_rate': 0.09066363915470495, 'epoch': 0.54} + 54%|█████▍ | 283/520 [17:54<15:29, 3.92s/it] 55%|█████▍ | 284/520 [17:58<15:25, 3.92s/it] {'loss': 1.3464, 'grad_norm': 0.0038657387608199145, 'learning_rate': 0.09004321534041836, 'epoch': 0.55} + 55%|█████▍ | 284/520 [17:58<15:25, 3.92s/it] 55%|█████▍ | 285/520 [18:02<15:20, 3.92s/it] {'loss': 1.2711, 'grad_norm': 0.0034830710389954656, 'learning_rate': 0.08942317838840624, 'epoch': 0.55} + 55%|█████▍ | 285/520 [18:02<15:20, 3.92s/it] 55%|█████▌ | 286/520 [18:06<15:15, 3.91s/it] {'loss': 1.1429, 'grad_norm': 0.003777955283124046, 'learning_rate': 0.08880355238966922, 'epoch': 0.55} + 55%|█████▌ | 286/520 [18:06<15:15, 3.91s/it] 55%|█████▌ | 287/520 [18:09<14:59, 3.86s/it] {'loss': 1.3927, 'grad_norm': 0.0035135271304367124, 'learning_rate': 0.08818436141924073, 'epoch': 0.55} + 55%|█████▌ | 287/520 [18:09<14:59, 3.86s/it] 55%|█████▌ | 288/520 [18:13<14:40, 3.80s/it] {'loss': 1.4403, 'grad_norm': 0.0033087333571394303, 'learning_rate': 0.08756562953525152, 'epoch': 0.55} + 55%|█████▌ | 288/520 [18:13<14:40, 3.80s/it] 56%|█████▌ | 289/520 [18:17<14:31, 3.77s/it] {'loss': 1.2868, 'grad_norm': 0.003105392067730659, 'learning_rate': 0.08694738077799487, 'epoch': 0.56} + 56%|█████▌ | 289/520 [18:17<14:31, 3.77s/it] 56%|█████▌ | 290/520 [18:20<14:21, 3.74s/it] {'loss': 1.2003, 'grad_norm': 0.003108177600042451, 'learning_rate': 0.08632963916899268, 'epoch': 0.56} + 56%|█████▌ | 290/520 [18:20<14:21, 3.74s/it] 56%|█████▌ | 291/520 [18:24<14:11, 3.72s/it] {'loss': 1.2624, 'grad_norm': 0.0037435525668899033, 'learning_rate': 0.08571242871006202, 'epoch': 0.56} + 56%|█████▌ | 291/520 [18:24<14:11, 3.72s/it] 56%|█████▌ | 292/520 [18:28<14:03, 3.70s/it] {'loss': 1.3056, 'grad_norm': 0.003147172329259371, 'learning_rate': 0.08509577338238256, 'epoch': 0.56} + 56%|█████▌ | 292/520 [18:28<14:03, 3.70s/it] 56%|█████▋ | 293/520 [18:31<13:56, 3.69s/it] {'loss': 1.2571, 'grad_norm': 0.003495006386188701, 'learning_rate': 0.08447969714556484, 'epoch': 0.56} + 56%|█████▋ | 293/520 [18:31<13:56, 3.69s/it] 57%|█████▋ | 294/520 [18:35<13:52, 3.68s/it] {'loss': 1.2906, 'grad_norm': 0.0034375939409897684, 'learning_rate': 0.08386422393671933, 'epoch': 0.57} + 57%|█████▋ | 294/520 [18:35<13:52, 3.68s/it] 57%|█████▋ | 295/520 [18:39<13:48, 3.68s/it] {'loss': 1.4369, 'grad_norm': 0.0037392069153639623, 'learning_rate': 0.08324937766952638, 'epoch': 0.57} + 57%|█████▋ | 295/520 [18:39<13:48, 3.68s/it] 57%|█████▋ | 296/520 [18:42<13:42, 3.67s/it] {'loss': 1.2254, 'grad_norm': 0.003405076907012863, 'learning_rate': 0.08263518223330697, 'epoch': 0.57} + 57%|█████▋ | 296/520 [18:42<13:42, 3.67s/it] 57%|█████▋ | 297/520 [18:46<13:40, 3.68s/it] {'loss': 1.3649, 'grad_norm': 0.0036062826274636866, 'learning_rate': 0.08202166149209474, 'epoch': 0.57} + 57%|█████▋ | 297/520 [18:46<13:40, 3.68s/it] 57%|█████▋ | 298/520 [18:50<13:35, 3.67s/it] {'loss': 1.3239, 'grad_norm': 0.002960011028922108, 'learning_rate': 0.08140883928370855, 'epoch': 0.57} + 57%|█████▋ | 298/520 [18:50<13:35, 3.67s/it] 57%|█████▊ | 299/520 [18:53<13:31, 3.67s/it] {'loss': 1.4312, 'grad_norm': 0.003872084016718616, 'learning_rate': 0.0807967394188264, 'epoch': 0.57} + 57%|█████▊ | 299/520 [18:53<13:31, 3.67s/it] 58%|█████▊ | 300/520 [18:57<13:29, 3.68s/it] {'loss': 1.385, 'grad_norm': 0.0032458213713354436, 'learning_rate': 0.08018538568006027, 'epoch': 0.58} + 58%|█████▊ | 300/520 [18:57<13:29, 3.68s/it] 58%|█████▊ | 301/520 [19:01<13:32, 3.71s/it] {'loss': 1.3557, 'grad_norm': 0.0033664224841518384, 'learning_rate': 0.07957480182103199, 'epoch': 0.58} + 58%|█████▊ | 301/520 [19:01<13:32, 3.71s/it] 58%|█████▊ | 302/520 [19:05<13:28, 3.71s/it] {'loss': 1.4397, 'grad_norm': 0.003408459710261295, 'learning_rate': 0.07896501156545044, 'epoch': 0.58} + 58%|█████▊ | 302/520 [19:05<13:28, 3.71s/it] 58%|█████▊ | 303/520 [19:08<13:23, 3.70s/it] {'loss': 1.2845, 'grad_norm': 0.003656059034593322, 'learning_rate': 0.07835603860618973, 'epoch': 0.58} + 58%|█████▊ | 303/520 [19:08<13:23, 3.70s/it] 58%|█████▊ | 304/520 [19:12<13:20, 3.70s/it] {'loss': 1.3412, 'grad_norm': 0.0036318252422635157, 'learning_rate': 0.07774790660436857, 'epoch': 0.58} + 58%|█████▊ | 304/520 [19:12<13:20, 3.70s/it] 59%|█████▊ | 305/520 [19:16<13:14, 3.70s/it] {'loss': 1.4036, 'grad_norm': 0.003756688916834322, 'learning_rate': 0.07714063918843106, 'epoch': 0.59} + 59%|█████▊ | 305/520 [19:16<13:14, 3.70s/it] 59%|█████▉ | 306/520 [19:19<13:09, 3.69s/it] {'loss': 1.325, 'grad_norm': 0.0032341477541887004, 'learning_rate': 0.0765342599532285, 'epoch': 0.59} + 59%|█████▉ | 306/520 [19:19<13:09, 3.69s/it] 59%|█████▉ | 307/520 [19:23<13:30, 3.80s/it] {'loss': 1.2669, 'grad_norm': 0.003092617984078062, 'learning_rate': 0.07592879245910272, 'epoch': 0.59} + 59%|█████▉ | 307/520 [19:23<13:30, 3.80s/it] 59%|█████▉ | 308/520 [19:27<13:16, 3.76s/it] {'loss': 1.3928, 'grad_norm': 0.0031786765838397054, 'learning_rate': 0.07532426023097064, 'epoch': 0.59} + 59%|█████▉ | 308/520 [19:27<13:16, 3.76s/it] 59%|█████▉ | 309/520 [19:31<13:06, 3.73s/it] {'loss': 1.2615, 'grad_norm': 0.003067952262848062, 'learning_rate': 0.07472068675741024, 'epoch': 0.59} + 59%|█████▉ | 309/520 [19:31<13:06, 3.73s/it] 60%|█████▉ | 310/520 [19:34<12:57, 3.70s/it] {'loss': 1.2454, 'grad_norm': 0.003232327275572514, 'learning_rate': 0.07411809548974792, 'epoch': 0.6} + 60%|█████▉ | 310/520 [19:34<12:57, 3.70s/it] 60%|█████▉ | 311/520 [19:38<12:49, 3.68s/it] {'loss': 1.2155, 'grad_norm': 0.003137967456681367, 'learning_rate': 0.07351650984114727, 'epoch': 0.6} + 60%|█████▉ | 311/520 [19:38<12:49, 3.68s/it] 60%|██████ | 312/520 [19:42<12:42, 3.67s/it] {'loss': 1.2035, 'grad_norm': 0.003657658986647462, 'learning_rate': 0.0729159531856995, 'epoch': 0.6} + 60%|██████ | 312/520 [19:42<12:42, 3.67s/it] 60%|██████ | 313/520 [19:45<12:36, 3.66s/it] {'loss': 1.1969, 'grad_norm': 0.0029751383912572614, 'learning_rate': 0.07231644885751508, 'epoch': 0.6} + 60%|██████ | 313/520 [19:45<12:36, 3.66s/it] 60%|██████ | 314/520 [19:49<12:54, 3.76s/it] {'loss': 1.2322, 'grad_norm': 0.003074055726919744, 'learning_rate': 0.07171802014981725, 'epoch': 0.6} + 60%|██████ | 314/520 [19:49<12:54, 3.76s/it] 61%|██████ | 315/520 [19:53<12:42, 3.72s/it] {'loss': 1.4205, 'grad_norm': 0.004536962499242739, 'learning_rate': 0.07112069031403703, 'epoch': 0.61} + 61%|██████ | 315/520 [19:53<12:42, 3.72s/it] 61%|██████ | 316/520 [19:57<13:01, 3.83s/it] {'loss': 1.2047, 'grad_norm': 0.0040846183392077955, 'learning_rate': 0.07052448255890957, 'epoch': 0.61} + 61%|██████ | 316/520 [19:57<13:01, 3.83s/it] 61%|██████ | 317/520 [20:01<12:46, 3.78s/it] {'loss': 1.2286, 'grad_norm': 0.0029473870595599938, 'learning_rate': 0.0699294200495727, 'epoch': 0.61} + 61%|██████ | 317/520 [20:01<12:46, 3.78s/it] 61%|██████ | 318/520 [20:04<12:41, 3.77s/it] {'loss': 1.3517, 'grad_norm': 0.003574080227865359, 'learning_rate': 0.06933552590666658, 'epoch': 0.61} + 61%|██████ | 318/520 [20:04<12:41, 3.77s/it] 61%|██████▏ | 319/520 [20:08<12:56, 3.87s/it] {'loss': 1.2165, 'grad_norm': 0.0033020347423976335, 'learning_rate': 0.06874282320543557, 'epoch': 0.61} + 61%|██████▏ | 319/520 [20:08<12:56, 3.87s/it] 62%|██████▏ | 320/520 [20:12<12:44, 3.82s/it] {'loss': 1.1571, 'grad_norm': 0.003421069849364812, 'learning_rate': 0.06815133497483157, 'epoch': 0.62} + 62%|██████▏ | 320/520 [20:12<12:44, 3.82s/it] 62%|██████▏ | 321/520 [20:16<12:34, 3.79s/it] {'loss': 1.3676, 'grad_norm': 0.0035654798616239812, 'learning_rate': 0.06756108419661931, 'epoch': 0.62} + 62%|██████▏ | 321/520 [20:16<12:34, 3.79s/it] 62%|██████▏ | 322/520 [50:00<29:34:59, 537.88s/it] {'loss': 1.2656, 'grad_norm': 0.003309531129170382, 'learning_rate': 0.06697209380448332, 'epoch': 0.62} + 62%|██████▏ | 322/520 [50:00<29:34:59, 537.88s/it] 62%|██████▏ | 323/520 [50:04<20:39:52, 377.63s/it] {'loss': 1.35, 'grad_norm': 0.0037552521780461613, 'learning_rate': 0.06638438668313694, 'epoch': 0.62} + 62%|██████▏ | 323/520 [50:04<20:39:52, 377.63s/it] 62%|██████▏ | 324/520 [50:07<14:27:07, 265.45s/it] {'loss': 1.2945, 'grad_norm': 0.003961905266950623, 'learning_rate': 0.06579798566743314, 'epoch': 0.62} + 62%|██████▏ | 324/520 [50:07<14:27:07, 265.45s/it] 62%|██████▎ | 325/520 [50:11<10:07:29, 186.92s/it] {'loss': 1.3146, 'grad_norm': 0.003683668965639517, 'learning_rate': 0.06521291354147728, 'epoch': 0.62} + 62%|██████▎ | 325/520 [50:11<10:07:29, 186.92s/it] 63%|██████▎ | 326/520 [50:15<7:06:38, 131.95s/it] {'loss': 1.2872, 'grad_norm': 0.003290016405630281, 'learning_rate': 0.06462919303774187, 'epoch': 0.63} + 63%|██████▎ | 326/520 [50:15<7:06:38, 131.95s/it] 63%|██████▎ | 327/520 [50:18<5:00:44, 93.50s/it] {'loss': 1.4219, 'grad_norm': 0.004104710130097114, 'learning_rate': 0.06404684683618325, 'epoch': 0.63} + 63%|██████▎ | 327/520 [50:18<5:00:44, 93.50s/it] 63%|██████▎ | 328/520 [50:22<3:32:57, 66.55s/it] {'loss': 1.3582, 'grad_norm': 0.0034466721112048826, 'learning_rate': 0.0634658975633605, 'epoch': 0.63} + 63%|██████▎ | 328/520 [50:22<3:32:57, 66.55s/it] 63%|██████▎ | 329/520 [50:26<2:31:47, 47.68s/it] {'loss': 1.2049, 'grad_norm': 0.002890932138166811, 'learning_rate': 0.06288636779155621, 'epoch': 0.63} + 63%|██████▎ | 329/520 [50:26<2:31:47, 47.68s/it] 63%|██████▎ | 330/520 [50:29<1:49:13, 34.49s/it] {'loss': 1.2778, 'grad_norm': 0.002843321268849721, 'learning_rate': 0.06230828003789948, 'epoch': 0.63} + 63%|██████▎ | 330/520 [50:29<1:49:13, 34.49s/it] 64%|██████▎ | 331/520 [50:33<1:19:30, 25.24s/it] {'loss': 1.2424, 'grad_norm': 0.0031120126214264497, 'learning_rate': 0.06173165676349103, 'epoch': 0.64} + 64%|██████▎ | 331/520 [50:33<1:19:30, 25.24s/it] 64%|██████▍ | 332/520 [50:37<58:48, 18.77s/it] {'loss': 1.4202, 'grad_norm': 0.003263942079138196, 'learning_rate': 0.06115652037253053, 'epoch': 0.64} + 64%|██████▍ | 332/520 [50:37<58:48, 18.77s/it] 64%|██████▍ | 333/520 [50:40<44:22, 14.24s/it] {'loss': 1.4115, 'grad_norm': 0.0035643071593927404, 'learning_rate': 0.06058289321144608, 'epoch': 0.64} + 64%|██████▍ | 333/520 [50:40<44:22, 14.24s/it] 64%|██████▍ | 334/520 [50:44<34:20, 11.08s/it] {'loss': 1.2963, 'grad_norm': 0.0038492420704627052, 'learning_rate': 0.06001079756802592, 'epoch': 0.64} + 64%|██████▍ | 334/520 [50:44<34:20, 11.08s/it] 64%|██████▍ | 335/520 [50:48<27:16, 8.85s/it] {'loss': 1.2852, 'grad_norm': 0.0030415397239888986, 'learning_rate': 0.059440255670552514, 'epoch': 0.64} + 64%|██████▍ | 335/520 [50:48<27:16, 8.85s/it] 65%|██████▍ | 336/520 [50:51<22:21, 7.29s/it] {'loss': 1.1788, 'grad_norm': 0.003514154216675056, 'learning_rate': 0.05887128968693887, 'epoch': 0.65} + 65%|██████▍ | 336/520 [50:51<22:21, 7.29s/it] 65%|██████▍ | 337/520 [50:55<18:56, 6.21s/it] {'loss': 1.171, 'grad_norm': 0.0031988184132868565, 'learning_rate': 0.058303921723867225, 'epoch': 0.65} + 65%|██████▍ | 337/520 [50:55<18:56, 6.21s/it] 65%|██████▌ | 338/520 [50:59<16:32, 5.45s/it] {'loss': 1.3081, 'grad_norm': 0.0033320798843520425, 'learning_rate': 0.05773817382593008, 'epoch': 0.65} + 65%|██████▌ | 338/520 [50:59<16:32, 5.45s/it] 65%|██████▌ | 339/520 [51:03<14:48, 4.91s/it] {'loss': 1.2445, 'grad_norm': 0.003263615952132255, 'learning_rate': 0.057174067974773715, 'epoch': 0.65} + 65%|██████▌ | 339/520 [51:03<14:48, 4.91s/it] 65%|██████▌ | 340/520 [51:06<13:37, 4.54s/it] {'loss': 1.2294, 'grad_norm': 0.0031931941040805145, 'learning_rate': 0.056611626088244195, 'epoch': 0.65} + 65%|██████▌ | 340/520 [51:06<13:37, 4.54s/it] 66%|██████▌ | 341/520 [51:10<12:43, 4.27s/it] {'loss': 1.2586, 'grad_norm': 0.003592808840271568, 'learning_rate': 0.056050870019535494, 'epoch': 0.66} + 66%|██████▌ | 341/520 [51:10<12:43, 4.27s/it] 66%|██████▌ | 342/520 [51:13<12:07, 4.08s/it] {'loss': 1.3813, 'grad_norm': 0.003734584494643109, 'learning_rate': 0.05549182155634076, 'epoch': 0.66} + 66%|██████▌ | 342/520 [51:13<12:07, 4.08s/it] 66%|██████▌ | 343/520 [51:17<11:41, 3.96s/it] {'loss': 1.3563, 'grad_norm': 0.003446670147348687, 'learning_rate': 0.054934502420005464, 'epoch': 0.66} + 66%|██████▌ | 343/520 [51:17<11:41, 3.96s/it] 66%|██████▌ | 344/520 [51:21<11:21, 3.87s/it] {'loss': 1.2015, 'grad_norm': 0.003371249060718611, 'learning_rate': 0.0543789342646837, 'epoch': 0.66} + 66%|██████▌ | 344/520 [51:21<11:21, 3.87s/it] 66%|██████▋ | 345/520 [51:25<11:08, 3.82s/it] {'loss': 1.3224, 'grad_norm': 0.0037860711059486555, 'learning_rate': 0.05382513867649663, 'epoch': 0.66} + 66%|██████▋ | 345/520 [51:25<11:08, 3.82s/it] 67%|██████▋ | 346/520 [51:28<10:56, 3.77s/it] {'loss': 1.34, 'grad_norm': 0.003154850973700134, 'learning_rate': 0.0532731371726938, 'epoch': 0.67} + 67%|██████▋ | 346/520 [51:28<10:56, 3.77s/it] 67%|██████▋ | 347/520 [51:32<10:47, 3.74s/it] {'loss': 1.22, 'grad_norm': 0.003092809735045898, 'learning_rate': 0.05272295120081732, 'epoch': 0.67} + 67%|██████▋ | 347/520 [51:32<10:47, 3.74s/it]Token indices sequence length is longer than the specified maximum sequence length for this model (2778 > 2048). Running this sequence through the model will result in indexing errors + 67%|██████▋ | 348/520 [51:35<10:38, 3.71s/it] {'loss': 1.1779, 'grad_norm': 0.003968131917835207, 'learning_rate': 0.05217460213786821, 'epoch': 0.67} + 67%|██████▋ | 348/520 [51:35<10:38, 3.71s/it] 67%|██████▋ | 349/520 [51:39<10:37, 3.73s/it] {'loss': 1.2253, 'grad_norm': 0.0035082083903847493, 'learning_rate': 0.051628111289476024, 'epoch': 0.67} + 67%|██████▋ | 349/520 [51:39<10:37, 3.73s/it] 67%|██████▋ | 350/520 [51:43<10:31, 3.72s/it] {'loss': 1.2589, 'grad_norm': 0.003429948795676369, 'learning_rate': 0.051083499889071106, 'epoch': 0.67} + 67%|██████▋ | 350/520 [51:43<10:31, 3.72s/it] 68%|██████▊ | 351/520 [51:47<10:27, 3.72s/it] {'loss': 1.1662, 'grad_norm': 0.0031273503819398895, 'learning_rate': 0.05054078909705926, 'epoch': 0.68} + 68%|██████▊ | 351/520 [51:47<10:27, 3.72s/it] 68%|██████▊ | 352/520 [51:50<10:21, 3.70s/it] {'loss': 1.2923, 'grad_norm': 0.0031804175047245526, 'learning_rate': 0.050000000000000024, 'epoch': 0.68} + 68%|██████▊ | 352/520 [51:50<10:21, 3.70s/it] 68%|██████▊ | 353/520 [51:54<10:19, 3.71s/it] {'loss': 1.2813, 'grad_norm': 0.0027466563637846106, 'learning_rate': 0.04946115360978696, 'epoch': 0.68} + 68%|██████▊ | 353/520 [51:54<10:19, 3.71s/it] 68%|██████▊ | 354/520 [51:58<10:15, 3.71s/it] {'loss': 1.4319, 'grad_norm': 0.0032840543182404403, 'learning_rate': 0.048924270862831465, 'epoch': 0.68} + 68%|██████▊ | 354/520 [51:58<10:15, 3.71s/it] 68%|██████▊ | 355/520 [52:01<10:09, 3.69s/it] {'loss': 1.2271, 'grad_norm': 0.0032644931365180723, 'learning_rate': 0.04838937261924933, 'epoch': 0.68} + 68%|██████▊ | 355/520 [52:01<10:09, 3.69s/it] 68%|██████▊ | 356/520 [52:05<10:05, 3.69s/it] {'loss': 1.2297, 'grad_norm': 0.003464117015177039, 'learning_rate': 0.0478564796620502, 'epoch': 0.68} + 68%|██████▊ | 356/520 [52:05<10:05, 3.69s/it] 69%|██████▊ | 357/520 [52:09<10:02, 3.70s/it] {'loss': 1.2514, 'grad_norm': 0.0028917516479444417, 'learning_rate': 0.04732561269632993, 'epoch': 0.69} + 69%|██████▊ | 357/520 [52:09<10:02, 3.70s/it] 69%|██████▉ | 358/520 [52:12<09:57, 3.69s/it] {'loss': 1.1777, 'grad_norm': 0.003396580415627929, 'learning_rate': 0.04679679234846636, 'epoch': 0.69} + 69%|██████▉ | 358/520 [52:12<09:57, 3.69s/it] 69%|██████▉ | 359/520 [52:16<09:52, 3.68s/it] {'loss': 1.362, 'grad_norm': 0.003638920800767125, 'learning_rate': 0.046270039165317606, 'epoch': 0.69} + 69%|██████▉ | 359/520 [52:16<09:52, 3.68s/it] 69%|██████▉ | 360/520 [52:20<09:48, 3.68s/it] {'loss': 1.3695, 'grad_norm': 0.003497773534314793, 'learning_rate': 0.04574537361342407, 'epoch': 0.69} + 69%|██████▉ | 360/520 [52:20<09:48, 3.68s/it] 69%|██████▉ | 361/520 [52:23<09:44, 3.68s/it] {'loss': 1.3622, 'grad_norm': 0.0029938283281656093, 'learning_rate': 0.04522281607821288, 'epoch': 0.69} + 69%|██████▉ | 361/520 [52:23<09:44, 3.68s/it] 70%|██████▉ | 362/520 [52:27<09:40, 3.67s/it] {'loss': 1.2447, 'grad_norm': 0.003392158266637909, 'learning_rate': 0.04470238686320606, 'epoch': 0.7} + 70%|██████▉ | 362/520 [52:27<09:40, 3.67s/it] 70%|██████▉ | 363/520 [52:31<09:35, 3.66s/it] {'loss': 1.2711, 'grad_norm': 0.003118451084010664, 'learning_rate': 0.044184106189231624, 'epoch': 0.7} + 70%|██████▉ | 363/520 [52:31<09:35, 3.66s/it] 70%|███████ | 364/520 [52:35<09:34, 3.68s/it] {'loss': 1.374, 'grad_norm': 0.0033458751111310127, 'learning_rate': 0.043667994193637795, 'epoch': 0.7} + 70%|███████ | 364/520 [52:35<09:34, 3.68s/it] 70%|███████ | 365/520 [52:38<09:32, 3.69s/it] {'loss': 1.3338, 'grad_norm': 0.003283597886272527, 'learning_rate': 0.043154070929510784, 'epoch': 0.7} + 70%|███████ | 365/520 [52:38<09:32, 3.69s/it] 70%|███████ | 366/520 [52:42<09:34, 3.73s/it] {'loss': 1.2861, 'grad_norm': 0.0028656984542982074, 'learning_rate': 0.04264235636489542, 'epoch': 0.7} + 70%|███████ | 366/520 [52:42<09:34, 3.73s/it] 71%|███████ | 367/520 [52:46<09:27, 3.71s/it] {'loss': 1.2854, 'grad_norm': 0.003084087254936327, 'learning_rate': 0.04213287038201943, 'epoch': 0.71} + 71%|███████ | 367/520 [52:46<09:27, 3.71s/it] 71%|███████ | 368/520 [52:49<09:23, 3.71s/it] {'loss': 1.1349, 'grad_norm': 0.0034754198565774223, 'learning_rate': 0.04162563277652104, 'epoch': 0.71} + 71%|███████ | 368/520 [52:49<09:23, 3.71s/it] 71%|███████ | 369/520 [52:53<09:17, 3.69s/it] {'loss': 1.3294, 'grad_norm': 0.0031060545230998317, 'learning_rate': 0.04112066325667954, 'epoch': 0.71} + 71%|███████ | 369/520 [52:53<09:17, 3.69s/it] 71%|███████ | 370/520 [52:57<09:11, 3.68s/it] {'loss': 1.196, 'grad_norm': 0.0029326229916906243, 'learning_rate': 0.04061798144264986, 'epoch': 0.71} + 71%|███████ | 370/520 [52:57<09:11, 3.68s/it] 71%|███████▏ | 371/520 [53:00<09:07, 3.67s/it] {'loss': 1.1964, 'grad_norm': 0.0032414734671384927, 'learning_rate': 0.04011760686569998, 'epoch': 0.71} + 71%|███████▏ | 371/520 [53:00<09:07, 3.67s/it] 72%|███████▏ | 372/520 [53:04<09:13, 3.74s/it] {'loss': 1.4276, 'grad_norm': 0.002954225047009287, 'learning_rate': 0.03961955896745224, 'epoch': 0.72} + 72%|███████▏ | 372/520 [53:04<09:13, 3.74s/it] 72%|███████▏ | 373/520 [53:08<09:11, 3.75s/it] {'loss': 1.3077, 'grad_norm': 0.003443840229960228, 'learning_rate': 0.03912385709912794, 'epoch': 0.72} + 72%|███████▏ | 373/520 [53:08<09:11, 3.75s/it] 72%|███████▏ | 374/520 [53:12<09:12, 3.79s/it] {'loss': 1.2776, 'grad_norm': 0.0031644598378342275, 'learning_rate': 0.038630520520795276, 'epoch': 0.72} + 72%|███████▏ | 374/520 [53:12<09:12, 3.79s/it] 72%|███████▏ | 375/520 [53:16<09:05, 3.76s/it] {'loss': 1.1873, 'grad_norm': 0.0033232916252222515, 'learning_rate': 0.03813956840062119, 'epoch': 0.72} + 72%|███████▏ | 375/520 [53:16<09:05, 3.76s/it] 72%|███████▏ | 376/520 [53:20<09:06, 3.80s/it] {'loss': 1.3071, 'grad_norm': 0.003003710193185398, 'learning_rate': 0.037651019814126656, 'epoch': 0.72} + 72%|███████▏ | 376/520 [53:20<09:06, 3.80s/it] 72%|███████▎ | 377/520 [53:23<09:02, 3.80s/it] {'loss': 1.2422, 'grad_norm': 0.003235032694942288, 'learning_rate': 0.037164893743445275, 'epoch': 0.72} + 72%|███████▎ | 377/520 [53:23<09:02, 3.80s/it] 73%|███████▎ | 378/520 [53:27<08:58, 3.80s/it] {'loss': 1.2967, 'grad_norm': 0.0031564343980277996, 'learning_rate': 0.03668120907658603, 'epoch': 0.73} + 73%|███████▎ | 378/520 [53:27<08:58, 3.80s/it] 73%|███████▎ | 379/520 [53:31<08:48, 3.75s/it] {'loss': 1.2824, 'grad_norm': 0.0029856934953282543, 'learning_rate': 0.036199984606699154, 'epoch': 0.73} + 73%|███████▎ | 379/520 [53:31<08:48, 3.75s/it] 73%|███████▎ | 380/520 [53:34<08:41, 3.72s/it] {'loss': 1.4037, 'grad_norm': 0.0037294408348633405, 'learning_rate': 0.035721239031346066, 'epoch': 0.73} + 73%|███████▎ | 380/520 [53:34<08:41, 3.72s/it] 73%|███████▎ | 381/520 [53:38<08:35, 3.71s/it] {'loss': 1.2777, 'grad_norm': 0.0031374158745987916, 'learning_rate': 0.03524499095177297, 'epoch': 0.73} + 73%|███████▎ | 381/520 [53:38<08:35, 3.71s/it] 73%|███████▎ | 382/520 [53:42<08:31, 3.71s/it] {'loss': 1.3425, 'grad_norm': 0.003101705559275741, 'learning_rate': 0.03477125887218792, 'epoch': 0.73} + 73%|███████▎ | 382/520 [53:42<08:31, 3.71s/it] 74%|███████▎ | 383/520 [53:46<08:28, 3.71s/it] {'loss': 1.1098, 'grad_norm': 0.003477337853890237, 'learning_rate': 0.03430006119904196, 'epoch': 0.74} + 74%|███████▎ | 383/520 [53:46<08:28, 3.71s/it] 74%|███████▍ | 384/520 [53:49<08:30, 3.75s/it] {'loss': 1.4647, 'grad_norm': 0.0035421705025240275, 'learning_rate': 0.033831416240314084, 'epoch': 0.74} + 74%|███████▍ | 384/520 [53:49<08:30, 3.75s/it] 74%|███████▍ | 385/520 [53:53<08:26, 3.75s/it] {'loss': 1.2552, 'grad_norm': 0.003073662593757523, 'learning_rate': 0.03336534220479961, 'epoch': 0.74} + 74%|███████▍ | 385/520 [53:53<08:26, 3.75s/it] 74%|███████▍ | 386/520 [53:57<08:24, 3.77s/it] {'loss': 1.2087, 'grad_norm': 0.0029247381661118735, 'learning_rate': 0.032901857201403005, 'epoch': 0.74} + 74%|███████▍ | 386/520 [53:57<08:24, 3.77s/it] 74%|███████▍ | 387/520 [54:01<08:19, 3.76s/it] {'loss': 1.4262, 'grad_norm': 0.0032040228342630488, 'learning_rate': 0.032440979238433976, 'epoch': 0.74} + 74%|███████▍ | 387/520 [54:01<08:19, 3.76s/it] 75%|███████▍ | 388/520 [54:04<08:19, 3.79s/it] {'loss': 1.1549, 'grad_norm': 0.002908827941966743, 'learning_rate': 0.03198272622290804, 'epoch': 0.75} + 75%|███████▍ | 388/520 [54:04<08:19, 3.79s/it] 75%|███████▍ | 389/520 [54:08<08:12, 3.76s/it] {'loss': 1.2155, 'grad_norm': 0.0035304523984517958, 'learning_rate': 0.03152711595985065, 'epoch': 0.75} + 75%|███████▍ | 389/520 [54:08<08:12, 3.76s/it] 75%|███████▌ | 390/520 [54:12<08:06, 3.74s/it] {'loss': 1.2694, 'grad_norm': 0.0030308942945358464, 'learning_rate': 0.031074166151605298, 'epoch': 0.75} + 75%|███████▌ | 390/520 [54:12<08:06, 3.74s/it] 75%|███████▌ | 391/520 [54:16<08:01, 3.74s/it] {'loss': 1.3577, 'grad_norm': 0.0031919840177463345, 'learning_rate': 0.030623894397145836, 'epoch': 0.75} + 75%|███████▌ | 391/520 [54:16<08:01, 3.74s/it] 75%|███████▌ | 392/520 [54:19<07:56, 3.72s/it] {'loss': 1.1625, 'grad_norm': 0.0029622273740964947, 'learning_rate': 0.03017631819139273, 'epoch': 0.75} + 75%|███████▌ | 392/520 [54:19<07:56, 3.72s/it] 76%|███████▌ | 393/520 [54:23<07:52, 3.72s/it] {'loss': 1.2254, 'grad_norm': 0.0028822034447103834, 'learning_rate': 0.029731454924533086, 'epoch': 0.76} + 76%|███████▌ | 393/520 [54:23<07:52, 3.72s/it] 76%|███████▌ | 394/520 [54:27<07:46, 3.70s/it] {'loss': 1.23, 'grad_norm': 0.0034482819190281688, 'learning_rate': 0.029289321881345254, 'epoch': 0.76} + 76%|███████▌ | 394/520 [54:27<07:46, 3.70s/it] 76%|███████▌ | 395/520 [54:30<07:42, 3.70s/it] {'loss': 1.192, 'grad_norm': 0.0033603259543095254, 'learning_rate': 0.028849936240527008, 'epoch': 0.76} + 76%|███████▌ | 395/520 [54:30<07:42, 3.70s/it] 76%|███████▌ | 396/520 [54:34<07:37, 3.69s/it] {'loss': 1.283, 'grad_norm': 0.003210570208483512, 'learning_rate': 0.028413315074028157, 'epoch': 0.76} + 76%|███████▌ | 396/520 [54:34<07:37, 3.69s/it] 76%|███████▋ | 397/520 [54:38<07:34, 3.69s/it] {'loss': 1.2573, 'grad_norm': 0.0029753127617753745, 'learning_rate': 0.027979475346387363, 'epoch': 0.76} + 76%|███████▋ | 397/520 [54:38<07:34, 3.69s/it] 77%|███████▋ | 398/520 [54:41<07:30, 3.69s/it] {'loss': 1.2513, 'grad_norm': 0.003256088834926698, 'learning_rate': 0.027548433914072735, 'epoch': 0.77} + 77%|███████▋ | 398/520 [54:41<07:30, 3.69s/it] 77%|███████▋ | 399/520 [54:45<07:28, 3.70s/it] {'loss': 1.2721, 'grad_norm': 0.0032141450076540825, 'learning_rate': 0.027120207524827168, 'epoch': 0.77} + 77%|███████▋ | 399/520 [54:45<07:28, 3.70s/it] 77%|███████▋ | 400/520 [54:49<07:27, 3.73s/it] {'loss': 1.3298, 'grad_norm': 0.003169247250265038, 'learning_rate': 0.02669481281701739, 'epoch': 0.77} + 77%|███████▋ | 400/520 [54:49<07:27, 3.73s/it] 77%|███████▋ | 401/520 [54:53<07:27, 3.76s/it] {'loss': 1.0835, 'grad_norm': 0.0033397919943273942, 'learning_rate': 0.026272266318987603, 'epoch': 0.77} + 77%|███████▋ | 401/520 [54:53<07:27, 3.76s/it] 77%|███████▋ | 402/520 [54:57<07:28, 3.80s/it] {'loss': 1.1888, 'grad_norm': 0.0032240022734702687, 'learning_rate': 0.02585258444841733, 'epoch': 0.77} + 77%|███████▋ | 402/520 [54:57<07:28, 3.80s/it] 78%|███████▊ | 403/520 [55:01<07:28, 3.83s/it] {'loss': 1.2293, 'grad_norm': 0.0034355142268003443, 'learning_rate': 0.025435783511683442, 'epoch': 0.78} + 78%|███████▊ | 403/520 [55:01<07:28, 3.83s/it] 78%|███████▊ | 404/520 [55:04<07:26, 3.85s/it] {'loss': 1.142, 'grad_norm': 0.003978356062276969, 'learning_rate': 0.02502187970322657, 'epoch': 0.78} + 78%|███████▊ | 404/520 [55:04<07:26, 3.85s/it] 78%|███████▊ | 405/520 [55:08<07:23, 3.86s/it] {'loss': 1.2782, 'grad_norm': 0.003069239238275929, 'learning_rate': 0.02461088910492202, 'epoch': 0.78} + 78%|███████▊ | 405/520 [55:08<07:23, 3.86s/it] 78%|███████▊ | 406/520 [55:12<07:20, 3.86s/it] {'loss': 1.2184, 'grad_norm': 0.003815120357735909, 'learning_rate': 0.02420282768545469, 'epoch': 0.78} + 78%|███████▊ | 406/520 [55:12<07:20, 3.86s/it] 78%|███████▊ | 407/520 [55:16<07:15, 3.86s/it] {'loss': 1.3277, 'grad_norm': 0.00324362912516234, 'learning_rate': 0.02379771129969892, 'epoch': 0.78} + 78%|███████▊ | 407/520 [55:16<07:15, 3.86s/it] 78%|███████▊ | 408/520 [55:20<07:13, 3.87s/it] {'loss': 1.2142, 'grad_norm': 0.0032960826257728043, 'learning_rate': 0.023395555688102213, 'epoch': 0.78} + 78%|███████▊ | 408/520 [55:20<07:13, 3.87s/it] 79%|███████▊ | 409/520 [55:24<07:08, 3.86s/it] {'loss': 1.3453, 'grad_norm': 0.0035935061996262164, 'learning_rate': 0.02299637647607372, 'epoch': 0.79} + 79%|███████▊ | 409/520 [55:24<07:08, 3.86s/it] 79%|███████▉ | 410/520 [55:28<07:05, 3.86s/it] {'loss': 1.0604, 'grad_norm': 0.003170001288754671, 'learning_rate': 0.022600189173377264, 'epoch': 0.79} + 79%|███████▉ | 410/520 [55:28<07:05, 3.86s/it] 79%|███████▉ | 411/520 [55:32<07:01, 3.87s/it] {'loss': 1.3118, 'grad_norm': 0.0034483393820476716, 'learning_rate': 0.022207009173528525, 'epoch': 0.79} + 79%|███████▉ | 411/520 [55:32<07:01, 3.87s/it] 79%|███████▉ | 412/520 [55:35<06:56, 3.86s/it] {'loss': 1.2319, 'grad_norm': 0.0031611383096917765, 'learning_rate': 0.02181685175319702, 'epoch': 0.79} + 79%|███████▉ | 412/520 [55:35<06:56, 3.86s/it] 79%|███████▉ | 413/520 [55:39<06:51, 3.85s/it] {'loss': 1.3129, 'grad_norm': 0.0031743118726295996, 'learning_rate': 0.021429732071612653, 'epoch': 0.79} + 79%|███████▉ | 413/520 [55:39<06:51, 3.85s/it] 80%|███████▉ | 414/520 [55:43<06:42, 3.80s/it] {'loss': 1.1028, 'grad_norm': 0.002899140789359117, 'learning_rate': 0.02104566516997647, 'epoch': 0.8} + 80%|███████▉ | 414/520 [55:43<06:42, 3.80s/it] 80%|███████▉ | 415/520 [55:47<06:34, 3.76s/it] {'loss': 1.2077, 'grad_norm': 0.0031001264022983933, 'learning_rate': 0.020664665970876496, 'epoch': 0.8} + 80%|███████▉ | 415/520 [55:47<06:34, 3.76s/it] 80%|████████ | 416/520 [55:50<06:28, 3.73s/it] {'loss': 1.1232, 'grad_norm': 0.003534341372636104, 'learning_rate': 0.020286749277707784, 'epoch': 0.8} + 80%|████████ | 416/520 [55:50<06:28, 3.73s/it] 80%|████████ | 417/520 [55:54<06:24, 3.73s/it] {'loss': 1.2969, 'grad_norm': 0.003578855814793782, 'learning_rate': 0.019911929774097215, 'epoch': 0.8} + 80%|████████ | 417/520 [55:54<06:24, 3.73s/it] 80%|████████ | 418/520 [55:58<06:18, 3.71s/it] {'loss': 1.2734, 'grad_norm': 0.002941943033815469, 'learning_rate': 0.019540222023333165, 'epoch': 0.8} + 80%|████████ | 418/520 [55:58<06:18, 3.71s/it] 81%|████████ | 419/520 [56:01<06:13, 3.70s/it] {'loss': 1.2575, 'grad_norm': 0.0034267512532424314, 'learning_rate': 0.01917164046779948, 'epoch': 0.81} + 81%|████████ | 419/520 [56:01<06:13, 3.70s/it] 81%|████████ | 420/520 [56:05<06:09, 3.69s/it] {'loss': 1.1447, 'grad_norm': 0.0033148159576007045, 'learning_rate': 0.018806199428414352, 'epoch': 0.81} + 81%|████████ | 420/520 [56:05<06:09, 3.69s/it] 81%|████████ | 421/520 [56:09<06:05, 3.69s/it] {'loss': 1.0727, 'grad_norm': 0.003323896304599482, 'learning_rate': 0.018443913104073985, 'epoch': 0.81} + 81%|████████ | 421/520 [56:09<06:05, 3.69s/it] 81%|████████ | 422/520 [56:12<06:00, 3.68s/it] {'loss': 1.2001, 'grad_norm': 0.0034134797770804743, 'learning_rate': 0.01808479557110081, 'epoch': 0.81} + 81%|████████ | 422/520 [56:12<06:00, 3.68s/it] 81%|████████▏ | 423/520 [56:16<05:55, 3.67s/it] {'loss': 1.1903, 'grad_norm': 0.003739042248627091, 'learning_rate': 0.017728860782696667, 'epoch': 0.81} + 81%|████████▏ | 423/520 [56:16<05:55, 3.67s/it] 82%|████████▏ | 424/520 [56:20<05:53, 3.68s/it] {'loss': 1.3996, 'grad_norm': 0.0034830154043275336, 'learning_rate': 0.017376122568400532, 'epoch': 0.82} + 82%|████████▏ | 424/520 [56:20<05:53, 3.68s/it] 82%|████████▏ | 425/520 [56:23<05:49, 3.68s/it] {'loss': 1.1959, 'grad_norm': 0.003058141965792347, 'learning_rate': 0.017026594633551252, 'epoch': 0.82} + 82%|████████▏ | 425/520 [56:23<05:49, 3.68s/it] 82%|████████▏ | 426/520 [56:27<05:45, 3.67s/it] {'loss': 1.2278, 'grad_norm': 0.004198254137402504, 'learning_rate': 0.01668029055875512, 'epoch': 0.82} + 82%|████████▏ | 426/520 [56:27<05:45, 3.67s/it] 82%|████████▏ | 427/520 [56:31<05:41, 3.67s/it] {'loss': 1.1291, 'grad_norm': 0.0030285184957581698, 'learning_rate': 0.016337223799358026, 'epoch': 0.82} + 82%|████████▏ | 427/520 [56:31<05:41, 3.67s/it] 82%|████████▏ | 428/520 [56:34<05:37, 3.67s/it] {'loss': 1.1086, 'grad_norm': 0.003186408197371642, 'learning_rate': 0.01599740768492286, 'epoch': 0.82} + 82%|████████▏ | 428/520 [56:34<05:37, 3.67s/it] 82%|████████▎ | 429/520 [56:38<05:34, 3.68s/it] {'loss': 1.2131, 'grad_norm': 0.0030650194980444898, 'learning_rate': 0.015660855418711452, 'epoch': 0.82} + 82%|████████▎ | 429/520 [56:38<05:34, 3.68s/it]Token indices sequence length is longer than the specified maximum sequence length for this model (2076 > 2048). Running this sequence through the model will result in indexing errors + 83%|████████▎ | 430/520 [56:42<05:31, 3.68s/it] {'loss': 1.2055, 'grad_norm': 0.002982861677956146, 'learning_rate': 0.015327580077171589, 'epoch': 0.83} + 83%|████████▎ | 430/520 [56:42<05:31, 3.68s/it] 83%|████████▎ | 431/520 [56:45<05:28, 3.69s/it] {'loss': 1.2796, 'grad_norm': 0.003405974211330956, 'learning_rate': 0.014997594609429088, 'epoch': 0.83} + 83%|████████▎ | 431/520 [56:45<05:28, 3.69s/it] 83%|████████▎ | 432/520 [56:49<05:25, 3.70s/it] {'loss': 1.1157, 'grad_norm': 0.0035487540147491484, 'learning_rate': 0.01467091183678444, 'epoch': 0.83} + 83%|████████▎ | 432/520 [56:49<05:25, 3.70s/it] 83%|████████▎ | 433/520 [56:53<05:20, 3.69s/it] {'loss': 1.2526, 'grad_norm': 0.003319907212154964, 'learning_rate': 0.014347544452214867, 'epoch': 0.83} + 83%|████████▎ | 433/520 [56:53<05:20, 3.69s/it] 83%|████████▎ | 434/520 [56:56<05:17, 3.69s/it] {'loss': 0.9933, 'grad_norm': 0.003066445373372604, 'learning_rate': 0.014027505019880971, 'epoch': 0.83} + 83%|████████▎ | 434/520 [56:57<05:17, 3.69s/it] 84%|████████▎ | 435/520 [57:00<05:12, 3.68s/it] {'loss': 1.2842, 'grad_norm': 0.0036045473347932405, 'learning_rate': 0.013710805974638696, 'epoch': 0.84} + 84%|████████▎ | 435/520 [57:00<05:12, 3.68s/it] 84%|████████▍ | 436/520 [57:04<05:11, 3.71s/it] {'loss': 1.0806, 'grad_norm': 0.003222194108981431, 'learning_rate': 0.01339745962155613, 'epoch': 0.84} + 84%|████████▍ | 436/520 [57:04<05:11, 3.71s/it] 84%|████████▍ | 437/520 [57:08<05:10, 3.74s/it] {'loss': 1.3155, 'grad_norm': 0.003149893152797403, 'learning_rate': 0.01308747813543536, 'epoch': 0.84} + 84%|████████▍ | 437/520 [57:08<05:10, 3.74s/it] 84%|████████▍ | 438/520 [57:12<05:09, 3.77s/it] {'loss': 1.1158, 'grad_norm': 0.0030417436598169107, 'learning_rate': 0.012780873560339467, 'epoch': 0.84} + 84%|████████▍ | 438/520 [57:12<05:09, 3.77s/it] 84%|████████▍ | 439/520 [57:15<05:05, 3.78s/it] {'loss': 1.2426, 'grad_norm': 0.0026483818320440114, 'learning_rate': 0.012477657809124632, 'epoch': 0.84} + 84%|████████▍ | 439/520 [57:15<05:05, 3.78s/it] 85%|████████▍ | 440/520 [57:19<05:03, 3.79s/it] {'loss': 1.1696, 'grad_norm': 0.003129196714593542, 'learning_rate': 0.012177842662977134, 'epoch': 0.85} + 85%|████████▍ | 440/520 [57:19<05:03, 3.79s/it] 85%|████████▍ | 441/520 [57:23<05:00, 3.80s/it] {'loss': 1.2736, 'grad_norm': 0.0032116937925355444, 'learning_rate': 0.01188143977095576, 'epoch': 0.85} + 85%|████████▍ | 441/520 [57:23<05:00, 3.80s/it] 85%|████████▌ | 442/520 [57:27<04:52, 3.75s/it] {'loss': 1.2212, 'grad_norm': 0.0036712251911801123, 'learning_rate': 0.011588460649539035, 'epoch': 0.85} + 85%|████████▌ | 442/520 [57:27<04:52, 3.75s/it] 85%|████████▌ | 443/520 [57:30<04:47, 3.73s/it] {'loss': 1.2431, 'grad_norm': 0.0033074445352635237, 'learning_rate': 0.011298916682177829, 'epoch': 0.85} + 85%|████████▌ | 443/520 [57:30<04:47, 3.73s/it] 85%|████████▌ | 444/520 [57:34<04:42, 3.72s/it] {'loss': 1.2127, 'grad_norm': 0.002876166608583892, 'learning_rate': 0.011012819118853146, 'epoch': 0.85} + 85%|████████▌ | 444/520 [57:34<04:42, 3.72s/it] 86%|████████▌ | 445/520 [57:38<04:37, 3.70s/it] {'loss': 1.1265, 'grad_norm': 0.003098558246981894, 'learning_rate': 0.01073017907563887, 'epoch': 0.86} + 86%|████████▌ | 445/520 [57:38<04:37, 3.70s/it] 86%|████████▌ | 446/520 [57:41<04:33, 3.70s/it] {'loss': 1.3514, 'grad_norm': 0.0030224825682291377, 'learning_rate': 0.010451007534269908, 'epoch': 0.86} + 86%|████████▌ | 446/520 [57:41<04:33, 3.70s/it] 86%|████████▌ | 447/520 [57:45<04:30, 3.70s/it] {'loss': 1.2172, 'grad_norm': 0.0031065149592809788, 'learning_rate': 0.010175315341715598, 'epoch': 0.86} + 86%|████████▌ | 447/520 [57:45<04:30, 3.70s/it] 86%|████████▌ | 448/520 [57:49<04:26, 3.69s/it] {'loss': 1.1936, 'grad_norm': 0.0033872634922416486, 'learning_rate': 0.009903113209758098, 'epoch': 0.86} + 86%|████████▌ | 448/520 [57:49<04:26, 3.69s/it] 86%|████████▋ | 449/520 [57:52<04:21, 3.69s/it] {'loss': 1.3117, 'grad_norm': 0.0032493675013364827, 'learning_rate': 0.009634411714576352, 'epoch': 0.86} + 86%|████████▋ | 449/520 [57:52<04:21, 3.69s/it] 87%|████████▋ | 450/520 [57:56<04:18, 3.69s/it] {'loss': 1.2376, 'grad_norm': 0.0031376127703661586, 'learning_rate': 0.009369221296335007, 'epoch': 0.87} + 87%|████████▋ | 450/520 [57:56<04:18, 3.69s/it] 87%|████████▋ | 451/520 [58:00<04:15, 3.71s/it] {'loss': 1.2327, 'grad_norm': 0.0032178667804046174, 'learning_rate': 0.009107552258778906, 'epoch': 0.87} + 87%|████████▋ | 451/520 [58:00<04:15, 3.71s/it] 87%|████████▋ | 452/520 [58:04<04:12, 3.71s/it] {'loss': 1.3379, 'grad_norm': 0.0030831995672261528, 'learning_rate': 0.008849414768832687, 'epoch': 0.87} + 87%|████████▋ | 452/520 [58:04<04:12, 3.71s/it] 87%|████████▋ | 453/520 [58:07<04:08, 3.71s/it] {'loss': 1.3155, 'grad_norm': 0.0032680630305238333, 'learning_rate': 0.008594818856205699, 'epoch': 0.87} + 87%|████████▋ | 453/520 [58:07<04:08, 3.71s/it] 87%|████████▋ | 454/520 [58:11<04:04, 3.71s/it] {'loss': 1.1438, 'grad_norm': 0.0034273397778481814, 'learning_rate': 0.00834377441300238, 'epoch': 0.87} + 87%|████████▋ | 454/520 [58:11<04:04, 3.71s/it] 88%|████████▊ | 455/520 [58:15<04:00, 3.70s/it] {'loss': 1.2791, 'grad_norm': 0.003171348983518716, 'learning_rate': 0.008096291193337934, 'epoch': 0.88} + 88%|████████▊ | 455/520 [58:15<04:00, 3.70s/it] 88%|████████▊ | 456/520 [58:18<03:56, 3.70s/it] {'loss': 1.1934, 'grad_norm': 0.003140317464477926, 'learning_rate': 0.007852378812959226, 'epoch': 0.88} + 88%|████████▊ | 456/520 [58:18<03:56, 3.70s/it] 88%|████████▊ | 457/520 [58:22<03:53, 3.71s/it] {'loss': 1.3, 'grad_norm': 0.0029258653688962882, 'learning_rate': 0.007612046748871327, 'epoch': 0.88} + 88%|████████▊ | 457/520 [58:22<03:53, 3.71s/it] 88%|████████▊ | 458/520 [58:26<03:51, 3.73s/it] {'loss': 1.3435, 'grad_norm': 0.003351254168327238, 'learning_rate': 0.007375304338969136, 'epoch': 0.88} + 88%|████████▊ | 458/520 [58:26<03:51, 3.73s/it] 88%|████████▊ | 459/520 [58:30<03:47, 3.73s/it] {'loss': 1.2717, 'grad_norm': 0.0031077674112747872, 'learning_rate': 0.007142160781674645, 'epoch': 0.88} + 88%|████████▊ | 459/520 [58:30<03:47, 3.73s/it] 88%|████████▊ | 460/520 [58:33<03:43, 3.72s/it] {'loss': 1.1376, 'grad_norm': 0.002994622714699063, 'learning_rate': 0.006912625135579587, 'epoch': 0.88} + 88%|████████▊ | 460/520 [58:33<03:43, 3.72s/it] 89%|████████▊ | 461/520 [58:37<03:39, 3.72s/it] {'loss': 1.3702, 'grad_norm': 0.0026341137135090773, 'learning_rate': 0.0066867063190933496, 'epoch': 0.89} + 89%|████████▊ | 461/520 [58:37<03:39, 3.72s/it] 89%|████████▉ | 462/520 [58:41<03:35, 3.72s/it] {'loss': 1.3882, 'grad_norm': 0.0030549796646143854, 'learning_rate': 0.006464413110096601, 'epoch': 0.89} + 89%|████████▉ | 462/520 [58:41<03:35, 3.72s/it] 89%|████████▉ | 463/520 [58:44<03:32, 3.72s/it] {'loss': 1.1089, 'grad_norm': 0.003387908686337034, 'learning_rate': 0.006245754145600091, 'epoch': 0.89} + 89%|████████▉ | 463/520 [58:44<03:32, 3.72s/it] 89%|████████▉ | 464/520 [58:48<03:28, 3.72s/it] {'loss': 1.2556, 'grad_norm': 0.003283797879911061, 'learning_rate': 0.006030737921409169, 'epoch': 0.89} + 89%|████████▉ | 464/520 [58:48<03:28, 3.72s/it] 89%|████████▉ | 465/520 [58:52<03:23, 3.70s/it] {'loss': 1.3736, 'grad_norm': 0.003732756348592813, 'learning_rate': 0.005819372791793654, 'epoch': 0.89} + 89%|████████▉ | 465/520 [58:52<03:23, 3.70s/it] 90%|████████▉ | 466/520 [58:56<03:20, 3.71s/it] {'loss': 1.2348, 'grad_norm': 0.0028398000848346406, 'learning_rate': 0.005611666969163243, 'epoch': 0.9} + 90%|████████▉ | 466/520 [58:56<03:20, 3.71s/it] 90%|████████▉ | 467/520 [58:59<03:16, 3.71s/it] {'loss': 1.2741, 'grad_norm': 0.0029474782178999345, 'learning_rate': 0.005407628523748398, 'epoch': 0.9} + 90%|████████▉ | 467/520 [58:59<03:16, 3.71s/it] 90%|█████████ | 468/520 [59:03<03:12, 3.71s/it] {'loss': 1.2155, 'grad_norm': 0.003544305444648752, 'learning_rate': 0.00520726538328683, 'epoch': 0.9} + 90%|█████████ | 468/520 [59:03<03:12, 3.71s/it] 90%|█████████ | 469/520 [59:07<03:08, 3.70s/it] {'loss': 1.2713, 'grad_norm': 0.00337567015246457, 'learning_rate': 0.005010585332715401, 'epoch': 0.9} + 90%|█████████ | 469/520 [59:07<03:08, 3.70s/it] 90%|█████████ | 470/520 [59:10<03:05, 3.71s/it] {'loss': 1.1493, 'grad_norm': 0.002797792223453487, 'learning_rate': 0.004817596013867765, 'epoch': 0.9} + 90%|█████████ | 470/520 [59:10<03:05, 3.71s/it] 91%|█████████ | 471/520 [59:14<03:01, 3.71s/it] {'loss': 1.1671, 'grad_norm': 0.0031334483425514625, 'learning_rate': 0.004628304925177318, 'epoch': 0.91} + 91%|█████████ | 471/520 [59:14<03:01, 3.71s/it] 91%|█████████ | 472/520 [59:18<02:58, 3.72s/it] {'loss': 1.1428, 'grad_norm': 0.0032368442450379786, 'learning_rate': 0.004442719421385921, 'epoch': 0.91} + 91%|█████████ | 472/520 [59:18<02:58, 3.72s/it] 91%|█████████ | 473/520 [59:22<02:54, 3.72s/it] {'loss': 1.1908, 'grad_norm': 0.0031790841847252798, 'learning_rate': 0.004260846713258193, 'epoch': 0.91} + 91%|█████████ | 473/520 [59:22<02:54, 3.72s/it] 91%|█████████ | 474/520 [59:25<02:51, 3.73s/it] {'loss': 1.3076, 'grad_norm': 0.0029507539499784856, 'learning_rate': 0.004082693867301224, 'epoch': 0.91} + 91%|█████████ | 474/520 [59:25<02:51, 3.73s/it] 91%|█████████▏| 475/520 [59:29<02:47, 3.72s/it] {'loss': 1.2177, 'grad_norm': 0.003010499069307706, 'learning_rate': 0.003908267805490051, 'epoch': 0.91} + 91%|█████████▏| 475/520 [59:29<02:47, 3.72s/it] 92%|█████████▏| 476/520 [59:33<02:43, 3.71s/it] {'loss': 1.1971, 'grad_norm': 0.0032754759790251686, 'learning_rate': 0.003737575304998797, 'epoch': 0.92} + 92%|█████████▏| 476/520 [59:33<02:43, 3.71s/it] 92%|█████████▏| 477/520 [59:36<02:39, 3.71s/it] {'loss': 1.178, 'grad_norm': 0.003577375350157732, 'learning_rate': 0.003570622997937234, 'epoch': 0.92} + 92%|█████████▏| 477/520 [59:36<02:39, 3.71s/it] 92%|█████████▏| 478/520 [59:40<02:35, 3.71s/it] {'loss': 1.1427, 'grad_norm': 0.0030011936432154263, 'learning_rate': 0.00340741737109318, 'epoch': 0.92} + 92%|█████████▏| 478/520 [59:40<02:35, 3.71s/it] 92%|█████████▏| 479/520 [59:44<02:31, 3.71s/it] {'loss': 1.2821, 'grad_norm': 0.003434746323204659, 'learning_rate': 0.003247964765680389, 'epoch': 0.92} + 92%|█████████▏| 479/520 [59:44<02:31, 3.71s/it] 92%|█████████▏| 480/520 [59:48<02:28, 3.72s/it] {'loss': 1.3103, 'grad_norm': 0.0030643166507188066, 'learning_rate': 0.0030922713770922153, 'epoch': 0.92} + 92%|█████████▏| 480/520 [59:48<02:28, 3.72s/it] 92%|█████████▎| 481/520 [59:51<02:24, 3.71s/it] {'loss': 1.3217, 'grad_norm': 0.0029423703824462454, 'learning_rate': 0.0029403432546609046, 'epoch': 0.93} + 92%|█████████▎| 481/520 [59:51<02:24, 3.71s/it] 93%|█████████▎| 482/520 [59:55<02:20, 3.70s/it] {'loss': 1.3183, 'grad_norm': 0.003109644122724853, 'learning_rate': 0.0027921863014225504, 'epoch': 0.93} + 93%|█████████▎| 482/520 [59:55<02:20, 3.70s/it] 93%|█████████▎| 483/520 [59:59<02:17, 3.72s/it] {'loss': 1.2174, 'grad_norm': 0.00334596254513209, 'learning_rate': 0.002647806273887665, 'epoch': 0.93} + 93%|█████████▎| 483/520 [59:59<02:17, 3.72s/it] 93%|█████████▎| 484/520 [1:00:02<02:14, 3.74s/it] {'loss': 1.2167, 'grad_norm': 0.003156399025683408, 'learning_rate': 0.0025072087818176383, 'epoch': 0.93} + 93%|█████████▎| 484/520 [1:00:02<02:14, 3.74s/it] 93%|█████████▎| 485/520 [1:00:06<02:10, 3.72s/it] {'loss': 1.1608, 'grad_norm': 0.003037836458931493, 'learning_rate': 0.002370399288006664, 'epoch': 0.93} + 93%|█████████▎| 485/520 [1:00:06<02:10, 3.72s/it] 93%|█████████▎| 486/520 [1:00:10<02:05, 3.71s/it] {'loss': 1.2841, 'grad_norm': 0.0032774994642006785, 'learning_rate': 0.0022373831080695463, 'epoch': 0.93} + 93%|█████████▎| 486/520 [1:00:10<02:05, 3.71s/it] 94%|█████████▎| 487/520 [1:00:14<02:03, 3.73s/it] {'loss': 1.136, 'grad_norm': 0.0029625664906410094, 'learning_rate': 0.0021081654102351635, 'epoch': 0.94} + 94%|█████████▎| 487/520 [1:00:14<02:03, 3.73s/it] 94%|█████████▍| 488/520 [1:00:17<02:00, 3.76s/it] {'loss': 1.0791, 'grad_norm': 0.0032330828765428764, 'learning_rate': 0.0019827512151456175, 'epoch': 0.94} + 94%|█████████▍| 488/520 [1:00:17<02:00, 3.76s/it] 94%|█████████▍| 489/520 [1:00:21<01:56, 3.75s/it] {'loss': 1.2969, 'grad_norm': 0.0027039527705939558, 'learning_rate': 0.0018611453956612345, 'epoch': 0.94} + 94%|█████████▍| 489/520 [1:00:21<01:56, 3.75s/it] 94%|█████████▍| 490/520 [1:00:25<01:51, 3.72s/it] {'loss': 1.2061, 'grad_norm': 0.003178582570346949, 'learning_rate': 0.0017433526766711727, 'epoch': 0.94} + 94%|█████████▍| 490/520 [1:00:25<01:51, 3.72s/it] 94%|█████████▍| 491/520 [1:00:29<01:47, 3.72s/it] {'loss': 1.1648, 'grad_norm': 0.0031709509507533173, 'learning_rate': 0.0016293776349098677, 'epoch': 0.94} + 94%|█████████▍| 491/520 [1:00:29<01:47, 3.72s/it] 95%|█████████▍| 492/520 [1:00:32<01:43, 3.71s/it] {'loss': 1.2895, 'grad_norm': 0.00333051435380077, 'learning_rate': 0.0015192246987791981, 'epoch': 0.95} + 95%|█████████▍| 492/520 [1:00:32<01:43, 3.71s/it] 95%|█████████▍| 493/520 [1:00:36<01:40, 3.73s/it] {'loss': 1.3547, 'grad_norm': 0.0032372870532858277, 'learning_rate': 0.0014128981481764114, 'epoch': 0.95} + 95%|█████████▍| 493/520 [1:00:36<01:40, 3.73s/it] 95%|█████████▌| 494/520 [1:00:40<01:38, 3.77s/it] {'loss': 1.22, 'grad_norm': 0.0029095243525324474, 'learning_rate': 0.0013104021143278911, 'epoch': 0.95} + 95%|█████████▌| 494/520 [1:00:40<01:38, 3.77s/it] 95%|█████████▌| 495/520 [1:00:44<01:34, 3.79s/it] {'loss': 1.1824, 'grad_norm': 0.0030631733690279824, 'learning_rate': 0.0012117405796285285, 'epoch': 0.95} + 95%|█████████▌| 495/520 [1:00:44<01:34, 3.79s/it] 95%|█████████▌| 496/520 [1:00:48<01:31, 3.80s/it] {'loss': 1.1001, 'grad_norm': 0.0031731301344953985, 'learning_rate': 0.0011169173774871477, 'epoch': 0.95} + 95%|█████████▌| 496/520 [1:00:48<01:31, 3.80s/it] 96%|█████████▌| 497/520 [1:00:51<01:27, 3.81s/it] {'loss': 1.2385, 'grad_norm': 0.002819912720891087, 'learning_rate': 0.0010259361921774012, 'epoch': 0.96} + 96%|█████████▌| 497/520 [1:00:51<01:27, 3.81s/it] 96%|█████████▌| 498/520 [1:00:55<01:23, 3.82s/it] {'loss': 1.1831, 'grad_norm': 0.0032359433888995838, 'learning_rate': 0.000938800558694719, 'epoch': 0.96} + 96%|█████████▌| 498/520 [1:00:55<01:23, 3.82s/it] 96%|█████████▌| 499/520 [1:00:59<01:20, 3.83s/it] {'loss': 1.3748, 'grad_norm': 0.003102383008663707, 'learning_rate': 0.0008555138626189618, 'epoch': 0.96} + 96%|█████████▌| 499/520 [1:00:59<01:20, 3.83s/it] 96%|█████████▌| 500/520 [1:01:03<01:16, 3.84s/it] {'loss': 1.305, 'grad_norm': 0.0035894047528910525, 'learning_rate': 0.0007760793399827937, 'epoch': 0.96} + 96%|█████████▌| 500/520 [1:01:03<01:16, 3.84s/it] 96%|█████████▋| 501/520 [1:01:07<01:12, 3.84s/it] {'loss': 1.3284, 'grad_norm': 0.005017204191182442, 'learning_rate': 0.000700500077146038, 'epoch': 0.96} + 96%|█████████▋| 501/520 [1:01:07<01:12, 3.84s/it] 97%|█████████▋| 502/520 [1:01:11<01:09, 3.84s/it] {'loss': 1.2147, 'grad_norm': 0.0029618065269610156, 'learning_rate': 0.0006287790106757397, 'epoch': 0.97} + 97%|█████████▋| 502/520 [1:01:11<01:09, 3.84s/it] 97%|█████████▋| 503/520 [1:01:14<01:05, 3.83s/it] {'loss': 1.266, 'grad_norm': 0.0032014013858814483, 'learning_rate': 0.0005609189272320237, 'epoch': 0.97} + 97%|█████████▋| 503/520 [1:01:14<01:05, 3.83s/it] 97%|█████████▋| 504/520 [1:01:18<01:01, 3.84s/it] {'loss': 1.2152, 'grad_norm': 0.003645078870951587, 'learning_rate': 0.000496922463459859, 'epoch': 0.97} + 97%|█████████▋| 504/520 [1:01:18<01:01, 3.84s/it] 97%|█████████▋| 505/520 [1:01:22<00:57, 3.85s/it] {'loss': 1.2597, 'grad_norm': 0.0031549334916656417, 'learning_rate': 0.0004367921058866187, 'epoch': 0.97} + 97%|█████████▋| 505/520 [1:01:22<00:57, 3.85s/it] 97%|█████████▋| 506/520 [1:01:26<00:53, 3.84s/it] {'loss': 1.1751, 'grad_norm': 0.003387894223172485, 'learning_rate': 0.0003805301908254455, 'epoch': 0.97} + 97%|█████████▋| 506/520 [1:01:26<00:53, 3.84s/it] 98%|█████████▊| 507/520 [1:01:30<00:50, 3.85s/it] {'loss': 1.4179, 'grad_norm': 0.003064695336747645, 'learning_rate': 0.0003281389042844918, 'epoch': 0.97} + 98%|█████████▊| 507/520 [1:01:30<00:50, 3.85s/it] 98%|█████████▊| 508/520 [1:01:34<00:46, 3.84s/it] {'loss': 1.287, 'grad_norm': 0.0031835783338036777, 'learning_rate': 0.00027962028188198705, 'epoch': 0.98} + 98%|█████████▊| 508/520 [1:01:34<00:46, 3.84s/it] 98%|█████████▊| 509/520 [1:01:37<00:42, 3.83s/it] {'loss': 1.2586, 'grad_norm': 0.0030307886738463186, 'learning_rate': 0.00023497620876711256, 'epoch': 0.98} + 98%|█████████▊| 509/520 [1:01:37<00:42, 3.83s/it] 98%|█████████▊| 510/520 [1:01:41<00:38, 3.83s/it] {'loss': 1.2176, 'grad_norm': 0.003056546804570777, 'learning_rate': 0.00019420841954681523, 'epoch': 0.98} + 98%|█████████▊| 510/520 [1:01:41<00:38, 3.83s/it] 98%|█████████▊| 511/520 [1:01:45<00:34, 3.82s/it] {'loss': 1.1876, 'grad_norm': 0.0029828982392071397, 'learning_rate': 0.00015731849821833956, 'epoch': 0.98} + 98%|█████████▊| 511/520 [1:01:45<00:34, 3.82s/it] 98%|█████████▊| 512/520 [1:01:49<00:30, 3.82s/it] {'loss': 1.0686, 'grad_norm': 0.002999816461991049, 'learning_rate': 0.00012430787810776555, 'epoch': 0.98} + 98%|█████████▊| 512/520 [1:01:49<00:30, 3.82s/it] 99%|█████████▊| 513/520 [1:01:53<00:26, 3.83s/it] {'loss': 1.2742, 'grad_norm': 0.0033435136023005707, 'learning_rate': 9.517784181422018e-05, 'epoch': 0.99} + 99%|█████████▊| 513/520 [1:01:53<00:26, 3.83s/it] 99%|█████████▉| 514/520 [1:01:57<00:22, 3.81s/it] {'loss': 1.2461, 'grad_norm': 0.0028376035365173156, 'learning_rate': 6.992952116013917e-05, 'epoch': 0.99} + 99%|█████████▉| 514/520 [1:01:57<00:22, 3.81s/it] 99%|█████████▉| 515/520 [1:02:00<00:18, 3.80s/it] {'loss': 1.2972, 'grad_norm': 0.0037109765104748876, 'learning_rate': 4.856389714723575e-05, 'epoch': 0.99} + 99%|█████████▉| 515/520 [1:02:00<00:18, 3.80s/it] 99%|█████████▉| 516/520 [1:02:04<00:15, 3.80s/it] {'loss': 1.1842, 'grad_norm': 0.003061884101549843, 'learning_rate': 3.108179991837545e-05, 'epoch': 0.99} + 99%|█████████▉| 516/520 [1:02:04<00:15, 3.80s/it] 99%|█████████▉| 517/520 [1:02:08<00:11, 3.79s/it] {'loss': 1.3271, 'grad_norm': 0.003031045543182312, 'learning_rate': 1.7483908725357544e-05, 'epoch': 0.99} + 99%|█████████▉| 517/520 [1:02:08<00:11, 3.79s/it] 100%|█████████▉| 518/520 [1:02:12<00:07, 3.79s/it] {'loss': 1.2035, 'grad_norm': 0.0031214401451076412, 'learning_rate': 7.770751902513862e-06, 'epoch': 1.0} + 100%|█████████▉| 518/520 [1:02:12<00:07, 3.79s/it] 100%|█████████▉| 519/520 [1:02:15<00:03, 3.79s/it] {'loss': 1.2679, 'grad_norm': 0.0031816485121019798, 'learning_rate': 1.9427068461808086e-06, 'epoch': 1.0} + 100%|█████████▉| 519/520 [1:02:15<00:03, 3.79s/it] 100%|██████████| 520/520 [1:02:20<00:00, 4.05s/it] {'loss': 1.3424, 'grad_norm': 0.003159642858614262, 'learning_rate': 0.0, 'epoch': 1.0} + 100%|██████████| 520/520 [1:02:20<00:00, 4.05s/it] {'train_runtime': 3740.6296, 'train_samples_per_second': 17.786, 'train_steps_per_second': 0.139, 'train_loss': 1.4996101328959832, 'epoch': 1.0} + 100%|██████████| 520/520 [1:02:20<00:00, 4.05s/it] 100%|██████████| 520/520 [1:02:20<00:00, 7.19s/it] +[2025-10-12 11:36:21,266] [INFO] [launch.py:348:main] Process 325947 exits successfully. +[2025-10-12 11:36:22,268] [INFO] [launch.py:348:main] Process 325951 exits successfully. +[2025-10-12 11:36:22,268] [INFO] [launch.py:348:main] Process 325945 exits successfully. +[2025-10-12 11:36:22,269] [INFO] [launch.py:348:main] Process 325946 exits successfully. +[2025-10-12 11:36:23,270] [INFO] [launch.py:348:main] Process 325949 exits successfully. +[2025-10-12 11:36:23,271] [INFO] [launch.py:348:main] Process 325950 exits successfully. +[2025-10-12 11:36:23,271] [INFO] [launch.py:348:main] Process 325948 exits successfully. +[2025-10-12 11:36:26,275] [INFO] [launch.py:348:main] Process 325944 exits successfully. +==== EXPERIMENT COMPLETED: qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.7_2e-1_connector-1.0_2.7_2e-1_ablation ==== +Log File: qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.7_2e-1_connector-1.0_2.7_2e-1_ablation_20251012_100457.log +Timestamp: 2025-10-12 11:36:28 +===================================== diff --git a/logs_oct11/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.9_2e-1_connector-1.0_2.9_2e-1_ablation_20251012_113628.log b/logs_oct11/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.9_2e-1_connector-1.0_2.9_2e-1_ablation_20251012_113628.log new file mode 100644 index 0000000000000000000000000000000000000000..783a20ba28880c365fb3b046d0013251cad16761 --- /dev/null +++ b/logs_oct11/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.9_2e-1_connector-1.0_2.9_2e-1_ablation_20251012_113628.log @@ -0,0 +1,1143 @@ +==== STARTING EXPERIMENT: qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.9_2e-1_connector-1.0_2.9_2e-1_ablation ==== +Log File: qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.9_2e-1_connector-1.0_2.9_2e-1_ablation_20251012_113628.log +Timestamp: 2025-10-12 11:36:28 +===================================== +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 11:36:31,634] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 11:36:35,301] [WARNING] [runner.py:202:fetch_hostfile] Unable to find hostfile, will proceed with training with local resources only. +[2025-10-12 11:36:35,303] [INFO] [runner.py:568:main] cmd = /opt/conda/envs/tinyllava/bin/python3.10 -u -m deepspeed.launcher.launch --world_info=eyJsb2NhbGhvc3QiOiBbMCwgMSwgMiwgMywgNCwgNSwgNiwgN119 --master_addr=127.0.0.1 --master_port=29501 --enable_each_rank_log=None tinyllava/train/train.py --deepspeed ./scripts/zero3.json --data_path /s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json --image_folder /s3-code/ywang29/datasets/tinyllava --is_multimodal True --conv_version qwen2_base --model_name_or_path Qwen/Qwen2.5-0.5B --vision_tower google/siglip-so400m-patch14-384 --vision_tower2 --connector_type mlp2x_gelu --mm_vision_select_layer -2 --image_aspect_ratio square --attn_implementation flash_attention_2 --bf16 True --training_recipe common --tune_type_llm full --tune_type_vision_tower frozen --tune_vision_tower_from_layer 0 --tune_type_connector full --group_by_modality_length True --pretrained_model_path /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain --output_dir /nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.9_2e-1_connector-1.0_2.9_2e-1_ablation --num_train_epochs 1 --per_device_train_batch_size 4 --per_device_eval_batch_size 4 --gradient_accumulation_steps 4 --evaluation_strategy no --learning_rate 2e-1 --weight_decay 0. --warmup_ratio 0.03 --lr_scheduler_type cosine --logging_steps 1 --tf32 False --model_max_length 2048 --gradient_checkpointing True --dataloader_num_workers 8 --lazy_preprocess True --report_to tensorboard --tokenizer_use_fast False --run_name tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune --subnet_mode_text both --subnet_type_text None --mask_type_text soft --init_mean_text 1.0 --temperature_attn_text 2.9 --temperature_mlp_text 2.9 --backward_type_text normal --masked_layers_text all --subnet_mode_vision both --subnet_type_vision None --mask_type_vision soft --init_mean_vision 1.0 --temperature_attn_vision 2.9 --temperature_mlp_vision 2.9 --backward_type_vision normal --masked_layers_vision all --subnet_type_connector global --mask_type_connector soft --init_mean_connector 1.0 --temperature_connector 2.9 --backward_type_connector normal --mm_projector_lr 2e-1 --seed 42 --mask_model llm-connector --save_strategy steps --save_steps 50000 --save_total_limit 1 --train_data_ratio 0.1 +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 11:36:37,944] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 11:36:38,960] [INFO] [launch.py:138:main] 0 NCCL_VERSION=2.21.5 +[2025-10-12 11:36:38,960] [INFO] [launch.py:138:main] 0 NCCL_SOCKET_IFNAME=eth +[2025-10-12 11:36:38,960] [INFO] [launch.py:145:main] WORLD INFO DICT: {'localhost': [0, 1, 2, 3, 4, 5, 6, 7]} +[2025-10-12 11:36:38,960] [INFO] [launch.py:151:main] nnodes=1, num_local_procs=8, node_rank=0 +[2025-10-12 11:36:38,960] [INFO] [launch.py:162:main] global_rank_mapping=defaultdict(, {'localhost': [0, 1, 2, 3, 4, 5, 6, 7]}) +[2025-10-12 11:36:38,960] [INFO] [launch.py:163:main] dist_world_size=8 +[2025-10-12 11:36:38,960] [INFO] [launch.py:165:main] Setting CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 +[2025-10-12 11:36:38,963] [INFO] [launch.py:253:main] process 374420 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=0', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.9_2e-1_connector-1.0_2.9_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '1.0', '--temperature_attn_text', '2.9', '--temperature_mlp_text', '2.9', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '1.0', '--temperature_attn_vision', '2.9', '--temperature_mlp_vision', '2.9', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '1.0', '--temperature_connector', '2.9', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 11:36:38,965] [INFO] [launch.py:253:main] process 374421 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=1', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.9_2e-1_connector-1.0_2.9_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '1.0', '--temperature_attn_text', '2.9', '--temperature_mlp_text', '2.9', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '1.0', '--temperature_attn_vision', '2.9', '--temperature_mlp_vision', '2.9', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '1.0', '--temperature_connector', '2.9', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 11:36:38,967] [INFO] [launch.py:253:main] process 374422 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=2', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.9_2e-1_connector-1.0_2.9_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '1.0', '--temperature_attn_text', '2.9', '--temperature_mlp_text', '2.9', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '1.0', '--temperature_attn_vision', '2.9', '--temperature_mlp_vision', '2.9', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '1.0', '--temperature_connector', '2.9', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 11:36:38,969] [INFO] [launch.py:253:main] process 374423 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=3', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.9_2e-1_connector-1.0_2.9_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '1.0', '--temperature_attn_text', '2.9', '--temperature_mlp_text', '2.9', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '1.0', '--temperature_attn_vision', '2.9', '--temperature_mlp_vision', '2.9', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '1.0', '--temperature_connector', '2.9', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 11:36:38,971] [INFO] [launch.py:253:main] process 374424 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=4', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.9_2e-1_connector-1.0_2.9_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '1.0', '--temperature_attn_text', '2.9', '--temperature_mlp_text', '2.9', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '1.0', '--temperature_attn_vision', '2.9', '--temperature_mlp_vision', '2.9', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '1.0', '--temperature_connector', '2.9', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 11:36:38,974] [INFO] [launch.py:253:main] process 374425 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=5', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.9_2e-1_connector-1.0_2.9_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '1.0', '--temperature_attn_text', '2.9', '--temperature_mlp_text', '2.9', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '1.0', '--temperature_attn_vision', '2.9', '--temperature_mlp_vision', '2.9', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '1.0', '--temperature_connector', '2.9', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 11:36:38,976] [INFO] [launch.py:253:main] process 374426 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=6', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.9_2e-1_connector-1.0_2.9_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '1.0', '--temperature_attn_text', '2.9', '--temperature_mlp_text', '2.9', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '1.0', '--temperature_attn_vision', '2.9', '--temperature_mlp_vision', '2.9', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '1.0', '--temperature_connector', '2.9', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 11:36:38,978] [INFO] [launch.py:253:main] process 374427 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=7', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.9_2e-1_connector-1.0_2.9_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '1.0', '--temperature_attn_text', '2.9', '--temperature_mlp_text', '2.9', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '1.0', '--temperature_attn_vision', '2.9', '--temperature_mlp_vision', '2.9', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '1.0', '--temperature_connector', '2.9', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 11:36:45,806] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 11:36:45,806] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 11:36:45,806] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 11:36:45,806] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 11:36:45,806] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 11:36:45,806] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 11:36:45,816] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 11:36:45,816] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 11:36:46,366] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 11:36:46,366] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 11:36:46,366] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 11:36:46,366] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 11:36:46,366] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 11:36:46,366] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 11:36:46,366] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 11:36:46,366] [INFO] [comm.py:668:init_distributed] Initializing TorchBackend in DeepSpeed with backend nccl +[2025-10-12 11:36:46,366] [INFO] [comm.py:637:init_distributed] cdb=None +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +{'llm': {'model_name_or_path': 'Qwen/Qwen2.5-0.5B', 'cache_dir': None, 'attn_implementation': 'flash_attention_2', 'subnet_mode': 'both', 'subnet_type': 'None', 'sparsity_attn': None, 'sparsity_mlp': None, 'threshold_attn': None, 'threshold_mlp': None, 'temperature_attn': 2.9, 'temperature_mlp': 2.9, 'masked_layers': 'all', 'mask_type': 'soft', 'backward_type': 'normal'}, 'vision_tower': {'model_name_or_path': 'google/siglip-so400m-patch14-384'}, 'connector': {'connector_type': 'mlp2x_gelu', 'subnet_type': 'global', 'threshold': None, 'sparsity': None, 'temperature': 2.9, 'mask_type': 'soft', 'backward_type': 'normal'}} +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +TinyLlavaConfig { + "backward_type_connector": "normal", + "cache_dir": null, + "connector_type": "mlp2x_gelu", + "hidden_size": 896, + "ignore_index": -100, + "image_aspect_ratio": "square", + "image_token_index": -200, + "llm_model_name_or_path": "Qwen/Qwen2.5-0.5B", + "mask_model": [ + "llm", + "connector" + ], + "mask_type_connector": "soft", + "model_type": "tinyllava", + "num_queries": 128, + "num_resampler_layers": 3, + "pad_token": null, + "resampler_hidden_size": 768, + "sparsity_connector": null, + "subnet_type_connector": "global", + "temperature_connector": 2.9, + "text_config": { + "_name_or_path": "Qwen/Qwen2.5-0.5B", + "architectures": [ + "Qwen2ForCausalLM" + ], + "backward_type": "normal", + "bos_token_id": 151643, + "eos_token_id": 151643, + "hidden_size": 896, + "intermediate_size": 4864, + "mask_type": "soft", + "masked_layers": "all", + "max_position_embeddings": 32768, + "max_window_layers": 24, + "model_type": "qwen2", + "num_attention_heads": 14, + "num_hidden_layers": 24, + "num_key_value_heads": 2, + "rope_theta": 1000000.0, + "sliding_window": 32768, + "subnet_mode": "both", + "subnet_type": "None", + "temperature_attn": 2.9, + "temperature_mlp": 2.9, + "tie_word_embeddings": true, + "torch_dtype": "bfloat16", + "use_mrope": false, + "use_sliding_window": false, + "vocab_size": 151936 + }, + "threshold_connector": null, + "tokenizer_model_max_length": 2048, + "tokenizer_name_or_path": "Qwen/Qwen2.5-0.5B", + "tokenizer_padding_side": "right", + "tokenizer_use_fast": false, + "transformers_version": "4.40.1", + "tune_type_connector": "frozen", + "tune_type_llm": "frozen", + "tune_type_vision_tower": "frozen", + "tune_vision_tower_from_layer": -1, + "use_cache": false, + "vision_config": { + "hidden_act": "gelu_pytorch_tanh", + "hidden_size": 1152, + "image_size": 384, + "intermediate_size": 4304, + "layer_norm_eps": 1e-06, + "model_name_or_path": "google/siglip-so400m-patch14-384", + "model_name_or_path2": "", + "model_type": "siglip_vision_model", + "num_attention_heads": 16, + "num_hidden_layers": 27, + "patch_size": 14 + }, + "vision_feature_layer": -2, + "vision_feature_select_strategy": "patch", + "vision_hidden_size": 1152, + "vision_model_name_or_path": "google/siglip-so400m-patch14-384", + "vision_model_name_or_path2": "", + "vocab_size": 151936 +} + +Apply masks for the following modules: ['llm', 'connector'] +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test1-worker-0:374420:374420 [0] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:374420:374420 [0] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:374420:374420 [0] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:374420:374420 [0] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:374420:374420 [0] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test1-worker-0:374420:374420 [0] NCCL INFO cudaDriverVersion 12040 +NCCL version 2.21.5+cuda12.1 +ywang29-vrdb-test1-worker-0:374425:374425 [5] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test1-worker-0:374425:374425 [5] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:374425:374425 [5] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:374425:374425 [5] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:374425:374425 [5] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:374425:374425 [5] NCCL INFO NET/Plugin: Using internal network plugin. +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +ywang29-vrdb-test1-worker-0:374420:376057 [0] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:374420:376057 [0] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:374420:376057 [0] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:374420:376057 [0] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:374420:376057 [0] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:374420:376057 [0] NCCL INFO Using network Socket +ywang29-vrdb-test1-worker-0:374425:376058 [5] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:374425:376058 [5] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:374425:376058 [5] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:374425:376058 [5] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:374425:376058 [5] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:374425:376058 [5] NCCL INFO Using network Socket +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test1-worker-0:374427:374427 [7] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test1-worker-0:374427:374427 [7] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:374427:374427 [7] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:374427:374427 [7] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:374427:374427 [7] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:374427:374427 [7] NCCL INFO NET/Plugin: Using internal network plugin. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test1-worker-0:374424:374424 [4] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test1-worker-0:374424:374424 [4] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:374424:374424 [4] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:374424:374424 [4] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:374424:374424 [4] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:374424:374424 [4] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test1-worker-0:374427:376059 [7] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:374427:376059 [7] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:374427:376059 [7] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:374427:376059 [7] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:374427:376059 [7] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:374427:376059 [7] NCCL INFO Using network Socket +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test1-worker-0:374424:376060 [4] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:374424:376060 [4] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:374424:376060 [4] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:374424:376060 [4] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:374424:376060 [4] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:374424:376060 [4] NCCL INFO Using network Socket +ywang29-vrdb-test1-worker-0:374426:374426 [6] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test1-worker-0:374426:374426 [6] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:374426:374426 [6] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:374426:374426 [6] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:374426:374426 [6] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:374426:374426 [6] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test1-worker-0:374426:376061 [6] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:374426:376061 [6] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:374426:376061 [6] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:374426:376061 [6] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:374426:376061 [6] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:374426:376061 [6] NCCL INFO Using network Socket +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test1-worker-0:374423:374423 [3] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test1-worker-0:374423:374423 [3] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:374423:374423 [3] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:374423:374423 [3] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:374423:374423 [3] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:374423:374423 [3] NCCL INFO NET/Plugin: Using internal network plugin. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test1-worker-0:374421:374421 [1] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test1-worker-0:374421:374421 [1] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:374421:374421 [1] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:374421:374421 [1] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:374421:374421 [1] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:374421:374421 [1] NCCL INFO NET/Plugin: Using internal network plugin. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test1-worker-0:374422:374422 [2] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test1-worker-0:374422:374422 [2] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:374422:374422 [2] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:374422:374422 [2] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:374422:374422 [2] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:374422:374422 [2] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test1-worker-0:374423:376095 [3] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:374423:376095 [3] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:374423:376095 [3] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:374423:376095 [3] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:374423:376095 [3] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:374423:376095 [3] NCCL INFO Using network Socket +ywang29-vrdb-test1-worker-0:374421:376096 [1] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:374421:376096 [1] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:374421:376096 [1] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:374421:376096 [1] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:374421:376096 [1] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:374421:376096 [1] NCCL INFO Using network Socket +ywang29-vrdb-test1-worker-0:374422:376097 [2] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:374422:376097 [2] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:374422:376097 [2] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:374422:376097 [2] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:374422:376097 [2] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:374422:376097 [2] NCCL INFO Using network Socket +ywang29-vrdb-test1-worker-0:374422:376097 [2] NCCL INFO ncclCommInitRank comm 0x562f3c7031f0 rank 2 nranks 8 cudaDev 2 nvmlDev 2 busId 201c0 commId 0x480161f7c3a0c00e - Init START +ywang29-vrdb-test1-worker-0:374426:376061 [6] NCCL INFO ncclCommInitRank comm 0x56550d4f8a40 rank 6 nranks 8 cudaDev 6 nvmlDev 6 busId a01c0 commId 0x480161f7c3a0c00e - Init START +ywang29-vrdb-test1-worker-0:374425:376058 [5] NCCL INFO ncclCommInitRank comm 0x55cbe7e1b780 rank 5 nranks 8 cudaDev 5 nvmlDev 5 busId 901d0 commId 0x480161f7c3a0c00e - Init START +ywang29-vrdb-test1-worker-0:374421:376096 [1] NCCL INFO ncclCommInitRank comm 0x56202cda2e20 rank 1 nranks 8 cudaDev 1 nvmlDev 1 busId 101d0 commId 0x480161f7c3a0c00e - Init START +ywang29-vrdb-test1-worker-0:374423:376095 [3] NCCL INFO ncclCommInitRank comm 0x55b6a72e3a60 rank 3 nranks 8 cudaDev 3 nvmlDev 3 busId 201d0 commId 0x480161f7c3a0c00e - Init START +ywang29-vrdb-test1-worker-0:374424:376060 [4] NCCL INFO ncclCommInitRank comm 0x5612fbc0c840 rank 4 nranks 8 cudaDev 4 nvmlDev 4 busId 901c0 commId 0x480161f7c3a0c00e - Init START +ywang29-vrdb-test1-worker-0:374427:376059 [7] NCCL INFO ncclCommInitRank comm 0x55fc30234e30 rank 7 nranks 8 cudaDev 7 nvmlDev 7 busId a01d0 commId 0x480161f7c3a0c00e - Init START +ywang29-vrdb-test1-worker-0:374420:376057 [0] NCCL INFO ncclCommInitRank comm 0x55717dff65a0 rank 0 nranks 8 cudaDev 0 nvmlDev 0 busId 101c0 commId 0x480161f7c3a0c00e - Init START +ywang29-vrdb-test1-worker-0:374420:376057 [0] NCCL INFO Setting affinity for GPU 0 to ff,ffff0000,00ffffff +ywang29-vrdb-test1-worker-0:374420:376057 [0] NCCL INFO NVLS multicast support is not available on dev 0 +ywang29-vrdb-test1-worker-0:374425:376058 [5] NCCL INFO Setting affinity for GPU 5 to ffffff00,0000ffff,ff000000 +ywang29-vrdb-test1-worker-0:374425:376058 [5] NCCL INFO NVLS multicast support is not available on dev 5 +ywang29-vrdb-test1-worker-0:374421:376096 [1] NCCL INFO Setting affinity for GPU 1 to ff,ffff0000,00ffffff +ywang29-vrdb-test1-worker-0:374421:376096 [1] NCCL INFO NVLS multicast support is not available on dev 1 +ywang29-vrdb-test1-worker-0:374427:376059 [7] NCCL INFO Setting affinity for GPU 7 to ffffff00,0000ffff,ff000000 +ywang29-vrdb-test1-worker-0:374427:376059 [7] NCCL INFO NVLS multicast support is not available on dev 7 +ywang29-vrdb-test1-worker-0:374422:376097 [2] NCCL INFO Setting affinity for GPU 2 to ff,ffff0000,00ffffff +ywang29-vrdb-test1-worker-0:374422:376097 [2] NCCL INFO NVLS multicast support is not available on dev 2 +ywang29-vrdb-test1-worker-0:374424:376060 [4] NCCL INFO Setting affinity for GPU 4 to ffffff00,0000ffff,ff000000 +ywang29-vrdb-test1-worker-0:374424:376060 [4] NCCL INFO NVLS multicast support is not available on dev 4 +ywang29-vrdb-test1-worker-0:374423:376095 [3] NCCL INFO Setting affinity for GPU 3 to ff,ffff0000,00ffffff +ywang29-vrdb-test1-worker-0:374423:376095 [3] NCCL INFO NVLS multicast support is not available on dev 3 +ywang29-vrdb-test1-worker-0:374426:376061 [6] NCCL INFO Setting affinity for GPU 6 to ffffff00,0000ffff,ff000000 +ywang29-vrdb-test1-worker-0:374426:376061 [6] NCCL INFO NVLS multicast support is not available on dev 6 +ywang29-vrdb-test1-worker-0:374427:376059 [7] NCCL INFO comm 0x55fc30234e30 rank 7 nRanks 8 nNodes 1 localRanks 8 localRank 7 MNNVL 0 +ywang29-vrdb-test1-worker-0:374422:376097 [2] NCCL INFO comm 0x562f3c7031f0 rank 2 nRanks 8 nNodes 1 localRanks 8 localRank 2 MNNVL 0 +ywang29-vrdb-test1-worker-0:374426:376061 [6] NCCL INFO comm 0x56550d4f8a40 rank 6 nRanks 8 nNodes 1 localRanks 8 localRank 6 MNNVL 0 +ywang29-vrdb-test1-worker-0:374421:376096 [1] NCCL INFO comm 0x56202cda2e20 rank 1 nRanks 8 nNodes 1 localRanks 8 localRank 1 MNNVL 0 +ywang29-vrdb-test1-worker-0:374420:376057 [0] NCCL INFO comm 0x55717dff65a0 rank 0 nRanks 8 nNodes 1 localRanks 8 localRank 0 MNNVL 0 +ywang29-vrdb-test1-worker-0:374424:376060 [4] NCCL INFO comm 0x5612fbc0c840 rank 4 nRanks 8 nNodes 1 localRanks 8 localRank 4 MNNVL 0 +ywang29-vrdb-test1-worker-0:374423:376095 [3] NCCL INFO comm 0x55b6a72e3a60 rank 3 nRanks 8 nNodes 1 localRanks 8 localRank 3 MNNVL 0 +ywang29-vrdb-test1-worker-0:374425:376058 [5] NCCL INFO comm 0x55cbe7e1b780 rank 5 nRanks 8 nNodes 1 localRanks 8 localRank 5 MNNVL 0 +ywang29-vrdb-test1-worker-0:374420:376057 [0] NCCL INFO Channel 00/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:374422:376097 [2] NCCL INFO Trees [0] 3/-1/-1->2->1 [1] 3/-1/-1->2->1 [2] 3/-1/-1->2->1 [3] 3/-1/-1->2->1 [4] 3/-1/-1->2->1 [5] 3/-1/-1->2->1 [6] 3/-1/-1->2->1 [7] 3/-1/-1->2->1 [8] 3/-1/-1->2->1 [9] 3/-1/-1->2->1 [10] 3/-1/-1->2->1 [11] 3/-1/-1->2->1 [12] 3/-1/-1->2->1 [13] 3/-1/-1->2->1 [14] 3/-1/-1->2->1 [15] 3/-1/-1->2->1 [16] 3/-1/-1->2->1 [17] 3/-1/-1->2->1 [18] 3/-1/-1->2->1 [19] 3/-1/-1->2->1 [20] 3/-1/-1->2->1 [21] 3/-1/-1->2->1 [22] 3/-1/-1->2->1 [23] 3/-1/-1->2->1 +ywang29-vrdb-test1-worker-0:374427:376059 [7] NCCL INFO Trees [0] -1/-1/-1->7->6 [1] -1/-1/-1->7->6 [2] -1/-1/-1->7->6 [3] -1/-1/-1->7->6 [4] -1/-1/-1->7->6 [5] -1/-1/-1->7->6 [6] -1/-1/-1->7->6 [7] -1/-1/-1->7->6 [8] -1/-1/-1->7->6 [9] -1/-1/-1->7->6 [10] -1/-1/-1->7->6 [11] -1/-1/-1->7->6 [12] -1/-1/-1->7->6 [13] -1/-1/-1->7->6 [14] -1/-1/-1->7->6 [15] -1/-1/-1->7->6 [16] -1/-1/-1->7->6 [17] -1/-1/-1->7->6 [18] -1/-1/-1->7->6 [19] -1/-1/-1->7->6 [20] -1/-1/-1->7->6 [21] -1/-1/-1->7->6 [22] -1/-1/-1->7->6 [23] -1/-1/-1->7->6 +ywang29-vrdb-test1-worker-0:374421:376096 [1] NCCL INFO Trees [0] 2/-1/-1->1->0 [1] 2/-1/-1->1->0 [2] 2/-1/-1->1->0 [3] 2/-1/-1->1->0 [4] 2/-1/-1->1->0 [5] 2/-1/-1->1->0 [6] 2/-1/-1->1->0 [7] 2/-1/-1->1->0 [8] 2/-1/-1->1->0 [9] 2/-1/-1->1->0 [10] 2/-1/-1->1->0 [11] 2/-1/-1->1->0 [12] 2/-1/-1->1->0 [13] 2/-1/-1->1->0 [14] 2/-1/-1->1->0 [15] 2/-1/-1->1->0 [16] 2/-1/-1->1->0 [17] 2/-1/-1->1->0 [18] 2/-1/-1->1->0 [19] 2/-1/-1->1->0 [20] 2/-1/-1->1->0 [21] 2/-1/-1->1->0 [22] 2/-1/-1->1->0 [23] 2/-1/-1->1->0 +ywang29-vrdb-test1-worker-0:374422:376097 [2] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:374427:376059 [7] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:374421:376096 [1] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:374420:376057 [0] NCCL INFO Channel 01/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:374424:376060 [4] NCCL INFO Trees [0] 5/-1/-1->4->3 [1] 5/-1/-1->4->3 [2] 5/-1/-1->4->3 [3] 5/-1/-1->4->3 [4] 5/-1/-1->4->3 [5] 5/-1/-1->4->3 [6] 5/-1/-1->4->3 [7] 5/-1/-1->4->3 [8] 5/-1/-1->4->3 [9] 5/-1/-1->4->3 [10] 5/-1/-1->4->3 [11] 5/-1/-1->4->3 [12] 5/-1/-1->4->3 [13] 5/-1/-1->4->3 [14] 5/-1/-1->4->3 [15] 5/-1/-1->4->3 [16] 5/-1/-1->4->3 [17] 5/-1/-1->4->3 [18] 5/-1/-1->4->3 [19] 5/-1/-1->4->3 [20] 5/-1/-1->4->3 [21] 5/-1/-1->4->3 [22] 5/-1/-1->4->3 [23] 5/-1/-1->4->3 +ywang29-vrdb-test1-worker-0:374420:376057 [0] NCCL INFO Channel 02/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:374426:376061 [6] NCCL INFO Trees [0] 7/-1/-1->6->5 [1] 7/-1/-1->6->5 [2] 7/-1/-1->6->5 [3] 7/-1/-1->6->5 [4] 7/-1/-1->6->5 [5] 7/-1/-1->6->5 [6] 7/-1/-1->6->5 [7] 7/-1/-1->6->5 [8] 7/-1/-1->6->5 [9] 7/-1/-1->6->5 [10] 7/-1/-1->6->5 [11] 7/-1/-1->6->5 [12] 7/-1/-1->6->5 [13] 7/-1/-1->6->5 [14] 7/-1/-1->6->5 [15] 7/-1/-1->6->5 [16] 7/-1/-1->6->5 [17] 7/-1/-1->6->5 [18] 7/-1/-1->6->5 [19] 7/-1/-1->6->5 [20] 7/-1/-1->6->5 [21] 7/-1/-1->6->5 [22] 7/-1/-1->6->5 [23] 7/-1/-1->6->5 +ywang29-vrdb-test1-worker-0:374420:376057 [0] NCCL INFO Channel 03/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:374424:376060 [4] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:374426:376061 [6] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:374420:376057 [0] NCCL INFO Channel 04/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:374420:376057 [0] NCCL INFO Channel 05/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:374423:376095 [3] NCCL INFO Trees [0] 4/-1/-1->3->2 [1] 4/-1/-1->3->2 [2] 4/-1/-1->3->2 [3] 4/-1/-1->3->2 [4] 4/-1/-1->3->2 [5] 4/-1/-1->3->2 [6] 4/-1/-1->3->2 [7] 4/-1/-1->3->2 [8] 4/-1/-1->3->2 [9] 4/-1/-1->3->2 [10] 4/-1/-1->3->2 [11] 4/-1/-1->3->2 [12] 4/-1/-1->3->2 [13] 4/-1/-1->3->2 [14] 4/-1/-1->3->2 [15] 4/-1/-1->3->2 [16] 4/-1/-1->3->2 [17] 4/-1/-1->3->2 [18] 4/-1/-1->3->2 [19] 4/-1/-1->3->2 [20] 4/-1/-1->3->2 [21] 4/-1/-1->3->2 [22] 4/-1/-1->3->2 [23] 4/-1/-1->3->2 +ywang29-vrdb-test1-worker-0:374420:376057 [0] NCCL INFO Channel 06/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:374425:376058 [5] NCCL INFO Trees [0] 6/-1/-1->5->4 [1] 6/-1/-1->5->4 [2] 6/-1/-1->5->4 [3] 6/-1/-1->5->4 [4] 6/-1/-1->5->4 [5] 6/-1/-1->5->4 [6] 6/-1/-1->5->4 [7] 6/-1/-1->5->4 [8] 6/-1/-1->5->4 [9] 6/-1/-1->5->4 [10] 6/-1/-1->5->4 [11] 6/-1/-1->5->4 [12] 6/-1/-1->5->4 [13] 6/-1/-1->5->4 [14] 6/-1/-1->5->4 [15] 6/-1/-1->5->4 [16] 6/-1/-1->5->4 [17] 6/-1/-1->5->4 [18] 6/-1/-1->5->4 [19] 6/-1/-1->5->4 [20] 6/-1/-1->5->4 [21] 6/-1/-1->5->4 [22] 6/-1/-1->5->4 [23] 6/-1/-1->5->4 +ywang29-vrdb-test1-worker-0:374420:376057 [0] NCCL INFO Channel 07/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:374423:376095 [3] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:374420:376057 [0] NCCL INFO Channel 08/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:374425:376058 [5] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:374420:376057 [0] NCCL INFO Channel 09/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:374420:376057 [0] NCCL INFO Channel 10/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:374420:376057 [0] NCCL INFO Channel 11/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:374420:376057 [0] NCCL INFO Channel 12/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:374420:376057 [0] NCCL INFO Channel 13/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:374420:376057 [0] NCCL INFO Channel 14/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:374420:376057 [0] NCCL INFO Channel 15/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:374420:376057 [0] NCCL INFO Channel 16/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:374420:376057 [0] NCCL INFO Channel 17/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:374420:376057 [0] NCCL INFO Channel 18/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:374420:376057 [0] NCCL INFO Channel 19/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:374420:376057 [0] NCCL INFO Channel 20/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:374420:376057 [0] NCCL INFO Channel 21/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:374420:376057 [0] NCCL INFO Channel 22/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:374420:376057 [0] NCCL INFO Channel 23/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:374420:376057 [0] NCCL INFO Trees [0] 1/-1/-1->0->-1 [1] 1/-1/-1->0->-1 [2] 1/-1/-1->0->-1 [3] 1/-1/-1->0->-1 [4] 1/-1/-1->0->-1 [5] 1/-1/-1->0->-1 [6] 1/-1/-1->0->-1 [7] 1/-1/-1->0->-1 [8] 1/-1/-1->0->-1 [9] 1/-1/-1->0->-1 [10] 1/-1/-1->0->-1 [11] 1/-1/-1->0->-1 [12] 1/-1/-1->0->-1 [13] 1/-1/-1->0->-1 [14] 1/-1/-1->0->-1 [15] 1/-1/-1->0->-1 [16] 1/-1/-1->0->-1 [17] 1/-1/-1->0->-1 [18] 1/-1/-1->0->-1 [19] 1/-1/-1->0->-1 [20] 1/-1/-1->0->-1 [21] 1/-1/-1->0->-1 [22] 1/-1/-1->0->-1 [23] 1/-1/-1->0->-1 +ywang29-vrdb-test1-worker-0:374420:376057 [0] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:374427:376059 [7] NCCL INFO Channel 00/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374422:376097 [2] NCCL INFO Channel 00/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374424:376060 [4] NCCL INFO Channel 00/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374427:376059 [7] NCCL INFO Channel 01/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374422:376097 [2] NCCL INFO Channel 01/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374424:376060 [4] NCCL INFO Channel 01/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374427:376059 [7] NCCL INFO Channel 02/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374422:376097 [2] NCCL INFO Channel 02/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374424:376060 [4] NCCL INFO Channel 02/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374427:376059 [7] NCCL INFO Channel 03/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374425:376058 [5] NCCL INFO Channel 00/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374422:376097 [2] NCCL INFO Channel 03/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374424:376060 [4] NCCL INFO Channel 03/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374427:376059 [7] NCCL INFO Channel 04/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374425:376058 [5] NCCL INFO Channel 01/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374422:376097 [2] NCCL INFO Channel 04/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374424:376060 [4] NCCL INFO Channel 04/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374420:376057 [0] NCCL INFO Channel 00/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374427:376059 [7] NCCL INFO Channel 05/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374423:376095 [3] NCCL INFO Channel 00/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374425:376058 [5] NCCL INFO Channel 02/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374422:376097 [2] NCCL INFO Channel 05/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374424:376060 [4] NCCL INFO Channel 05/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374420:376057 [0] NCCL INFO Channel 01/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374427:376059 [7] NCCL INFO Channel 06/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374423:376095 [3] NCCL INFO Channel 01/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374425:376058 [5] NCCL INFO Channel 03/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374422:376097 [2] NCCL INFO Channel 06/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374424:376060 [4] NCCL INFO Channel 06/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374420:376057 [0] NCCL INFO Channel 02/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374427:376059 [7] NCCL INFO Channel 07/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374423:376095 [3] NCCL INFO Channel 02/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374422:376097 [2] NCCL INFO Channel 07/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374424:376060 [4] NCCL INFO Channel 07/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374425:376058 [5] NCCL INFO Channel 04/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374427:376059 [7] NCCL INFO Channel 08/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374420:376057 [0] NCCL INFO Channel 03/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374422:376097 [2] NCCL INFO Channel 08/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374424:376060 [4] NCCL INFO Channel 08/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374423:376095 [3] NCCL INFO Channel 03/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374425:376058 [5] NCCL INFO Channel 05/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374427:376059 [7] NCCL INFO Channel 09/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374420:376057 [0] NCCL INFO Channel 04/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374422:376097 [2] NCCL INFO Channel 09/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374424:376060 [4] NCCL INFO Channel 09/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374423:376095 [3] NCCL INFO Channel 04/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374427:376059 [7] NCCL INFO Channel 10/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374425:376058 [5] NCCL INFO Channel 06/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374420:376057 [0] NCCL INFO Channel 05/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374422:376097 [2] NCCL INFO Channel 10/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374424:376060 [4] NCCL INFO Channel 10/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374427:376059 [7] NCCL INFO Channel 11/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374423:376095 [3] NCCL INFO Channel 05/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374425:376058 [5] NCCL INFO Channel 07/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374420:376057 [0] NCCL INFO Channel 06/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374422:376097 [2] NCCL INFO Channel 11/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374424:376060 [4] NCCL INFO Channel 11/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374427:376059 [7] NCCL INFO Channel 12/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374423:376095 [3] NCCL INFO Channel 06/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374425:376058 [5] NCCL INFO Channel 08/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374422:376097 [2] NCCL INFO Channel 12/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374420:376057 [0] NCCL INFO Channel 07/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374424:376060 [4] NCCL INFO Channel 12/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374427:376059 [7] NCCL INFO Channel 13/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374423:376095 [3] NCCL INFO Channel 07/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374421:376096 [1] NCCL INFO Channel 00/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374425:376058 [5] NCCL INFO Channel 09/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374420:376057 [0] NCCL INFO Channel 08/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374422:376097 [2] NCCL INFO Channel 13/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374424:376060 [4] NCCL INFO Channel 13/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374427:376059 [7] NCCL INFO Channel 14/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374423:376095 [3] NCCL INFO Channel 08/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374421:376096 [1] NCCL INFO Channel 01/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374425:376058 [5] NCCL INFO Channel 10/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374420:376057 [0] NCCL INFO Channel 09/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374422:376097 [2] NCCL INFO Channel 14/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374424:376060 [4] NCCL INFO Channel 14/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374427:376059 [7] NCCL INFO Channel 15/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374423:376095 [3] NCCL INFO Channel 09/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374421:376096 [1] NCCL INFO Channel 02/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374425:376058 [5] NCCL INFO Channel 11/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374420:376057 [0] NCCL INFO Channel 10/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374422:376097 [2] NCCL INFO Channel 15/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374424:376060 [4] NCCL INFO Channel 15/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374427:376059 [7] NCCL INFO Channel 16/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374423:376095 [3] NCCL INFO Channel 10/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374421:376096 [1] NCCL INFO Channel 03/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374425:376058 [5] NCCL INFO Channel 12/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374420:376057 [0] NCCL INFO Channel 11/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374422:376097 [2] NCCL INFO Channel 16/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374424:376060 [4] NCCL INFO Channel 16/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374427:376059 [7] NCCL INFO Channel 17/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374423:376095 [3] NCCL INFO Channel 11/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374421:376096 [1] NCCL INFO Channel 04/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374422:376097 [2] NCCL INFO Channel 17/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374425:376058 [5] NCCL INFO Channel 13/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374424:376060 [4] NCCL INFO Channel 17/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374420:376057 [0] NCCL INFO Channel 12/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374426:376061 [6] NCCL INFO Channel 00/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374427:376059 [7] NCCL INFO Channel 18/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374423:376095 [3] NCCL INFO Channel 12/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374421:376096 [1] NCCL INFO Channel 05/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374422:376097 [2] NCCL INFO Channel 18/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374424:376060 [4] NCCL INFO Channel 18/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374425:376058 [5] NCCL INFO Channel 14/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374420:376057 [0] NCCL INFO Channel 13/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374426:376061 [6] NCCL INFO Channel 01/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374427:376059 [7] NCCL INFO Channel 19/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374423:376095 [3] NCCL INFO Channel 13/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374421:376096 [1] NCCL INFO Channel 06/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374422:376097 [2] NCCL INFO Channel 19/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374424:376060 [4] NCCL INFO Channel 19/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374425:376058 [5] NCCL INFO Channel 15/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374420:376057 [0] NCCL INFO Channel 14/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374426:376061 [6] NCCL INFO Channel 02/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374427:376059 [7] NCCL INFO Channel 20/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374423:376095 [3] NCCL INFO Channel 14/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374421:376096 [1] NCCL INFO Channel 07/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374422:376097 [2] NCCL INFO Channel 20/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374424:376060 [4] NCCL INFO Channel 20/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374425:376058 [5] NCCL INFO Channel 16/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374420:376057 [0] NCCL INFO Channel 15/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374427:376059 [7] NCCL INFO Channel 21/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374426:376061 [6] NCCL INFO Channel 03/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374423:376095 [3] NCCL INFO Channel 15/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374421:376096 [1] NCCL INFO Channel 08/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374422:376097 [2] NCCL INFO Channel 21/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374424:376060 [4] NCCL INFO Channel 21/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374425:376058 [5] NCCL INFO Channel 17/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374420:376057 [0] NCCL INFO Channel 16/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374427:376059 [7] NCCL INFO Channel 22/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374426:376061 [6] NCCL INFO Channel 04/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374422:376097 [2] NCCL INFO Channel 22/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374421:376096 [1] NCCL INFO Channel 09/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374423:376095 [3] NCCL INFO Channel 16/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374424:376060 [4] NCCL INFO Channel 22/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374425:376058 [5] NCCL INFO Channel 18/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374427:376059 [7] NCCL INFO Channel 23/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374420:376057 [0] NCCL INFO Channel 17/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374423:376095 [3] NCCL INFO Channel 17/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374422:376097 [2] NCCL INFO Channel 23/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374426:376061 [6] NCCL INFO Channel 05/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374421:376096 [1] NCCL INFO Channel 10/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374424:376060 [4] NCCL INFO Channel 23/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374425:376058 [5] NCCL INFO Channel 19/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374420:376057 [0] NCCL INFO Channel 18/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374423:376095 [3] NCCL INFO Channel 18/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374421:376096 [1] NCCL INFO Channel 11/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374426:376061 [6] NCCL INFO Channel 06/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374425:376058 [5] NCCL INFO Channel 20/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374420:376057 [0] NCCL INFO Channel 19/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374423:376095 [3] NCCL INFO Channel 19/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374421:376096 [1] NCCL INFO Channel 12/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374426:376061 [6] NCCL INFO Channel 07/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374425:376058 [5] NCCL INFO Channel 21/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374420:376057 [0] NCCL INFO Channel 20/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374421:376096 [1] NCCL INFO Channel 13/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374423:376095 [3] NCCL INFO Channel 20/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374426:376061 [6] NCCL INFO Channel 08/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374425:376058 [5] NCCL INFO Channel 22/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374420:376057 [0] NCCL INFO Channel 21/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374421:376096 [1] NCCL INFO Channel 14/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374423:376095 [3] NCCL INFO Channel 21/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374426:376061 [6] NCCL INFO Channel 09/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374425:376058 [5] NCCL INFO Channel 23/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374420:376057 [0] NCCL INFO Channel 22/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374421:376096 [1] NCCL INFO Channel 15/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374423:376095 [3] NCCL INFO Channel 22/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374426:376061 [6] NCCL INFO Channel 10/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374420:376057 [0] NCCL INFO Channel 23/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374421:376096 [1] NCCL INFO Channel 16/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374423:376095 [3] NCCL INFO Channel 23/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374426:376061 [6] NCCL INFO Channel 11/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374421:376096 [1] NCCL INFO Channel 17/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374426:376061 [6] NCCL INFO Channel 12/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374421:376096 [1] NCCL INFO Channel 18/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374426:376061 [6] NCCL INFO Channel 13/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374421:376096 [1] NCCL INFO Channel 19/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374426:376061 [6] NCCL INFO Channel 14/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374421:376096 [1] NCCL INFO Channel 20/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374421:376096 [1] NCCL INFO Channel 21/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374426:376061 [6] NCCL INFO Channel 15/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374421:376096 [1] NCCL INFO Channel 22/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374426:376061 [6] NCCL INFO Channel 16/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374421:376096 [1] NCCL INFO Channel 23/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374426:376061 [6] NCCL INFO Channel 17/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374426:376061 [6] NCCL INFO Channel 18/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374426:376061 [6] NCCL INFO Channel 19/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374426:376061 [6] NCCL INFO Channel 20/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374426:376061 [6] NCCL INFO Channel 21/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374426:376061 [6] NCCL INFO Channel 22/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374426:376061 [6] NCCL INFO Channel 23/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374422:376097 [2] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:374421:376096 [1] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:374423:376095 [3] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:374420:376057 [0] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:374422:376097 [2] NCCL INFO Channel 00/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374422:376097 [2] NCCL INFO Channel 01/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374422:376097 [2] NCCL INFO Channel 02/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374422:376097 [2] NCCL INFO Channel 03/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374422:376097 [2] NCCL INFO Channel 04/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374422:376097 [2] NCCL INFO Channel 05/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374422:376097 [2] NCCL INFO Channel 06/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374422:376097 [2] NCCL INFO Channel 07/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374422:376097 [2] NCCL INFO Channel 08/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374422:376097 [2] NCCL INFO Channel 09/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374422:376097 [2] NCCL INFO Channel 10/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374422:376097 [2] NCCL INFO Channel 11/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374422:376097 [2] NCCL INFO Channel 12/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374422:376097 [2] NCCL INFO Channel 13/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374424:376060 [4] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:374422:376097 [2] NCCL INFO Channel 14/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374421:376096 [1] NCCL INFO Channel 00/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374422:376097 [2] NCCL INFO Channel 15/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374421:376096 [1] NCCL INFO Channel 01/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374422:376097 [2] NCCL INFO Channel 16/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374421:376096 [1] NCCL INFO Channel 02/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374421:376096 [1] NCCL INFO Channel 03/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374423:376095 [3] NCCL INFO Channel 00/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374422:376097 [2] NCCL INFO Channel 17/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374425:376058 [5] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:374426:376061 [6] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:374421:376096 [1] NCCL INFO Channel 04/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374427:376059 [7] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:374423:376095 [3] NCCL INFO Channel 01/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374427:376059 [7] NCCL INFO Channel 00/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374422:376097 [2] NCCL INFO Channel 18/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374421:376096 [1] NCCL INFO Channel 05/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374423:376095 [3] NCCL INFO Channel 02/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374422:376097 [2] NCCL INFO Channel 19/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374427:376059 [7] NCCL INFO Channel 01/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374421:376096 [1] NCCL INFO Channel 06/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374423:376095 [3] NCCL INFO Channel 03/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374422:376097 [2] NCCL INFO Channel 20/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374427:376059 [7] NCCL INFO Channel 02/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374421:376096 [1] NCCL INFO Channel 07/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374423:376095 [3] NCCL INFO Channel 04/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374422:376097 [2] NCCL INFO Channel 21/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374427:376059 [7] NCCL INFO Channel 03/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374421:376096 [1] NCCL INFO Channel 08/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374423:376095 [3] NCCL INFO Channel 05/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374422:376097 [2] NCCL INFO Channel 22/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374427:376059 [7] NCCL INFO Channel 04/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374421:376096 [1] NCCL INFO Channel 09/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374423:376095 [3] NCCL INFO Channel 06/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374422:376097 [2] NCCL INFO Channel 23/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374427:376059 [7] NCCL INFO Channel 05/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374421:376096 [1] NCCL INFO Channel 10/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374423:376095 [3] NCCL INFO Channel 07/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374427:376059 [7] NCCL INFO Channel 06/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374421:376096 [1] NCCL INFO Channel 11/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374423:376095 [3] NCCL INFO Channel 08/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374427:376059 [7] NCCL INFO Channel 07/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374421:376096 [1] NCCL INFO Channel 12/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374423:376095 [3] NCCL INFO Channel 09/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374427:376059 [7] NCCL INFO Channel 08/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374421:376096 [1] NCCL INFO Channel 13/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374423:376095 [3] NCCL INFO Channel 10/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374427:376059 [7] NCCL INFO Channel 09/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374421:376096 [1] NCCL INFO Channel 14/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374423:376095 [3] NCCL INFO Channel 11/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374427:376059 [7] NCCL INFO Channel 10/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374421:376096 [1] NCCL INFO Channel 15/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374423:376095 [3] NCCL INFO Channel 12/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374427:376059 [7] NCCL INFO Channel 11/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374421:376096 [1] NCCL INFO Channel 16/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374423:376095 [3] NCCL INFO Channel 13/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374427:376059 [7] NCCL INFO Channel 12/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374421:376096 [1] NCCL INFO Channel 17/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374423:376095 [3] NCCL INFO Channel 14/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374427:376059 [7] NCCL INFO Channel 13/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374421:376096 [1] NCCL INFO Channel 18/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374423:376095 [3] NCCL INFO Channel 15/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374427:376059 [7] NCCL INFO Channel 14/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374421:376096 [1] NCCL INFO Channel 19/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374423:376095 [3] NCCL INFO Channel 16/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374427:376059 [7] NCCL INFO Channel 15/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374421:376096 [1] NCCL INFO Channel 20/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374423:376095 [3] NCCL INFO Channel 17/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374427:376059 [7] NCCL INFO Channel 16/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374421:376096 [1] NCCL INFO Channel 21/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374423:376095 [3] NCCL INFO Channel 18/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374427:376059 [7] NCCL INFO Channel 17/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374421:376096 [1] NCCL INFO Channel 22/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374423:376095 [3] NCCL INFO Channel 19/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374421:376096 [1] NCCL INFO Channel 23/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374427:376059 [7] NCCL INFO Channel 18/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374423:376095 [3] NCCL INFO Channel 20/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374427:376059 [7] NCCL INFO Channel 19/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374423:376095 [3] NCCL INFO Channel 21/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374427:376059 [7] NCCL INFO Channel 20/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374423:376095 [3] NCCL INFO Channel 22/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374427:376059 [7] NCCL INFO Channel 21/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374423:376095 [3] NCCL INFO Channel 23/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374427:376059 [7] NCCL INFO Channel 22/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374425:376058 [5] NCCL INFO Channel 00/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374427:376059 [7] NCCL INFO Channel 23/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374425:376058 [5] NCCL INFO Channel 01/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374426:376061 [6] NCCL INFO Channel 00/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374425:376058 [5] NCCL INFO Channel 02/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374425:376058 [5] NCCL INFO Channel 03/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374426:376061 [6] NCCL INFO Channel 01/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374426:376061 [6] NCCL INFO Channel 02/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374426:376061 [6] NCCL INFO Channel 03/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374425:376058 [5] NCCL INFO Channel 04/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374426:376061 [6] NCCL INFO Channel 04/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374425:376058 [5] NCCL INFO Channel 05/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374425:376058 [5] NCCL INFO Channel 06/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374425:376058 [5] NCCL INFO Channel 07/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374426:376061 [6] NCCL INFO Channel 05/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374426:376061 [6] NCCL INFO Channel 06/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374424:376060 [4] NCCL INFO Channel 00/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374426:376061 [6] NCCL INFO Channel 07/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374425:376058 [5] NCCL INFO Channel 08/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374424:376060 [4] NCCL INFO Channel 01/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374425:376058 [5] NCCL INFO Channel 09/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374426:376061 [6] NCCL INFO Channel 08/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374424:376060 [4] NCCL INFO Channel 02/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374425:376058 [5] NCCL INFO Channel 10/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374424:376060 [4] NCCL INFO Channel 03/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374425:376058 [5] NCCL INFO Channel 11/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374424:376060 [4] NCCL INFO Channel 04/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374426:376061 [6] NCCL INFO Channel 09/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374425:376058 [5] NCCL INFO Channel 12/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374426:376061 [6] NCCL INFO Channel 10/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374424:376060 [4] NCCL INFO Channel 05/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374426:376061 [6] NCCL INFO Channel 11/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374425:376058 [5] NCCL INFO Channel 13/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374426:376061 [6] NCCL INFO Channel 12/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374426:376061 [6] NCCL INFO Channel 13/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374424:376060 [4] NCCL INFO Channel 06/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374424:376060 [4] NCCL INFO Channel 07/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374424:376060 [4] NCCL INFO Channel 08/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374424:376060 [4] NCCL INFO Channel 09/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374424:376060 [4] NCCL INFO Channel 10/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374424:376060 [4] NCCL INFO Channel 11/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374424:376060 [4] NCCL INFO Channel 12/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374424:376060 [4] NCCL INFO Channel 13/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374424:376060 [4] NCCL INFO Channel 14/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374424:376060 [4] NCCL INFO Channel 15/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374424:376060 [4] NCCL INFO Channel 16/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374425:376058 [5] NCCL INFO Channel 14/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374426:376061 [6] NCCL INFO Channel 14/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374424:376060 [4] NCCL INFO Channel 17/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374426:376061 [6] NCCL INFO Channel 15/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374425:376058 [5] NCCL INFO Channel 15/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374426:376061 [6] NCCL INFO Channel 16/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374424:376060 [4] NCCL INFO Channel 18/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374425:376058 [5] NCCL INFO Channel 16/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374426:376061 [6] NCCL INFO Channel 17/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374424:376060 [4] NCCL INFO Channel 19/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374425:376058 [5] NCCL INFO Channel 17/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374426:376061 [6] NCCL INFO Channel 18/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374424:376060 [4] NCCL INFO Channel 20/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374425:376058 [5] NCCL INFO Channel 18/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374426:376061 [6] NCCL INFO Channel 19/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374424:376060 [4] NCCL INFO Channel 21/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374425:376058 [5] NCCL INFO Channel 19/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374426:376061 [6] NCCL INFO Channel 20/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374424:376060 [4] NCCL INFO Channel 22/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374425:376058 [5] NCCL INFO Channel 20/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374426:376061 [6] NCCL INFO Channel 21/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374424:376060 [4] NCCL INFO Channel 23/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374425:376058 [5] NCCL INFO Channel 21/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374426:376061 [6] NCCL INFO Channel 22/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374425:376058 [5] NCCL INFO Channel 22/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374426:376061 [6] NCCL INFO Channel 23/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374425:376058 [5] NCCL INFO Channel 23/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:374420:376057 [0] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:374420:376057 [0] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:374420:376057 [0] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:374421:376096 [1] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:374421:376096 [1] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:374421:376096 [1] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:374422:376097 [2] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:374422:376097 [2] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:374422:376097 [2] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:374426:376061 [6] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:374426:376061 [6] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:374426:376061 [6] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:374427:376059 [7] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:374427:376059 [7] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:374427:376059 [7] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:374423:376095 [3] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:374423:376095 [3] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:374423:376095 [3] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:374425:376058 [5] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:374424:376060 [4] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:374425:376058 [5] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:374425:376058 [5] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:374424:376060 [4] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:374424:376060 [4] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:374423:376095 [3] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:374427:376059 [7] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:374427:376059 [7] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:374423:376095 [3] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:374427:376059 [7] NCCL INFO ncclCommInitRank comm 0x55fc30234e30 rank 7 nranks 8 cudaDev 7 nvmlDev 7 busId a01d0 commId 0x480161f7c3a0c00e - Init COMPLETE +ywang29-vrdb-test1-worker-0:374423:376095 [3] NCCL INFO ncclCommInitRank comm 0x55b6a72e3a60 rank 3 nranks 8 cudaDev 3 nvmlDev 3 busId 201d0 commId 0x480161f7c3a0c00e - Init COMPLETE +ywang29-vrdb-test1-worker-0:374424:376060 [4] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:374424:376060 [4] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:374424:376060 [4] NCCL INFO ncclCommInitRank comm 0x5612fbc0c840 rank 4 nranks 8 cudaDev 4 nvmlDev 4 busId 901c0 commId 0x480161f7c3a0c00e - Init COMPLETE +ywang29-vrdb-test1-worker-0:374426:376061 [6] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:374426:376061 [6] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:374426:376061 [6] NCCL INFO ncclCommInitRank comm 0x56550d4f8a40 rank 6 nranks 8 cudaDev 6 nvmlDev 6 busId a01c0 commId 0x480161f7c3a0c00e - Init COMPLETE +ywang29-vrdb-test1-worker-0:374422:376097 [2] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:374422:376097 [2] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:374422:376097 [2] NCCL INFO ncclCommInitRank comm 0x562f3c7031f0 rank 2 nranks 8 cudaDev 2 nvmlDev 2 busId 201c0 commId 0x480161f7c3a0c00e - Init COMPLETE +ywang29-vrdb-test1-worker-0:374425:376058 [5] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:374425:376058 [5] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:374425:376058 [5] NCCL INFO ncclCommInitRank comm 0x55cbe7e1b780 rank 5 nranks 8 cudaDev 5 nvmlDev 5 busId 901d0 commId 0x480161f7c3a0c00e - Init COMPLETE +ywang29-vrdb-test1-worker-0:374420:376057 [0] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:374420:376057 [0] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:374420:376057 [0] NCCL INFO ncclCommInitRank comm 0x55717dff65a0 rank 0 nranks 8 cudaDev 0 nvmlDev 0 busId 101c0 commId 0x480161f7c3a0c00e - Init COMPLETE +ywang29-vrdb-test1-worker-0:374421:376096 [1] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:374421:376096 [1] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:374421:376096 [1] NCCL INFO ncclCommInitRank comm 0x56202cda2e20 rank 1 nranks 8 cudaDev 1 nvmlDev 1 busId 101d0 commId 0x480161f7c3a0c00e - Init COMPLETE +[2025-10-12 11:37:18,836] [INFO] [partition_parameters.py:343:__exit__] finished initializing model - num_params = 459, num_elems = 0.99B +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +[2025-10-12 11:37:20,615] [INFO] [partition_parameters.py:343:__exit__] finished initializing model - num_params = 907, num_elems = 1.42B +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +TinyLlavaForConditionalGeneration( + (language_model): Qwen2ForCausalLM( + (model): Qwen2Model( + (embed_tokens): Embedding(151936, 896) + (layers): ModuleList( + (0-23): 24 x Qwen2DecoderLayer( + (self_attn): Qwen2FlashAttention2( + (q_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=True) + (k_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=128, bias=True) + (v_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=128, bias=True) + (o_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=False) + (rotary_emb): Qwen2RotaryEmbedding() + ) + (mlp): Qwen2MLP( + (gate_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=4864, bias=False) + (up_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=4864, bias=False) + (down_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=4864, out_features=896, bias=False) + (act_fn): SiLU() + ) + (input_layernorm): Qwen2RMSNorm() + (post_attention_layernorm): Qwen2RMSNorm() + ) + ) + (norm): Qwen2RMSNorm() + ) + (lm_head): Linear(in_features=896, out_features=151936, bias=False) + ) + (vision_tower): SIGLIPVisionTower( + (_vision_tower): SiglipVisionModel( + (vision_model): SiglipVisionTransformer( + (embeddings): SiglipVisionEmbeddings( + (patch_embedding): Conv2d(3, 1152, kernel_size=(14, 14), stride=(14, 14), padding=valid) + (position_embedding): Embedding(729, 1152) + ) + (encoder): SiglipEncoder( + (layers): ModuleList( + (0-26): 27 x SiglipEncoderLayer( + (self_attn): SiglipAttention( + (k_proj): Linear(in_features=1152, out_features=1152, bias=True) + (v_proj): Linear(in_features=1152, out_features=1152, bias=True) + (q_proj): Linear(in_features=1152, out_features=1152, bias=True) + (out_proj): Linear(in_features=1152, out_features=1152, bias=True) + ) + (layer_norm1): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + (layer_norm2): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + ) + ) + ) + (post_layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (head): SiglipMultiheadAttentionPoolingHead( + (attention): MultiheadAttention( + (out_proj): NonDynamicallyQuantizableLinear(in_features=1152, out_features=1152, bias=True) + ) + (layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + ) + ) + ) + ) + (connector): MLPConnector( + (_connector): Sequential( + (0): SupermaskLinearSparsity_SoftForward_Normal(in_features=1152, out_features=896, bias=True) + (1): GELU(approximate='none') + (2): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=True) + ) + ) +) +Pre-training init language_model.model.layers.0.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.0.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.0.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.0.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.0.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.0.mlp.up_proj.scores: Mean=1.000000[E ProcessGroupNCCL.cpp:474] [Rank 1] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=3660, OpType=BROADCAST, NumelIn=136134656, NumelOut=136134656, Timeout(ms)=1800000) ran for 1800000 milliseconds before timing out. + +[E ProcessGroupNCCL.cpp:474] [Rank 4] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=3660, OpType=BROADCAST, NumelIn=136134656, NumelOut=136134656, Timeout(ms)=1800000) ran for 1800041 milliseconds before timing out. +[E ProcessGroupNCCL.cpp:474] [Rank 6] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=3660, OpType=BROADCAST, NumelIn=136134656, NumelOut=136134656, Timeout(ms)=1800000) ran for 1800046 milliseconds before timing out. +[E ProcessGroupNCCL.cpp:474] [Rank 5] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=3660, OpType=BROADCAST, NumelIn=136134656, NumelOut=136134656, Timeout(ms)=1800000) ran for 1800691 milliseconds before timing out. +[E ProcessGroupNCCL.cpp:474] [Rank 3] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=3660, OpType=BROADCAST, NumelIn=136134656, NumelOut=136134656, Timeout(ms)=1800000) ran for 1800775 milliseconds before timing out. +[E ProcessGroupNCCL.cpp:474] [Rank 2] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=3660, OpType=BROADCAST, NumelIn=136134656, NumelOut=136134656, Timeout(ms)=1800000) ran for 1800824 milliseconds before timing out. +[E ProcessGroupNCCL.cpp:474] [Rank 7] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=3660, OpType=BROADCAST, NumelIn=136134656, NumelOut=136134656, Timeout(ms)=1800000) ran for 1800919 milliseconds before timing out. +Pre-training init language_model.model.layers.0.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.1.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.1.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.1.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.1.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.1.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.1.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.1.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.2.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.2.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.2.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.2.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.2.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.2.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.2.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.3.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.3.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.3.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.3.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.3.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.3.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.3.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.4.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.4.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.4.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.4.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.4.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.4.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.4.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.5.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.5.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.5.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.5.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.5.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.5.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.5.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.6.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.6.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.6.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.6.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.6.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.6.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.6.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.7.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.7.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.7.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.7.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.7.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.7.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.7.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.8.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.8.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.8.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.8.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.8.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.8.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.8.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.9.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.9.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.9.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.9.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.9.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.9.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.9.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.10.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.10.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.10.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.10.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.10.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.10.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.10.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.11.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.11.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.11.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.11.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.11.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.11.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.11.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.12.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.12.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.12.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.12.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.12.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.12.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.12.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.13.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.13.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.13.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.13.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.13.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.13.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.13.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.14.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.14.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.14.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.14.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.14.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.14.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.14.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.15.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.15.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.15.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.15.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.15.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.15.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.15.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.16.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.16.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.16.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.16.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.16.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.16.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.16.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.17.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.17.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.17.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.17.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.17.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.17.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.17.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.18.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.18.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.18.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.18.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.18.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.18.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.18.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.19.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.19.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.19.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.19.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.19.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.19.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.19.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.20.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.20.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.20.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.20.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.20.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.20.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.20.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.21.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.21.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.21.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.21.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.21.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.21.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.21.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.22.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.22.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.22.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.22.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.22.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.22.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.22.mlp.down_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.23.self_attn.q_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.23.self_attn.k_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.23.self_attn.v_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.23.self_attn.o_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.23.mlp.gate_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.23.mlp.up_proj.scores: Mean=1.000000 +Pre-training init language_model.model.layers.23.mlp.down_proj.scores: Mean=1.000000 +ywang29-vrdb-test1-worker-0:374423:376128 [3] NCCL INFO [Service thread] Connection closed by localRank 3 +ywang29-vrdb-test1-worker-0:374422:376115 [2] NCCL INFO [Service thread] Connection closed by localRank 2 +ywang29-vrdb-test1-worker-0:374425:376123 [5] NCCL INFO [Service thread] Connection closed by localRank 5 +ywang29-vrdb-test1-worker-0:374426:376122 [6] NCCL INFO [Service thread] Connection closed by localRank 6 +ywang29-vrdb-test1-worker-0:374424:376120 [4] NCCL INFO [Service thread] Connection closed by localRank 4 +ywang29-vrdb-test1-worker-0:374427:376118 [7] NCCL INFO [Service thread] Connection closed by localRank 7 +ywang29-vrdb-test1-worker-0:374421:376114 [1] NCCL INFO [Service thread] Connection closed by localRank 1 +Pre-training init connector._connector.0.scores: Mean=1.000005 +Pre-training init connector._connector.2.scores: Mean=0.999970 +ywang29-vrdb-test1-worker-0:374427:375530 [7] NCCL INFO comm 0x55fc30234e30 rank 7 nranks 8 cudaDev 7 busId a01d0 - Abort COMPLETE +[E ProcessGroupNCCL.cpp:488] Some NCCL operations have failed or timed out. Due to the asynchronous nature of CUDA kernels, subsequent GPU operations might run on corrupted/incomplete data. +[E ProcessGroupNCCL.cpp:494] To avoid data inconsistency, we are taking the entire process down. +[E ProcessGroupNCCL.cpp:915] [Rank 7] NCCL watchdog thread terminated with exception: [Rank 7] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=3660, OpType=BROADCAST, NumelIn=136134656, NumelOut=136134656, Timeout(ms)=1800000) ran for 1800919 milliseconds before timing out. +terminate called after throwing an instance of 'std::runtime_error' + what(): [Rank 7] NCCL watchdog thread terminated with exception: [Rank 7] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=3660, OpType=BROADCAST, NumelIn=136134656, NumelOut=136134656, Timeout(ms)=1800000) ran for 1800919 milliseconds before timing out. +ywang29-vrdb-test1-worker-0:374424:375527 [4] NCCL INFO comm 0x5612fbc0c840 rank 4 nranks 8 cudaDev 4 busId 901c0 - Abort COMPLETE +[E ProcessGroupNCCL.cpp:488] Some NCCL operations have failed or timed out. Due to the asynchronous nature of CUDA kernels, subsequent GPU operations might run on corrupted/incomplete data. +[E ProcessGroupNCCL.cpp:494] To avoid data inconsistency, we are taking the entire process down. +[E ProcessGroupNCCL.cpp:915] [Rank 4] NCCL watchdog thread terminated with exception: [Rank 4] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=3660, OpType=BROADCAST, NumelIn=136134656, NumelOut=136134656, Timeout(ms)=1800000) ran for 1800041 milliseconds before timing out. +terminate called after throwing an instance of 'std::runtime_error' + what(): [Rank 4] NCCL watchdog thread terminated with exception: [Rank 4] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=3660, OpType=BROADCAST, NumelIn=136134656, NumelOut=136134656, Timeout(ms)=1800000) ran for 1800041 milliseconds before timing out. +ywang29-vrdb-test1-worker-0:374422:375529 [2] NCCL INFO comm 0x562f3c7031f0 rank 2 nranks 8 cudaDev 2 busId 201c0 - Abort COMPLETE +[E ProcessGroupNCCL.cpp:488] Some NCCL operations have failed or timed out. Due to the asynchronous nature of CUDA kernels, subsequent GPU operations might run on corrupted/incomplete data. +[E ProcessGroupNCCL.cpp:494] To avoid data inconsistency, we are taking the entire process down. +[E ProcessGroupNCCL.cpp:915] [Rank 2] NCCL watchdog thread terminated with exception: [Rank 2] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=3660, OpType=BROADCAST, NumelIn=136134656, NumelOut=136134656, Timeout(ms)=1800000) ran for 1800824 milliseconds before timing out. +terminate called after throwing an instance of 'std::runtime_error' + what(): [Rank 2] NCCL watchdog thread terminated with exception: [Rank 2] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=3660, OpType=BROADCAST, NumelIn=136134656, NumelOut=136134656, Timeout(ms)=1800000) ran for 1800824 milliseconds before timing out. +ywang29-vrdb-test1-worker-0:374421:375525 [1] NCCL INFO comm 0x56202cda2e20 rank 1 nranks 8 cudaDev 1 busId 101d0 - Abort COMPLETE +[E ProcessGroupNCCL.cpp:488] Some NCCL operations have failed or timed out. Due to the asynchronous nature of CUDA kernels, subsequent GPU operations might run on corrupted/incomplete data. +[E ProcessGroupNCCL.cpp:494] To avoid data inconsistency, we are taking the entire process down. +[E ProcessGroupNCCL.cpp:915] [Rank 1] NCCL watchdog thread terminated with exception: [Rank 1] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=3660, OpType=BROADCAST, NumelIn=136134656, NumelOut=136134656, Timeout(ms)=1800000) ran for 1800000 milliseconds before timing out. +terminate called after throwing an instance of 'std::runtime_error' + what(): [Rank 1] NCCL watchdog thread terminated with exception: [Rank 1] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=3660, OpType=BROADCAST, NumelIn=136134656, NumelOut=136134656, Timeout(ms)=1800000) ran for 1800000 milliseconds before timing out. +ywang29-vrdb-test1-worker-0:374425:375520 [5] NCCL INFO comm 0x55cbe7e1b780 rank 5 nranks 8 cudaDev 5 busId 901d0 - Abort COMPLETE +[E ProcessGroupNCCL.cpp:488] Some NCCL operations have failed or timed out. Due to the asynchronous nature of CUDA kernels, subsequent GPU operations might run on corrupted/incomplete data. +[E ProcessGroupNCCL.cpp:494] To avoid data inconsistency, we are taking the entire process down. +[E ProcessGroupNCCL.cpp:915] [Rank 5] NCCL watchdog thread terminated with exception: [Rank 5] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=3660, OpType=BROADCAST, NumelIn=136134656, NumelOut=136134656, Timeout(ms)=1800000) ran for 1800691 milliseconds before timing out. +terminate called after throwing an instance of 'std::runtime_error' +ywang29-vrdb-test1-worker-0:374426:375528 [6] NCCL INFO comm 0x56550d4f8a40 rank 6 nranks 8 cudaDev 6 busId a01c0 - Abort COMPLETE + what(): [Rank 5] NCCL watchdog thread terminated with exception: [Rank 5] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=3660, OpType=BROADCAST, NumelIn=136134656, NumelOut=136134656, Timeout(ms)=1800000) ran for 1800691 milliseconds before timing out.[E ProcessGroupNCCL.cpp:488] Some NCCL operations have failed or timed out. Due to the asynchronous nature of CUDA kernels, subsequent GPU operations might run on corrupted/incomplete data. +[E ProcessGroupNCCL.cpp:494] To avoid data inconsistency, we are taking the entire process down. + +ywang29-vrdb-test1-worker-0:374423:375526 [3] NCCL INFO comm 0x55b6a72e3a60 rank 3 nranks 8 cudaDev 3 busId 201d0 - Abort COMPLETE +[E ProcessGroupNCCL.cpp:915] [Rank 6] NCCL watchdog thread terminated with exception: [Rank 6] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=3660, OpType=BROADCAST, NumelIn=136134656, NumelOut=136134656, Timeout(ms)=1800000) ran for 1800046 milliseconds before timing out. +[E ProcessGroupNCCL.cpp:488] Some NCCL operations have failed or timed out. Due to the asynchronous nature of CUDA kernels, subsequent GPU operations might run on corrupted/incomplete data. +terminate called after throwing an instance of 'std::runtime_error' +[E ProcessGroupNCCL.cpp:494] To avoid data inconsistency, we are taking the entire process down. + what(): [Rank 6] NCCL watchdog thread terminated with exception: [Rank 6] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=3660, OpType=BROADCAST, NumelIn=136134656, NumelOut=136134656, Timeout(ms)=1800000) ran for 1800046 milliseconds before timing out. +[E ProcessGroupNCCL.cpp:915] [Rank 3] NCCL watchdog thread terminated with exception: [Rank 3] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=3660, OpType=BROADCAST, NumelIn=136134656, NumelOut=136134656, Timeout(ms)=1800000) ran for 1800775 milliseconds before timing out. +terminate called after throwing an instance of 'std::runtime_error' + what(): [Rank 3] NCCL watchdog thread terminated with exception: [Rank 3] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=3660, OpType=BROADCAST, NumelIn=136134656, NumelOut=136134656, Timeout(ms)=1800000) ran for 1800775 milliseconds before timing out. +[2025-10-12 12:35:44,752] [INFO] [launch.py:316:sigkill_handler] Killing subprocess 374420 +[2025-10-12 12:35:45,371] [INFO] [launch.py:316:sigkill_handler] Killing subprocess 374421 +[2025-10-12 12:35:45,373] [INFO] [launch.py:316:sigkill_handler] Killing subprocess 374422 +[2025-10-12 12:35:45,586] [INFO] [launch.py:316:sigkill_handler] Killing subprocess 374423 +[2025-10-12 12:35:45,586] [INFO] [launch.py:316:sigkill_handler] Killing subprocess 374424 +[2025-10-12 12:35:45,588] [INFO] [launch.py:316:sigkill_handler] Killing subprocess 374425 +[2025-10-12 12:35:45,721] [INFO] [launch.py:316:sigkill_handler] Killing subprocess 374426 +[2025-10-12 12:35:45,722] [INFO] [launch.py:316:sigkill_handler] Killing subprocess 374427 +[2025-10-12 12:35:45,723] [ERROR] [launch.py:322:sigkill_handler] ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=7', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.9_2e-1_connector-1.0_2.9_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '1.0', '--temperature_attn_text', '2.9', '--temperature_mlp_text', '2.9', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '1.0', '--temperature_attn_vision', '2.9', '--temperature_mlp_vision', '2.9', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '1.0', '--temperature_connector', '2.9', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] exits with return code = -6 +==== EXPERIMENT COMPLETED: qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.9_2e-1_connector-1.0_2.9_2e-1_ablation ==== +Log File: qwen2.5-0_5b_base_masktune_42_llm-connector_text-1.0_2.9_2e-1_connector-1.0_2.9_2e-1_ablation_20251012_113628.log +Timestamp: 2025-10-12 12:35:46 +===================================== diff --git a/logs_oct11/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_1e-2_connector-3.0_0.5_1e-2_ablation_20251012_091538.log b/logs_oct11/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_1e-2_connector-3.0_0.5_1e-2_ablation_20251012_091538.log new file mode 100644 index 0000000000000000000000000000000000000000..bdef7ea02f2ceaf40bc13da0106898a858e5a14a --- /dev/null +++ b/logs_oct11/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_1e-2_connector-3.0_0.5_1e-2_ablation_20251012_091538.log @@ -0,0 +1,2312 @@ +==== STARTING EXPERIMENT: qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_1e-2_connector-3.0_0.5_1e-2_ablation ==== +Log File: qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_1e-2_connector-3.0_0.5_1e-2_ablation_20251012_091538.log +Timestamp: 2025-10-12 09:15:38 +===================================== +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 09:15:41,048] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 09:15:43,940] [WARNING] [runner.py:202:fetch_hostfile] Unable to find hostfile, will proceed with training with local resources only. +[2025-10-12 09:15:43,942] [INFO] [runner.py:568:main] cmd = /opt/conda/envs/tinyllava/bin/python3.10 -u -m deepspeed.launcher.launch --world_info=eyJsb2NhbGhvc3QiOiBbMCwgMSwgMiwgMywgNCwgNSwgNiwgN119 --master_addr=127.0.0.1 --master_port=29501 --enable_each_rank_log=None tinyllava/train/train.py --deepspeed ./scripts/zero3.json --data_path /s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json --image_folder /s3-code/ywang29/datasets/tinyllava --is_multimodal True --conv_version qwen2_base --model_name_or_path Qwen/Qwen2.5-0.5B --vision_tower google/siglip-so400m-patch14-384 --vision_tower2 --connector_type mlp2x_gelu --mm_vision_select_layer -2 --image_aspect_ratio square --attn_implementation flash_attention_2 --bf16 True --training_recipe common --tune_type_llm full --tune_type_vision_tower frozen --tune_vision_tower_from_layer 0 --tune_type_connector full --group_by_modality_length True --pretrained_model_path /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain --output_dir /nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_1e-2_connector-3.0_0.5_1e-2_ablation --num_train_epochs 1 --per_device_train_batch_size 4 --per_device_eval_batch_size 4 --gradient_accumulation_steps 4 --evaluation_strategy no --learning_rate 1e-2 --weight_decay 0. --warmup_ratio 0.03 --lr_scheduler_type cosine --logging_steps 1 --tf32 False --model_max_length 2048 --gradient_checkpointing True --dataloader_num_workers 8 --lazy_preprocess True --report_to tensorboard --tokenizer_use_fast False --run_name tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune --subnet_mode_text both --subnet_type_text None --mask_type_text soft --init_mean_text 3.0 --temperature_attn_text 0.5 --temperature_mlp_text 0.5 --backward_type_text normal --masked_layers_text all --subnet_mode_vision both --subnet_type_vision None --mask_type_vision soft --init_mean_vision 3.0 --temperature_attn_vision 0.5 --temperature_mlp_vision 0.5 --backward_type_vision normal --masked_layers_vision all --subnet_type_connector global --mask_type_connector soft --init_mean_connector 3.0 --temperature_connector 0.5 --backward_type_connector normal --mm_projector_lr 1e-2 --seed 42 --mask_model llm-connector --save_strategy steps --save_steps 50000 --save_total_limit 1 --train_data_ratio 0.1 +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 09:15:46,574] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 09:15:47,615] [INFO] [launch.py:138:main] 0 NCCL_VERSION=2.21.5 +[2025-10-12 09:15:47,615] [INFO] [launch.py:138:main] 0 NCCL_SOCKET_IFNAME=eth +[2025-10-12 09:15:47,615] [INFO] [launch.py:145:main] WORLD INFO DICT: {'localhost': [0, 1, 2, 3, 4, 5, 6, 7]} +[2025-10-12 09:15:47,615] [INFO] [launch.py:151:main] nnodes=1, num_local_procs=8, node_rank=0 +[2025-10-12 09:15:47,615] [INFO] [launch.py:162:main] global_rank_mapping=defaultdict(, {'localhost': [0, 1, 2, 3, 4, 5, 6, 7]}) +[2025-10-12 09:15:47,615] [INFO] [launch.py:163:main] dist_world_size=8 +[2025-10-12 09:15:47,615] [INFO] [launch.py:165:main] Setting CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 +[2025-10-12 09:15:47,617] [INFO] [launch.py:253:main] process 579480 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=0', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_1e-2_connector-3.0_0.5_1e-2_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '1e-2', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '1e-2', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 09:15:47,619] [INFO] [launch.py:253:main] process 579481 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=1', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_1e-2_connector-3.0_0.5_1e-2_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '1e-2', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '1e-2', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 09:15:47,621] [INFO] [launch.py:253:main] process 579482 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=2', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_1e-2_connector-3.0_0.5_1e-2_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '1e-2', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '1e-2', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 09:15:47,623] [INFO] [launch.py:253:main] process 579483 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=3', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_1e-2_connector-3.0_0.5_1e-2_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '1e-2', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '1e-2', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 09:15:47,625] [INFO] [launch.py:253:main] process 579484 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=4', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_1e-2_connector-3.0_0.5_1e-2_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '1e-2', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '1e-2', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 09:15:47,627] [INFO] [launch.py:253:main] process 579485 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=5', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_1e-2_connector-3.0_0.5_1e-2_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '1e-2', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '1e-2', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 09:15:47,628] [INFO] [launch.py:253:main] process 579486 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=6', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_1e-2_connector-3.0_0.5_1e-2_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '1e-2', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '1e-2', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 09:15:47,630] [INFO] [launch.py:253:main] process 579487 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=7', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_1e-2_connector-3.0_0.5_1e-2_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '1e-2', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '1e-2', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 09:15:54,237] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 09:15:54,478] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 09:15:54,508] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 09:15:54,640] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 09:15:54,640] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 09:15:54,640] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 09:15:54,641] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 09:15:54,643] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 09:15:54,660] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 09:15:54,893] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 09:15:54,916] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 09:15:55,047] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 09:15:55,049] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 09:15:55,052] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 09:15:55,061] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 09:15:55,061] [INFO] [comm.py:668:init_distributed] Initializing TorchBackend in DeepSpeed with backend nccl +[2025-10-12 09:15:55,073] [INFO] [comm.py:637:init_distributed] cdb=None +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +{'llm': {'model_name_or_path': 'Qwen/Qwen2.5-0.5B', 'cache_dir': None, 'attn_implementation': 'flash_attention_2', 'subnet_mode': 'both', 'subnet_type': 'None', 'sparsity_attn': None, 'sparsity_mlp': None, 'threshold_attn': None, 'threshold_mlp': None, 'temperature_attn': 0.5, 'temperature_mlp': 0.5, 'masked_layers': 'all', 'mask_type': 'soft', 'backward_type': 'normal'}, 'vision_tower': {'model_name_or_path': 'google/siglip-so400m-patch14-384'}, 'connector': {'connector_type': 'mlp2x_gelu', 'subnet_type': 'global', 'threshold': None, 'sparsity': None, 'temperature': 0.5, 'mask_type': 'soft', 'backward_type': 'normal'}} +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +TinyLlavaConfig { + "backward_type_connector": "normal", + "cache_dir": null, + "connector_type": "mlp2x_gelu", + "hidden_size": 896, + "ignore_index": -100, + "image_aspect_ratio": "square", + "image_token_index": -200, + "llm_model_name_or_path": "Qwen/Qwen2.5-0.5B", + "mask_model": [ + "llm", + "connector" + ], + "mask_type_connector": "soft", + "model_type": "tinyllava", + "num_queries": 128, + "num_resampler_layers": 3, + "pad_token": null, + "resampler_hidden_size": 768, + "sparsity_connector": null, + "subnet_type_connector": "global", + "temperature_connector": 0.5, + "text_config": { + "_name_or_path": "Qwen/Qwen2.5-0.5B", + "architectures": [ + "Qwen2ForCausalLM" + ], + "backward_type": "normal", + "bos_token_id": 151643, + "eos_token_id": 151643, + "hidden_size": 896, + "intermediate_size": 4864, + "mask_type": "soft", + "masked_layers": "all", + "max_position_embeddings": 32768, + "max_window_layers": 24, + "model_type": "qwen2", + "num_attention_heads": 14, + "num_hidden_layers": 24, + "num_key_value_heads": 2, + "rope_theta": 1000000.0, + "sliding_window": 32768, + "subnet_mode": "both", + "subnet_type": "None", + "temperature_attn": 0.5, + "temperature_mlp": 0.5, + "tie_word_embeddings": true, + "torch_dtype": "bfloat16", + "use_mrope": false, + "use_sliding_window": false, + "vocab_size": 151936 + }, + "threshold_connector": null, + "tokenizer_model_max_length": 2048, + "tokenizer_name_or_path": "Qwen/Qwen2.5-0.5B", + "tokenizer_padding_side": "right", + "tokenizer_use_fast": false, + "transformers_version": "4.40.1", + "tune_type_connector": "frozen", + "tune_type_llm": "frozen", + "tune_type_vision_tower": "frozen", + "tune_vision_tower_from_layer": -1, + "use_cache": false, + "vision_config": { + "hidden_act": "gelu_pytorch_tanh", + "hidden_size": 1152, + "image_size": 384, + "intermediate_size": 4304, + "layer_norm_eps": 1e-06, + "model_name_or_path": "google/siglip-so400m-patch14-384", + "model_name_or_path2": "", + "model_type": "siglip_vision_model", + "num_attention_heads": 16, + "num_hidden_layers": 27, + "patch_size": 14 + }, + "vision_feature_layer": -2, + "vision_feature_select_strategy": "patch", + "vision_hidden_size": 1152, + "vision_model_name_or_path": "google/siglip-so400m-patch14-384", + "vision_model_name_or_path2": "", + "vocab_size": 151936 +} + +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +ywang29-vrdb-test2-worker-0:579480:579480 [0] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:579480:579480 [0] NCCL INFO Bootstrap : Using eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:579480:579480 [0] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test2-worker-0:579480:579480 [0] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test2-worker-0:579480:579480 [0] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test2-worker-0:579480:579480 [0] NCCL INFO cudaDriverVersion 12040 +NCCL version 2.21.5+cuda12.1 +ywang29-vrdb-test2-worker-0:579482:579482 [2] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test2-worker-0:579482:579482 [2] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:579482:579482 [2] NCCL INFO Bootstrap : Using eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:579482:579482 [2] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test2-worker-0:579482:579482 [2] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test2-worker-0:579482:579482 [2] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test2-worker-0:579480:581103 [0] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:579480:581103 [0] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test2-worker-0:579480:581103 [0] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:579480:581103 [0] NCCL INFO NET/Socket : Using [0]eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:579480:581103 [0] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test2-worker-0:579480:581103 [0] NCCL INFO Using network Socket +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test2-worker-0:579482:581104 [2] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:579482:581104 [2] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test2-worker-0:579482:581104 [2] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:579482:581104 [2] NCCL INFO NET/Socket : Using [0]eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:579482:581104 [2] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test2-worker-0:579482:581104 [2] NCCL INFO Using network Socket +ywang29-vrdb-test2-worker-0:579483:579483 [3] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test2-worker-0:579483:579483 [3] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:579483:579483 [3] NCCL INFO Bootstrap : Using eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:579483:579483 [3] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test2-worker-0:579483:579483 [3] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test2-worker-0:579483:579483 [3] NCCL INFO NET/Plugin: Using internal network plugin. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test2-worker-0:579483:581113 [3] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:579483:581113 [3] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test2-worker-0:579483:581113 [3] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:579483:581113 [3] NCCL INFO NET/Socket : Using [0]eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:579483:581113 [3] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test2-worker-0:579483:581113 [3] NCCL INFO Using network Socket +ywang29-vrdb-test2-worker-0:579485:579485 [5] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test2-worker-0:579485:579485 [5] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:579485:579485 [5] NCCL INFO Bootstrap : Using eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:579485:579485 [5] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test2-worker-0:579485:579485 [5] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test2-worker-0:579485:579485 [5] NCCL INFO NET/Plugin: Using internal network plugin. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test2-worker-0:579481:579481 [1] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test2-worker-0:579481:579481 [1] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:579481:579481 [1] NCCL INFO Bootstrap : Using eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:579481:579481 [1] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test2-worker-0:579481:579481 [1] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test2-worker-0:579481:579481 [1] NCCL INFO NET/Plugin: Using internal network plugin. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test2-worker-0:579484:579484 [4] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test2-worker-0:579484:579484 [4] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:579484:579484 [4] NCCL INFO Bootstrap : Using eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:579484:579484 [4] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test2-worker-0:579484:579484 [4] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test2-worker-0:579484:579484 [4] NCCL INFO NET/Plugin: Using internal network plugin. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test2-worker-0:579485:581123 [5] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:579485:581123 [5] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test2-worker-0:579485:581123 [5] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:579485:581123 [5] NCCL INFO NET/Socket : Using [0]eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:579485:581123 [5] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test2-worker-0:579485:581123 [5] NCCL INFO Using network Socket +ywang29-vrdb-test2-worker-0:579487:579487 [7] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test2-worker-0:579487:579487 [7] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:579487:579487 [7] NCCL INFO Bootstrap : Using eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:579487:579487 [7] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test2-worker-0:579487:579487 [7] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test2-worker-0:579487:579487 [7] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test2-worker-0:579481:581124 [1] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:579481:581124 [1] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test2-worker-0:579481:581124 [1] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:579481:581124 [1] NCCL INFO NET/Socket : Using [0]eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:579481:581124 [1] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test2-worker-0:579481:581124 [1] NCCL INFO Using network Socket +ywang29-vrdb-test2-worker-0:579484:581125 [4] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:579484:581125 [4] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test2-worker-0:579484:581125 [4] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:579484:581125 [4] NCCL INFO NET/Socket : Using [0]eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:579484:581125 [4] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test2-worker-0:579484:581125 [4] NCCL INFO Using network Socket +ywang29-vrdb-test2-worker-0:579487:581126 [7] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:579487:581126 [7] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test2-worker-0:579487:581126 [7] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:579487:581126 [7] NCCL INFO NET/Socket : Using [0]eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:579487:581126 [7] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test2-worker-0:579487:581126 [7] NCCL INFO Using network Socket +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test2-worker-0:579486:579486 [6] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test2-worker-0:579486:579486 [6] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:579486:579486 [6] NCCL INFO Bootstrap : Using eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:579486:579486 [6] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test2-worker-0:579486:579486 [6] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test2-worker-0:579486:579486 [6] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test2-worker-0:579486:581127 [6] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:579486:581127 [6] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test2-worker-0:579486:581127 [6] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:579486:581127 [6] NCCL INFO NET/Socket : Using [0]eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:579486:581127 [6] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test2-worker-0:579486:581127 [6] NCCL INFO Using network Socket +ywang29-vrdb-test2-worker-0:579486:581127 [6] NCCL INFO ncclCommInitRank comm 0x558d01cd3e90 rank 6 nranks 8 cudaDev 6 nvmlDev 6 busId a01c0 commId 0x15505c6ef4db9e88 - Init START +ywang29-vrdb-test2-worker-0:579487:581126 [7] NCCL INFO ncclCommInitRank comm 0x557d0ee2e8b0 rank 7 nranks 8 cudaDev 7 nvmlDev 7 busId a01d0 commId 0x15505c6ef4db9e88 - Init START +ywang29-vrdb-test2-worker-0:579482:581104 [2] NCCL INFO ncclCommInitRank comm 0x555f512662c0 rank 2 nranks 8 cudaDev 2 nvmlDev 2 busId 201c0 commId 0x15505c6ef4db9e88 - Init START +ywang29-vrdb-test2-worker-0:579485:581123 [5] NCCL INFO ncclCommInitRank comm 0x560499df5860 rank 5 nranks 8 cudaDev 5 nvmlDev 5 busId 901d0 commId 0x15505c6ef4db9e88 - Init START +ywang29-vrdb-test2-worker-0:579484:581125 [4] NCCL INFO ncclCommInitRank comm 0x562a7f098240 rank 4 nranks 8 cudaDev 4 nvmlDev 4 busId 901c0 commId 0x15505c6ef4db9e88 - Init START +ywang29-vrdb-test2-worker-0:579480:581103 [0] NCCL INFO ncclCommInitRank comm 0x564da1f117d0 rank 0 nranks 8 cudaDev 0 nvmlDev 0 busId 101c0 commId 0x15505c6ef4db9e88 - Init START +ywang29-vrdb-test2-worker-0:579481:581124 [1] NCCL INFO ncclCommInitRank comm 0x56071e245920 rank 1 nranks 8 cudaDev 1 nvmlDev 1 busId 101d0 commId 0x15505c6ef4db9e88 - Init START +ywang29-vrdb-test2-worker-0:579483:581113 [3] NCCL INFO ncclCommInitRank comm 0x55ca472b8f50 rank 3 nranks 8 cudaDev 3 nvmlDev 3 busId 201d0 commId 0x15505c6ef4db9e88 - Init START +ywang29-vrdb-test2-worker-0:579480:581103 [0] NCCL INFO Setting affinity for GPU 0 to ff,ffff0000,00ffffff +ywang29-vrdb-test2-worker-0:579480:581103 [0] NCCL INFO NVLS multicast support is not available on dev 0 +ywang29-vrdb-test2-worker-0:579487:581126 [7] NCCL INFO Setting affinity for GPU 7 to ffffff00,0000ffff,ff000000 +ywang29-vrdb-test2-worker-0:579487:581126 [7] NCCL INFO NVLS multicast support is not available on dev 7 +ywang29-vrdb-test2-worker-0:579482:581104 [2] NCCL INFO Setting affinity for GPU 2 to ff,ffff0000,00ffffff +ywang29-vrdb-test2-worker-0:579482:581104 [2] NCCL INFO NVLS multicast support is not available on dev 2 +ywang29-vrdb-test2-worker-0:579486:581127 [6] NCCL INFO Setting affinity for GPU 6 to ffffff00,0000ffff,ff000000 +ywang29-vrdb-test2-worker-0:579484:581125 [4] NCCL INFO Setting affinity for GPU 4 to ffffff00,0000ffff,ff000000 +ywang29-vrdb-test2-worker-0:579484:581125 [4] NCCL INFO NVLS multicast support is not available on dev 4 +ywang29-vrdb-test2-worker-0:579486:581127 [6] NCCL INFO NVLS multicast support is not available on dev 6 +ywang29-vrdb-test2-worker-0:579485:581123 [5] NCCL INFO Setting affinity for GPU 5 to ffffff00,0000ffff,ff000000 +ywang29-vrdb-test2-worker-0:579485:581123 [5] NCCL INFO NVLS multicast support is not available on dev 5 +ywang29-vrdb-test2-worker-0:579481:581124 [1] NCCL INFO Setting affinity for GPU 1 to ff,ffff0000,00ffffff +ywang29-vrdb-test2-worker-0:579481:581124 [1] NCCL INFO NVLS multicast support is not available on dev 1 +ywang29-vrdb-test2-worker-0:579483:581113 [3] NCCL INFO Setting affinity for GPU 3 to ff,ffff0000,00ffffff +ywang29-vrdb-test2-worker-0:579483:581113 [3] NCCL INFO NVLS multicast support is not available on dev 3 +ywang29-vrdb-test2-worker-0:579487:581126 [7] NCCL INFO comm 0x557d0ee2e8b0 rank 7 nRanks 8 nNodes 1 localRanks 8 localRank 7 MNNVL 0 +ywang29-vrdb-test2-worker-0:579486:581127 [6] NCCL INFO comm 0x558d01cd3e90 rank 6 nRanks 8 nNodes 1 localRanks 8 localRank 6 MNNVL 0 +ywang29-vrdb-test2-worker-0:579485:581123 [5] NCCL INFO comm 0x560499df5860 rank 5 nRanks 8 nNodes 1 localRanks 8 localRank 5 MNNVL 0 +ywang29-vrdb-test2-worker-0:579483:581113 [3] NCCL INFO comm 0x55ca472b8f50 rank 3 nRanks 8 nNodes 1 localRanks 8 localRank 3 MNNVL 0 +ywang29-vrdb-test2-worker-0:579484:581125 [4] NCCL INFO comm 0x562a7f098240 rank 4 nRanks 8 nNodes 1 localRanks 8 localRank 4 MNNVL 0 +ywang29-vrdb-test2-worker-0:579482:581104 [2] NCCL INFO comm 0x555f512662c0 rank 2 nRanks 8 nNodes 1 localRanks 8 localRank 2 MNNVL 0 +ywang29-vrdb-test2-worker-0:579480:581103 [0] NCCL INFO comm 0x564da1f117d0 rank 0 nRanks 8 nNodes 1 localRanks 8 localRank 0 MNNVL 0 +ywang29-vrdb-test2-worker-0:579481:581124 [1] NCCL INFO comm 0x56071e245920 rank 1 nRanks 8 nNodes 1 localRanks 8 localRank 1 MNNVL 0 +ywang29-vrdb-test2-worker-0:579486:581127 [6] NCCL INFO Trees [0] 7/-1/-1->6->5 [1] 7/-1/-1->6->5 [2] 7/-1/-1->6->5 [3] 7/-1/-1->6->5 [4] 7/-1/-1->6->5 [5] 7/-1/-1->6->5 [6] 7/-1/-1->6->5 [7] 7/-1/-1->6->5 [8] 7/-1/-1->6->5 [9] 7/-1/-1->6->5 [10] 7/-1/-1->6->5 [11] 7/-1/-1->6->5 [12] 7/-1/-1->6->5 [13] 7/-1/-1->6->5 [14] 7/-1/-1->6->5 [15] 7/-1/-1->6->5 [16] 7/-1/-1->6->5 [17] 7/-1/-1->6->5 [18] 7/-1/-1->6->5 [19] 7/-1/-1->6->5 [20] 7/-1/-1->6->5 [21] 7/-1/-1->6->5 [22] 7/-1/-1->6->5 [23] 7/-1/-1->6->5 +ywang29-vrdb-test2-worker-0:579485:581123 [5] NCCL INFO Trees [0] 6/-1/-1->5->4 [1] 6/-1/-1->5->4 [2] 6/-1/-1->5->4 [3] 6/-1/-1->5->4 [4] 6/-1/-1->5->4 [5] 6/-1/-1->5->4 [6] 6/-1/-1->5->4 [7] 6/-1/-1->5->4 [8] 6/-1/-1->5->4 [9] 6/-1/-1->5->4 [10] 6/-1/-1->5->4 [11] 6/-1/-1->5->4 [12] 6/-1/-1->5->4 [13] 6/-1/-1->5->4 [14] 6/-1/-1->5->4 [15] 6/-1/-1->5->4 [16] 6/-1/-1->5->4 [17] 6/-1/-1->5->4 [18] 6/-1/-1->5->4 [19] 6/-1/-1->5->4 [20] 6/-1/-1->5->4 [21] 6/-1/-1->5->4 [22] 6/-1/-1->5->4 [23] 6/-1/-1->5->4 +ywang29-vrdb-test2-worker-0:579487:581126 [7] NCCL INFO Trees [0] -1/-1/-1->7->6 [1] -1/-1/-1->7->6 [2] -1/-1/-1->7->6 [3] -1/-1/-1->7->6 [4] -1/-1/-1->7->6 [5] -1/-1/-1->7->6 [6] -1/-1/-1->7->6 [7] -1/-1/-1->7->6 [8] -1/-1/-1->7->6 [9] -1/-1/-1->7->6 [10] -1/-1/-1->7->6 [11] -1/-1/-1->7->6 [12] -1/-1/-1->7->6 [13] -1/-1/-1->7->6 [14] -1/-1/-1->7->6 [15] -1/-1/-1->7->6 [16] -1/-1/-1->7->6 [17] -1/-1/-1->7->6 [18] -1/-1/-1->7->6 [19] -1/-1/-1->7->6 [20] -1/-1/-1->7->6 [21] -1/-1/-1->7->6 [22] -1/-1/-1->7->6 [23] -1/-1/-1->7->6 +ywang29-vrdb-test2-worker-0:579485:581123 [5] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:579487:581126 [7] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:579483:581113 [3] NCCL INFO Trees [0] 4/-1/-1->3->2 [1] 4/-1/-1->3->2 [2] 4/-1/-1->3->2 [3] 4/-1/-1->3->2 [4] 4/-1/-1->3->2 [5] 4/-1/-1->3->2 [6] 4/-1/-1->3->2 [7] 4/-1/-1->3->2 [8] 4/-1/-1->3->2 [9] 4/-1/-1->3->2 [10] 4/-1/-1->3->2 [11] 4/-1/-1->3->2 [12] 4/-1/-1->3->2 [13] 4/-1/-1->3->2 [14] 4/-1/-1->3->2 [15] 4/-1/-1->3->2 [16] 4/-1/-1->3->2 [17] 4/-1/-1->3->2 [18] 4/-1/-1->3->2 [19] 4/-1/-1->3->2 [20] 4/-1/-1->3->2 [21] 4/-1/-1->3->2 [22] 4/-1/-1->3->2 [23] 4/-1/-1->3->2 +ywang29-vrdb-test2-worker-0:579480:581103 [0] NCCL INFO Channel 00/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:579484:581125 [4] NCCL INFO Trees [0] 5/-1/-1->4->3 [1] 5/-1/-1->4->3 [2] 5/-1/-1->4->3 [3] 5/-1/-1->4->3 [4] 5/-1/-1->4->3 [5] 5/-1/-1->4->3 [6] 5/-1/-1->4->3 [7] 5/-1/-1->4->3 [8] 5/-1/-1->4->3 [9] 5/-1/-1->4->3 [10] 5/-1/-1->4->3 [11] 5/-1/-1->4->3 [12] 5/-1/-1->4->3 [13] 5/-1/-1->4->3 [14] 5/-1/-1->4->3 [15] 5/-1/-1->4->3 [16] 5/-1/-1->4->3 [17] 5/-1/-1->4->3 [18] 5/-1/-1->4->3 [19] 5/-1/-1->4->3 [20] 5/-1/-1->4->3 [21] 5/-1/-1->4->3 [22] 5/-1/-1->4->3 [23] 5/-1/-1->4->3 +ywang29-vrdb-test2-worker-0:579486:581127 [6] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:579483:581113 [3] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:579480:581103 [0] NCCL INFO Channel 01/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:579482:581104 [2] NCCL INFO Trees [0] 3/-1/-1->2->1 [1] 3/-1/-1->2->1 [2] 3/-1/-1->2->1 [3] 3/-1/-1->2->1 [4] 3/-1/-1->2->1 [5] 3/-1/-1->2->1 [6] 3/-1/-1->2->1 [7] 3/-1/-1->2->1 [8] 3/-1/-1->2->1 [9] 3/-1/-1->2->1 [10] 3/-1/-1->2->1 [11] 3/-1/-1->2->1 [12] 3/-1/-1->2->1 [13] 3/-1/-1->2->1 [14] 3/-1/-1->2->1 [15] 3/-1/-1->2->1 [16] 3/-1/-1->2->1 [17] 3/-1/-1->2->1 [18] 3/-1/-1->2->1 [19] 3/-1/-1->2->1 [20] 3/-1/-1->2->1 [21] 3/-1/-1->2->1 [22] 3/-1/-1->2->1 [23] 3/-1/-1->2->1 +ywang29-vrdb-test2-worker-0:579484:581125 [4] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:579480:581103 [0] NCCL INFO Channel 02/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:579482:581104 [2] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:579480:581103 [0] NCCL INFO Channel 03/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:579480:581103 [0] NCCL INFO Channel 04/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:579480:581103 [0] NCCL INFO Channel 05/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:579480:581103 [0] NCCL INFO Channel 06/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:579480:581103 [0] NCCL INFO Channel 07/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:579480:581103 [0] NCCL INFO Channel 08/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:579481:581124 [1] NCCL INFO Trees [0] 2/-1/-1->1->0 [1] 2/-1/-1->1->0 [2] 2/-1/-1->1->0 [3] 2/-1/-1->1->0 [4] 2/-1/-1->1->0 [5] 2/-1/-1->1->0 [6] 2/-1/-1->1->0 [7] 2/-1/-1->1->0 [8] 2/-1/-1->1->0 [9] 2/-1/-1->1->0 [10] 2/-1/-1->1->0 [11] 2/-1/-1->1->0 [12] 2/-1/-1->1->0 [13] 2/-1/-1->1->0 [14] 2/-1/-1->1->0 [15] 2/-1/-1->1->0 [16] 2/-1/-1->1->0 [17] 2/-1/-1->1->0 [18] 2/-1/-1->1->0 [19] 2/-1/-1->1->0 [20] 2/-1/-1->1->0 [21] 2/-1/-1->1->0 [22] 2/-1/-1->1->0 [23] 2/-1/-1->1->0 +ywang29-vrdb-test2-worker-0:579480:581103 [0] NCCL INFO Channel 09/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:579480:581103 [0] NCCL INFO Channel 10/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:579481:581124 [1] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:579480:581103 [0] NCCL INFO Channel 11/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:579480:581103 [0] NCCL INFO Channel 12/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:579480:581103 [0] NCCL INFO Channel 13/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:579480:581103 [0] NCCL INFO Channel 14/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:579480:581103 [0] NCCL INFO Channel 15/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:579480:581103 [0] NCCL INFO Channel 16/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:579480:581103 [0] NCCL INFO Channel 17/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:579480:581103 [0] NCCL INFO Channel 18/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:579480:581103 [0] NCCL INFO Channel 19/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:579480:581103 [0] NCCL INFO Channel 20/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:579480:581103 [0] NCCL INFO Channel 21/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:579480:581103 [0] NCCL INFO Channel 22/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:579480:581103 [0] NCCL INFO Channel 23/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:579480:581103 [0] NCCL INFO Trees [0] 1/-1/-1->0->-1 [1] 1/-1/-1->0->-1 [2] 1/-1/-1->0->-1 [3] 1/-1/-1->0->-1 [4] 1/-1/-1->0->-1 [5] 1/-1/-1->0->-1 [6] 1/-1/-1->0->-1 [7] 1/-1/-1->0->-1 [8] 1/-1/-1->0->-1 [9] 1/-1/-1->0->-1 [10] 1/-1/-1->0->-1 [11] 1/-1/-1->0->-1 [12] 1/-1/-1->0->-1 [13] 1/-1/-1->0->-1 [14] 1/-1/-1->0->-1 [15] 1/-1/-1->0->-1 [16] 1/-1/-1->0->-1 [17] 1/-1/-1->0->-1 [18] 1/-1/-1->0->-1 [19] 1/-1/-1->0->-1 [20] 1/-1/-1->0->-1 [21] 1/-1/-1->0->-1 [22] 1/-1/-1->0->-1 [23] 1/-1/-1->0->-1 +ywang29-vrdb-test2-worker-0:579480:581103 [0] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:579485:581123 [5] NCCL INFO Channel 00/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:581104 [2] NCCL INFO Channel 00/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:581125 [4] NCCL INFO Channel 00/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:581123 [5] NCCL INFO Channel 01/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:581126 [7] NCCL INFO Channel 00/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:581113 [3] NCCL INFO Channel 00/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:581104 [2] NCCL INFO Channel 01/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:581125 [4] NCCL INFO Channel 01/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:581123 [5] NCCL INFO Channel 02/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:581126 [7] NCCL INFO Channel 01/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:581113 [3] NCCL INFO Channel 01/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:581104 [2] NCCL INFO Channel 02/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:581124 [1] NCCL INFO Channel 00/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:581125 [4] NCCL INFO Channel 02/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:581123 [5] NCCL INFO Channel 03/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:581126 [7] NCCL INFO Channel 02/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:581113 [3] NCCL INFO Channel 02/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579480:581103 [0] NCCL INFO Channel 00/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:581104 [2] NCCL INFO Channel 03/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:581125 [4] NCCL INFO Channel 03/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:581124 [1] NCCL INFO Channel 01/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:581126 [7] NCCL INFO Channel 03/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:581123 [5] NCCL INFO Channel 04/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:581113 [3] NCCL INFO Channel 03/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579480:581103 [0] NCCL INFO Channel 01/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:581104 [2] NCCL INFO Channel 04/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:581125 [4] NCCL INFO Channel 04/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:581126 [7] NCCL INFO Channel 04/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:581124 [1] NCCL INFO Channel 02/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:581123 [5] NCCL INFO Channel 05/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:581113 [3] NCCL INFO Channel 04/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579480:581103 [0] NCCL INFO Channel 02/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:581104 [2] NCCL INFO Channel 05/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:581125 [4] NCCL INFO Channel 05/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:581126 [7] NCCL INFO Channel 05/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:581124 [1] NCCL INFO Channel 03/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:581123 [5] NCCL INFO Channel 06/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:581113 [3] NCCL INFO Channel 05/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579480:581103 [0] NCCL INFO Channel 03/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:581104 [2] NCCL INFO Channel 06/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:581125 [4] NCCL INFO Channel 06/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:581126 [7] NCCL INFO Channel 06/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:581124 [1] NCCL INFO Channel 04/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:581123 [5] NCCL INFO Channel 07/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:581113 [3] NCCL INFO Channel 06/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579480:581103 [0] NCCL INFO Channel 04/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:581104 [2] NCCL INFO Channel 07/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:581125 [4] NCCL INFO Channel 07/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:581126 [7] NCCL INFO Channel 07/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:581123 [5] NCCL INFO Channel 08/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:581124 [1] NCCL INFO Channel 05/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:581113 [3] NCCL INFO Channel 07/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579480:581103 [0] NCCL INFO Channel 05/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:581104 [2] NCCL INFO Channel 08/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:581125 [4] NCCL INFO Channel 08/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:581126 [7] NCCL INFO Channel 08/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:581123 [5] NCCL INFO Channel 09/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:581124 [1] NCCL INFO Channel 06/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:581113 [3] NCCL INFO Channel 08/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579480:581103 [0] NCCL INFO Channel 06/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:581104 [2] NCCL INFO Channel 09/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:581125 [4] NCCL INFO Channel 09/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:581126 [7] NCCL INFO Channel 09/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:581123 [5] NCCL INFO Channel 10/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:581124 [1] NCCL INFO Channel 07/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:581113 [3] NCCL INFO Channel 09/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579480:581103 [0] NCCL INFO Channel 07/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:581104 [2] NCCL INFO Channel 10/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:581125 [4] NCCL INFO Channel 10/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:581126 [7] NCCL INFO Channel 10/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:581123 [5] NCCL INFO Channel 11/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:581124 [1] NCCL INFO Channel 08/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:581113 [3] NCCL INFO Channel 10/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579480:581103 [0] NCCL INFO Channel 08/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:581104 [2] NCCL INFO Channel 11/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:581125 [4] NCCL INFO Channel 11/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:581126 [7] NCCL INFO Channel 11/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:581123 [5] NCCL INFO Channel 12/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:581127 [6] NCCL INFO Channel 00/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:581124 [1] NCCL INFO Channel 09/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:581113 [3] NCCL INFO Channel 11/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579480:581103 [0] NCCL INFO Channel 09/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:581104 [2] NCCL INFO Channel 12/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:581125 [4] NCCL INFO Channel 12/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:581126 [7] NCCL INFO Channel 12/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:581123 [5] NCCL INFO Channel 13/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:581127 [6] NCCL INFO Channel 01/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:581124 [1] NCCL INFO Channel 10/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:581113 [3] NCCL INFO Channel 12/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579480:581103 [0] NCCL INFO Channel 10/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:581104 [2] NCCL INFO Channel 13/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:581125 [4] NCCL INFO Channel 13/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:581126 [7] NCCL INFO Channel 13/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:581123 [5] NCCL INFO Channel 14/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:581127 [6] NCCL INFO Channel 02/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:581113 [3] NCCL INFO Channel 13/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:581124 [1] NCCL INFO Channel 11/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579480:581103 [0] NCCL INFO Channel 11/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:581104 [2] NCCL INFO Channel 14/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:581125 [4] NCCL INFO Channel 14/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:581126 [7] NCCL INFO Channel 14/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:581123 [5] NCCL INFO Channel 15/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:581127 [6] NCCL INFO Channel 03/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:581113 [3] NCCL INFO Channel 14/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:581124 [1] NCCL INFO Channel 12/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579480:581103 [0] NCCL INFO Channel 12/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:581104 [2] NCCL INFO Channel 15/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:581125 [4] NCCL INFO Channel 15/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:581126 [7] NCCL INFO Channel 15/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:581123 [5] NCCL INFO Channel 16/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:581127 [6] NCCL INFO Channel 04/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:581113 [3] NCCL INFO Channel 15/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:581124 [1] NCCL INFO Channel 13/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579480:581103 [0] NCCL INFO Channel 13/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:581104 [2] NCCL INFO Channel 16/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:581125 [4] NCCL INFO Channel 16/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:581126 [7] NCCL INFO Channel 16/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:581123 [5] NCCL INFO Channel 17/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:581127 [6] NCCL INFO Channel 05/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:581113 [3] NCCL INFO Channel 16/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:581124 [1] NCCL INFO Channel 14/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:581104 [2] NCCL INFO Channel 17/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579480:581103 [0] NCCL INFO Channel 14/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:581125 [4] NCCL INFO Channel 17/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:581126 [7] NCCL INFO Channel 17/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:581123 [5] NCCL INFO Channel 18/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:581127 [6] NCCL INFO Channel 06/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:581113 [3] NCCL INFO Channel 17/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579480:581103 [0] NCCL INFO Channel 15/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:581104 [2] NCCL INFO Channel 18/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:581124 [1] NCCL INFO Channel 15/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:581125 [4] NCCL INFO Channel 18/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:581126 [7] NCCL INFO Channel 18/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:581123 [5] NCCL INFO Channel 19/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:581127 [6] NCCL INFO Channel 07/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:581113 [3] NCCL INFO Channel 18/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:581104 [2] NCCL INFO Channel 19/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579480:581103 [0] NCCL INFO Channel 16/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:581124 [1] NCCL INFO Channel 16/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:581125 [4] NCCL INFO Channel 19/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:581126 [7] NCCL INFO Channel 19/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:581123 [5] NCCL INFO Channel 20/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:581127 [6] NCCL INFO Channel 08/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:581113 [3] NCCL INFO Channel 19/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:581104 [2] NCCL INFO Channel 20/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579480:581103 [0] NCCL INFO Channel 17/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:581124 [1] NCCL INFO Channel 17/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:581125 [4] NCCL INFO Channel 20/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:581123 [5] NCCL INFO Channel 21/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:581126 [7] NCCL INFO Channel 20/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:581127 [6] NCCL INFO Channel 09/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:581113 [3] NCCL INFO Channel 20/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:581104 [2] NCCL INFO Channel 21/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579480:581103 [0] NCCL INFO Channel 18/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:581124 [1] NCCL INFO Channel 18/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:581125 [4] NCCL INFO Channel 21/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:581123 [5] NCCL INFO Channel 22/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:581126 [7] NCCL INFO Channel 21/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:581127 [6] NCCL INFO Channel 10/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:581113 [3] NCCL INFO Channel 21/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:581104 [2] NCCL INFO Channel 22/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:581124 [1] NCCL INFO Channel 19/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579480:581103 [0] NCCL INFO Channel 19/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:581125 [4] NCCL INFO Channel 22/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:581127 [6] NCCL INFO Channel 11/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:581123 [5] NCCL INFO Channel 23/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:581126 [7] NCCL INFO Channel 22/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:581113 [3] NCCL INFO Channel 22/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:581104 [2] NCCL INFO Channel 23/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:581124 [1] NCCL INFO Channel 20/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579480:581103 [0] NCCL INFO Channel 20/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:581125 [4] NCCL INFO Channel 23/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:581127 [6] NCCL INFO Channel 12/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:581126 [7] NCCL INFO Channel 23/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:581113 [3] NCCL INFO Channel 23/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:581124 [1] NCCL INFO Channel 21/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579480:581103 [0] NCCL INFO Channel 21/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:581127 [6] NCCL INFO Channel 13/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:581124 [1] NCCL INFO Channel 22/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579480:581103 [0] NCCL INFO Channel 22/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:581127 [6] NCCL INFO Channel 14/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579480:581103 [0] NCCL INFO Channel 23/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:581127 [6] NCCL INFO Channel 15/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:581124 [1] NCCL INFO Channel 23/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:581127 [6] NCCL INFO Channel 16/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:581127 [6] NCCL INFO Channel 17/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:581127 [6] NCCL INFO Channel 18/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:581127 [6] NCCL INFO Channel 19/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:581127 [6] NCCL INFO Channel 20/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:581127 [6] NCCL INFO Channel 21/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:581127 [6] NCCL INFO Channel 22/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:581127 [6] NCCL INFO Channel 23/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:581104 [2] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:579483:581113 [3] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:579484:581125 [4] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:579480:581103 [0] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:579481:581124 [1] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:579485:581123 [5] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:579486:581127 [6] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:579483:581113 [3] NCCL INFO Channel 00/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:581126 [7] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:579487:581126 [7] NCCL INFO Channel 00/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:581113 [3] NCCL INFO Channel 01/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:581126 [7] NCCL INFO Channel 01/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:581113 [3] NCCL INFO Channel 02/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:581126 [7] NCCL INFO Channel 02/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:581113 [3] NCCL INFO Channel 03/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:581113 [3] NCCL INFO Channel 04/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:581126 [7] NCCL INFO Channel 03/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:581113 [3] NCCL INFO Channel 05/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:581104 [2] NCCL INFO Channel 00/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:581126 [7] NCCL INFO Channel 04/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:581113 [3] NCCL INFO Channel 06/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:581104 [2] NCCL INFO Channel 01/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:581113 [3] NCCL INFO Channel 07/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:581104 [2] NCCL INFO Channel 02/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:581113 [3] NCCL INFO Channel 08/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:581104 [2] NCCL INFO Channel 03/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:581113 [3] NCCL INFO Channel 09/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:581104 [2] NCCL INFO Channel 04/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:581113 [3] NCCL INFO Channel 10/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:581104 [2] NCCL INFO Channel 05/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:581113 [3] NCCL INFO Channel 11/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:581126 [7] NCCL INFO Channel 05/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:581104 [2] NCCL INFO Channel 06/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:581113 [3] NCCL INFO Channel 12/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:581126 [7] NCCL INFO Channel 06/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:581104 [2] NCCL INFO Channel 07/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:581113 [3] NCCL INFO Channel 13/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:581126 [7] NCCL INFO Channel 07/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:581104 [2] NCCL INFO Channel 08/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:581113 [3] NCCL INFO Channel 14/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:581126 [7] NCCL INFO Channel 08/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:581104 [2] NCCL INFO Channel 09/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:581113 [3] NCCL INFO Channel 15/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:581126 [7] NCCL INFO Channel 09/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:581123 [5] NCCL INFO Channel 00/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:581104 [2] NCCL INFO Channel 10/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:581113 [3] NCCL INFO Channel 16/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:581126 [7] NCCL INFO Channel 10/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:581123 [5] NCCL INFO Channel 01/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:581104 [2] NCCL INFO Channel 11/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:581113 [3] NCCL INFO Channel 17/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:581126 [7] NCCL INFO Channel 11/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:581123 [5] NCCL INFO Channel 02/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:581125 [4] NCCL INFO Channel 00/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:581104 [2] NCCL INFO Channel 12/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:581113 [3] NCCL INFO Channel 18/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:581126 [7] NCCL INFO Channel 12/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:581123 [5] NCCL INFO Channel 03/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:581125 [4] NCCL INFO Channel 01/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:581104 [2] NCCL INFO Channel 13/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:581113 [3] NCCL INFO Channel 19/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:581126 [7] NCCL INFO Channel 13/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:581123 [5] NCCL INFO Channel 04/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:581125 [4] NCCL INFO Channel 02/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:581104 [2] NCCL INFO Channel 14/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:581113 [3] NCCL INFO Channel 20/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:581126 [7] NCCL INFO Channel 14/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:581123 [5] NCCL INFO Channel 05/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:581125 [4] NCCL INFO Channel 03/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:581113 [3] NCCL INFO Channel 21/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:581104 [2] NCCL INFO Channel 15/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:581126 [7] NCCL INFO Channel 15/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:581125 [4] NCCL INFO Channel 04/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:581113 [3] NCCL INFO Channel 22/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:581104 [2] NCCL INFO Channel 16/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:581125 [4] NCCL INFO Channel 05/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:581113 [3] NCCL INFO Channel 23/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:581104 [2] NCCL INFO Channel 17/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:581125 [4] NCCL INFO Channel 06/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:581124 [1] NCCL INFO Channel 00/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:581104 [2] NCCL INFO Channel 18/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:581125 [4] NCCL INFO Channel 07/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:581124 [1] NCCL INFO Channel 01/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:581125 [4] NCCL INFO Channel 08/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:581104 [2] NCCL INFO Channel 19/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:581124 [1] NCCL INFO Channel 02/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:581125 [4] NCCL INFO Channel 09/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:581104 [2] NCCL INFO Channel 20/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:581123 [5] NCCL INFO Channel 06/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:581124 [1] NCCL INFO Channel 03/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:581125 [4] NCCL INFO Channel 10/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:581104 [2] NCCL INFO Channel 21/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:581123 [5] NCCL INFO Channel 07/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:581125 [4] NCCL INFO Channel 11/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:581124 [1] NCCL INFO Channel 04/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:581104 [2] NCCL INFO Channel 22/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:581123 [5] NCCL INFO Channel 08/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:581125 [4] NCCL INFO Channel 12/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:581124 [1] NCCL INFO Channel 05/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:581104 [2] NCCL INFO Channel 23/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:581123 [5] NCCL INFO Channel 09/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:581125 [4] NCCL INFO Channel 13/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:581124 [1] NCCL INFO Channel 06/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:581127 [6] NCCL INFO Channel 00/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:581123 [5] NCCL INFO Channel 10/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:581125 [4] NCCL INFO Channel 14/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:581126 [7] NCCL INFO Channel 16/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:581127 [6] NCCL INFO Channel 01/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:581124 [1] NCCL INFO Channel 07/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:581123 [5] NCCL INFO Channel 11/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:581125 [4] NCCL INFO Channel 15/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:581126 [7] NCCL INFO Channel 17/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:581127 [6] NCCL INFO Channel 02/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:581124 [1] NCCL INFO Channel 08/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:581123 [5] NCCL INFO Channel 12/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:581125 [4] NCCL INFO Channel 16/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:581126 [7] NCCL INFO Channel 18/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:581127 [6] NCCL INFO Channel 03/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:581124 [1] NCCL INFO Channel 09/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:581123 [5] NCCL INFO Channel 13/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:581125 [4] NCCL INFO Channel 17/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:581126 [7] NCCL INFO Channel 19/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:581127 [6] NCCL INFO Channel 04/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:581123 [5] NCCL INFO Channel 14/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:581124 [1] NCCL INFO Channel 10/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:581125 [4] NCCL INFO Channel 18/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:581126 [7] NCCL INFO Channel 20/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:581127 [6] NCCL INFO Channel 05/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:581123 [5] NCCL INFO Channel 15/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:581124 [1] NCCL INFO Channel 11/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:581125 [4] NCCL INFO Channel 19/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:581126 [7] NCCL INFO Channel 21/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:581127 [6] NCCL INFO Channel 06/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:581123 [5] NCCL INFO Channel 16/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:581124 [1] NCCL INFO Channel 12/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:581125 [4] NCCL INFO Channel 20/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:581126 [7] NCCL INFO Channel 22/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:581127 [6] NCCL INFO Channel 07/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:581123 [5] NCCL INFO Channel 17/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:581124 [1] NCCL INFO Channel 13/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:581125 [4] NCCL INFO Channel 21/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:581126 [7] NCCL INFO Channel 23/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:581127 [6] NCCL INFO Channel 08/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:581123 [5] NCCL INFO Channel 18/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:581125 [4] NCCL INFO Channel 22/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:581124 [1] NCCL INFO Channel 14/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:581127 [6] NCCL INFO Channel 09/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:581123 [5] NCCL INFO Channel 19/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:581125 [4] NCCL INFO Channel 23/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:581124 [1] NCCL INFO Channel 15/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:581127 [6] NCCL INFO Channel 10/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:581123 [5] NCCL INFO Channel 20/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:581124 [1] NCCL INFO Channel 16/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:581127 [6] NCCL INFO Channel 11/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:581123 [5] NCCL INFO Channel 21/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:581124 [1] NCCL INFO Channel 17/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:581123 [5] NCCL INFO Channel 22/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:581127 [6] NCCL INFO Channel 12/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:581123 [5] NCCL INFO Channel 23/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:581127 [6] NCCL INFO Channel 13/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:581124 [1] NCCL INFO Channel 18/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:581127 [6] NCCL INFO Channel 14/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:581124 [1] NCCL INFO Channel 19/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:581127 [6] NCCL INFO Channel 15/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:581124 [1] NCCL INFO Channel 20/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:581127 [6] NCCL INFO Channel 16/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:581124 [1] NCCL INFO Channel 21/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:581127 [6] NCCL INFO Channel 17/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:581124 [1] NCCL INFO Channel 22/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:581127 [6] NCCL INFO Channel 18/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:581124 [1] NCCL INFO Channel 23/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:581127 [6] NCCL INFO Channel 19/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:581127 [6] NCCL INFO Channel 20/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:581127 [6] NCCL INFO Channel 21/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:581127 [6] NCCL INFO Channel 22/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:581127 [6] NCCL INFO Channel 23/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579480:581103 [0] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:579480:581103 [0] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:579480:581103 [0] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:579481:581124 [1] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:579481:581124 [1] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:579481:581124 [1] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:579482:581104 [2] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:579482:581104 [2] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:579482:581104 [2] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:579483:581113 [3] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:579483:581113 [3] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:579483:581113 [3] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:579487:581126 [7] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:579487:581126 [7] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:579487:581126 [7] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:579484:581125 [4] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:579484:581125 [4] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:579484:581125 [4] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:579486:581127 [6] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:579485:581123 [5] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:579486:581127 [6] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:579486:581127 [6] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:579485:581123 [5] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:579485:581123 [5] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:579487:581126 [7] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test2-worker-0:579486:581127 [6] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test2-worker-0:579485:581123 [5] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test2-worker-0:579483:581113 [3] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test2-worker-0:579486:581127 [6] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test2-worker-0:579487:581126 [7] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test2-worker-0:579485:581123 [5] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test2-worker-0:579484:581125 [4] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test2-worker-0:579483:581113 [3] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test2-worker-0:579484:581125 [4] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test2-worker-0:579487:581126 [7] NCCL INFO ncclCommInitRank comm 0x557d0ee2e8b0 rank 7 nranks 8 cudaDev 7 nvmlDev 7 busId a01d0 commId 0x15505c6ef4db9e88 - Init COMPLETE +ywang29-vrdb-test2-worker-0:579485:581123 [5] NCCL INFO ncclCommInitRank comm 0x560499df5860 rank 5 nranks 8 cudaDev 5 nvmlDev 5 busId 901d0 commId 0x15505c6ef4db9e88 - Init COMPLETE +ywang29-vrdb-test2-worker-0:579486:581127 [6] NCCL INFO ncclCommInitRank comm 0x558d01cd3e90 rank 6 nranks 8 cudaDev 6 nvmlDev 6 busId a01c0 commId 0x15505c6ef4db9e88 - Init COMPLETE +ywang29-vrdb-test2-worker-0:579483:581113 [3] NCCL INFO ncclCommInitRank comm 0x55ca472b8f50 rank 3 nranks 8 cudaDev 3 nvmlDev 3 busId 201d0 commId 0x15505c6ef4db9e88 - Init COMPLETE +ywang29-vrdb-test2-worker-0:579484:581125 [4] NCCL INFO ncclCommInitRank comm 0x562a7f098240 rank 4 nranks 8 cudaDev 4 nvmlDev 4 busId 901c0 commId 0x15505c6ef4db9e88 - Init COMPLETE +ywang29-vrdb-test2-worker-0:579481:581124 [1] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test2-worker-0:579481:581124 [1] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test2-worker-0:579481:581124 [1] NCCL INFO ncclCommInitRank comm 0x56071e245920 rank 1 nranks 8 cudaDev 1 nvmlDev 1 busId 101d0 commId 0x15505c6ef4db9e88 - Init COMPLETE +ywang29-vrdb-test2-worker-0:579480:581103 [0] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test2-worker-0:579480:581103 [0] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test2-worker-0:579480:581103 [0] NCCL INFO ncclCommInitRank comm 0x564da1f117d0 rank 0 nranks 8 cudaDev 0 nvmlDev 0 busId 101c0 commId 0x15505c6ef4db9e88 - Init COMPLETE +ywang29-vrdb-test2-worker-0:579482:581104 [2] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test2-worker-0:579482:581104 [2] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test2-worker-0:579482:581104 [2] NCCL INFO ncclCommInitRank comm 0x555f512662c0 rank 2 nranks 8 cudaDev 2 nvmlDev 2 busId 201c0 commId 0x15505c6ef4db9e88 - Init COMPLETE +[2025-10-12 09:16:38,165] [INFO] [partition_parameters.py:343:__exit__] finished initializing model - num_params = 459, num_elems = 0.99B +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.laSome weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +loading language model from loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +yers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +[2025-10-12 09:32:26,961] [INFO] [partition_parameters.py:343:__exit__] finished initializing model - num_params = 907, num_elems = 1.42B +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +TinyLlavaForConditionalGeneration( + (language_model): Qwen2ForCausalLM( + (model): Qwen2Model( + (embed_tokens): Embedding(151936, 896) + (layers): ModuleList( + (0-23): 24 x Qwen2DecoderLayer( + (self_attn): Qwen2FlashAttention2( + (q_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=True) + (k_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=128, bias=True) + (v_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=128, bias=True) + (o_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=False) + (rotary_emb): Qwen2RotaryEmbedding() + ) + (mlp): Qwen2MLP( + (gate_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=4864, bias=False) + (up_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=4864, bias=False) + (down_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=4864, out_features=896, bias=False) + (act_fn): SiLU() + ) + (input_layernorm): Qwen2RMSNorm() + (post_attention_layernorm): Qwen2RMSNorm() + ) + ) + (norm): Qwen2RMSNorm() + ) + (lm_head): Linear(in_features=896, out_features=151936, bias=False) + ) + (vision_tower): SIGLIPVisionTower( + (_vision_tower): SiglipVisionModel( + (vision_model): SiglipVisionTransformer( + (embeddings): SiglipVisionEmbeddings( + (patch_embedding): Conv2d(3, 1152, kernel_size=(14, 14), stride=(14, 14), padding=valid) + (position_embedding): Embedding(729, 1152) + ) + (encoder): SiglipEncoder( + (layers): ModuleList( + (0-26): 27 x SiglipEncoderLayer( + (self_attn): SiglipAttention( + (k_proj): Linear(in_features=1152, out_features=1152, bias=True) + (v_proj): Linear(in_features=1152, out_features=1152, bias=True) + (q_proj): Linear(in_features=1152, out_features=1152, bias=True) + (out_proj): Linear(in_features=1152, out_features=1152, bias=True) + ) + (layer_norm1): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + (layer_norm2): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + ) + ) + ) + (post_layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (head): SiglipMultiheadAttentionPoolingHead( + (attention): MultiheadAttention( + (out_proj): NonDynamicallyQuantizableLinear(in_features=1152, out_features=1152, bias=True) + ) + (layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + ) + ) + ) + ) + (connector): MLPConnector( + (_connector): Sequential( + (0): SupermaskLinearSparsity_SoftForward_Normal(in_features=1152, out_features=896, bias=True) + (1): GELU(approximate='none') + (2): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=True) + ) + ) +) +Pre-training init language_model.model.layers.0.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.0.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.0.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.0.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.0.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.0.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.0.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.mlp.down_proj.scores: Mean=3.000000 +Pre-training init connector._connector.0.scores: Mean=3.000005 +Pre-training init connector._connector.2.scores: Mean=2.999970 +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +2025-10-12 09:32:44,794 | INFO: Total Parameters: 1283756736, Total Trainable Parameters: 359661568 +2025-10-12 09:32:44,799 | INFO: Trainable Parameters: +language_model.model.layers.0.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.0.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.0.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.0.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.0.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.0.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.0.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.1.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.1.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.1.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.1.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.1.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.1.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.1.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.2.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.2.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.2.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.2.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.2.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.2.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.2.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.3.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.3.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.3.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.3.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.3.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.3.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.3.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.4.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.4.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.4.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.4.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.4.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.4.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.4.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.5.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.5.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.5.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.5.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.5.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.5.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.5.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.6.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.6.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.6.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.6.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.6.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.6.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.6.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.7.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.7.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.7.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.7.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.7.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.7.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.7.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.8.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.8.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.8.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.8.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.8.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.8.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.8.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.9.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.9.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.9.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.9.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.9.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.9.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.9.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.10.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.10.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.10.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.10.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.10.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.10.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.10.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.11.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.11.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.11.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.11.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.11.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.11.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.11.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.12.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.12.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.12.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.12.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.12.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.12.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.12.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.13.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.13.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.13.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.13.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.13.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.13.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.13.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.14.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.14.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.14.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.14.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.14.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.14.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.14.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.15.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.15.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.15.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.15.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.15.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.15.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.15.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.16.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.16.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.16.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.16.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.16.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.16.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.16.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.17.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.17.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.17.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.17.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.17.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.17.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.17.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.18.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.18.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.18.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.18.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.18.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.18.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.18.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.19.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.19.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.19.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.19.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.19.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.19.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.19.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.20.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.20.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.20.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.20.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.20.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.20.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.20.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.21.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.21.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.21.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.21.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.21.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.21.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.21.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.22.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.22.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.22.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.22.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.22.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.22.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.22.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.23.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.23.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.23.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.23.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.23.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.23.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.23.mlp.down_proj.scores: 4358144 parameters +connector._connector.0.scores: 1032192 parameters +connector._connector.2.scores: 802816 parameters +Parameter Offload: Total persistent parameters: 486464 in 403 params + 0%| | 0/520 [00:003->2 [1] 4/-1/-1->3->2 [2] 4/-1/-1->3->2 [3] 4/-1/-1->3->2 [4] 4/-1/-1->3->2 [5] 4/-1/-1->3->2 [6] 4/-1/-1->3->2 [7] 4/-1/-1->3->2 [8] 4/-1/-1->3->2 [9] 4/-1/-1->3->2 [10] 4/-1/-1->3->2 [11] 4/-1/-1->3->2 [12] 4/-1/-1->3->2 [13] 4/-1/-1->3->2 [14] 4/-1/-1->3->2 [15] 4/-1/-1->3->2 [16] 4/-1/-1->3->2 [17] 4/-1/-1->3->2 [18] 4/-1/-1->3->2 [19] 4/-1/-1->3->2 [20] 4/-1/-1->3->2 [21] 4/-1/-1->3->2 [22] 4/-1/-1->3->2 [23] 4/-1/-1->3->2 +ywang29-vrdb-test2-worker-0:579483:586521 [3] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:579480:586519 [0] NCCL INFO Channel 01/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:579487:586523 [7] NCCL INFO Trees [0] -1/-1/-1->7->6 [1] -1/-1/-1->7->6 [2] -1/-1/-1->7->6 [3] -1/-1/-1->7->6 [4] -1/-1/-1->7->6 [5] -1/-1/-1->7->6 [6] -1/-1/-1->7->6 [7] -1/-1/-1->7->6 [8] -1/-1/-1->7->6 [9] -1/-1/-1->7->6 [10] -1/-1/-1->7->6 [11] -1/-1/-1->7->6 [12] -1/-1/-1->7->6 [13] -1/-1/-1->7->6 [14] -1/-1/-1->7->6 [15] -1/-1/-1->7->6 [16] -1/-1/-1->7->6 [17] -1/-1/-1->7->6 [18] -1/-1/-1->7->6 [19] -1/-1/-1->7->6 [20] -1/-1/-1->7->6 [21] -1/-1/-1->7->6 [22] -1/-1/-1->7->6 [23] -1/-1/-1->7->6 +ywang29-vrdb-test2-worker-0:579481:586524 [1] NCCL INFO Trees [0] 2/-1/-1->1->0 [1] 2/-1/-1->1->0 [2] 2/-1/-1->1->0 [3] 2/-1/-1->1->0 [4] 2/-1/-1->1->0 [5] 2/-1/-1->1->0 [6] 2/-1/-1->1->0 [7] 2/-1/-1->1->0 [8] 2/-1/-1->1->0 [9] 2/-1/-1->1->0 [10] 2/-1/-1->1->0 [11] 2/-1/-1->1->0 [12] 2/-1/-1->1->0 [13] 2/-1/-1->1->0 [14] 2/-1/-1->1->0 [15] 2/-1/-1->1->0 [16] 2/-1/-1->1->0 [17] 2/-1/-1->1->0 [18] 2/-1/-1->1->0 [19] 2/-1/-1->1->0 [20] 2/-1/-1->1->0 [21] 2/-1/-1->1->0 [22] 2/-1/-1->1->0 [23] 2/-1/-1->1->0 +ywang29-vrdb-test2-worker-0:579480:586519 [0] NCCL INFO Channel 02/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:579486:586520 [6] NCCL INFO Trees [0] 7/-1/-1->6->5 [1] 7/-1/-1->6->5 [2] 7/-1/-1->6->5 [3] 7/-1/-1->6->5 [4] 7/-1/-1->6->5 [5] 7/-1/-1->6->5 [6] 7/-1/-1->6->5 [7] 7/-1/-1->6->5 [8] 7/-1/-1->6->5 [9] 7/-1/-1->6->5 [10] 7/-1/-1->6->5 [11] 7/-1/-1->6->5 [12] 7/-1/-1->6->5 [13] 7/-1/-1->6->5 [14] 7/-1/-1->6->5 [15] 7/-1/-1->6->5 [16] 7/-1/-1->6->5 [17] 7/-1/-1->6->5 [18] 7/-1/-1->6->5 [19] 7/-1/-1->6->5 [20] 7/-1/-1->6->5 [21] 7/-1/-1->6->5 [22] 7/-1/-1->6->5 [23] 7/-1/-1->6->5 +ywang29-vrdb-test2-worker-0:579484:586525 [4] NCCL INFO Trees [0] 5/-1/-1->4->3 [1] 5/-1/-1->4->3 [2] 5/-1/-1->4->3 [3] 5/-1/-1->4->3 [4] 5/-1/-1->4->3 [5] 5/-1/-1->4->3 [6] 5/-1/-1->4->3 [7] 5/-1/-1->4->3 [8] 5/-1/-1->4->3 [9] 5/-1/-1->4->3 [10] 5/-1/-1->4->3 [11] 5/-1/-1->4->3 [12] 5/-1/-1->4->3 [13] 5/-1/-1->4->3 [14] 5/-1/-1->4->3 [15] 5/-1/-1->4->3 [16] 5/-1/-1->4->3 [17] 5/-1/-1->4->3 [18] 5/-1/-1->4->3 [19] 5/-1/-1->4->3 [20] 5/-1/-1->4->3 [21] 5/-1/-1->4->3 [22] 5/-1/-1->4->3 [23] 5/-1/-1->4->3 +ywang29-vrdb-test2-worker-0:579481:586524 [1] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:579487:586523 [7] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:579485:586526 [5] NCCL INFO Trees [0] 6/-1/-1->5->4 [1] 6/-1/-1->5->4 [2] 6/-1/-1->5->4 [3] 6/-1/-1->5->4 [4] 6/-1/-1->5->4 [5] 6/-1/-1->5->4 [6] 6/-1/-1->5->4 [7] 6/-1/-1->5->4 [8] 6/-1/-1->5->4 [9] 6/-1/-1->5->4 [10] 6/-1/-1->5->4 [11] 6/-1/-1->5->4 [12] 6/-1/-1->5->4 [13] 6/-1/-1->5->4 [14] 6/-1/-1->5->4 [15] 6/-1/-1->5->4 [16] 6/-1/-1->5->4 [17] 6/-1/-1->5->4 [18] 6/-1/-1->5->4 [19] 6/-1/-1->5->4 [20] 6/-1/-1->5->4 [21] 6/-1/-1->5->4 [22] 6/-1/-1->5->4 [23] 6/-1/-1->5->4 +ywang29-vrdb-test2-worker-0:579486:586520 [6] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:579484:586525 [4] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:579480:586519 [0] NCCL INFO Channel 03/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:579485:586526 [5] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:579480:586519 [0] NCCL INFO Channel 04/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:579480:586519 [0] NCCL INFO Channel 05/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:579482:586522 [2] NCCL INFO Trees [0] 3/-1/-1->2->1 [1] 3/-1/-1->2->1 [2] 3/-1/-1->2->1 [3] 3/-1/-1->2->1 [4] 3/-1/-1->2->1 [5] 3/-1/-1->2->1 [6] 3/-1/-1->2->1 [7] 3/-1/-1->2->1 [8] 3/-1/-1->2->1 [9] 3/-1/-1->2->1 [10] 3/-1/-1->2->1 [11] 3/-1/-1->2->1 [12] 3/-1/-1->2->1 [13] 3/-1/-1->2->1 [14] 3/-1/-1->2->1 [15] 3/-1/-1->2->1 [16] 3/-1/-1->2->1 [17] 3/-1/-1->2->1 [18] 3/-1/-1->2->1 [19] 3/-1/-1->2->1 [20] 3/-1/-1->2->1 [21] 3/-1/-1->2->1 [22] 3/-1/-1->2->1 [23] 3/-1/-1->2->1 +ywang29-vrdb-test2-worker-0:579480:586519 [0] NCCL INFO Channel 06/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:579482:586522 [2] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:579480:586519 [0] NCCL INFO Channel 07/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:579480:586519 [0] NCCL INFO Channel 08/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:579480:586519 [0] NCCL INFO Channel 09/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:579480:586519 [0] NCCL INFO Channel 10/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:579480:586519 [0] NCCL INFO Channel 11/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:579480:586519 [0] NCCL INFO Channel 12/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:579480:586519 [0] NCCL INFO Channel 13/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:579480:586519 [0] NCCL INFO Channel 14/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:579480:586519 [0] NCCL INFO Channel 15/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:579480:586519 [0] NCCL INFO Channel 16/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:579480:586519 [0] NCCL INFO Channel 17/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:579480:586519 [0] NCCL INFO Channel 18/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:579480:586519 [0] NCCL INFO Channel 19/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:579480:586519 [0] NCCL INFO Channel 20/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:579480:586519 [0] NCCL INFO Channel 21/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:579480:586519 [0] NCCL INFO Channel 22/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:579480:586519 [0] NCCL INFO Channel 23/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:579480:586519 [0] NCCL INFO Trees [0] 1/-1/-1->0->-1 [1] 1/-1/-1->0->-1 [2] 1/-1/-1->0->-1 [3] 1/-1/-1->0->-1 [4] 1/-1/-1->0->-1 [5] 1/-1/-1->0->-1 [6] 1/-1/-1->0->-1 [7] 1/-1/-1->0->-1 [8] 1/-1/-1->0->-1 [9] 1/-1/-1->0->-1 [10] 1/-1/-1->0->-1 [11] 1/-1/-1->0->-1 [12] 1/-1/-1->0->-1 [13] 1/-1/-1->0->-1 [14] 1/-1/-1->0->-1 [15] 1/-1/-1->0->-1 [16] 1/-1/-1->0->-1 [17] 1/-1/-1->0->-1 [18] 1/-1/-1->0->-1 [19] 1/-1/-1->0->-1 [20] 1/-1/-1->0->-1 [21] 1/-1/-1->0->-1 [22] 1/-1/-1->0->-1 [23] 1/-1/-1->0->-1 +ywang29-vrdb-test2-worker-0:579480:586519 [0] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:579484:586525 [4] NCCL INFO Channel 00/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:586523 [7] NCCL INFO Channel 00/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:586525 [4] NCCL INFO Channel 01/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:586523 [7] NCCL INFO Channel 01/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:586525 [4] NCCL INFO Channel 02/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:586523 [7] NCCL INFO Channel 02/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:586521 [3] NCCL INFO Channel 00/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:586525 [4] NCCL INFO Channel 03/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:586523 [7] NCCL INFO Channel 03/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579480:586519 [0] NCCL INFO Channel 00/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:586521 [3] NCCL INFO Channel 01/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:586525 [4] NCCL INFO Channel 04/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:586523 [7] NCCL INFO Channel 04/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579480:586519 [0] NCCL INFO Channel 01/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:586521 [3] NCCL INFO Channel 02/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:586525 [4] NCCL INFO Channel 05/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:586523 [7] NCCL INFO Channel 05/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579480:586519 [0] NCCL INFO Channel 02/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:586521 [3] NCCL INFO Channel 03/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:586525 [4] NCCL INFO Channel 06/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:586523 [7] NCCL INFO Channel 06/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579480:586519 [0] NCCL INFO Channel 03/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:586521 [3] NCCL INFO Channel 04/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:586525 [4] NCCL INFO Channel 07/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:586523 [7] NCCL INFO Channel 07/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579480:586519 [0] NCCL INFO Channel 04/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:586521 [3] NCCL INFO Channel 05/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:586525 [4] NCCL INFO Channel 08/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:586523 [7] NCCL INFO Channel 08/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579480:586519 [0] NCCL INFO Channel 05/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:586521 [3] NCCL INFO Channel 06/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:586525 [4] NCCL INFO Channel 09/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:586523 [7] NCCL INFO Channel 09/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579480:586519 [0] NCCL INFO Channel 06/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:586521 [3] NCCL INFO Channel 07/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:586525 [4] NCCL INFO Channel 10/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:586523 [7] NCCL INFO Channel 10/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579480:586519 [0] NCCL INFO Channel 07/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:586521 [3] NCCL INFO Channel 08/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:586525 [4] NCCL INFO Channel 11/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:586523 [7] NCCL INFO Channel 11/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579480:586519 [0] NCCL INFO Channel 08/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:586521 [3] NCCL INFO Channel 09/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:586525 [4] NCCL INFO Channel 12/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:586523 [7] NCCL INFO Channel 12/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579480:586519 [0] NCCL INFO Channel 09/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:586521 [3] NCCL INFO Channel 10/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:586525 [4] NCCL INFO Channel 13/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:586523 [7] NCCL INFO Channel 13/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579480:586519 [0] NCCL INFO Channel 10/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:586521 [3] NCCL INFO Channel 11/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:586525 [4] NCCL INFO Channel 14/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:586523 [7] NCCL INFO Channel 14/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579480:586519 [0] NCCL INFO Channel 11/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:586521 [3] NCCL INFO Channel 12/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:586525 [4] NCCL INFO Channel 15/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:586523 [7] NCCL INFO Channel 15/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579480:586519 [0] NCCL INFO Channel 12/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:586521 [3] NCCL INFO Channel 13/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:586525 [4] NCCL INFO Channel 16/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:586523 [7] NCCL INFO Channel 16/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579480:586519 [0] NCCL INFO Channel 13/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:586521 [3] NCCL INFO Channel 14/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:586525 [4] NCCL INFO Channel 17/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:586523 [7] NCCL INFO Channel 17/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579480:586519 [0] NCCL INFO Channel 14/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:586521 [3] NCCL INFO Channel 15/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:586525 [4] NCCL INFO Channel 18/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:586523 [7] NCCL INFO Channel 18/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579480:586519 [0] NCCL INFO Channel 15/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:586521 [3] NCCL INFO Channel 16/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:586525 [4] NCCL INFO Channel 19/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:586523 [7] NCCL INFO Channel 19/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579480:586519 [0] NCCL INFO Channel 16/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:586521 [3] NCCL INFO Channel 17/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:586525 [4] NCCL INFO Channel 20/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:586523 [7] NCCL INFO Channel 20/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579480:586519 [0] NCCL INFO Channel 17/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:586521 [3] NCCL INFO Channel 18/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:586525 [4] NCCL INFO Channel 21/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:586523 [7] NCCL INFO Channel 21/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579480:586519 [0] NCCL INFO Channel 18/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:586521 [3] NCCL INFO Channel 19/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:586525 [4] NCCL INFO Channel 22/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:586524 [1] NCCL INFO Channel 00/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:586520 [6] NCCL INFO Channel 00/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:586523 [7] NCCL INFO Channel 22/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579480:586519 [0] NCCL INFO Channel 19/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:586521 [3] NCCL INFO Channel 20/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:586525 [4] NCCL INFO Channel 23/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:586526 [5] NCCL INFO Channel 00/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:586524 [1] NCCL INFO Channel 01/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:586520 [6] NCCL INFO Channel 01/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:586523 [7] NCCL INFO Channel 23/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579480:586519 [0] NCCL INFO Channel 20/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:586521 [3] NCCL INFO Channel 21/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:586526 [5] NCCL INFO Channel 01/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:586524 [1] NCCL INFO Channel 02/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:586520 [6] NCCL INFO Channel 02/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579480:586519 [0] NCCL INFO Channel 21/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:586521 [3] NCCL INFO Channel 22/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:586526 [5] NCCL INFO Channel 02/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:586524 [1] NCCL INFO Channel 03/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:586520 [6] NCCL INFO Channel 03/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579480:586519 [0] NCCL INFO Channel 22/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:586521 [3] NCCL INFO Channel 23/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:586526 [5] NCCL INFO Channel 03/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:586524 [1] NCCL INFO Channel 04/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:586520 [6] NCCL INFO Channel 04/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579480:586519 [0] NCCL INFO Channel 23/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:586522 [2] NCCL INFO Channel 00/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:586526 [5] NCCL INFO Channel 04/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:586524 [1] NCCL INFO Channel 05/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:586520 [6] NCCL INFO Channel 05/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:586522 [2] NCCL INFO Channel 01/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:586526 [5] NCCL INFO Channel 05/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:586524 [1] NCCL INFO Channel 06/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:586520 [6] NCCL INFO Channel 06/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:586522 [2] NCCL INFO Channel 02/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:586526 [5] NCCL INFO Channel 06/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:586524 [1] NCCL INFO Channel 07/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:586520 [6] NCCL INFO Channel 07/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:586522 [2] NCCL INFO Channel 03/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:586526 [5] NCCL INFO Channel 07/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:586524 [1] NCCL INFO Channel 08/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:586520 [6] NCCL INFO Channel 08/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:586522 [2] NCCL INFO Channel 04/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:586526 [5] NCCL INFO Channel 08/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:586524 [1] NCCL INFO Channel 09/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:586520 [6] NCCL INFO Channel 09/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:586522 [2] NCCL INFO Channel 05/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:586526 [5] NCCL INFO Channel 09/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:586524 [1] NCCL INFO Channel 10/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:586520 [6] NCCL INFO Channel 10/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:586522 [2] NCCL INFO Channel 06/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:586526 [5] NCCL INFO Channel 10/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:586524 [1] NCCL INFO Channel 11/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:586526 [5] NCCL INFO Channel 11/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:586520 [6] NCCL INFO Channel 11/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:586522 [2] NCCL INFO Channel 07/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:586524 [1] NCCL INFO Channel 12/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:586526 [5] NCCL INFO Channel 12/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:586520 [6] NCCL INFO Channel 12/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:586522 [2] NCCL INFO Channel 08/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:586524 [1] NCCL INFO Channel 13/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:586526 [5] NCCL INFO Channel 13/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:586520 [6] NCCL INFO Channel 13/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:586524 [1] NCCL INFO Channel 14/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:586522 [2] NCCL INFO Channel 09/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:586526 [5] NCCL INFO Channel 14/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:586520 [6] NCCL INFO Channel 14/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:586524 [1] NCCL INFO Channel 15/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:586526 [5] NCCL INFO Channel 15/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:586522 [2] NCCL INFO Channel 10/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:586520 [6] NCCL INFO Channel 15/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:586524 [1] NCCL INFO Channel 16/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:586522 [2] NCCL INFO Channel 11/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:586520 [6] NCCL INFO Channel 16/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:586526 [5] NCCL INFO Channel 16/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:586522 [2] NCCL INFO Channel 12/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:586520 [6] NCCL INFO Channel 17/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:586524 [1] NCCL INFO Channel 17/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:586526 [5] NCCL INFO Channel 17/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:586522 [2] NCCL INFO Channel 13/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:586520 [6] NCCL INFO Channel 18/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:586524 [1] NCCL INFO Channel 18/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:586526 [5] NCCL INFO Channel 18/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:586522 [2] NCCL INFO Channel 14/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:586520 [6] NCCL INFO Channel 19/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:586524 [1] NCCL INFO Channel 19/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:586526 [5] NCCL INFO Channel 19/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:586520 [6] NCCL INFO Channel 20/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:586522 [2] NCCL INFO Channel 15/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:586524 [1] NCCL INFO Channel 20/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:586526 [5] NCCL INFO Channel 20/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:586520 [6] NCCL INFO Channel 21/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:586522 [2] NCCL INFO Channel 16/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:586524 [1] NCCL INFO Channel 21/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:586526 [5] NCCL INFO Channel 21/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:586520 [6] NCCL INFO Channel 22/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:586522 [2] NCCL INFO Channel 17/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:586524 [1] NCCL INFO Channel 22/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:586526 [5] NCCL INFO Channel 22/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:586520 [6] NCCL INFO Channel 23/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:586522 [2] NCCL INFO Channel 18/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:586524 [1] NCCL INFO Channel 23/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:586526 [5] NCCL INFO Channel 23/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:586522 [2] NCCL INFO Channel 19/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:586522 [2] NCCL INFO Channel 20/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:586522 [2] NCCL INFO Channel 21/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:586522 [2] NCCL INFO Channel 22/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:586522 [2] NCCL INFO Channel 23/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:586522 [2] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:579481:586524 [1] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:579480:586519 [0] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:579483:586521 [3] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:579484:586525 [4] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:579487:586523 [7] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:579487:586523 [7] NCCL INFO Channel 00/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:586526 [5] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:579486:586520 [6] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:579487:586523 [7] NCCL INFO Channel 01/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:586523 [7] NCCL INFO Channel 02/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:586523 [7] NCCL INFO Channel 03/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:586523 [7] NCCL INFO Channel 04/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:586523 [7] NCCL INFO Channel 05/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:586523 [7] NCCL INFO Channel 06/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:586523 [7] NCCL INFO Channel 07/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:586523 [7] NCCL INFO Channel 08/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:586522 [2] NCCL INFO Channel 00/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:586523 [7] NCCL INFO Channel 09/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:586522 [2] NCCL INFO Channel 01/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:586523 [7] NCCL INFO Channel 10/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:586522 [2] NCCL INFO Channel 02/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:586523 [7] NCCL INFO Channel 11/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:586522 [2] NCCL INFO Channel 03/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:586525 [4] NCCL INFO Channel 00/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:586525 [4] NCCL INFO Channel 01/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:586525 [4] NCCL INFO Channel 02/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:586525 [4] NCCL INFO Channel 03/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:586524 [1] NCCL INFO Channel 00/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:586525 [4] NCCL INFO Channel 04/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:586521 [3] NCCL INFO Channel 00/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:586524 [1] NCCL INFO Channel 01/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:586525 [4] NCCL INFO Channel 05/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:586521 [3] NCCL INFO Channel 01/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:586524 [1] NCCL INFO Channel 02/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:586525 [4] NCCL INFO Channel 06/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:586521 [3] NCCL INFO Channel 02/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:586524 [1] NCCL INFO Channel 03/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:586525 [4] NCCL INFO Channel 07/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:586524 [1] NCCL INFO Channel 04/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:586525 [4] NCCL INFO Channel 08/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:586524 [1] NCCL INFO Channel 05/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:586525 [4] NCCL INFO Channel 09/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:586524 [1] NCCL INFO Channel 06/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:586523 [7] NCCL INFO Channel 12/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:586525 [4] NCCL INFO Channel 10/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:586524 [1] NCCL INFO Channel 07/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:586523 [7] NCCL INFO Channel 13/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:586525 [4] NCCL INFO Channel 11/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:586520 [6] NCCL INFO Channel 00/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:586524 [1] NCCL INFO Channel 08/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:586523 [7] NCCL INFO Channel 14/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:586525 [4] NCCL INFO Channel 12/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:586522 [2] NCCL INFO Channel 04/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:586520 [6] NCCL INFO Channel 01/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:586524 [1] NCCL INFO Channel 09/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:586523 [7] NCCL INFO Channel 15/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:586525 [4] NCCL INFO Channel 13/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:586522 [2] NCCL INFO Channel 05/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:586520 [6] NCCL INFO Channel 02/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:586524 [1] NCCL INFO Channel 10/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:586523 [7] NCCL INFO Channel 16/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:586525 [4] NCCL INFO Channel 14/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:586522 [2] NCCL INFO Channel 06/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:586520 [6] NCCL INFO Channel 03/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:586524 [1] NCCL INFO Channel 11/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:586523 [7] NCCL INFO Channel 17/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:586525 [4] NCCL INFO Channel 15/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:586522 [2] NCCL INFO Channel 07/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:586520 [6] NCCL INFO Channel 04/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:586524 [1] NCCL INFO Channel 12/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:586523 [7] NCCL INFO Channel 18/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:586525 [4] NCCL INFO Channel 16/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:586522 [2] NCCL INFO Channel 08/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:586520 [6] NCCL INFO Channel 05/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:586524 [1] NCCL INFO Channel 13/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:586523 [7] NCCL INFO Channel 19/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:586525 [4] NCCL INFO Channel 17/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:586522 [2] NCCL INFO Channel 09/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:586520 [6] NCCL INFO Channel 06/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:586525 [4] NCCL INFO Channel 18/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:586522 [2] NCCL INFO Channel 10/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:586520 [6] NCCL INFO Channel 07/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:586525 [4] NCCL INFO Channel 19/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:586522 [2] NCCL INFO Channel 11/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:586520 [6] NCCL INFO Channel 08/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:586525 [4] NCCL INFO Channel 20/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:586522 [2] NCCL INFO Channel 12/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:586520 [6] NCCL INFO Channel 09/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:586525 [4] NCCL INFO Channel 21/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:586522 [2] NCCL INFO Channel 13/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:586520 [6] NCCL INFO Channel 10/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:586525 [4] NCCL INFO Channel 22/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:586522 [2] NCCL INFO Channel 14/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:586520 [6] NCCL INFO Channel 11/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579484:586525 [4] NCCL INFO Channel 23/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:586522 [2] NCCL INFO Channel 15/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:586520 [6] NCCL INFO Channel 12/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:586522 [2] NCCL INFO Channel 16/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:586520 [6] NCCL INFO Channel 13/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:586522 [2] NCCL INFO Channel 17/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:586520 [6] NCCL INFO Channel 14/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:586522 [2] NCCL INFO Channel 18/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:586521 [3] NCCL INFO Channel 03/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:586520 [6] NCCL INFO Channel 15/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:586522 [2] NCCL INFO Channel 19/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:586521 [3] NCCL INFO Channel 04/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:586526 [5] NCCL INFO Channel 00/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:586520 [6] NCCL INFO Channel 16/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:586522 [2] NCCL INFO Channel 20/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:586521 [3] NCCL INFO Channel 05/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:586526 [5] NCCL INFO Channel 01/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:586520 [6] NCCL INFO Channel 17/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:586522 [2] NCCL INFO Channel 21/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:586521 [3] NCCL INFO Channel 06/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:586526 [5] NCCL INFO Channel 02/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:586520 [6] NCCL INFO Channel 18/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:586522 [2] NCCL INFO Channel 22/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:586521 [3] NCCL INFO Channel 07/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:586526 [5] NCCL INFO Channel 03/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:586520 [6] NCCL INFO Channel 19/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579482:586522 [2] NCCL INFO Channel 23/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:586521 [3] NCCL INFO Channel 08/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:586524 [1] NCCL INFO Channel 14/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:586526 [5] NCCL INFO Channel 04/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:586520 [6] NCCL INFO Channel 20/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:586521 [3] NCCL INFO Channel 09/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:586524 [1] NCCL INFO Channel 15/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:586526 [5] NCCL INFO Channel 05/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:586520 [6] NCCL INFO Channel 21/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:586521 [3] NCCL INFO Channel 10/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:586526 [5] NCCL INFO Channel 06/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:586524 [1] NCCL INFO Channel 16/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:586520 [6] NCCL INFO Channel 22/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:586521 [3] NCCL INFO Channel 11/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:586526 [5] NCCL INFO Channel 07/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579486:586520 [6] NCCL INFO Channel 23/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:586524 [1] NCCL INFO Channel 17/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:586521 [3] NCCL INFO Channel 12/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:586526 [5] NCCL INFO Channel 08/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:586524 [1] NCCL INFO Channel 18/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:586521 [3] NCCL INFO Channel 13/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:586526 [5] NCCL INFO Channel 09/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:586524 [1] NCCL INFO Channel 19/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:586521 [3] NCCL INFO Channel 14/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:586526 [5] NCCL INFO Channel 10/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:586524 [1] NCCL INFO Channel 20/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:586521 [3] NCCL INFO Channel 15/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:586526 [5] NCCL INFO Channel 11/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:586524 [1] NCCL INFO Channel 21/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:586521 [3] NCCL INFO Channel 16/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:586526 [5] NCCL INFO Channel 12/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:586524 [1] NCCL INFO Channel 22/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:586521 [3] NCCL INFO Channel 17/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:586526 [5] NCCL INFO Channel 13/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579481:586524 [1] NCCL INFO Channel 23/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:586521 [3] NCCL INFO Channel 18/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:586526 [5] NCCL INFO Channel 14/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:586523 [7] NCCL INFO Channel 20/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:586521 [3] NCCL INFO Channel 19/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:586526 [5] NCCL INFO Channel 15/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:586523 [7] NCCL INFO Channel 21/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:586521 [3] NCCL INFO Channel 20/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:586526 [5] NCCL INFO Channel 16/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:586523 [7] NCCL INFO Channel 22/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:586526 [5] NCCL INFO Channel 17/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579487:586523 [7] NCCL INFO Channel 23/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:586521 [3] NCCL INFO Channel 21/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:586521 [3] NCCL INFO Channel 22/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:586526 [5] NCCL INFO Channel 18/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579483:586521 [3] NCCL INFO Channel 23/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:586526 [5] NCCL INFO Channel 19/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:586526 [5] NCCL INFO Channel 20/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:586526 [5] NCCL INFO Channel 21/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:586526 [5] NCCL INFO Channel 22/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579485:586526 [5] NCCL INFO Channel 23/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:579480:586519 [0] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:579480:586519 [0] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:579480:586519 [0] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:579481:586524 [1] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:579481:586524 [1] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:579481:586524 [1] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:579482:586522 [2] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:579482:586522 [2] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:579482:586522 [2] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:579483:586521 [3] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:579483:586521 [3] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:579483:586521 [3] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:579484:586525 [4] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:579484:586525 [4] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:579484:586525 [4] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:579485:586526 [5] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:579485:586526 [5] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:579485:586526 [5] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:579487:586523 [7] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:579487:586523 [7] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:579486:586520 [6] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:579487:586523 [7] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:579486:586520 [6] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:579486:586520 [6] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:579482:586522 [2] NCCL INFO ncclCommInitRank comm 0x7f4b9c06aa00 rank 2 nranks 8 cudaDev 2 nvmlDev 2 busId 201c0 commId 0x19e8e646f31205 - Init COMPLETE +ywang29-vrdb-test2-worker-0:579486:586520 [6] NCCL INFO ncclCommInitRank comm 0x7f2b4c06b080 rank 6 nranks 8 cudaDev 6 nvmlDev 6 busId a01c0 commId 0x19e8e646f31205 - Init COMPLETE +ywang29-vrdb-test2-worker-0:579480:586519 [0] NCCL INFO ncclCommInitRank comm 0x7f6cf806b780 rank 0 nranks 8 cudaDev 0 nvmlDev 0 busId 101c0 commId 0x19e8e646f31205 - Init COMPLETE +ywang29-vrdb-test2-worker-0:579487:586523 [7] NCCL INFO ncclCommInitRank comm 0x7f7ffc06a290 rank 7 nranks 8 cudaDev 7 nvmlDev 7 busId a01d0 commId 0x19e8e646f31205 - Init COMPLETE +ywang29-vrdb-test2-worker-0:579484:586525 [4] NCCL INFO ncclCommInitRank comm 0x7fdea406abd0 rank 4 nranks 8 cudaDev 4 nvmlDev 4 busId 901c0 commId 0x19e8e646f31205 - Init COMPLETE +ywang29-vrdb-test2-worker-0:579483:586521 [3] NCCL INFO ncclCommInitRank comm 0x7f60f806b640 rank 3 nranks 8 cudaDev 3 nvmlDev 3 busId 201d0 commId 0x19e8e646f31205 - Init COMPLETE +ywang29-vrdb-test2-worker-0:579485:586526 [5] NCCL INFO ncclCommInitRank comm 0x7f02b806aac0 rank 5 nranks 8 cudaDev 5 nvmlDev 5 busId 901d0 commId 0x19e8e646f31205 - Init COMPLETE +ywang29-vrdb-test2-worker-0:579481:586524 [1] NCCL INFO ncclCommInitRank comm 0x7f8e1806b160 rank 1 nranks 8 cudaDev 1 nvmlDev 1 busId 101d0 commId 0x19e8e646f31205 - Init COMPLETE + 0%| | 1/520 [00:14<2:01:28, 14.04s/it] {'loss': 2.0453, 'grad_norm': 0.004834215790793869, 'learning_rate': 0.000625, 'epoch': 0.0} + 0%| | 1/520 [00:14<2:01:28, 14.04s/it] 0%| | 2/520 [00:17<1:08:40, 7.96s/it] {'loss': 2.0549, 'grad_norm': 0.005249115840462403, 'learning_rate': 0.00125, 'epoch': 0.0} + 0%| | 2/520 [00:17<1:08:40, 7.96s/it] 1%| | 3/520 [00:21<51:33, 5.98s/it] {'loss': 2.1899, 'grad_norm': 0.0060071466655826095, 'learning_rate': 0.001875, 'epoch': 0.01} + 1%| | 3/520 [00:21<51:33, 5.98s/it] 1%| | 4/520 [00:25<43:37, 5.07s/it] {'loss': 2.0656, 'grad_norm': 0.004963473954214743, 'learning_rate': 0.0025, 'epoch': 0.01} + 1%| | 4/520 [00:25<43:37, 5.07s/it] 1%| | 5/520 [00:28<39:08, 4.56s/it] {'loss': 2.2333, 'grad_norm': 0.005482443975898494, 'learning_rate': 0.003125, 'epoch': 0.01} + 1%| | 5/520 [00:28<39:08, 4.56s/it] 1%| | 6/520 [00:32<36:35, 4.27s/it] {'loss': 1.6754, 'grad_norm': 0.0028034200924157864, 'learning_rate': 0.00375, 'epoch': 0.01} + 1%| | 6/520 [00:32<36:35, 4.27s/it] 1%|▏ | 7/520 [00:36<34:51, 4.08s/it] {'loss': 2.0776, 'grad_norm': 0.00541338890259897, 'learning_rate': 0.004375, 'epoch': 0.01} + 1%|▏ | 7/520 [00:36<34:51, 4.08s/it] 2%|▏ | 8/520 [00:40<35:19, 4.14s/it] {'loss': 2.0541, 'grad_norm': 0.004571814498953211, 'learning_rate': 0.005, 'epoch': 0.02} + 2%|▏ | 8/520 [00:40<35:19, 4.14s/it] 2%|▏ | 9/520 [00:44<35:17, 4.14s/it] {'loss': 2.19, 'grad_norm': 0.005019034751788077, 'learning_rate': 0.005625, 'epoch': 0.02} + 2%|▏ | 9/520 [00:44<35:17, 4.14s/it] 2%|▏ | 10/520 [00:48<33:53, 3.99s/it] {'loss': 2.0841, 'grad_norm': 0.005625156216572168, 'learning_rate': 0.00625, 'epoch': 0.02} + 2%|▏ | 10/520 [00:48<33:53, 3.99s/it] 2%|▏ | 11/520 [00:51<33:23, 3.94s/it] {'loss': 2.0582, 'grad_norm': 0.004956577675044716, 'learning_rate': 0.006875, 'epoch': 0.02} + 2%|▏ | 11/520 [00:51<33:23, 3.94s/it] 2%|▏ | 12/520 [00:55<32:39, 3.86s/it] {'loss': 1.879, 'grad_norm': 0.004359547240079958, 'learning_rate': 0.0075, 'epoch': 0.02} + 2%|▏ | 12/520 [00:55<32:39, 3.86s/it][2025-10-12 09:33:49,986] [WARNING] [stage3.py:2069:step] 1 pytorch allocator cache flushes since last step. this happens when there is high memory pressure and is detrimental to performance. if this is happening frequently consider adjusting settings to reduce memory consumption. If you are unable to make the cache flushes go away consider adding get_accelerator().empty_cache() calls in your training loop to ensure that all ranks flush their caches at the same time + 2%|▎ | 13/520 [00:59<33:46, 4.00s/it] {'loss': 2.0676, 'grad_norm': 0.00488581096049291, 'learning_rate': 0.008125, 'epoch': 0.03} + 2%|▎ | 13/520 [00:59<33:46, 4.00s/it] 3%|▎ | 14/520 [01:03<32:47, 3.89s/it] {'loss': 2.1081, 'grad_norm': 0.00498595024813787, 'learning_rate': 0.00875, 'epoch': 0.03} + 3%|▎ | 14/520 [01:03<32:47, 3.89s/it] 3%|▎ | 15/520 [01:07<32:24, 3.85s/it] {'loss': 1.7457, 'grad_norm': 0.002705656498577133, 'learning_rate': 0.009375, 'epoch': 0.03} + 3%|▎ | 15/520 [01:07<32:24, 3.85s/it] 3%|▎ | 16/520 [01:11<32:09, 3.83s/it] {'loss': 1.8915, 'grad_norm': 0.0041614506933553685, 'learning_rate': 0.01, 'epoch': 0.03} + 3%|▎ | 16/520 [01:11<32:09, 3.83s/it] 3%|▎ | 17/520 [01:14<31:57, 3.81s/it] {'loss': 2.1108, 'grad_norm': 0.00462694874851067, 'learning_rate': 0.00999990286465769, 'epoch': 0.03} + 3%|▎ | 17/520 [01:14<31:57, 3.81s/it] 3%|▎ | 18/520 [01:18<31:45, 3.80s/it] {'loss': 2.1647, 'grad_norm': 0.006087178667876502, 'learning_rate': 0.009999611462404873, 'epoch': 0.03} + 3%|▎ | 18/520 [01:18<31:45, 3.80s/it] 4%|▎ | 19/520 [01:22<31:36, 3.79s/it] {'loss': 1.603, 'grad_norm': 0.0013970674084673638, 'learning_rate': 0.009999125804563733, 'epoch': 0.04} + 4%|▎ | 19/520 [01:22<31:36, 3.79s/it] 4%|▍ | 20/520 [01:26<31:31, 3.78s/it] {'loss': 1.7046, 'grad_norm': 0.003357344515858102, 'learning_rate': 0.00999844591000408, 'epoch': 0.04} + 4%|▍ | 20/520 [01:26<31:31, 3.78s/it] 4%|▍ | 21/520 [01:29<31:24, 3.78s/it] {'loss': 1.584, 'grad_norm': 0.002425661034048465, 'learning_rate': 0.009997571805142639, 'epoch': 0.04} + 4%|▍ | 21/520 [01:29<31:24, 3.78s/it] 4%|▍ | 22/520 [01:33<31:31, 3.80s/it] {'loss': 1.648, 'grad_norm': 0.0011154353372775183, 'learning_rate': 0.009996503523941993, 'epoch': 0.04} + 4%|▍ | 22/520 [01:33<31:31, 3.80s/it] 4%|▍ | 23/520 [01:37<31:32, 3.81s/it] {'loss': 1.5705, 'grad_norm': 0.0008971002025416439, 'learning_rate': 0.00999524110790929, 'epoch': 0.04} + 4%|▍ | 23/520 [01:37<31:32, 3.81s/it] 5%|▍ | 24/520 [01:41<31:26, 3.80s/it] {'loss': 1.4762, 'grad_norm': 0.0007937698707290141, 'learning_rate': 0.00999378460609461, 'epoch': 0.05} + 5%|▍ | 24/520 [01:41<31:26, 3.80s/it] 5%|▍ | 25/520 [01:45<31:22, 3.80s/it] {'loss': 1.6096, 'grad_norm': 0.000900334852911342, 'learning_rate': 0.009992134075089084, 'epoch': 0.05} + 5%|▍ | 25/520 [01:45<31:22, 3.80s/it] 5%|▌ | 26/520 [01:49<31:18, 3.80s/it] {'loss': 1.4908, 'grad_norm': 0.0005935868495090288, 'learning_rate': 0.00999028957902266, 'epoch': 0.05} + 5%|▌ | 26/520 [01:49<31:18, 3.80s/it] 5%|▌ | 27/520 [01:52<31:12, 3.80s/it] {'loss': 1.4275, 'grad_norm': 0.0005596112275528387, 'learning_rate': 0.009988251189561644, 'epoch': 0.05} + 5%|▌ | 27/520 [01:52<31:12, 3.80s/it] 5%|▌ | 28/520 [01:56<31:06, 3.79s/it] {'loss': 1.4663, 'grad_norm': 0.0005686849391418991, 'learning_rate': 0.0099860189859059, 'epoch': 0.05} + 5%|▌ | 28/520 [01:56<31:06, 3.79s/it] 6%|▌ | 29/520 [02:00<31:03, 3.80s/it] {'loss': 1.4573, 'grad_norm': 0.000504282174257353, 'learning_rate': 0.009983593054785776, 'epoch': 0.06} + 6%|▌ | 29/520 [02:00<31:03, 3.80s/it] 6%|▌ | 30/520 [02:04<30:58, 3.79s/it] {'loss': 1.5062, 'grad_norm': 0.0004206802079201103, 'learning_rate': 0.009980973490458728, 'epoch': 0.06} + 6%|▌ | 30/520 [02:04<30:58, 3.79s/it] 6%|▌ | 31/520 [02:08<30:55, 3.79s/it] {'loss': 1.4272, 'grad_norm': 0.0004388605047781606, 'learning_rate': 0.009978160394705669, 'epoch': 0.06} + 6%|▌ | 31/520 [02:08<30:55, 3.79s/it] 6%|▌ | 32/520 [02:11<30:49, 3.79s/it] {'loss': 1.2891, 'grad_norm': 0.0003741459912809133, 'learning_rate': 0.009975153876827007, 'epoch': 0.06} + 6%|▌ | 32/520 [02:11<30:49, 3.79s/it] 6%|▋ | 33/520 [02:15<30:49, 3.80s/it] {'loss': 1.4431, 'grad_norm': 0.0004825917627172464, 'learning_rate': 0.0099719540536384, 'epoch': 0.06} + 6%|▋ | 33/520 [02:15<30:49, 3.80s/it] 7%|▋ | 34/520 [02:19<30:41, 3.79s/it] {'loss': 1.431, 'grad_norm': 0.0004918370287247212, 'learning_rate': 0.009968561049466213, 'epoch': 0.07} + 7%|▋ | 34/520 [02:19<30:41, 3.79s/it] 7%|▋ | 35/520 [02:23<30:35, 3.78s/it] {'loss': 1.4279, 'grad_norm': 0.0004631879446108963, 'learning_rate': 0.009964974996142698, 'epoch': 0.07} + 7%|▋ | 35/520 [02:23<30:35, 3.78s/it] 7%|▋ | 36/520 [02:26<30:36, 3.79s/it] {'loss': 1.5166, 'grad_norm': 0.00042047304766866197, 'learning_rate': 0.009961196033000861, 'epoch': 0.07} + 7%|▋ | 36/520 [02:26<30:36, 3.79s/it] 7%|▋ | 37/520 [02:30<30:34, 3.80s/it] {'loss': 1.462, 'grad_norm': 0.00034262540035649746, 'learning_rate': 0.009957224306869053, 'epoch': 0.07} + 7%|▋ | 37/520 [02:30<30:34, 3.80s/it] 7%|▋ | 38/520 [02:34<30:30, 3.80s/it] {'loss': 1.5766, 'grad_norm': 0.00038862721331337267, 'learning_rate': 0.009953059972065264, 'epoch': 0.07} + 7%|▋ | 38/520 [02:34<30:30, 3.80s/it] 8%|▊ | 39/520 [02:38<30:25, 3.80s/it] {'loss': 1.482, 'grad_norm': 0.0004711527938499943, 'learning_rate': 0.009948703190391131, 'epoch': 0.07} + 8%|▊ | 39/520 [02:38<30:25, 3.80s/it] 8%|▊ | 40/520 [02:42<30:19, 3.79s/it] {'loss': 1.4748, 'grad_norm': 0.00035429779352372703, 'learning_rate': 0.009944154131125642, 'epoch': 0.08} + 8%|▊ | 40/520 [02:42<30:19, 3.79s/it] 8%|▊ | 41/520 [02:45<30:07, 3.77s/it] {'loss': 1.4655, 'grad_norm': 0.00038448074106788537, 'learning_rate': 0.009939412971018573, 'epoch': 0.08} + 8%|▊ | 41/520 [02:45<30:07, 3.77s/it] 8%|▊ | 42/520 [02:49<29:42, 3.73s/it] {'loss': 1.4612, 'grad_norm': 0.0004299687238397088, 'learning_rate': 0.009934479894283605, 'epoch': 0.08} + 8%|▊ | 42/520 [02:49<29:42, 3.73s/it] 8%|▊ | 43/520 [02:53<29:27, 3.71s/it] {'loss': 1.3606, 'grad_norm': 0.0003454808087383135, 'learning_rate': 0.009929355092591179, 'epoch': 0.08} + 8%|▊ | 43/520 [02:53<29:27, 3.71s/it] 8%|▊ | 44/520 [02:56<29:19, 3.70s/it] {'loss': 1.46, 'grad_norm': 0.0003422278265820586, 'learning_rate': 0.00992403876506104, 'epoch': 0.08} + 8%|▊ | 44/520 [02:56<29:19, 3.70s/it] 9%|▊ | 45/520 [03:00<29:04, 3.67s/it] {'loss': 1.4829, 'grad_norm': 0.00039220516436831997, 'learning_rate': 0.009918531118254506, 'epoch': 0.09} + 9%|▊ | 45/520 [03:00<29:04, 3.67s/it] 9%|▉ | 46/520 [03:04<28:52, 3.66s/it] {'loss': 1.4802, 'grad_norm': 0.00030663617784567395, 'learning_rate': 0.009912832366166443, 'epoch': 0.09} + 9%|▉ | 46/520 [03:04<28:52, 3.66s/it] 9%|▉ | 47/520 [03:07<28:45, 3.65s/it] {'loss': 1.4441, 'grad_norm': 0.0003635511886713411, 'learning_rate': 0.00990694273021694, 'epoch': 0.09} + 9%|▉ | 47/520 [03:07<28:45, 3.65s/it] 9%|▉ | 48/520 [03:11<28:39, 3.64s/it] {'loss': 1.4809, 'grad_norm': 0.00041726319631894723, 'learning_rate': 0.00990086243924272, 'epoch': 0.09} + 9%|▉ | 48/520 [03:11<28:39, 3.64s/it] 9%|▉ | 49/520 [03:14<28:38, 3.65s/it] {'loss': 1.4906, 'grad_norm': 0.00035968811775258296, 'learning_rate': 0.009894591729488242, 'epoch': 0.09} + 9%|▉ | 49/520 [03:14<28:38, 3.65s/it] 10%|▉ | 50/520 [03:18<28:35, 3.65s/it] {'loss': 1.4923, 'grad_norm': 0.00036824175796084844, 'learning_rate': 0.009888130844596523, 'epoch': 0.1} + 10%|▉ | 50/520 [03:18<28:35, 3.65s/it] 10%|▉ | 51/520 [03:22<28:44, 3.68s/it] {'loss': 1.4196, 'grad_norm': 0.00038053202657458454, 'learning_rate': 0.009881480035599667, 'epoch': 0.1} + 10%|▉ | 51/520 [03:22<28:44, 3.68s/it] 10%|█ | 52/520 [03:26<28:42, 3.68s/it] {'loss': 1.5449, 'grad_norm': 0.00035116817688625034, 'learning_rate': 0.009874639560909117, 'epoch': 0.1} + 10%|█ | 52/520 [03:26<28:42, 3.68s/it] 10%|█ | 53/520 [03:29<28:31, 3.66s/it] {'loss': 1.5152, 'grad_norm': 0.00034240941806520697, 'learning_rate': 0.009867609686305616, 'epoch': 0.1} + 10%|█ | 53/520 [03:29<28:31, 3.66s/it] 10%|█ | 54/520 [03:33<28:26, 3.66s/it] {'loss': 1.4599, 'grad_norm': 0.0003374562655031863, 'learning_rate': 0.009860390684928872, 'epoch': 0.1} + 10%|█ | 54/520 [03:33<28:26, 3.66s/it] 11%|█ | 55/520 [03:37<28:26, 3.67s/it] {'loss': 1.4041, 'grad_norm': 0.0003569032874863551, 'learning_rate': 0.009852982837266955, 'epoch': 0.11} + 11%|█ | 55/520 [03:37<28:26, 3.67s/it] 11%|█ | 56/520 [03:40<28:20, 3.67s/it] {'loss': 1.5285, 'grad_norm': 0.0003407983361801125, 'learning_rate': 0.00984538643114539, 'epoch': 0.11} + 11%|█ | 56/520 [03:40<28:20, 3.67s/it] 11%|█ | 57/520 [03:44<28:14, 3.66s/it] {'loss': 1.4325, 'grad_norm': 0.00036398674196995277, 'learning_rate': 0.009837601761715982, 'epoch': 0.11} + 11%|█ | 57/520 [03:44<28:14, 3.66s/it] 11%|█ | 58/520 [03:47<28:09, 3.66s/it] {'loss': 1.5254, 'grad_norm': 0.00029002436731556295, 'learning_rate': 0.009829629131445342, 'epoch': 0.11} + 11%|█ | 58/520 [03:47<28:09, 3.66s/it] 11%|█▏ | 59/520 [03:51<28:04, 3.65s/it] {'loss': 1.3445, 'grad_norm': 0.000310865545166773, 'learning_rate': 0.009821468850103139, 'epoch': 0.11} + 11%|█▏ | 59/520 [03:51<28:04, 3.65s/it] 12%|█▏ | 60/520 [03:55<28:09, 3.67s/it] {'loss': 1.4483, 'grad_norm': 0.00031450869354591944, 'learning_rate': 0.009813121234750059, 'epoch': 0.12} + 12%|█▏ | 60/520 [03:55<28:09, 3.67s/it] 12%|█▏ | 61/520 [03:59<28:19, 3.70s/it] {'loss': 1.4077, 'grad_norm': 0.0003295745062539448, 'learning_rate': 0.009804586609725498, 'epoch': 0.12} + 12%|█▏ | 61/520 [03:59<28:19, 3.70s/it] 12%|█▏ | 62/520 [04:02<28:10, 3.69s/it] {'loss': 1.4302, 'grad_norm': 0.00033240383480787896, 'learning_rate': 0.009795865306634939, 'epoch': 0.12} + 12%|█▏ | 62/520 [04:02<28:10, 3.69s/it] 12%|█▏ | 63/520 [04:06<28:01, 3.68s/it] {'loss': 1.4414, 'grad_norm': 0.0003065941623040795, 'learning_rate': 0.009786957664337091, 'epoch': 0.12} + 12%|█▏ | 63/520 [04:06<28:01, 3.68s/it] 12%|█▏ | 64/520 [04:10<27:57, 3.68s/it] {'loss': 1.4646, 'grad_norm': 0.0003116321938236012, 'learning_rate': 0.009777864028930704, 'epoch': 0.12} + 12%|█▏ | 64/520 [04:10<27:57, 3.68s/it] 12%|█▎ | 65/520 [04:13<28:00, 3.69s/it] {'loss': 1.4753, 'grad_norm': 0.0003458615930817743, 'learning_rate': 0.009768584753741134, 'epoch': 0.12} + 12%|█▎ | 65/520 [04:13<28:00, 3.69s/it] 13%|█▎ | 66/520 [04:17<27:49, 3.68s/it] {'loss': 1.4044, 'grad_norm': 0.00031047648382440276, 'learning_rate': 0.009759120199306612, 'epoch': 0.13} + 13%|█▎ | 66/520 [04:17<27:49, 3.68s/it] 13%|█▎ | 67/520 [04:21<27:38, 3.66s/it] {'loss': 1.3353, 'grad_norm': 0.00030880943567941296, 'learning_rate': 0.00974947073336423, 'epoch': 0.13} + 13%|█▎ | 67/520 [04:21<27:38, 3.66s/it] 13%|█▎ | 68/520 [04:24<27:39, 3.67s/it] {'loss': 1.4037, 'grad_norm': 0.0003013026714396219, 'learning_rate': 0.009739636730835659, 'epoch': 0.13} + 13%|█▎ | 68/520 [04:24<27:39, 3.67s/it] 13%|█▎ | 69/520 [04:28<27:51, 3.71s/it] {'loss': 1.3985, 'grad_norm': 0.0003443801859134488, 'learning_rate': 0.00972961857381258, 'epoch': 0.13} + 13%|█▎ | 69/520 [04:28<27:51, 3.71s/it] 13%|█▎ | 70/520 [04:32<27:40, 3.69s/it] {'loss': 1.3969, 'grad_norm': 0.0002980219718030262, 'learning_rate': 0.009719416651541838, 'epoch': 0.13} + 13%|█▎ | 70/520 [04:32<27:40, 3.69s/it] 14%|█▎ | 71/520 [04:35<27:28, 3.67s/it] {'loss': 1.3531, 'grad_norm': 0.0003016235866208048, 'learning_rate': 0.009709031360410318, 'epoch': 0.14} + 14%|█▎ | 71/520 [04:35<27:28, 3.67s/it] 14%|█▍ | 72/520 [04:39<27:17, 3.66s/it] {'loss': 1.5072, 'grad_norm': 0.0003294946508100881, 'learning_rate': 0.009698463103929543, 'epoch': 0.14} + 14%|█▍ | 72/520 [04:39<27:17, 3.66s/it] 14%|█▍ | 73/520 [04:43<27:08, 3.64s/it] {'loss': 1.3158, 'grad_norm': 0.00030151372375898, 'learning_rate': 0.009687712292719997, 'epoch': 0.14} + 14%|█▍ | 73/520 [04:43<27:08, 3.64s/it] 14%|█▍ | 74/520 [04:46<27:06, 3.65s/it] {'loss': 1.4215, 'grad_norm': 0.0003371393715468394, 'learning_rate': 0.00967677934449517, 'epoch': 0.14} + 14%|█▍ | 74/520 [04:46<27:06, 3.65s/it] 14%|█▍ | 75/520 [04:50<26:57, 3.63s/it] {'loss': 1.3462, 'grad_norm': 0.000270273483920802, 'learning_rate': 0.009665664684045332, 'epoch': 0.14} + 14%|█▍ | 75/520 [04:50<26:57, 3.63s/it] 15%|█▍ | 76/520 [04:53<26:54, 3.64s/it] {'loss': 1.4252, 'grad_norm': 0.0002585114490822128, 'learning_rate': 0.009654368743221021, 'epoch': 0.15} + 15%|█▍ | 76/520 [04:54<26:54, 3.64s/it] 15%|█▍ | 77/520 [04:57<26:49, 3.63s/it] {'loss': 1.3012, 'grad_norm': 0.00035489179473574697, 'learning_rate': 0.009642891960916267, 'epoch': 0.15} + 15%|█▍ | 77/520 [04:57<26:49, 3.63s/it] 15%|█▌ | 78/520 [05:01<26:45, 3.63s/it] {'loss': 1.3809, 'grad_norm': 0.0003077204655698174, 'learning_rate': 0.009631234783051543, 'epoch': 0.15} + 15%|█▌ | 78/520 [05:01<26:45, 3.63s/it] 15%|█▌ | 79/520 [05:04<26:44, 3.64s/it] {'loss': 1.3576, 'grad_norm': 0.0002894278108181348, 'learning_rate': 0.009619397662556433, 'epoch': 0.15} + 15%|█▌ | 79/520 [05:04<26:44, 3.64s/it] 15%|█▌ | 80/520 [05:08<26:40, 3.64s/it] {'loss': 1.4088, 'grad_norm': 0.00029517850525809003, 'learning_rate': 0.009607381059352039, 'epoch': 0.15} + 15%|█▌ | 80/520 [05:08<26:40, 3.64s/it] 16%|█▌ | 81/520 [05:12<26:32, 3.63s/it] {'loss': 1.5237, 'grad_norm': 0.00037505275550736003, 'learning_rate': 0.009595185440333102, 'epoch': 0.16} + 16%|█▌ | 81/520 [05:12<26:32, 3.63s/it] 16%|█▌ | 82/520 [05:15<26:28, 3.63s/it] {'loss': 1.4342, 'grad_norm': 0.00030542296827305523, 'learning_rate': 0.009582811279349882, 'epoch': 0.16} + 16%|█▌ | 82/520 [05:15<26:28, 3.63s/it] 16%|█▌ | 83/520 [05:19<26:21, 3.62s/it] {'loss': 1.4277, 'grad_norm': 0.0003030444754510099, 'learning_rate': 0.009570259057189716, 'epoch': 0.16} + 16%|█▌ | 83/520 [05:19<26:21, 3.62s/it] 16%|█▌ | 84/520 [05:23<26:30, 3.65s/it] {'loss': 1.4571, 'grad_norm': 0.0003199324463365123, 'learning_rate': 0.009557529261558366, 'epoch': 0.16} + 16%|█▌ | 84/520 [05:23<26:30, 3.65s/it] 16%|█▋ | 85/520 [05:26<26:46, 3.69s/it] {'loss': 1.5068, 'grad_norm': 0.0002810738784247904, 'learning_rate': 0.009544622387061054, 'epoch': 0.16} + 16%|█▋ | 85/520 [05:26<26:46, 3.69s/it] 17%|█▋ | 86/520 [05:30<26:58, 3.73s/it] {'loss': 1.4758, 'grad_norm': 0.00029023224954142867, 'learning_rate': 0.009531538935183251, 'epoch': 0.17} + 17%|█▋ | 86/520 [05:30<26:58, 3.73s/it] 17%|█▋ | 87/520 [05:34<27:01, 3.75s/it] {'loss': 1.3742, 'grad_norm': 0.00027760144844236385, 'learning_rate': 0.009518279414271184, 'epoch': 0.17} + 17%|█▋ | 87/520 [05:34<27:01, 3.75s/it] 17%|█▋ | 88/520 [05:38<27:08, 3.77s/it] {'loss': 1.2614, 'grad_norm': 0.0002491647662007423, 'learning_rate': 0.009504844339512096, 'epoch': 0.17} + 17%|█▋ | 88/520 [05:38<27:08, 3.77s/it] 17%|█▋ | 89/520 [05:41<26:50, 3.74s/it] {'loss': 1.4458, 'grad_norm': 0.0003172206710919181, 'learning_rate': 0.00949123423291422, 'epoch': 0.17} + 17%|█▋ | 89/520 [05:41<26:50, 3.74s/it] 17%|█▋ | 90/520 [05:45<26:39, 3.72s/it] {'loss': 1.3865, 'grad_norm': 0.0003185082386909959, 'learning_rate': 0.009477449623286505, 'epoch': 0.17} + 17%|█▋ | 90/520 [05:45<26:39, 3.72s/it] 18%|█▊ | 91/520 [05:49<26:22, 3.69s/it] {'loss': 1.4304, 'grad_norm': 0.0002757721573990466, 'learning_rate': 0.009463491046218058, 'epoch': 0.17} + 18%|█▊ | 91/520 [05:49<26:22, 3.69s/it] 18%|█▊ | 92/520 [05:52<26:13, 3.68s/it] {'loss': 1.37, 'grad_norm': 0.0003126374465060212, 'learning_rate': 0.009449359044057344, 'epoch': 0.18} + 18%|█▊ | 92/520 [05:52<26:13, 3.68s/it] 18%|█▊ | 93/520 [05:56<26:03, 3.66s/it] {'loss': 1.3962, 'grad_norm': 0.00031847983569381063, 'learning_rate': 0.009435054165891108, 'epoch': 0.18} + 18%|█▊ | 93/520 [05:56<26:03, 3.66s/it] 18%|█▊ | 94/520 [06:00<25:57, 3.66s/it] {'loss': 1.4721, 'grad_norm': 0.00037285869561422703, 'learning_rate': 0.00942057696752305, 'epoch': 0.18} + 18%|█▊ | 94/520 [06:00<25:57, 3.66s/it] 18%|█▊ | 95/520 [06:03<25:45, 3.64s/it] {'loss': 1.4044, 'grad_norm': 0.00034441779528408416, 'learning_rate': 0.009405928011452212, 'epoch': 0.18} + 18%|█▊ | 95/520 [06:03<25:45, 3.64s/it] 18%|█▊ | 96/520 [06:07<25:42, 3.64s/it] {'loss': 1.393, 'grad_norm': 0.00030403937214568057, 'learning_rate': 0.009391107866851142, 'epoch': 0.18} + 18%|█▊ | 96/520 [06:07<25:42, 3.64s/it] 19%|█▊ | 97/520 [06:11<25:43, 3.65s/it] {'loss': 1.3868, 'grad_norm': 0.0003572976290006637, 'learning_rate': 0.00937611710954377, 'epoch': 0.19} + 19%|█▊ | 97/520 [06:11<25:43, 3.65s/it] 19%|█▉ | 98/520 [06:14<25:40, 3.65s/it] {'loss': 1.3608, 'grad_norm': 0.00025988825550890684, 'learning_rate': 0.009360956321983027, 'epoch': 0.19} + 19%|█▉ | 98/520 [06:14<25:40, 3.65s/it] 19%|█▉ | 99/520 [06:18<25:36, 3.65s/it] {'loss': 1.3661, 'grad_norm': 0.00031572138280138064, 'learning_rate': 0.009345626093228232, 'epoch': 0.19} + 19%|█▉ | 99/520 [06:18<25:36, 3.65s/it] 19%|█▉ | 100/520 [06:22<25:30, 3.64s/it] {'loss': 1.3217, 'grad_norm': 0.0002740401398149855, 'learning_rate': 0.009330127018922194, 'epoch': 0.19} + 19%|█▉ | 100/520 [06:22<25:30, 3.64s/it] 19%|█▉ | 101/520 [06:25<25:22, 3.63s/it] {'loss': 1.3953, 'grad_norm': 0.0003036934009241118, 'learning_rate': 0.009314459701268065, 'epoch': 0.19} + 19%|█▉ | 101/520 [06:25<25:22, 3.63s/it] 20%|█▉ | 102/520 [06:29<25:16, 3.63s/it] {'loss': 1.4203, 'grad_norm': 0.00033675187375961164, 'learning_rate': 0.009298624749005951, 'epoch': 0.2} + 20%|█▉ | 102/520 [06:29<25:16, 3.63s/it] 20%|█▉ | 103/520 [06:32<25:13, 3.63s/it] {'loss': 1.3273, 'grad_norm': 0.00029794372732567983, 'learning_rate': 0.009282622777389258, 'epoch': 0.2} + 20%|█▉ | 103/520 [06:32<25:13, 3.63s/it] 20%|██ | 104/520 [06:36<25:16, 3.65s/it] {'loss': 1.4196, 'grad_norm': 0.0003112281153993208, 'learning_rate': 0.009266454408160778, 'epoch': 0.2} + 20%|██ | 104/520 [06:36<25:16, 3.65s/it] 20%|██ | 105/520 [06:40<25:20, 3.66s/it] {'loss': 1.3872, 'grad_norm': 0.0002784303949640728, 'learning_rate': 0.009250120269528546, 'epoch': 0.2} + 20%|██ | 105/520 [06:40<25:20, 3.66s/it] 20%|██ | 106/520 [06:43<25:15, 3.66s/it] {'loss': 1.3371, 'grad_norm': 0.0002494769285374635, 'learning_rate': 0.00923362099614142, 'epoch': 0.2} + 20%|██ | 106/520 [06:43<25:15, 3.66s/it] 21%|██ | 107/520 [06:47<25:14, 3.67s/it] {'loss': 1.3339, 'grad_norm': 0.0002898173325019802, 'learning_rate': 0.009216957229064428, 'epoch': 0.21} + 21%|██ | 107/520 [06:47<25:14, 3.67s/it] 21%|██ | 108/520 [06:51<25:13, 3.67s/it] {'loss': 1.3641, 'grad_norm': 0.00032031349686365273, 'learning_rate': 0.009200129615753859, 'epoch': 0.21} + 21%|██ | 108/520 [06:51<25:13, 3.67s/it] 21%|██ | 109/520 [06:54<25:08, 3.67s/it] {'loss': 1.2909, 'grad_norm': 0.0002577108748836273, 'learning_rate': 0.009183138810032099, 'epoch': 0.21} + 21%|██ | 109/520 [06:54<25:08, 3.67s/it] 21%|██ | 110/520 [06:58<25:06, 3.67s/it] {'loss': 1.5144, 'grad_norm': 0.00030020533247569005, 'learning_rate': 0.009165985472062245, 'epoch': 0.21} + 21%|██ | 110/520 [06:58<25:06, 3.67s/it] 21%|██▏ | 111/520 [07:02<25:07, 3.68s/it] {'loss': 1.5274, 'grad_norm': 0.000312793646819477, 'learning_rate': 0.009148670268322438, 'epoch': 0.21} + 21%|██▏ | 111/520 [07:02<25:07, 3.68s/it] 22%|██▏ | 112/520 [07:06<25:02, 3.68s/it] {'loss': 1.3836, 'grad_norm': 0.0002772909176120394, 'learning_rate': 0.009131193871579974, 'epoch': 0.22} + 22%|██▏ | 112/520 [07:06<25:02, 3.68s/it] 22%|██▏ | 113/520 [07:09<25:00, 3.69s/it] {'loss': 1.3089, 'grad_norm': 0.0002691476866070299, 'learning_rate': 0.009113556960865167, 'epoch': 0.22} + 22%|██▏ | 113/520 [07:09<25:00, 3.69s/it] 22%|██▏ | 114/520 [07:13<24:56, 3.69s/it] {'loss': 1.4157, 'grad_norm': 0.0003094172389664706, 'learning_rate': 0.00909576022144496, 'epoch': 0.22} + 22%|██▏ | 114/520 [07:13<24:56, 3.69s/it] 22%|██▏ | 115/520 [07:17<24:56, 3.69s/it] {'loss': 1.4699, 'grad_norm': 0.0003237654090902668, 'learning_rate': 0.009077804344796301, 'epoch': 0.22} + 22%|██▏ | 115/520 [07:17<24:56, 3.69s/it] 22%|██▏ | 116/520 [07:20<24:55, 3.70s/it] {'loss': 1.4917, 'grad_norm': 0.00029207953385605116, 'learning_rate': 0.009059690028579283, 'epoch': 0.22} + 22%|██▏ | 116/520 [07:20<24:55, 3.70s/it] 22%|██▎ | 117/520 [07:24<24:48, 3.69s/it] {'loss': 1.4597, 'grad_norm': 0.0003154373965650443, 'learning_rate': 0.009041417976610027, 'epoch': 0.23} + 22%|██▎ | 117/520 [07:24<24:48, 3.69s/it] 23%|██▎ | 118/520 [07:28<24:38, 3.68s/it] {'loss': 1.4074, 'grad_norm': 0.0002896675308553919, 'learning_rate': 0.009022988898833342, 'epoch': 0.23} + 23%|██▎ | 118/520 [07:28<24:38, 3.68s/it] 23%|██▎ | 119/520 [07:31<24:33, 3.67s/it] {'loss': 1.3355, 'grad_norm': 0.00029257107312441775, 'learning_rate': 0.009004403511295141, 'epoch': 0.23} + 23%|██▎ | 119/520 [07:31<24:33, 3.67s/it] 23%|██▎ | 120/520 [07:35<24:31, 3.68s/it] {'loss': 1.3729, 'grad_norm': 0.0003274315223751983, 'learning_rate': 0.008985662536114613, 'epoch': 0.23} + 23%|██▎ | 120/520 [07:35<24:31, 3.68s/it] 23%|██▎ | 121/520 [07:39<24:25, 3.67s/it] {'loss': 1.4074, 'grad_norm': 0.00031492142962194524, 'learning_rate': 0.008966766701456175, 'epoch': 0.23} + 23%|██▎ | 121/520 [07:39<24:25, 3.67s/it] 23%|██▎ | 122/520 [07:42<24:18, 3.67s/it] {'loss': 1.3358, 'grad_norm': 0.00030386053113815774, 'learning_rate': 0.008947716741501178, 'epoch': 0.23} + 23%|██▎ | 122/520 [07:42<24:18, 3.67s/it] 24%|██▎ | 123/520 [07:46<24:22, 3.68s/it] {'loss': 1.3669, 'grad_norm': 0.00026814341555306904, 'learning_rate': 0.008928513396419369, 'epoch': 0.24} + 24%|██▎ | 123/520 [07:46<24:22, 3.68s/it] 24%|██▍ | 124/520 [07:50<24:14, 3.67s/it] {'loss': 1.3756, 'grad_norm': 0.00033334246696235966, 'learning_rate': 0.008909157412340149, 'epoch': 0.24} + 24%|██▍ | 124/520 [07:50<24:14, 3.67s/it] 24%|██▍ | 125/520 [07:53<24:15, 3.68s/it] {'loss': 1.3716, 'grad_norm': 0.0003115454552448618, 'learning_rate': 0.008889649541323574, 'epoch': 0.24} + 24%|██▍ | 125/520 [07:53<24:15, 3.68s/it] 24%|██▍ | 126/520 [07:58<25:32, 3.89s/it] {'loss': 1.2968, 'grad_norm': 0.00026184754924993227, 'learning_rate': 0.008869990541331137, 'epoch': 0.24} + 24%|██▍ | 126/520 [07:58<25:32, 3.89s/it] 24%|██▍ | 127/520 [08:01<25:02, 3.82s/it] {'loss': 1.3755, 'grad_norm': 0.0003514569316146674, 'learning_rate': 0.008850181176196315, 'epoch': 0.24} + 24%|██▍ | 127/520 [08:01<25:02, 3.82s/it] 25%|██▍ | 128/520 [08:05<25:17, 3.87s/it] {'loss': 1.3915, 'grad_norm': 0.000321135389229085, 'learning_rate': 0.00883022221559489, 'epoch': 0.25} + 25%|██▍ | 128/520 [08:05<25:17, 3.87s/it] 25%|██▍ | 129/520 [08:09<25:17, 3.88s/it] {'loss': 1.373, 'grad_norm': 0.0002720205072945954, 'learning_rate': 0.008810114435015054, 'epoch': 0.25} + 25%|██▍ | 129/520 [08:09<25:17, 3.88s/it] 25%|██▌ | 130/520 [08:13<25:13, 3.88s/it] {'loss': 1.3759, 'grad_norm': 0.00027335067664140204, 'learning_rate': 0.008789858615727265, 'epoch': 0.25} + 25%|██▌ | 130/520 [08:13<25:13, 3.88s/it] 25%|██▌ | 131/520 [08:17<25:08, 3.88s/it] {'loss': 1.2643, 'grad_norm': 0.00026447201045865177, 'learning_rate': 0.008769455544753899, 'epoch': 0.25} + 25%|██▌ | 131/520 [08:17<25:08, 3.88s/it] 25%|██▌ | 132/520 [08:21<24:59, 3.86s/it] {'loss': 1.4492, 'grad_norm': 0.0003527253896990239, 'learning_rate': 0.008748906014838671, 'epoch': 0.25} + 25%|██▌ | 132/520 [08:21<24:59, 3.86s/it] 26%|██▌ | 133/520 [08:25<24:55, 3.87s/it] {'loss': 1.3698, 'grad_norm': 0.00037828976282186516, 'learning_rate': 0.008728210824415827, 'epoch': 0.26} + 26%|██▌ | 133/520 [08:25<24:55, 3.87s/it] 26%|██▌ | 134/520 [08:29<25:14, 3.92s/it] {'loss': 1.4275, 'grad_norm': 0.0003174387996481141, 'learning_rate': 0.008707370777579134, 'epoch': 0.26} + 26%|██▌ | 134/520 [08:29<25:14, 3.92s/it] 26%|██▌ | 135/520 [08:33<25:10, 3.92s/it] {'loss': 1.4811, 'grad_norm': 0.00031373075053569546, 'learning_rate': 0.00868638668405062, 'epoch': 0.26} + 26%|██▌ | 135/520 [08:33<25:10, 3.92s/it] 26%|██▌ | 136/520 [08:36<24:44, 3.87s/it] {'loss': 1.4418, 'grad_norm': 0.00030480847510918, 'learning_rate': 0.008665259359149132, 'epoch': 0.26} + 26%|██▌ | 136/520 [08:36<24:44, 3.87s/it] 26%|██▋ | 137/520 [08:40<24:17, 3.81s/it] {'loss': 1.369, 'grad_norm': 0.0003602678529457946, 'learning_rate': 0.008643989623758643, 'epoch': 0.26} + 26%|██▋ | 137/520 [08:40<24:17, 3.81s/it] 27%|██▋ | 138/520 [08:44<23:54, 3.76s/it] {'loss': 1.3626, 'grad_norm': 0.0002740770442490755, 'learning_rate': 0.008622578304296363, 'epoch': 0.27} + 27%|██▋ | 138/520 [08:44<23:54, 3.76s/it] 27%|██▋ | 139/520 [08:47<23:38, 3.72s/it] {'loss': 1.1943, 'grad_norm': 0.00028449173723487086, 'learning_rate': 0.008601026232680633, 'epoch': 0.27} + 27%|██▋ | 139/520 [08:47<23:38, 3.72s/it] 27%|██▋ | 140/520 [08:51<23:25, 3.70s/it] {'loss': 1.3274, 'grad_norm': 0.00024577637918773716, 'learning_rate': 0.008579334246298592, 'epoch': 0.27} + 27%|██▋ | 140/520 [08:51<23:25, 3.70s/it] 27%|██▋ | 141/520 [08:55<23:17, 3.69s/it] {'loss': 1.4371, 'grad_norm': 0.00028711609488188796, 'learning_rate': 0.008557503187973652, 'epoch': 0.27} + 27%|██▋ | 141/520 [08:55<23:17, 3.69s/it] 27%|██▋ | 142/520 [08:58<23:08, 3.67s/it] {'loss': 1.3405, 'grad_norm': 0.0002857471022483226, 'learning_rate': 0.008535533905932738, 'epoch': 0.27} + 27%|██▋ | 142/520 [08:58<23:08, 3.67s/it] 28%|██▊ | 143/520 [09:02<22:59, 3.66s/it] {'loss': 1.3695, 'grad_norm': 0.00033134500614292366, 'learning_rate': 0.008513427253773347, 'epoch': 0.28} + 28%|██▊ | 143/520 [09:02<22:59, 3.66s/it] 28%|██▊ | 144/520 [09:06<22:55, 3.66s/it] {'loss': 1.3898, 'grad_norm': 0.00034876929264290483, 'learning_rate': 0.008491184090430363, 'epoch': 0.28} + 28%|██▊ | 144/520 [09:06<22:55, 3.66s/it] 28%|██▊ | 145/520 [09:09<22:53, 3.66s/it] {'loss': 1.2891, 'grad_norm': 0.0003380244529169648, 'learning_rate': 0.008468805280142709, 'epoch': 0.28} + 28%|██▊ | 145/520 [09:09<22:53, 3.66s/it] 28%|██▊ | 146/520 [09:13<22:47, 3.66s/it] {'loss': 1.3868, 'grad_norm': 0.0003107889147221958, 'learning_rate': 0.008446291692419736, 'epoch': 0.28} + 28%|██▊ | 146/520 [09:13<22:47, 3.66s/it] 28%|██▊ | 147/520 [09:17<22:41, 3.65s/it] {'loss': 1.3304, 'grad_norm': 0.000355955851198305, 'learning_rate': 0.008423644202007467, 'epoch': 0.28} + 28%|██▊ | 147/520 [09:17<22:41, 3.65s/it] 28%|██▊ | 148/520 [09:20<22:38, 3.65s/it] {'loss': 1.3767, 'grad_norm': 0.0003140888897222954, 'learning_rate': 0.008400863688854596, 'epoch': 0.28} + 28%|██▊ | 148/520 [09:20<22:38, 3.65s/it] 29%|██▊ | 149/520 [09:24<22:34, 3.65s/it] {'loss': 1.3011, 'grad_norm': 0.00033224949589374426, 'learning_rate': 0.008377951038078302, 'epoch': 0.29} + 29%|██▊ | 149/520 [09:24<22:34, 3.65s/it] 29%|██▉ | 150/520 [09:28<22:31, 3.65s/it] {'loss': 1.5058, 'grad_norm': 0.00033889219971937123, 'learning_rate': 0.00835490713992985, 'epoch': 0.29} + 29%|██▉ | 150/520 [09:28<22:31, 3.65s/it] 29%|██▉ | 151/520 [09:31<22:24, 3.64s/it] {'loss': 1.3602, 'grad_norm': 0.0003321010871123408, 'learning_rate': 0.00833173288976002, 'epoch': 0.29} + 29%|██▉ | 151/520 [09:31<22:24, 3.64s/it] 29%|██▉ | 152/520 [09:35<22:20, 3.64s/it] {'loss': 1.3321, 'grad_norm': 0.0003368995364096976, 'learning_rate': 0.008308429187984297, 'epoch': 0.29} + 29%|██▉ | 152/520 [09:35<22:20, 3.64s/it] 29%|██▉ | 153/520 [09:38<22:15, 3.64s/it] {'loss': 1.3526, 'grad_norm': 0.0003308038034124047, 'learning_rate': 0.008284996940047903, 'epoch': 0.29} + 29%|██▉ | 153/520 [09:38<22:15, 3.64s/it] 30%|██▉ | 154/520 [09:42<22:10, 3.63s/it] {'loss': 1.4258, 'grad_norm': 0.00029904314055525334, 'learning_rate': 0.008261437056390606, 'epoch': 0.3} + 30%|██▉ | 154/520 [09:42<22:10, 3.63s/it] 30%|██▉ | 155/520 [09:46<22:11, 3.65s/it] {'loss': 1.3481, 'grad_norm': 0.0003476391129528294, 'learning_rate': 0.008237750452411352, 'epoch': 0.3} + 30%|██▉ | 155/520 [09:46<22:11, 3.65s/it] 30%|███ | 156/520 [09:49<22:06, 3.64s/it] {'loss': 1.3765, 'grad_norm': 0.00036214944019093067, 'learning_rate': 0.008213938048432696, 'epoch': 0.3} + 30%|███ | 156/520 [09:49<22:06, 3.64s/it] 30%|███ | 157/520 [09:53<22:02, 3.64s/it] {'loss': 1.3924, 'grad_norm': 0.000315958698368938, 'learning_rate': 0.008190000769665043, 'epoch': 0.3} + 30%|███ | 157/520 [09:53<22:02, 3.64s/it] 30%|███ | 158/520 [09:57<22:03, 3.66s/it] {'loss': 1.3619, 'grad_norm': 0.00030851735152880147, 'learning_rate': 0.0081659395461707, 'epoch': 0.3} + 30%|███ | 158/520 [09:57<22:03, 3.66s/it] 31%|███ | 159/520 [10:00<22:01, 3.66s/it] {'loss': 1.4324, 'grad_norm': 0.0003354304903030481, 'learning_rate': 0.008141755312827736, 'epoch': 0.31} + 31%|███ | 159/520 [10:00<22:01, 3.66s/it] 31%|███ | 160/520 [10:04<22:19, 3.72s/it] {'loss': 1.4056, 'grad_norm': 0.0003471779342086279, 'learning_rate': 0.008117449009293669, 'epoch': 0.31} + 31%|███ | 160/520 [10:04<22:19, 3.72s/it] 31%|███ | 161/520 [10:08<22:32, 3.77s/it] {'loss': 1.4029, 'grad_norm': 0.0003336126932354692, 'learning_rate': 0.008093021579968942, 'epoch': 0.31} + 31%|███ | 161/520 [10:08<22:32, 3.77s/it] 31%|███ | 162/520 [10:12<22:39, 3.80s/it] {'loss': 1.3224, 'grad_norm': 0.0003084459853866861, 'learning_rate': 0.008068473973960237, 'epoch': 0.31} + 31%|███ | 162/520 [10:12<22:39, 3.80s/it] 31%|███▏ | 163/520 [10:16<22:44, 3.82s/it] {'loss': 1.3077, 'grad_norm': 0.000425591538011843, 'learning_rate': 0.008043807145043604, 'epoch': 0.31} + 31%|███▏ | 163/520 [10:16<22:44, 3.82s/it] 32%|███▏ | 164/520 [10:20<22:43, 3.83s/it] {'loss': 1.2534, 'grad_norm': 0.0003382114269677801, 'learning_rate': 0.008019022051627387, 'epoch': 0.32} + 32%|███▏ | 164/520 [10:20<22:43, 3.83s/it] 32%|███▏ | 165/520 [10:24<22:40, 3.83s/it] {'loss': 1.4366, 'grad_norm': 0.0003505735925327373, 'learning_rate': 0.007994119656715003, 'epoch': 0.32} + 32%|███▏ | 165/520 [10:24<22:40, 3.83s/it] 32%|███▏ | 166/520 [10:27<22:36, 3.83s/it] {'loss': 1.3843, 'grad_norm': 0.0003840619334354264, 'learning_rate': 0.007969100927867508, 'epoch': 0.32} + 32%|███▏ | 166/520 [10:27<22:36, 3.83s/it] 32%|███▏ | 167/520 [10:31<22:39, 3.85s/it] {'loss': 1.3666, 'grad_norm': 0.00035028570691400524, 'learning_rate': 0.007943966837166024, 'epoch': 0.32} + 32%|███▏ | 167/520 [10:31<22:39, 3.85s/it] 32%|███▏ | 168/520 [10:35<22:35, 3.85s/it] {'loss': 1.304, 'grad_norm': 0.0003205202900512857, 'learning_rate': 0.00791871836117395, 'epoch': 0.32} + 32%|███▏ | 168/520 [10:35<22:35, 3.85s/it] 32%|███▎ | 169/520 [10:39<22:36, 3.86s/it] {'loss': 1.3806, 'grad_norm': 0.00032420994791974693, 'learning_rate': 0.00789335648089903, 'epoch': 0.33} + 32%|███▎ | 169/520 [10:39<22:36, 3.86s/it] 33%|███▎ | 170/520 [10:43<22:32, 3.86s/it] {'loss': 1.2746, 'grad_norm': 0.00026367440140688303, 'learning_rate': 0.00786788218175523, 'epoch': 0.33} + 33%|███▎ | 170/520 [10:43<22:32, 3.86s/it] 33%|███▎ | 171/520 [10:47<22:34, 3.88s/it] {'loss': 1.3348, 'grad_norm': 0.0004136857075455587, 'learning_rate': 0.007842296453524463, 'epoch': 0.33} + 33%|███▎ | 171/520 [10:47<22:34, 3.88s/it] 33%|███▎ | 172/520 [10:51<22:26, 3.87s/it] {'loss': 1.3888, 'grad_norm': 0.0003496751325397971, 'learning_rate': 0.00781660029031811, 'epoch': 0.33} + 33%|███▎ | 172/520 [10:51<22:26, 3.87s/it] 33%|███▎ | 173/520 [10:55<22:18, 3.86s/it] {'loss': 1.3224, 'grad_norm': 0.0003570730370384107, 'learning_rate': 0.007790794690538421, 'epoch': 0.33} + 33%|███▎ | 173/520 [10:55<22:18, 3.86s/it] 33%|███▎ | 174/520 [10:58<21:56, 3.81s/it] {'loss': 1.3803, 'grad_norm': 0.00037574448956461347, 'learning_rate': 0.007764880656839696, 'epoch': 0.33} + 33%|███▎ | 174/520 [10:58<21:56, 3.81s/it] 34%|███▎ | 175/520 [11:02<21:37, 3.76s/it] {'loss': 1.287, 'grad_norm': 0.0003969489668879925, 'learning_rate': 0.007738859196089357, 'epoch': 0.34} + 34%|███▎ | 175/520 [11:02<21:37, 3.76s/it] 34%|███▍ | 176/520 [11:06<21:25, 3.74s/it] {'loss': 1.3364, 'grad_norm': 0.0003558522559069598, 'learning_rate': 0.007712731319328797, 'epoch': 0.34} + 34%|███▍ | 176/520 [11:06<21:25, 3.74s/it] 34%|███▍ | 177/520 [11:09<21:12, 3.71s/it] {'loss': 1.2364, 'grad_norm': 0.00031160630087074664, 'learning_rate': 0.0076864980417341204, 'epoch': 0.34} + 34%|███▍ | 177/520 [11:09<21:12, 3.71s/it] 34%|███▍ | 178/520 [11:13<21:01, 3.69s/it] {'loss': 1.3483, 'grad_norm': 0.00035248077338266847, 'learning_rate': 0.007660160382576683, 'epoch': 0.34} + 34%|███▍ | 178/520 [11:13<21:01, 3.69s/it] 34%|███▍ | 179/520 [11:16<20:53, 3.68s/it] {'loss': 1.4334, 'grad_norm': 0.00040817116842128563, 'learning_rate': 0.007633719365183504, 'epoch': 0.34} + 34%|███▍ | 179/520 [11:16<20:53, 3.68s/it] 35%|███▍ | 180/520 [11:20<20:45, 3.66s/it] {'loss': 1.3548, 'grad_norm': 0.00032581428693999505, 'learning_rate': 0.007607176016897491, 'epoch': 0.35} + 35%|███▍ | 180/520 [11:20<20:45, 3.66s/it] 35%|███▍ | 181/520 [11:24<20:40, 3.66s/it] {'loss': 1.3224, 'grad_norm': 0.0003046644785322441, 'learning_rate': 0.007580531369037533, 'epoch': 0.35} + 35%|███▍ | 181/520 [11:24<20:40, 3.66s/it] 35%|███▌ | 182/520 [11:27<20:35, 3.66s/it] {'loss': 1.3512, 'grad_norm': 0.00035791518431764187, 'learning_rate': 0.007553786456858429, 'epoch': 0.35} + 35%|███▌ | 182/520 [11:27<20:35, 3.66s/it] 35%|███▌ | 183/520 [11:31<20:33, 3.66s/it] {'loss': 1.3521, 'grad_norm': 0.0002892450010280373, 'learning_rate': 0.007526942319510655, 'epoch': 0.35} + 35%|███▌ | 183/520 [11:31<20:33, 3.66s/it] 35%|███▌ | 184/520 [11:35<20:25, 3.65s/it] {'loss': 1.3501, 'grad_norm': 0.0004161640206470938, 'learning_rate': 0.0075, 'epoch': 0.35} + 35%|███▌ | 184/520 [11:35<20:25, 3.65s/it] 36%|███▌ | 185/520 [11:38<20:21, 3.65s/it] {'loss': 1.401, 'grad_norm': 0.0003477383108531302, 'learning_rate': 0.007472960545147038, 'epoch': 0.36} + 36%|███▌ | 185/520 [11:38<20:21, 3.65s/it] 36%|███▌ | 186/520 [11:42<20:17, 3.64s/it] {'loss': 1.3545, 'grad_norm': 0.0003914856127471674, 'learning_rate': 0.0074458250055464475, 'epoch': 0.36} + 36%|███▌ | 186/520 [11:42<20:17, 3.64s/it] 36%|███▌ | 187/520 [11:46<20:15, 3.65s/it] {'loss': 1.3429, 'grad_norm': 0.0004905237560106394, 'learning_rate': 0.007418594435526199, 'epoch': 0.36} + 36%|███▌ | 187/520 [11:46<20:15, 3.65s/it] 36%|███▌ | 188/520 [11:49<20:08, 3.64s/it] {'loss': 1.4378, 'grad_norm': 0.0004148799541704753, 'learning_rate': 0.007391269893106591, 'epoch': 0.36} + 36%|███▌ | 188/520 [11:49<20:08, 3.64s/it] 36%|███▋ | 189/520 [11:53<20:12, 3.66s/it] {'loss': 1.4093, 'grad_norm': 0.0003195799867926795, 'learning_rate': 0.007363852439959136, 'epoch': 0.36} + 36%|███▋ | 189/520 [11:53<20:12, 3.66s/it] 37%|███▋ | 190/520 [11:57<20:20, 3.70s/it] {'loss': 1.3384, 'grad_norm': 0.0004374962836378826, 'learning_rate': 0.0073363431413653105, 'epoch': 0.37} + 37%|███▋ | 190/520 [11:57<20:20, 3.70s/it] 37%|███▋ | 191/520 [12:01<20:24, 3.72s/it] {'loss': 1.3178, 'grad_norm': 0.0003612803565215959, 'learning_rate': 0.007308743066175171, 'epoch': 0.37} + 37%|███▋ | 191/520 [12:01<20:24, 3.72s/it] 37%|███▋ | 192/520 [12:04<20:28, 3.75s/it] {'loss': 1.366, 'grad_norm': 0.0003207480151175892, 'learning_rate': 0.007281053286765815, 'epoch': 0.37} + 37%|███▋ | 192/520 [12:04<20:28, 3.75s/it] 37%|███▋ | 193/520 [12:08<20:31, 3.77s/it] {'loss': 1.2935, 'grad_norm': 0.0003514755532738094, 'learning_rate': 0.007253274878999727, 'epoch': 0.37} + 37%|███▋ | 193/520 [12:08<20:31, 3.77s/it] 37%|███▋ | 194/520 [12:12<20:33, 3.78s/it] {'loss': 1.1868, 'grad_norm': 0.00039861441816729505, 'learning_rate': 0.007225408922182961, 'epoch': 0.37} + 37%|███▋ | 194/520 [12:12<20:33, 3.78s/it] 38%|███▊ | 195/520 [12:16<20:29, 3.78s/it] {'loss': 1.3726, 'grad_norm': 0.00030846476644939, 'learning_rate': 0.007197456499023225, 'epoch': 0.38} + 38%|███▊ | 195/520 [12:16<20:29, 3.78s/it] 38%|███▊ | 196/520 [12:20<20:25, 3.78s/it] {'loss': 1.3983, 'grad_norm': 0.0003760115855576193, 'learning_rate': 0.007169418695587791, 'epoch': 0.38} + 38%|███▊ | 196/520 [12:20<20:25, 3.78s/it] 38%|███▊ | 197/520 [12:23<20:21, 3.78s/it] {'loss': 1.3217, 'grad_norm': 0.00044923827068069826, 'learning_rate': 0.007141296601261313, 'epoch': 0.38} + 38%|███▊ | 197/520 [12:23<20:21, 3.78s/it] 38%|███▊ | 198/520 [12:27<20:19, 3.79s/it] {'loss': 1.4143, 'grad_norm': 0.0003806470479756277, 'learning_rate': 0.007113091308703498, 'epoch': 0.38} + 38%|███▊ | 198/520 [12:27<20:19, 3.79s/it] 38%|███▊ | 199/520 [12:31<20:15, 3.79s/it] {'loss': 1.3148, 'grad_norm': 0.00042359478616503244, 'learning_rate': 0.007084803913806641, 'epoch': 0.38} + 38%|███▊ | 199/520 [12:31<20:15, 3.79s/it] 38%|███▊ | 200/520 [12:35<20:11, 3.79s/it] {'loss': 1.2468, 'grad_norm': 0.00038227016233417484, 'learning_rate': 0.007056435515653059, 'epoch': 0.38} + 38%|███▊ | 200/520 [12:35<20:11, 3.79s/it] 39%|███▊ | 201/520 [12:38<20:09, 3.79s/it] {'loss': 1.2595, 'grad_norm': 0.00034933212494368316, 'learning_rate': 0.007027987216472376, 'epoch': 0.39} + 39%|███▊ | 201/520 [12:38<20:09, 3.79s/it] 39%|███▉ | 202/520 [12:42<20:09, 3.80s/it] {'loss': 1.3551, 'grad_norm': 0.00038672102849766226, 'learning_rate': 0.006999460121598704, 'epoch': 0.39} + 39%|███▉ | 202/520 [12:42<20:09, 3.80s/it] 39%|███▉ | 203/520 [12:46<20:00, 3.79s/it] {'loss': 1.3638, 'grad_norm': 0.00036802043064536983, 'learning_rate': 0.0069708553394276975, 'epoch': 0.39} + 39%|███▉ | 203/520 [12:46<20:00, 3.79s/it] 39%|███▉ | 204/520 [12:50<19:44, 3.75s/it] {'loss': 1.3933, 'grad_norm': 0.0003968486910333053, 'learning_rate': 0.006942173981373474, 'epoch': 0.39} + 39%|███▉ | 204/520 [12:50<19:44, 3.75s/it] 39%|███▉ | 205/520 [12:53<19:35, 3.73s/it] {'loss': 1.2609, 'grad_norm': 0.0003319255043737861, 'learning_rate': 0.00691341716182545, 'epoch': 0.39} + 39%|███▉ | 205/520 [12:53<19:35, 3.73s/it] 40%|███▉ | 206/520 [12:57<19:25, 3.71s/it] {'loss': 1.3928, 'grad_norm': 0.0004368699306176923, 'learning_rate': 0.006884585998105026, 'epoch': 0.4} + 40%|███▉ | 206/520 [12:57<19:25, 3.71s/it] 40%|███▉ | 207/520 [13:01<19:18, 3.70s/it] {'loss': 1.2105, 'grad_norm': 0.0003087606684169389, 'learning_rate': 0.00685568161042219, 'epoch': 0.4} + 40%|███▉ | 207/520 [13:01<19:18, 3.70s/it] 40%|████ | 208/520 [13:04<19:10, 3.69s/it] {'loss': 1.3998, 'grad_norm': 0.0005736893961915962, 'learning_rate': 0.0068267051218319766, 'epoch': 0.4} + 40%|████ | 208/520 [13:04<19:10, 3.69s/it] 40%|████ | 209/520 [13:08<19:05, 3.68s/it] {'loss': 1.3098, 'grad_norm': 0.00038840659808171067, 'learning_rate': 0.006797657658190838, 'epoch': 0.4} + 40%|████ | 209/520 [13:08<19:05, 3.68s/it] 40%|████ | 210/520 [13:12<18:58, 3.67s/it] {'loss': 1.3746, 'grad_norm': 0.0003816904673885101, 'learning_rate': 0.006768540348112907, 'epoch': 0.4} + 40%|████ | 210/520 [13:12<18:58, 3.67s/it] 41%|████ | 211/520 [13:15<18:55, 3.67s/it] {'loss': 1.3674, 'grad_norm': 0.0004570078985228414, 'learning_rate': 0.006739354322926136, 'epoch': 0.41} + 41%|████ | 211/520 [13:15<18:55, 3.67s/it] 41%|████ | 212/520 [13:19<18:48, 3.66s/it] {'loss': 1.3814, 'grad_norm': 0.0004007420723847314, 'learning_rate': 0.006710100716628344, 'epoch': 0.41} + 41%|████ | 212/520 [13:19<18:48, 3.66s/it] 41%|████ | 213/520 [13:23<18:46, 3.67s/it] {'loss': 1.3336, 'grad_norm': 0.00041718937974087746, 'learning_rate': 0.006680780665843155, 'epoch': 0.41} + 41%|████ | 213/520 [13:23<18:46, 3.67s/it] 41%|████ | 214/520 [13:26<18:43, 3.67s/it] {'loss': 1.3424, 'grad_norm': 0.0004348102694601859, 'learning_rate': 0.006651395309775836, 'epoch': 0.41} + 41%|████ | 214/520 [13:26<18:43, 3.67s/it] 41%|████▏ | 215/520 [13:30<18:41, 3.68s/it] {'loss': 1.2007, 'grad_norm': 0.00036614370423521846, 'learning_rate': 0.006621945790169036, 'epoch': 0.41} + 41%|████▏ | 215/520 [13:30<18:41, 3.68s/it] 42%|████▏ | 216/520 [13:34<18:38, 3.68s/it] {'loss': 1.2448, 'grad_norm': 0.0004103482233242634, 'learning_rate': 0.006592433251258423, 'epoch': 0.42} + 42%|████▏ | 216/520 [13:34<18:38, 3.68s/it] 42%|████▏ | 217/520 [13:37<18:35, 3.68s/it] {'loss': 1.3744, 'grad_norm': 0.0004553716979897022, 'learning_rate': 0.006562858839728223, 'epoch': 0.42} + 42%|████▏ | 217/520 [13:37<18:35, 3.68s/it] 42%|████▏ | 218/520 [13:41<18:43, 3.72s/it] {'loss': 1.3554, 'grad_norm': 0.0004955348347744916, 'learning_rate': 0.006533223704666672, 'epoch': 0.42} + 42%|████▏ | 218/520 [13:41<18:43, 3.72s/it] 42%|████▏ | 219/520 [13:45<18:55, 3.77s/it] {'loss': 1.3742, 'grad_norm': 0.0003678708737847412, 'learning_rate': 0.006503528997521366, 'epoch': 0.42} + 42%|████▏ | 219/520 [13:45<18:55, 3.77s/it] 42%|████▏ | 220/520 [13:49<19:02, 3.81s/it] {'loss': 1.233, 'grad_norm': 0.00038240509641695574, 'learning_rate': 0.006473775872054522, 'epoch': 0.42} + 42%|████▏ | 220/520 [13:49<19:02, 3.81s/it] 42%|████▎ | 221/520 [13:53<19:06, 3.83s/it] {'loss': 1.3634, 'grad_norm': 0.0004255512487725784, 'learning_rate': 0.00644396548429815, 'epoch': 0.42} + 42%|████▎ | 221/520 [13:53<19:06, 3.83s/it] 43%|████▎ | 222/520 [13:57<19:09, 3.86s/it] {'loss': 1.2964, 'grad_norm': 0.0003930356434738575, 'learning_rate': 0.006414098992509137, 'epoch': 0.43} + 43%|████▎ | 222/520 [13:57<19:09, 3.86s/it] 43%|████▎ | 223/520 [14:01<19:07, 3.86s/it] {'loss': 1.313, 'grad_norm': 0.00042565292226225246, 'learning_rate': 0.0063841775571242465, 'epoch': 0.43} + 43%|████▎ | 223/520 [14:01<19:07, 3.86s/it] 43%|████▎ | 224/520 [14:05<19:07, 3.88s/it] {'loss': 1.2591, 'grad_norm': 0.00036315763161385146, 'learning_rate': 0.006354202340715026, 'epoch': 0.43} + 43%|████▎ | 224/520 [14:05<19:07, 3.88s/it] 43%|████▎ | 225/520 [14:09<19:07, 3.89s/it] {'loss': 1.3048, 'grad_norm': 0.0003919110713986617, 'learning_rate': 0.006324174507942636, 'epoch': 0.43} + 43%|████▎ | 225/520 [14:09<19:07, 3.89s/it] 43%|████▎ | 226/520 [14:12<19:02, 3.88s/it] {'loss': 1.4358, 'grad_norm': 0.00047437429745165804, 'learning_rate': 0.006294095225512604, 'epoch': 0.43} + 43%|████▎ | 226/520 [14:12<19:02, 3.88s/it] 44%|████▎ | 227/520 [14:16<18:58, 3.89s/it] {'loss': 1.3821, 'grad_norm': 0.0003829104818432132, 'learning_rate': 0.006263965662129487, 'epoch': 0.44} + 44%|████▎ | 227/520 [14:16<18:58, 3.89s/it] 44%|████▍ | 228/520 [14:20<18:55, 3.89s/it] {'loss': 1.3234, 'grad_norm': 0.00036436967759465564, 'learning_rate': 0.006233786988451467, 'epoch': 0.44} + 44%|████▍ | 228/520 [14:20<18:55, 3.89s/it] 44%|████▍ | 229/520 [14:24<18:50, 3.88s/it] {'loss': 1.3527, 'grad_norm': 0.00037202802599703986, 'learning_rate': 0.006203560377044865, 'epoch': 0.44} + 44%|████▍ | 229/520 [14:24<18:50, 3.88s/it] 44%|████▍ | 230/520 [14:28<18:48, 3.89s/it] {'loss': 1.263, 'grad_norm': 0.0004094704745796341, 'learning_rate': 0.006173287002338577, 'epoch': 0.44} + 44%|████▍ | 230/520 [14:28<18:48, 3.89s/it] 44%|████▍ | 231/520 [14:32<18:41, 3.88s/it] {'loss': 1.334, 'grad_norm': 0.0004313411660004114, 'learning_rate': 0.006142968040578448, 'epoch': 0.44} + 44%|████▍ | 231/520 [14:32<18:41, 3.88s/it] 45%|████▍ | 232/520 [14:36<18:34, 3.87s/it] {'loss': 1.3684, 'grad_norm': 0.0003885973847628582, 'learning_rate': 0.006112604669781572, 'epoch': 0.45} + 45%|████▍ | 232/520 [14:36<18:34, 3.87s/it] 45%|████▍ | 233/520 [14:40<18:29, 3.86s/it] {'loss': 1.2882, 'grad_norm': 0.0004643791191684537, 'learning_rate': 0.006082198069690514, 'epoch': 0.45} + 45%|████▍ | 233/520 [14:40<18:29, 3.86s/it] 45%|████▌ | 234/520 [14:43<18:24, 3.86s/it] {'loss': 1.2775, 'grad_norm': 0.0005222088884487007, 'learning_rate': 0.00605174942172748, 'epoch': 0.45} + 45%|████▌ | 234/520 [14:43<18:24, 3.86s/it] 45%|████▌ | 235/520 [14:47<18:20, 3.86s/it] {'loss': 1.3494, 'grad_norm': 0.0004690592911038061, 'learning_rate': 0.006021259908948402, 'epoch': 0.45} + 45%|████▌ | 235/520 [14:47<18:20, 3.86s/it] 45%|████▌ | 236/520 [14:51<18:19, 3.87s/it] {'loss': 1.3979, 'grad_norm': 0.0004612995650291491, 'learning_rate': 0.005990730715996988, 'epoch': 0.45} + 45%|████▌ | 236/520 [14:51<18:19, 3.87s/it] 46%|████▌ | 237/520 [14:55<18:16, 3.88s/it] {'loss': 1.3947, 'grad_norm': 0.0004196459490923467, 'learning_rate': 0.005960163029058682, 'epoch': 0.46} + 46%|████▌ | 237/520 [14:55<18:16, 3.88s/it] 46%|████▌ | 238/520 [14:59<18:10, 3.87s/it] {'loss': 1.3436, 'grad_norm': 0.00043265850262843176, 'learning_rate': 0.005929558035814574, 'epoch': 0.46} + 46%|████▌ | 238/520 [14:59<18:10, 3.87s/it] 46%|████▌ | 239/520 [15:03<18:09, 3.88s/it] {'loss': 1.3912, 'grad_norm': 0.0004234907084108366, 'learning_rate': 0.005898916925395264, 'epoch': 0.46} + 46%|████▌ | 239/520 [15:03<18:09, 3.88s/it] 46%|████▌ | 240/520 [15:07<17:54, 3.84s/it] {'loss': 1.2325, 'grad_norm': 0.0004725635457880219, 'learning_rate': 0.005868240888334653, 'epoch': 0.46} + 46%|████▌ | 240/520 [15:07<17:54, 3.84s/it] 46%|████▋ | 241/520 [15:10<17:35, 3.78s/it] {'loss': 1.325, 'grad_norm': 0.0003977501131936027, 'learning_rate': 0.005837531116523682, 'epoch': 0.46} + 46%|████▋ | 241/520 [15:10<17:35, 3.78s/it] 47%|████▋ | 242/520 [15:14<17:20, 3.74s/it] {'loss': 1.3067, 'grad_norm': 0.0003540761397514252, 'learning_rate': 0.005806788803164034, 'epoch': 0.47} + 47%|████▋ | 242/520 [15:14<17:20, 3.74s/it] 47%|████▋ | 243/520 [15:17<17:08, 3.71s/it] {'loss': 1.3171, 'grad_norm': 0.00038345510732708433, 'learning_rate': 0.005776015142721758, 'epoch': 0.47} + 47%|████▋ | 243/520 [15:17<17:08, 3.71s/it] 47%|████▋ | 244/520 [15:21<16:56, 3.68s/it] {'loss': 1.3846, 'grad_norm': 0.0004305866868903771, 'learning_rate': 0.005745211330880871, 'epoch': 0.47} + 47%|████▋ | 244/520 [15:21<16:56, 3.68s/it] 47%|████▋ | 245/520 [15:25<16:52, 3.68s/it] {'loss': 1.3057, 'grad_norm': 0.0004250289250104083, 'learning_rate': 0.005714378564496901, 'epoch': 0.47} + 47%|████▋ | 245/520 [15:25<16:52, 3.68s/it] 47%|████▋ | 246/520 [15:29<16:52, 3.69s/it] {'loss': 1.3564, 'grad_norm': 0.00039501022243078684, 'learning_rate': 0.0056835180415503676, 'epoch': 0.47} + 47%|████▋ | 246/520 [15:29<16:52, 3.69s/it] 48%|████▊ | 247/520 [15:32<16:45, 3.68s/it] {'loss': 1.4662, 'grad_norm': 0.0004554939096870537, 'learning_rate': 0.005652630961100259, 'epoch': 0.47} + 48%|████▊ | 247/520 [15:32<16:45, 3.68s/it] 48%|████▊ | 248/520 [15:36<16:38, 3.67s/it] {'loss': 1.2753, 'grad_norm': 0.00041016937147796557, 'learning_rate': 0.005621718523237427, 'epoch': 0.48} + 48%|████▊ | 248/520 [15:36<16:38, 3.67s/it] 48%|████▊ | 249/520 [15:39<16:34, 3.67s/it] {'loss': 1.3697, 'grad_norm': 0.0004349171365181869, 'learning_rate': 0.005590781929037965, 'epoch': 0.48} + 48%|████▊ | 249/520 [15:39<16:34, 3.67s/it] 48%|████▊ | 250/520 [15:43<16:33, 3.68s/it] {'loss': 1.3352, 'grad_norm': 0.0005204216312356105, 'learning_rate': 0.005559822380516539, 'epoch': 0.48} + 48%|████▊ | 250/520 [15:43<16:33, 3.68s/it] 48%|████▊ | 251/520 [15:47<16:29, 3.68s/it] {'loss': 1.3897, 'grad_norm': 0.00039019622617010634, 'learning_rate': 0.005528841080579689, 'epoch': 0.48} + 48%|████▊ | 251/520 [15:47<16:29, 3.68s/it] 48%|████▊ | 252/520 [15:51<16:24, 3.67s/it] {'loss': 1.2754, 'grad_norm': 0.0003735118791566679, 'learning_rate': 0.005497839232979084, 'epoch': 0.48} + 48%|████▊ | 252/520 [15:51<16:24, 3.67s/it] 49%|████▊ | 253/520 [15:54<16:20, 3.67s/it] {'loss': 1.362, 'grad_norm': 0.0004991796246349679, 'learning_rate': 0.0054668180422647525, 'epoch': 0.49} + 49%|████▊ | 253/520 [15:54<16:20, 3.67s/it] 49%|████▉ | 254/520 [15:58<16:17, 3.67s/it] {'loss': 1.3194, 'grad_norm': 0.00039391925936661874, 'learning_rate': 0.005435778713738292, 'epoch': 0.49} + 49%|████▉ | 254/520 [15:58<16:17, 3.67s/it] 49%|████▉ | 255/520 [16:02<16:21, 3.70s/it] {'loss': 1.3254, 'grad_norm': 0.0004923388938691125, 'learning_rate': 0.0054047224534060165, 'epoch': 0.49} + 49%|████▉ | 255/520 [16:02<16:21, 3.70s/it] 49%|████▉ | 256/520 [16:05<16:17, 3.70s/it] {'loss': 1.3988, 'grad_norm': 0.0004335550491843458, 'learning_rate': 0.005373650467932122, 'epoch': 0.49} + 49%|████▉ | 256/520 [16:05<16:17, 3.70s/it] 49%|████▉ | 257/520 [16:09<16:08, 3.68s/it] {'loss': 1.3609, 'grad_norm': 0.0004320266102492869, 'learning_rate': 0.005342563964591783, 'epoch': 0.49} + 49%|████▉ | 257/520 [16:09<16:08, 3.68s/it] 50%|████▉ | 258/520 [16:13<16:04, 3.68s/it] {'loss': 1.3608, 'grad_norm': 0.00038149015779216796, 'learning_rate': 0.0053114641512242615, 'epoch': 0.5} + 50%|████▉ | 258/520 [16:13<16:04, 3.68s/it] 50%|████▉ | 259/520 [16:16<16:00, 3.68s/it] {'loss': 1.4364, 'grad_norm': 0.0005250775678679614, 'learning_rate': 0.005280352236185959, 'epoch': 0.5} + 50%|████▉ | 259/520 [16:16<16:00, 3.68s/it] 50%|█████ | 260/520 [16:20<15:55, 3.68s/it] {'loss': 1.3137, 'grad_norm': 0.0003344794800234466, 'learning_rate': 0.005249229428303486, 'epoch': 0.5} + 50%|█████ | 260/520 [16:20<15:55, 3.68s/it] 50%|█████ | 261/520 [16:24<15:50, 3.67s/it] {'loss': 1.2671, 'grad_norm': 0.0004004462470468709, 'learning_rate': 0.005218096936826681, 'epoch': 0.5} + 50%|█████ | 261/520 [16:24<15:50, 3.67s/it] 50%|█████ | 262/520 [16:27<15:46, 3.67s/it] {'loss': 1.3084, 'grad_norm': 0.00046335083610826467, 'learning_rate': 0.00518695597138163, 'epoch': 0.5} + 50%|█████ | 262/520 [16:27<15:46, 3.67s/it] 51%|█████ | 263/520 [16:31<15:38, 3.65s/it] {'loss': 1.2979, 'grad_norm': 0.00043546928434412346, 'learning_rate': 0.005155807741923666, 'epoch': 0.51} + 51%|█████ | 263/520 [16:31<15:38, 3.65s/it] 51%|█████ | 264/520 [16:35<15:35, 3.65s/it] {'loss': 1.3818, 'grad_norm': 0.0003989601710459832, 'learning_rate': 0.005124653458690365, 'epoch': 0.51} + 51%|█████ | 264/520 [16:35<15:35, 3.65s/it] 51%|█████ | 265/520 [16:38<15:34, 3.66s/it] {'loss': 1.3462, 'grad_norm': 0.0005070148786447978, 'learning_rate': 0.005093494332154511, 'epoch': 0.51} + 51%|█████ | 265/520 [16:38<15:34, 3.66s/it] 51%|█████ | 266/520 [16:42<15:30, 3.66s/it] {'loss': 1.1997, 'grad_norm': 0.00043589511836962535, 'learning_rate': 0.005062331572977076, 'epoch': 0.51} + 51%|█████ | 266/520 [16:42<15:30, 3.66s/it] 51%|█████▏ | 267/520 [16:46<15:29, 3.67s/it] {'loss': 1.3092, 'grad_norm': 0.0005054433016104456, 'learning_rate': 0.005031166391960168, 'epoch': 0.51} + 51%|█████▏ | 267/520 [16:46<15:29, 3.67s/it] 52%|█████▏ | 268/520 [16:49<15:25, 3.67s/it] {'loss': 1.3993, 'grad_norm': 0.0005061150635406651, 'learning_rate': 0.005, 'epoch': 0.52} + 52%|█████▏ | 268/520 [16:49<15:25, 3.67s/it] 52%|█████▏ | 269/520 [16:53<15:20, 3.67s/it] {'loss': 1.4209, 'grad_norm': 0.00043659066177101515, 'learning_rate': 0.004968833608039832, 'epoch': 0.52} + 52%|█████▏ | 269/520 [16:53<15:20, 3.67s/it] 52%|█████▏ | 270/520 [16:57<15:20, 3.68s/it] {'loss': 1.2191, 'grad_norm': 0.000409702036333269, 'learning_rate': 0.004937668427022925, 'epoch': 0.52} + 52%|█████▏ | 270/520 [16:57<15:20, 3.68s/it] 52%|█████▏ | 271/520 [17:00<15:18, 3.69s/it] {'loss': 1.3935, 'grad_norm': 0.00046536654316933876, 'learning_rate': 0.00490650566784549, 'epoch': 0.52} + 52%|█████▏ | 271/520 [17:00<15:18, 3.69s/it] 52%|█████▏ | 272/520 [17:04<15:16, 3.69s/it] {'loss': 1.2501, 'grad_norm': 0.0004576112630985902, 'learning_rate': 0.004875346541309637, 'epoch': 0.52} + 52%|█████▏ | 272/520 [17:04<15:16, 3.69s/it] 52%|█████▎ | 273/520 [17:08<15:13, 3.70s/it] {'loss': 1.3214, 'grad_norm': 0.00045804048434895244, 'learning_rate': 0.004844192258076335, 'epoch': 0.53} + 52%|█████▎ | 273/520 [17:08<15:13, 3.70s/it] 53%|█████▎ | 274/520 [17:11<15:07, 3.69s/it] {'loss': 1.4029, 'grad_norm': 0.0006368182775964924, 'learning_rate': 0.004813044028618372, 'epoch': 0.53} + 53%|█████▎ | 274/520 [17:11<15:07, 3.69s/it] 53%|█████▎ | 275/520 [17:15<15:03, 3.69s/it] {'loss': 1.3219, 'grad_norm': 0.0004781289790057249, 'learning_rate': 0.00478190306317332, 'epoch': 0.53} + 53%|█████▎ | 275/520 [17:15<15:03, 3.69s/it] 53%|█████▎ | 276/520 [17:19<14:59, 3.69s/it] {'loss': 1.3843, 'grad_norm': 0.0004956414620791874, 'learning_rate': 0.0047507705716965136, 'epoch': 0.53} + 53%|█████▎ | 276/520 [17:19<14:59, 3.69s/it] 53%|█████▎ | 277/520 [17:23<14:56, 3.69s/it] {'loss': 1.3447, 'grad_norm': 0.00041847291762619934, 'learning_rate': 0.00471964776381404, 'epoch': 0.53} + 53%|█████▎ | 277/520 [17:23<14:56, 3.69s/it] 53%|█████▎ | 278/520 [17:26<14:50, 3.68s/it] {'loss': 1.2846, 'grad_norm': 0.00041030450608799595, 'learning_rate': 0.0046885358487757395, 'epoch': 0.53} + 53%|█████▎ | 278/520 [17:26<14:50, 3.68s/it] 54%|█████▎ | 279/520 [17:30<14:44, 3.67s/it] {'loss': 1.2509, 'grad_norm': 0.000594662333200628, 'learning_rate': 0.004657436035408217, 'epoch': 0.54} + 54%|█████▎ | 279/520 [17:30<14:44, 3.67s/it] 54%|█████▍ | 280/520 [17:33<14:40, 3.67s/it] {'loss': 1.328, 'grad_norm': 0.000503596340798819, 'learning_rate': 0.0046263495320678786, 'epoch': 0.54} + 54%|█████▍ | 280/520 [17:33<14:40, 3.67s/it] 54%|█████▍ | 281/520 [17:37<14:45, 3.71s/it] {'loss': 1.4096, 'grad_norm': 0.0005113355449874817, 'learning_rate': 0.004595277546593984, 'epoch': 0.54} + 54%|█████▍ | 281/520 [17:37<14:45, 3.71s/it] 54%|█████▍ | 282/520 [17:41<14:51, 3.75s/it] {'loss': 1.2962, 'grad_norm': 0.00044618612829297054, 'learning_rate': 0.004564221286261709, 'epoch': 0.54} + 54%|█████▍ | 282/520 [17:41<14:51, 3.75s/it] 54%|█████▍ | 283/520 [17:45<14:59, 3.79s/it] {'loss': 1.408, 'grad_norm': 0.0004463006731948915, 'learning_rate': 0.004533181957735248, 'epoch': 0.54} + 54%|█████▍ | 283/520 [17:45<14:59, 3.79s/it] 55%|█████▍ | 284/520 [17:49<15:00, 3.81s/it] {'loss': 1.2486, 'grad_norm': 0.0004836839321862439, 'learning_rate': 0.004502160767020917, 'epoch': 0.55} + 55%|█████▍ | 284/520 [17:49<15:00, 3.81s/it] 55%|█████▍ | 285/520 [17:53<14:59, 3.83s/it] {'loss': 1.3193, 'grad_norm': 0.0004261389249305528, 'learning_rate': 0.004471158919420312, 'epoch': 0.55} + 55%|█████▍ | 285/520 [17:53<14:59, 3.83s/it] 55%|█████▌ | 286/520 [17:57<14:57, 3.84s/it] {'loss': 1.1999, 'grad_norm': 0.0005314910859199066, 'learning_rate': 0.004440177619483461, 'epoch': 0.55} + 55%|█████▌ | 286/520 [17:57<14:57, 3.84s/it] 55%|█████▌ | 287/520 [18:00<14:55, 3.84s/it] {'loss': 1.425, 'grad_norm': 0.0005224367031617024, 'learning_rate': 0.004409218070962036, 'epoch': 0.55} + 55%|█████▌ | 287/520 [18:00<14:55, 3.84s/it] 55%|█████▌ | 288/520 [18:04<14:53, 3.85s/it] {'loss': 1.4472, 'grad_norm': 0.00041447444158558826, 'learning_rate': 0.004378281476762576, 'epoch': 0.55} + 55%|█████▌ | 288/520 [18:04<14:53, 3.85s/it] 56%|█████▌ | 289/520 [18:08<14:51, 3.86s/it] {'loss': 1.3545, 'grad_norm': 0.00047787404755839286, 'learning_rate': 0.004347369038899743, 'epoch': 0.56} + 56%|█████▌ | 289/520 [18:08<14:51, 3.86s/it] 56%|█████▌ | 290/520 [18:12<14:47, 3.86s/it] {'loss': 1.2554, 'grad_norm': 0.0004393286967452656, 'learning_rate': 0.004316481958449634, 'epoch': 0.56} + 56%|█████▌ | 290/520 [18:12<14:47, 3.86s/it] 56%|█████▌ | 291/520 [18:16<14:48, 3.88s/it] {'loss': 1.2847, 'grad_norm': 0.0005819173136679916, 'learning_rate': 0.004285621435503101, 'epoch': 0.56} + 56%|█████▌ | 291/520 [18:16<14:48, 3.88s/it] 56%|█████▌ | 292/520 [18:20<14:45, 3.88s/it] {'loss': 1.3621, 'grad_norm': 0.0005350637180637572, 'learning_rate': 0.004254788669119127, 'epoch': 0.56} + 56%|█████▌ | 292/520 [18:20<14:45, 3.88s/it] 56%|█████▋ | 293/520 [18:24<14:38, 3.87s/it] {'loss': 1.31, 'grad_norm': 0.0004865669097144128, 'learning_rate': 0.0042239848572782415, 'epoch': 0.56} + 56%|█████▋ | 293/520 [18:24<14:38, 3.87s/it] 57%|█████▋ | 294/520 [18:28<14:34, 3.87s/it] {'loss': 1.3326, 'grad_norm': 0.0004734349838445147, 'learning_rate': 0.0041932111968359664, 'epoch': 0.57} + 57%|█████▋ | 294/520 [18:28<14:34, 3.87s/it] 57%|█████▋ | 295/520 [18:31<14:27, 3.86s/it] {'loss': 1.2602, 'grad_norm': 0.0003902186356475534, 'learning_rate': 0.004162468883476319, 'epoch': 0.57} + 57%|█████▋ | 295/520 [18:31<14:27, 3.86s/it] 57%|█████▋ | 296/520 [18:35<14:20, 3.84s/it] {'loss': 1.2833, 'grad_norm': 0.00047734816509690847, 'learning_rate': 0.0041317591116653484, 'epoch': 0.57} + 57%|█████▋ | 296/520 [18:35<14:20, 3.84s/it] 57%|█████▋ | 297/520 [18:39<14:14, 3.83s/it] {'loss': 1.4199, 'grad_norm': 0.0005087400876824277, 'learning_rate': 0.004101083074604737, 'epoch': 0.57} + 57%|█████▋ | 297/520 [18:39<14:14, 3.83s/it] 57%|█████▋ | 298/520 [18:43<14:10, 3.83s/it] {'loss': 1.3658, 'grad_norm': 0.00045906417447843156, 'learning_rate': 0.004070441964185427, 'epoch': 0.57} + 57%|█████▋ | 298/520 [18:43<14:10, 3.83s/it] 57%|█████▊ | 299/520 [18:47<14:05, 3.83s/it] {'loss': 1.324, 'grad_norm': 0.0004408732437965751, 'learning_rate': 0.004039836970941319, 'epoch': 0.57} + 57%|█████▊ | 299/520 [18:47<14:05, 3.83s/it] 58%|█████▊ | 300/520 [18:51<14:04, 3.84s/it] {'loss': 1.4106, 'grad_norm': 0.0004752029409234229, 'learning_rate': 0.004009269284003013, 'epoch': 0.58} + 58%|█████▊ | 300/520 [18:51<14:04, 3.84s/it] 58%|█████▊ | 301/520 [18:54<13:59, 3.83s/it] {'loss': 1.4105, 'grad_norm': 0.00047291505179692853, 'learning_rate': 0.003978740091051599, 'epoch': 0.58} + 58%|█████▊ | 301/520 [18:54<13:59, 3.83s/it] 58%|█████▊ | 302/520 [18:58<13:55, 3.83s/it] {'loss': 1.3186, 'grad_norm': 0.000484182380466303, 'learning_rate': 0.003948250578272522, 'epoch': 0.58} + 58%|█████▊ | 302/520 [18:58<13:55, 3.83s/it] 58%|█████▊ | 303/520 [19:02<13:51, 3.83s/it] {'loss': 1.3297, 'grad_norm': 0.0005271325568331032, 'learning_rate': 0.003917801930309486, 'epoch': 0.58} + 58%|█████▊ | 303/520 [19:02<13:51, 3.83s/it] 58%|█████▊ | 304/520 [19:06<13:49, 3.84s/it] {'loss': 1.2484, 'grad_norm': 0.000508695830655986, 'learning_rate': 0.003887395330218428, 'epoch': 0.58} + 58%|█████▊ | 304/520 [19:06<13:49, 3.84s/it] 59%|█████▊ | 305/520 [19:10<13:43, 3.83s/it] {'loss': 1.4198, 'grad_norm': 0.0006219803166967275, 'learning_rate': 0.003857031959421553, 'epoch': 0.59} + 59%|█████▊ | 305/520 [19:10<13:43, 3.83s/it] 59%|█████▉ | 306/520 [19:14<13:40, 3.83s/it] {'loss': 1.3719, 'grad_norm': 0.000498501695233995, 'learning_rate': 0.003826712997661425, 'epoch': 0.59} + 59%|█████▉ | 306/520 [19:14<13:40, 3.83s/it] 59%|█████▉ | 307/520 [19:18<13:57, 3.93s/it] {'loss': 1.3052, 'grad_norm': 0.0005710758562673814, 'learning_rate': 0.0037964396229551362, 'epoch': 0.59} + 59%|█████▉ | 307/520 [19:18<13:57, 3.93s/it] 59%|█████▉ | 308/520 [19:22<13:47, 3.90s/it] {'loss': 1.4272, 'grad_norm': 0.00045023868609871316, 'learning_rate': 0.0037662130115485317, 'epoch': 0.59} + 59%|█████▉ | 308/520 [19:22<13:47, 3.90s/it] 59%|█████▉ | 309/520 [19:25<13:41, 3.89s/it] {'loss': 1.3226, 'grad_norm': 0.00043399713712825345, 'learning_rate': 0.003736034337870512, 'epoch': 0.59} + 59%|█████▉ | 309/520 [19:25<13:41, 3.89s/it] 60%|█████▉ | 310/520 [19:29<13:34, 3.88s/it] {'loss': 1.2803, 'grad_norm': 0.0004579212191040479, 'learning_rate': 0.003705904774487396, 'epoch': 0.6} + 60%|█████▉ | 310/520 [19:29<13:34, 3.88s/it] 60%|█████▉ | 311/520 [19:33<13:25, 3.85s/it] {'loss': 1.3009, 'grad_norm': 0.0005051659473786666, 'learning_rate': 0.003675825492057364, 'epoch': 0.6} + 60%|█████▉ | 311/520 [19:33<13:25, 3.85s/it] 60%|██████ | 312/520 [19:37<13:09, 3.79s/it] {'loss': 1.2766, 'grad_norm': 0.0004665911413269327, 'learning_rate': 0.0036457976592849752, 'epoch': 0.6} + 60%|██████ | 312/520 [19:37<13:09, 3.79s/it] 60%|██████ | 313/520 [19:40<12:57, 3.76s/it] {'loss': 1.2332, 'grad_norm': 0.0004199642804581971, 'learning_rate': 0.0036158224428757537, 'epoch': 0.6} + 60%|██████ | 313/520 [19:40<12:57, 3.76s/it] 60%|██████ | 314/520 [19:44<13:16, 3.87s/it] {'loss': 1.3035, 'grad_norm': 0.0005251113758978477, 'learning_rate': 0.003585901007490863, 'epoch': 0.6} + 60%|██████ | 314/520 [19:45<13:16, 3.87s/it] 61%|██████ | 315/520 [19:48<13:01, 3.81s/it] {'loss': 1.3072, 'grad_norm': 0.0005659750653833617, 'learning_rate': 0.0035560345157018515, 'epoch': 0.61} + 61%|██████ | 315/520 [19:48<13:01, 3.81s/it] 61%|██████ | 316/520 [19:52<13:18, 3.91s/it] {'loss': 1.2901, 'grad_norm': 0.0005462046570816072, 'learning_rate': 0.0035262241279454785, 'epoch': 0.61} + 61%|██████ | 316/520 [19:52<13:18, 3.91s/it] 61%|██████ | 317/520 [19:56<12:59, 3.84s/it] {'loss': 1.2656, 'grad_norm': 0.00047982823010562836, 'learning_rate': 0.003496471002478635, 'epoch': 0.61} + 61%|██████ | 317/520 [19:56<12:59, 3.84s/it] 61%|██████ | 318/520 [20:00<12:45, 3.79s/it] {'loss': 1.3988, 'grad_norm': 0.0005864391270649116, 'learning_rate': 0.0034667762953333294, 'epoch': 0.61} + 61%|██████ | 318/520 [20:00<12:45, 3.79s/it] 61%|██████▏ | 319/520 [20:04<12:51, 3.84s/it] {'loss': 1.261, 'grad_norm': 0.0004011612962661122, 'learning_rate': 0.0034371411602717784, 'epoch': 0.61} + 61%|██████▏ | 319/520 [20:04<12:51, 3.84s/it] 62%|██████▏ | 320/520 [20:07<12:36, 3.78s/it] {'loss': 1.1995, 'grad_norm': 0.0005690854296846111, 'learning_rate': 0.0034075667487415786, 'epoch': 0.62} + 62%|██████▏ | 320/520 [20:07<12:36, 3.78s/it] 62%|██████▏ | 321/520 [20:11<12:23, 3.74s/it] {'loss': 1.4174, 'grad_norm': 0.0004555007157227816, 'learning_rate': 0.0033780542098309652, 'epoch': 0.62} + 62%|██████▏ | 321/520 [20:11<12:23, 3.74s/it] 62%|██████▏ | 322/520 [20:15<12:15, 3.71s/it] {'loss': 1.1803, 'grad_norm': 0.0004931391528758471, 'learning_rate': 0.0033486046902241663, 'epoch': 0.62} + 62%|██████▏ | 322/520 [20:15<12:15, 3.71s/it] 62%|██████▏ | 323/520 [20:18<12:09, 3.70s/it] {'loss': 1.2761, 'grad_norm': 0.0005211527449973844, 'learning_rate': 0.003319219334156847, 'epoch': 0.62} + 62%|██████▏ | 323/520 [20:18<12:09, 3.70s/it] 62%|██████▏ | 324/520 [20:22<12:05, 3.70s/it] {'loss': 1.3574, 'grad_norm': 0.00045695917173147214, 'learning_rate': 0.0032898992833716566, 'epoch': 0.62} + 62%|██████▏ | 324/520 [20:22<12:05, 3.70s/it] 62%|██████▎ | 325/520 [20:26<12:10, 3.74s/it] {'loss': 1.3449, 'grad_norm': 0.0006431604641047298, 'learning_rate': 0.0032606456770738635, 'epoch': 0.62} + 62%|██████▎ | 325/520 [20:26<12:10, 3.74s/it] 63%|██████▎ | 326/520 [20:30<12:11, 3.77s/it] {'loss': 1.3569, 'grad_norm': 0.0005325096980171527, 'learning_rate': 0.0032314596518870932, 'epoch': 0.63} + 63%|██████▎ | 326/520 [20:30<12:11, 3.77s/it] 63%|██████▎ | 327/520 [20:33<12:10, 3.79s/it] {'loss': 1.2984, 'grad_norm': 0.0004950885108437451, 'learning_rate': 0.0032023423418091625, 'epoch': 0.63} + 63%|██████▎ | 327/520 [20:33<12:10, 3.79s/it] 63%|██████▎ | 328/520 [20:37<12:08, 3.79s/it] {'loss': 1.3969, 'grad_norm': 0.0005189321785775017, 'learning_rate': 0.003173294878168025, 'epoch': 0.63} + 63%|██████▎ | 328/520 [20:37<12:08, 3.79s/it] 63%|██████▎ | 329/520 [20:41<12:07, 3.81s/it] {'loss': 1.2713, 'grad_norm': 0.0004932365171969041, 'learning_rate': 0.0031443183895778107, 'epoch': 0.63} + 63%|██████▎ | 329/520 [20:41<12:07, 3.81s/it] 63%|██████▎ | 330/520 [20:45<12:12, 3.85s/it] {'loss': 1.3727, 'grad_norm': 0.0004846709254343869, 'learning_rate': 0.003115414001894974, 'epoch': 0.63} + 63%|██████▎ | 330/520 [20:45<12:12, 3.85s/it] 64%|██████▎ | 331/520 [20:49<12:07, 3.85s/it] {'loss': 1.308, 'grad_norm': 0.00045857764849418424, 'learning_rate': 0.0030865828381745515, 'epoch': 0.64} + 64%|██████▎ | 331/520 [20:49<12:07, 3.85s/it] 64%|██████▍ | 332/520 [20:53<12:02, 3.85s/it] {'loss': 1.3184, 'grad_norm': 0.0004241548418031458, 'learning_rate': 0.0030578260186265266, 'epoch': 0.64} + 64%|██████▍ | 332/520 [20:53<12:02, 3.85s/it] 64%|██████▍ | 333/520 [20:57<11:59, 3.85s/it] {'loss': 1.433, 'grad_norm': 0.0005043075412366843, 'learning_rate': 0.003029144660572304, 'epoch': 0.64} + 64%|██████▍ | 333/520 [20:57<11:59, 3.85s/it] 64%|██████▍ | 334/520 [21:00<11:53, 3.84s/it] {'loss': 1.3685, 'grad_norm': 0.0005406660971907602, 'learning_rate': 0.003000539878401296, 'epoch': 0.64} + 64%|██████▍ | 334/520 [21:00<11:53, 3.84s/it] 64%|██████▍ | 335/520 [21:04<11:48, 3.83s/it] {'loss': 1.353, 'grad_norm': 0.00048790808346506906, 'learning_rate': 0.0029720127835276256, 'epoch': 0.64} + 64%|██████▍ | 335/520 [21:04<11:48, 3.83s/it] 65%|██████▍ | 336/520 [21:08<11:45, 3.83s/it] {'loss': 1.2771, 'grad_norm': 0.0005965375449727902, 'learning_rate': 0.0029435644843469433, 'epoch': 0.65} + 65%|██████▍ | 336/520 [21:08<11:45, 3.83s/it] 65%|██████▍ | 337/520 [21:12<11:41, 3.83s/it] {'loss': 1.2692, 'grad_norm': 0.0005301012246136225, 'learning_rate': 0.002915196086193361, 'epoch': 0.65} + 65%|██████▍ | 337/520 [21:12<11:41, 3.83s/it] 65%|██████▌ | 338/520 [21:16<11:38, 3.84s/it] {'loss': 1.361, 'grad_norm': 0.0005829437815718346, 'learning_rate': 0.0028869086912965036, 'epoch': 0.65} + 65%|██████▌ | 338/520 [21:16<11:38, 3.84s/it] 65%|██████▌ | 339/520 [21:20<11:34, 3.83s/it] {'loss': 1.3116, 'grad_norm': 0.0005047131924041319, 'learning_rate': 0.0028587033987386855, 'epoch': 0.65} + 65%|██████▌ | 339/520 [21:20<11:34, 3.83s/it] 65%|██████▌ | 340/520 [21:23<11:30, 3.84s/it] {'loss': 1.2879, 'grad_norm': 0.0005346089937518355, 'learning_rate': 0.00283058130441221, 'epoch': 0.65} + 65%|██████▌ | 340/520 [21:23<11:30, 3.84s/it] 66%|██████▌ | 341/520 [21:27<11:26, 3.84s/it] {'loss': 1.3334, 'grad_norm': 0.0005454338611116659, 'learning_rate': 0.0028025435009767746, 'epoch': 0.66} + 66%|██████▌ | 341/520 [21:27<11:26, 3.84s/it] 66%|██████▌ | 342/520 [21:31<11:21, 3.83s/it] {'loss': 1.3313, 'grad_norm': 0.0005323154677515926, 'learning_rate': 0.002774591077817038, 'epoch': 0.66} + 66%|██████▌ | 342/520 [21:31<11:21, 3.83s/it] 66%|██████▌ | 343/520 [21:35<11:18, 3.83s/it] {'loss': 1.2302, 'grad_norm': 0.00044895396684055606, 'learning_rate': 0.002746725121000273, 'epoch': 0.66} + 66%|██████▌ | 343/520 [21:35<11:18, 3.83s/it] 66%|██████▌ | 344/520 [21:39<11:15, 3.84s/it] {'loss': 1.2676, 'grad_norm': 0.0005603158907172586, 'learning_rate': 0.002718946713234185, 'epoch': 0.66} + 66%|██████▌ | 344/520 [21:39<11:15, 3.84s/it] 66%|██████▋ | 345/520 [21:43<11:13, 3.85s/it] {'loss': 1.387, 'grad_norm': 0.0005270447220280124, 'learning_rate': 0.0026912569338248316, 'epoch': 0.66} + 66%|██████▋ | 345/520 [21:43<11:13, 3.85s/it] 67%|██████▋ | 346/520 [21:46<11:09, 3.85s/it] {'loss': 1.271, 'grad_norm': 0.0005434478963204814, 'learning_rate': 0.0026636568586346897, 'epoch': 0.67} + 67%|██████▋ | 346/520 [21:46<11:09, 3.85s/it] 67%|██████▋ | 347/520 [21:50<11:04, 3.84s/it] {'loss': 1.2972, 'grad_norm': 0.00045247630238323525, 'learning_rate': 0.002636147560040866, 'epoch': 0.67} + 67%|██████▋ | 347/520 [21:50<11:04, 3.84s/it]Token indices sequence length is longer than the specified maximum sequence length for this model (2778 > 2048). Running this sequence through the model will result in indexing errors + 67%|██████▋ | 348/520 [21:54<10:59, 3.83s/it] {'loss': 1.2555, 'grad_norm': 0.0006455542529512759, 'learning_rate': 0.0026087301068934104, 'epoch': 0.67} + 67%|██████▋ | 348/520 [21:54<10:59, 3.83s/it] 67%|██████▋ | 349/520 [21:58<10:57, 3.84s/it] {'loss': 1.2924, 'grad_norm': 0.0006033088824084121, 'learning_rate': 0.002581405564473801, 'epoch': 0.67} + 67%|██████▋ | 349/520 [21:58<10:57, 3.84s/it] 67%|██████▋ | 350/520 [22:02<10:51, 3.83s/it] {'loss': 1.3318, 'grad_norm': 0.0005183718701990698, 'learning_rate': 0.0025541749944535553, 'epoch': 0.67} + 67%|██████▋ | 350/520 [22:02<10:51, 3.83s/it] 68%|██████▊ | 351/520 [22:06<10:48, 3.84s/it] {'loss': 1.2402, 'grad_norm': 0.0004620239064107316, 'learning_rate': 0.002527039454852963, 'epoch': 0.68} + 68%|██████▊ | 351/520 [22:06<10:48, 3.84s/it] 68%|██████▊ | 352/520 [22:09<10:43, 3.83s/it] {'loss': 1.3609, 'grad_norm': 0.0004991583777423947, 'learning_rate': 0.0025000000000000014, 'epoch': 0.68} + 68%|██████▊ | 352/520 [22:09<10:43, 3.83s/it] 68%|██████▊ | 353/520 [22:13<10:37, 3.82s/it] {'loss': 1.2482, 'grad_norm': 0.0003989138197256084, 'learning_rate': 0.002473057680489348, 'epoch': 0.68} + 68%|██████▊ | 353/520 [22:13<10:37, 3.82s/it] 68%|██████▊ | 354/520 [22:17<10:30, 3.80s/it] {'loss': 1.3414, 'grad_norm': 0.00044062091235865663, 'learning_rate': 0.0024462135431415734, 'epoch': 0.68} + 68%|██████▊ | 354/520 [22:17<10:30, 3.80s/it] 68%|██████▊ | 355/520 [22:21<10:17, 3.74s/it] {'loss': 1.3195, 'grad_norm': 0.00046760865838666664, 'learning_rate': 0.0024194686309624664, 'epoch': 0.68} + 68%|██████▊ | 355/520 [22:21<10:17, 3.74s/it] 68%|██████▊ | 356/520 [22:24<10:07, 3.70s/it] {'loss': 1.3298, 'grad_norm': 0.0005384200336628944, 'learning_rate': 0.00239282398310251, 'epoch': 0.68} + 68%|██████▊ | 356/520 [22:24<10:07, 3.70s/it] 69%|██████▊ | 357/520 [22:28<09:59, 3.68s/it] {'loss': 1.3528, 'grad_norm': 0.0004523362566209494, 'learning_rate': 0.002366280634816496, 'epoch': 0.69} + 69%|██████▊ | 357/520 [22:28<09:59, 3.68s/it] 69%|██████▉ | 358/520 [22:31<09:54, 3.67s/it] {'loss': 1.273, 'grad_norm': 0.0004886138157574162, 'learning_rate': 0.0023398396174233176, 'epoch': 0.69} + 69%|██████▉ | 358/520 [22:31<09:54, 3.67s/it] 69%|██████▉ | 359/520 [22:35<09:51, 3.68s/it] {'loss': 1.282, 'grad_norm': 0.00049433845472355, 'learning_rate': 0.00231350195826588, 'epoch': 0.69} + 69%|██████▉ | 359/520 [22:35<09:51, 3.68s/it] 69%|██████▉ | 360/520 [22:39<09:46, 3.67s/it] {'loss': 1.2889, 'grad_norm': 0.00048427238144461174, 'learning_rate': 0.0022872686806712033, 'epoch': 0.69} + 69%|██████▉ | 360/520 [22:39<09:46, 3.67s/it] 69%|██████▉ | 361/520 [22:42<09:42, 3.67s/it] {'loss': 1.3094, 'grad_norm': 0.00044800265468065216, 'learning_rate': 0.0022611408039106442, 'epoch': 0.69} + 69%|██████▉ | 361/520 [22:42<09:42, 3.67s/it] 70%|██████▉ | 362/520 [22:46<09:38, 3.66s/it] {'loss': 1.3134, 'grad_norm': 0.0005127582909688809, 'learning_rate': 0.002235119343160303, 'epoch': 0.7} + 70%|██████▉ | 362/520 [22:46<09:38, 3.66s/it] 70%|██████▉ | 363/520 [22:50<09:34, 3.66s/it] {'loss': 1.3482, 'grad_norm': 0.00047679845104823724, 'learning_rate': 0.002209205309461581, 'epoch': 0.7} + 70%|██████▉ | 363/520 [22:50<09:34, 3.66s/it] 70%|███████ | 364/520 [22:53<09:33, 3.68s/it] {'loss': 1.3144, 'grad_norm': 0.000505246518225009, 'learning_rate': 0.0021833997096818896, 'epoch': 0.7} + 70%|███████ | 364/520 [22:53<09:33, 3.68s/it] 70%|███████ | 365/520 [22:57<09:28, 3.67s/it] {'loss': 1.389, 'grad_norm': 0.0004955891905701308, 'learning_rate': 0.002157703546475539, 'epoch': 0.7} + 70%|███████ | 365/520 [22:57<09:28, 3.67s/it] 70%|███████ | 366/520 [23:01<09:23, 3.66s/it] {'loss': 1.3807, 'grad_norm': 0.0004548172245899306, 'learning_rate': 0.0021321178182447708, 'epoch': 0.7} + 70%|███████ | 366/520 [23:01<09:23, 3.66s/it] 71%|███████ | 367/520 [23:04<09:19, 3.65s/it] {'loss': 1.3784, 'grad_norm': 0.0005667042658797965, 'learning_rate': 0.0021066435191009715, 'epoch': 0.71} + 71%|███████ | 367/520 [23:04<09:19, 3.65s/it] 71%|███████ | 368/520 [23:08<09:14, 3.65s/it] {'loss': 1.2117, 'grad_norm': 0.0005180137375122901, 'learning_rate': 0.002081281638826052, 'epoch': 0.71} + 71%|███████ | 368/520 [23:08<09:14, 3.65s/it] 71%|███████ | 369/520 [23:12<09:11, 3.65s/it] {'loss': 1.2786, 'grad_norm': 0.0004345257193144398, 'learning_rate': 0.002056033162833977, 'epoch': 0.71} + 71%|███████ | 369/520 [23:12<09:11, 3.65s/it] 71%|███████ | 370/520 [23:15<09:06, 3.64s/it] {'loss': 1.2784, 'grad_norm': 0.0005306723198998351, 'learning_rate': 0.0020308990721324928, 'epoch': 0.71} + 71%|███████ | 370/520 [23:15<09:06, 3.64s/it] 71%|███████▏ | 371/520 [23:19<09:03, 3.64s/it] {'loss': 1.2678, 'grad_norm': 0.0005639284614785732, 'learning_rate': 0.0020058803432849988, 'epoch': 0.71} + 71%|███████▏ | 371/520 [23:19<09:03, 3.64s/it] 72%|███████▏ | 372/520 [23:23<08:58, 3.64s/it] {'loss': 1.3273, 'grad_norm': 0.0004031138732291661, 'learning_rate': 0.001980977948372612, 'epoch': 0.72} + 72%|███████▏ | 372/520 [23:23<08:58, 3.64s/it] 72%|███████▏ | 373/520 [23:26<08:55, 3.64s/it] {'loss': 1.2375, 'grad_norm': 0.0005076417776017189, 'learning_rate': 0.0019561928549563967, 'epoch': 0.72} + 72%|███████▏ | 373/520 [23:26<08:55, 3.64s/it] 72%|███████▏ | 374/520 [23:30<08:51, 3.64s/it] {'loss': 1.36, 'grad_norm': 0.00048677604499043647, 'learning_rate': 0.0019315260260397637, 'epoch': 0.72} + 72%|███████▏ | 374/520 [23:30<08:51, 3.64s/it] 72%|███████▏ | 375/520 [23:34<08:47, 3.64s/it] {'loss': 1.2929, 'grad_norm': 0.0005173221580110328, 'learning_rate': 0.0019069784200310591, 'epoch': 0.72} + 72%|███████▏ | 375/520 [23:34<08:47, 3.64s/it] 72%|███████▏ | 376/520 [23:37<08:44, 3.64s/it] {'loss': 1.3872, 'grad_norm': 0.0004630789618872067, 'learning_rate': 0.0018825509907063327, 'epoch': 0.72} + 72%|███████▏ | 376/520 [23:37<08:44, 3.64s/it] 72%|███████▎ | 377/520 [23:41<08:41, 3.64s/it] {'loss': 1.2993, 'grad_norm': 0.0005711397926784965, 'learning_rate': 0.0018582446871722635, 'epoch': 0.72} + 72%|███████▎ | 377/520 [23:41<08:41, 3.64s/it] 73%|███████▎ | 378/520 [23:45<08:48, 3.72s/it] {'loss': 1.3794, 'grad_norm': 0.0004693684054049413, 'learning_rate': 0.0018340604538293016, 'epoch': 0.73} + 73%|███████▎ | 378/520 [23:45<08:48, 3.72s/it] 73%|███████▎ | 379/520 [23:49<08:52, 3.78s/it] {'loss': 1.3277, 'grad_norm': 0.00042542273102989277, 'learning_rate': 0.0018099992303349578, 'epoch': 0.73} + 73%|███████▎ | 379/520 [23:49<08:52, 3.78s/it] 73%|███████▎ | 380/520 [23:53<08:53, 3.81s/it] {'loss': 1.3199, 'grad_norm': 0.0005331318174118777, 'learning_rate': 0.0017860619515673033, 'epoch': 0.73} + 73%|███████▎ | 380/520 [23:53<08:53, 3.81s/it] 73%|███████▎ | 381/520 [23:56<08:52, 3.83s/it] {'loss': 1.3552, 'grad_norm': 0.000512486929900927, 'learning_rate': 0.0017622495475886485, 'epoch': 0.73} + 73%|███████▎ | 381/520 [23:56<08:52, 3.83s/it] 73%|███████▎ | 382/520 [24:00<08:52, 3.86s/it] {'loss': 1.2955, 'grad_norm': 0.00046471970538795285, 'learning_rate': 0.0017385629436093958, 'epoch': 0.73} + 73%|███████▎ | 382/520 [24:00<08:52, 3.86s/it] 74%|███████▎ | 383/520 [24:04<08:50, 3.87s/it] {'loss': 1.1962, 'grad_norm': 0.0006382016030622869, 'learning_rate': 0.0017150030599520983, 'epoch': 0.74} + 74%|███████▎ | 383/520 [24:04<08:50, 3.87s/it] 74%|███████▍ | 384/520 [24:08<08:46, 3.87s/it] {'loss': 1.2914, 'grad_norm': 0.000599099659028267, 'learning_rate': 0.0016915708120157041, 'epoch': 0.74} + 74%|███████▍ | 384/520 [24:08<08:46, 3.87s/it] 74%|███████▍ | 385/520 [24:12<08:43, 3.88s/it] {'loss': 1.3411, 'grad_norm': 0.00045566869934542185, 'learning_rate': 0.0016682671102399805, 'epoch': 0.74} + 74%|███████▍ | 385/520 [24:12<08:43, 3.88s/it] 74%|███████▍ | 386/520 [24:16<08:40, 3.88s/it] {'loss': 1.2948, 'grad_norm': 0.0004252311673154342, 'learning_rate': 0.0016450928600701504, 'epoch': 0.74} + 74%|███████▍ | 386/520 [24:16<08:40, 3.88s/it] 74%|███████▍ | 387/520 [24:20<08:38, 3.90s/it] {'loss': 1.3292, 'grad_norm': 0.00047860583126992833, 'learning_rate': 0.0016220489619216988, 'epoch': 0.74} + 74%|███████▍ | 387/520 [24:20<08:38, 3.90s/it] 75%|███████▍ | 388/520 [24:24<08:35, 3.90s/it] {'loss': 1.2576, 'grad_norm': 0.0005474372872311499, 'learning_rate': 0.0015991363111454021, 'epoch': 0.75} + 75%|███████▍ | 388/520 [24:24<08:35, 3.90s/it] 75%|███████▍ | 389/520 [24:28<08:30, 3.90s/it] {'loss': 1.3242, 'grad_norm': 0.0006457528050396208, 'learning_rate': 0.0015763557979925325, 'epoch': 0.75} + 75%|███████▍ | 389/520 [24:28<08:30, 3.90s/it] 75%|███████▌ | 390/520 [24:31<08:24, 3.88s/it] {'loss': 1.3703, 'grad_norm': 0.00046188650891372654, 'learning_rate': 0.0015537083075802649, 'epoch': 0.75} + 75%|███████▌ | 390/520 [24:31<08:24, 3.88s/it] 75%|███████▌ | 391/520 [24:35<08:11, 3.81s/it] {'loss': 1.4183, 'grad_norm': 0.0005114804616321118, 'learning_rate': 0.0015311947198572917, 'epoch': 0.75} + 75%|███████▌ | 391/520 [24:35<08:11, 3.81s/it] 75%|███████▌ | 392/520 [24:39<08:01, 3.76s/it] {'loss': 1.2716, 'grad_norm': 0.00044845728443202256, 'learning_rate': 0.0015088159095696363, 'epoch': 0.75} + 75%|███████▌ | 392/520 [24:39<08:01, 3.76s/it] 76%|███████▌ | 393/520 [24:42<07:53, 3.73s/it] {'loss': 1.2016, 'grad_norm': 0.00043060073264547706, 'learning_rate': 0.0014865727462266543, 'epoch': 0.76} + 76%|███████▌ | 393/520 [24:42<07:53, 3.73s/it] 76%|███████▌ | 394/520 [24:46<07:47, 3.71s/it] {'loss': 1.3454, 'grad_norm': 0.0005008539197560767, 'learning_rate': 0.0014644660940672626, 'epoch': 0.76} + 76%|███████▌ | 394/520 [24:46<07:47, 3.71s/it] 76%|███████▌ | 395/520 [24:50<07:41, 3.69s/it] {'loss': 1.308, 'grad_norm': 0.0005440993917548199, 'learning_rate': 0.0014424968120263504, 'epoch': 0.76} + 76%|███████▌ | 395/520 [24:50<07:41, 3.69s/it] 76%|███████▌ | 396/520 [24:53<07:35, 3.67s/it] {'loss': 1.3649, 'grad_norm': 0.0005244495294018863, 'learning_rate': 0.001420665753701408, 'epoch': 0.76} + 76%|███████▌ | 396/520 [24:53<07:35, 3.67s/it] 76%|███████▋ | 397/520 [24:57<07:31, 3.67s/it] {'loss': 1.3376, 'grad_norm': 0.0004662920645036704, 'learning_rate': 0.0013989737673193682, 'epoch': 0.76} + 76%|███████▋ | 397/520 [24:57<07:31, 3.67s/it] 77%|███████▋ | 398/520 [25:01<07:27, 3.67s/it] {'loss': 1.3535, 'grad_norm': 0.0005631135431592066, 'learning_rate': 0.0013774216957036368, 'epoch': 0.77} + 77%|███████▋ | 398/520 [25:01<07:27, 3.67s/it] 77%|███████▋ | 399/520 [25:04<07:23, 3.66s/it] {'loss': 1.2276, 'grad_norm': 0.00047900443750976854, 'learning_rate': 0.0013560103762413583, 'epoch': 0.77} + 77%|███████▋ | 399/520 [25:04<07:23, 3.66s/it] 77%|███████▋ | 400/520 [25:08<07:18, 3.66s/it] {'loss': 1.2764, 'grad_norm': 0.00044695749979880943, 'learning_rate': 0.0013347406408508694, 'epoch': 0.77} + 77%|███████▋ | 400/520 [25:08<07:18, 3.66s/it] 77%|███████▋ | 401/520 [25:12<07:15, 3.66s/it] {'loss': 1.1797, 'grad_norm': 0.0006158918588519499, 'learning_rate': 0.00131361331594938, 'epoch': 0.77} + 77%|███████▋ | 401/520 [25:12<07:15, 3.66s/it] 77%|███████▋ | 402/520 [25:15<07:10, 3.65s/it] {'loss': 1.3114, 'grad_norm': 0.0005260600122620925, 'learning_rate': 0.0012926292224208662, 'epoch': 0.77} + 77%|███████▋ | 402/520 [25:15<07:10, 3.65s/it] 78%|███████▊ | 403/520 [25:19<07:07, 3.65s/it] {'loss': 1.331, 'grad_norm': 0.0005685914495488634, 'learning_rate': 0.0012717891755841722, 'epoch': 0.78} + 78%|███████▊ | 403/520 [25:19<07:07, 3.65s/it] 78%|███████▊ | 404/520 [25:23<07:03, 3.65s/it] {'loss': 1.2643, 'grad_norm': 0.0006072192467456486, 'learning_rate': 0.0012510939851613284, 'epoch': 0.78} + 78%|███████▊ | 404/520 [25:23<07:03, 3.65s/it] 78%|███████▊ | 405/520 [25:26<06:59, 3.65s/it] {'loss': 1.253, 'grad_norm': 0.0006406188258090838, 'learning_rate': 0.0012305444552461009, 'epoch': 0.78} + 78%|███████▊ | 405/520 [25:26<06:59, 3.65s/it] 78%|███████▊ | 406/520 [25:30<06:56, 3.65s/it] {'loss': 1.1996, 'grad_norm': 0.0006491846708390076, 'learning_rate': 0.0012101413842727344, 'epoch': 0.78} + 78%|███████▊ | 406/520 [25:30<06:56, 3.65s/it] 78%|███████▊ | 407/520 [25:34<06:52, 3.65s/it] {'loss': 1.397, 'grad_norm': 0.0005383619834483161, 'learning_rate': 0.001189885564984946, 'epoch': 0.78} + 78%|███████▊ | 407/520 [25:34<06:52, 3.65s/it] 78%|███████▊ | 408/520 [25:37<06:49, 3.65s/it] {'loss': 1.3375, 'grad_norm': 0.0005777494110534441, 'learning_rate': 0.0011697777844051104, 'epoch': 0.78} + 78%|███████▊ | 408/520 [25:37<06:49, 3.65s/it] 79%|███████▊ | 409/520 [25:41<06:44, 3.65s/it] {'loss': 1.4709, 'grad_norm': 0.0005812569929647872, 'learning_rate': 0.001149818823803686, 'epoch': 0.79} + 79%|███████▊ | 409/520 [25:41<06:44, 3.65s/it] 79%|███████▉ | 410/520 [25:44<06:41, 3.65s/it] {'loss': 1.1955, 'grad_norm': 0.000572268773933964, 'learning_rate': 0.001130009458668863, 'epoch': 0.79} + 79%|███████▉ | 410/520 [25:44<06:41, 3.65s/it] 79%|███████▉ | 411/520 [25:48<06:37, 3.65s/it] {'loss': 1.4278, 'grad_norm': 0.0005754637369573751, 'learning_rate': 0.0011103504586764262, 'epoch': 0.79} + 79%|███████▉ | 411/520 [25:48<06:37, 3.65s/it] 79%|███████▉ | 412/520 [25:52<06:33, 3.65s/it] {'loss': 1.3326, 'grad_norm': 0.0004892864436726338, 'learning_rate': 0.001090842587659851, 'epoch': 0.79} + 79%|███████▉ | 412/520 [25:52<06:33, 3.65s/it] 79%|███████▉ | 413/520 [25:55<06:30, 3.65s/it] {'loss': 1.2815, 'grad_norm': 0.0006278470855687946, 'learning_rate': 0.0010714866035806325, 'epoch': 0.79} + 79%|███████▉ | 413/520 [25:55<06:30, 3.65s/it] 80%|███████▉ | 414/520 [25:59<06:27, 3.65s/it] {'loss': 1.0808, 'grad_norm': 0.0005245439392164385, 'learning_rate': 0.0010522832584988235, 'epoch': 0.8} + 80%|███████▉ | 414/520 [25:59<06:27, 3.65s/it] 80%|███████▉ | 415/520 [26:03<06:22, 3.65s/it] {'loss': 1.3169, 'grad_norm': 0.000552430634681182, 'learning_rate': 0.0010332332985438248, 'epoch': 0.8} + 80%|███████▉ | 415/520 [26:03<06:22, 3.65s/it] 80%|████████ | 416/520 [26:06<06:21, 3.67s/it] {'loss': 1.2399, 'grad_norm': 0.0005987433633236129, 'learning_rate': 0.0010143374638853892, 'epoch': 0.8} + 80%|████████ | 416/520 [26:06<06:21, 3.67s/it] 80%|████████ | 417/520 [26:10<06:22, 3.71s/it] {'loss': 1.3772, 'grad_norm': 0.0005125633138353078, 'learning_rate': 0.0009955964887048607, 'epoch': 0.8} + 80%|████████ | 417/520 [26:10<06:22, 3.71s/it] 80%|████████ | 418/520 [26:14<06:20, 3.73s/it] {'loss': 1.3757, 'grad_norm': 0.0005295549631524602, 'learning_rate': 0.0009770111011666582, 'epoch': 0.8} + 80%|████████ | 418/520 [26:14<06:20, 3.73s/it] 81%|████████ | 419/520 [26:18<06:18, 3.75s/it] {'loss': 1.39, 'grad_norm': 0.0005480437016091615, 'learning_rate': 0.0009585820233899739, 'epoch': 0.81} + 81%|████████ | 419/520 [26:18<06:18, 3.75s/it] 81%|████████ | 420/520 [26:22<06:16, 3.76s/it] {'loss': 1.2664, 'grad_norm': 0.0006385821218951522, 'learning_rate': 0.0009403099714207175, 'epoch': 0.81} + 81%|████████ | 420/520 [26:22<06:16, 3.76s/it] 81%|████████ | 421/520 [26:25<06:13, 3.77s/it] {'loss': 1.2084, 'grad_norm': 0.0006597205920922653, 'learning_rate': 0.0009221956552036992, 'epoch': 0.81} + 81%|████████ | 421/520 [26:25<06:13, 3.77s/it] 81%|████████ | 422/520 [26:29<06:09, 3.77s/it] {'loss': 1.328, 'grad_norm': 0.0005928915640309399, 'learning_rate': 0.0009042397785550405, 'epoch': 0.81} + 81%|████████ | 422/520 [26:29<06:09, 3.77s/it] 81%|████████▏ | 423/520 [26:33<06:05, 3.77s/it] {'loss': 1.3141, 'grad_norm': 0.000613005289935912, 'learning_rate': 0.0008864430391348333, 'epoch': 0.81} + 81%|████████▏ | 423/520 [26:33<06:05, 3.77s/it] 82%|████████▏ | 424/520 [26:37<06:02, 3.78s/it] {'loss': 1.3362, 'grad_norm': 0.00044746322964834813, 'learning_rate': 0.0008688061284200266, 'epoch': 0.82} + 82%|████████▏ | 424/520 [26:37<06:02, 3.78s/it] 82%|████████▏ | 425/520 [26:40<05:58, 3.78s/it] {'loss': 1.2843, 'grad_norm': 0.0005596974401272676, 'learning_rate': 0.0008513297316775626, 'epoch': 0.82} + 82%|████████▏ | 425/520 [26:41<05:58, 3.78s/it] 82%|████████▏ | 426/520 [26:44<05:52, 3.75s/it] {'loss': 1.3644, 'grad_norm': 0.0007346607932287136, 'learning_rate': 0.0008340145279377559, 'epoch': 0.82} + 82%|████████▏ | 426/520 [26:44<05:52, 3.75s/it] 82%|████████▏ | 427/520 [26:48<05:46, 3.72s/it] {'loss': 1.2216, 'grad_norm': 0.0005223022222029871, 'learning_rate': 0.0008168611899679012, 'epoch': 0.82} + 82%|████████▏ | 427/520 [26:48<05:46, 3.72s/it] 82%|████████▏ | 428/520 [26:51<05:40, 3.70s/it] {'loss': 1.2387, 'grad_norm': 0.0005175235567613663, 'learning_rate': 0.000799870384246143, 'epoch': 0.82} + 82%|████████▏ | 428/520 [26:51<05:40, 3.70s/it] 82%|████████▎ | 429/520 [26:55<05:34, 3.68s/it] {'loss': 1.3552, 'grad_norm': 0.000534537917723012, 'learning_rate': 0.0007830427709355725, 'epoch': 0.82} + 82%|████████▎ | 429/520 [26:55<05:34, 3.68s/it]Token indices sequence length is longer than the specified maximum sequence length for this model (2076 > 2048). Running this sequence through the model will result in indexing errors + 83%|████████▎ | 430/520 [26:59<05:33, 3.70s/it] {'loss': 1.3375, 'grad_norm': 0.00047335273751669324, 'learning_rate': 0.0007663790038585794, 'epoch': 0.83} + 83%|████████▎ | 430/520 [26:59<05:33, 3.70s/it] 83%|████████▎ | 431/520 [27:03<05:31, 3.72s/it] {'loss': 1.2372, 'grad_norm': 0.0006026562983080904, 'learning_rate': 0.0007498797304714544, 'epoch': 0.83} + 83%|████████▎ | 431/520 [27:03<05:31, 3.72s/it] 83%|████████▎ | 432/520 [27:06<05:29, 3.74s/it] {'loss': 1.2547, 'grad_norm': 0.0005169247947701919, 'learning_rate': 0.000733545591839222, 'epoch': 0.83} + 83%|████████▎ | 432/520 [27:06<05:29, 3.74s/it] 83%|████████▎ | 433/520 [27:10<05:26, 3.75s/it] {'loss': 1.372, 'grad_norm': 0.0005377617274835417, 'learning_rate': 0.0007173772226107434, 'epoch': 0.83} + 83%|████████▎ | 433/520 [27:10<05:26, 3.75s/it] 83%|████████▎ | 434/520 [27:14<05:23, 3.76s/it] {'loss': 1.1469, 'grad_norm': 0.0005199402788760457, 'learning_rate': 0.0007013752509940485, 'epoch': 0.83} + 83%|████████▎ | 434/520 [27:14<05:23, 3.76s/it] 84%|████████▎ | 435/520 [27:18<05:20, 3.77s/it] {'loss': 1.4245, 'grad_norm': 0.0005498365890772362, 'learning_rate': 0.0006855402987319348, 'epoch': 0.84} + 84%|████████▎ | 435/520 [27:18<05:20, 3.77s/it] 84%|████████▍ | 436/520 [27:22<05:16, 3.77s/it] {'loss': 1.2196, 'grad_norm': 0.0005828847723073406, 'learning_rate': 0.0006698729810778065, 'epoch': 0.84} + 84%|████████▍ | 436/520 [27:22<05:16, 3.77s/it] 84%|████████▍ | 437/520 [27:25<05:13, 3.77s/it] {'loss': 1.4282, 'grad_norm': 0.0004913617488701365, 'learning_rate': 0.000654373906771768, 'epoch': 0.84} + 84%|████████▍ | 437/520 [27:25<05:13, 3.77s/it] 84%|████████▍ | 438/520 [27:29<05:09, 3.78s/it] {'loss': 1.2509, 'grad_norm': 0.0005572901626110226, 'learning_rate': 0.0006390436780169733, 'epoch': 0.84} + 84%|████████▍ | 438/520 [27:29<05:09, 3.78s/it] 84%|████████▍ | 439/520 [27:33<05:06, 3.78s/it] {'loss': 1.2332, 'grad_norm': 0.0004067622776459641, 'learning_rate': 0.0006238828904562316, 'epoch': 0.84} + 84%|████████▍ | 439/520 [27:33<05:06, 3.78s/it] 85%|████████▍ | 440/520 [27:37<05:02, 3.78s/it] {'loss': 1.2904, 'grad_norm': 0.0005297785539707133, 'learning_rate': 0.0006088921331488567, 'epoch': 0.85} + 85%|████████▍ | 440/520 [27:37<05:02, 3.78s/it] 85%|████████▍ | 441/520 [27:41<05:00, 3.80s/it] {'loss': 1.2351, 'grad_norm': 0.0004974690948224327, 'learning_rate': 0.000594071988547788, 'epoch': 0.85} + 85%|████████▍ | 441/520 [27:41<05:00, 3.80s/it] 85%|████████▌ | 442/520 [27:44<04:55, 3.79s/it] {'loss': 1.3539, 'grad_norm': 0.0005650686199988829, 'learning_rate': 0.0005794230324769517, 'epoch': 0.85} + 85%|████████▌ | 442/520 [27:44<04:55, 3.79s/it] 85%|████████▌ | 443/520 [27:48<04:52, 3.80s/it] {'loss': 1.357, 'grad_norm': 0.0004970695056834997, 'learning_rate': 0.0005649458341088914, 'epoch': 0.85} + 85%|████████▌ | 443/520 [27:48<04:52, 3.80s/it] 85%|████████▌ | 444/520 [27:52<04:49, 3.80s/it] {'loss': 1.3041, 'grad_norm': 0.0004888198688219263, 'learning_rate': 0.0005506409559426573, 'epoch': 0.85} + 85%|████████▌ | 444/520 [27:52<04:49, 3.80s/it] 86%|████████▌ | 445/520 [27:56<04:46, 3.81s/it] {'loss': 1.226, 'grad_norm': 0.0005339355821514227, 'learning_rate': 0.0005365089537819434, 'epoch': 0.86} + 86%|████████▌ | 445/520 [27:56<04:46, 3.81s/it] 86%|████████▌ | 446/520 [28:00<04:41, 3.80s/it] {'loss': 1.3227, 'grad_norm': 0.0004365255583205874, 'learning_rate': 0.0005225503767134953, 'epoch': 0.86} + 86%|████████▌ | 446/520 [28:00<04:41, 3.80s/it] 86%|████████▌ | 447/520 [28:03<04:38, 3.81s/it] {'loss': 1.3118, 'grad_norm': 0.0004942171053503861, 'learning_rate': 0.0005087657670857798, 'epoch': 0.86} + 86%|████████▌ | 447/520 [28:03<04:38, 3.81s/it] 86%|████████▌ | 448/520 [28:07<04:34, 3.81s/it] {'loss': 1.3064, 'grad_norm': 0.0005372927656296006, 'learning_rate': 0.0004951556604879049, 'epoch': 0.86} + 86%|████████▌ | 448/520 [28:07<04:34, 3.81s/it] 86%|████████▋ | 449/520 [28:11<04:30, 3.81s/it] {'loss': 1.2904, 'grad_norm': 0.0005108871752901569, 'learning_rate': 0.0004817205857288176, 'epoch': 0.86} + 86%|████████▋ | 449/520 [28:11<04:30, 3.81s/it] 87%|████████▋ | 450/520 [28:15<04:26, 3.80s/it] {'loss': 1.3421, 'grad_norm': 0.0005908264987147664, 'learning_rate': 0.0004684610648167503, 'epoch': 0.87} + 87%|████████▋ | 450/520 [28:15<04:26, 3.80s/it] 87%|████████▋ | 451/520 [28:19<04:21, 3.80s/it] {'loss': 1.3358, 'grad_norm': 0.0005312423414321853, 'learning_rate': 0.0004553776129389453, 'epoch': 0.87} + 87%|████████▋ | 451/520 [28:19<04:21, 3.80s/it] 87%|████████▋ | 452/520 [28:22<04:17, 3.79s/it] {'loss': 1.3183, 'grad_norm': 0.00048455004040605577, 'learning_rate': 0.0004424707384416343, 'epoch': 0.87} + 87%|████████▋ | 452/520 [28:22<04:17, 3.79s/it] 87%|████████▋ | 453/520 [28:26<04:14, 3.80s/it] {'loss': 1.3195, 'grad_norm': 0.0005190426096770675, 'learning_rate': 0.000429740942810285, 'epoch': 0.87} + 87%|████████▋ | 453/520 [28:26<04:14, 3.80s/it] 87%|████████▋ | 454/520 [28:30<04:11, 3.81s/it] {'loss': 1.2384, 'grad_norm': 0.0004795013938887458, 'learning_rate': 0.00041718872065011904, 'epoch': 0.87} + 87%|████████▋ | 454/520 [28:30<04:11, 3.81s/it] 88%|████████▊ | 455/520 [28:34<04:07, 3.80s/it] {'loss': 1.3742, 'grad_norm': 0.0005209222116330576, 'learning_rate': 0.00040481455966689673, 'epoch': 0.88} + 88%|████████▊ | 455/520 [28:34<04:07, 3.80s/it] 88%|████████▊ | 456/520 [28:38<04:03, 3.80s/it] {'loss': 1.337, 'grad_norm': 0.0005850828592552455, 'learning_rate': 0.00039261894064796134, 'epoch': 0.88} + 88%|████████▊ | 456/520 [28:38<04:03, 3.80s/it] 88%|████████▊ | 457/520 [28:41<04:00, 3.81s/it] {'loss': 1.1511, 'grad_norm': 0.000503318819000637, 'learning_rate': 0.0003806023374435663, 'epoch': 0.88} + 88%|████████▊ | 457/520 [28:41<04:00, 3.81s/it] 88%|████████▊ | 458/520 [28:45<03:56, 3.82s/it] {'loss': 1.4502, 'grad_norm': 0.000510927018266583, 'learning_rate': 0.00036876521694845677, 'epoch': 0.88} + 88%|████████▊ | 458/520 [28:45<03:56, 3.82s/it] 88%|████████▊ | 459/520 [28:49<03:52, 3.81s/it] {'loss': 1.3624, 'grad_norm': 0.0005483328261077069, 'learning_rate': 0.00035710803908373226, 'epoch': 0.88} + 88%|████████▊ | 459/520 [28:49<03:52, 3.81s/it] 88%|████████▊ | 460/520 [28:53<03:48, 3.81s/it] {'loss': 1.2679, 'grad_norm': 0.0005765332291143138, 'learning_rate': 0.0003456312567789793, 'epoch': 0.88} + 88%|████████▊ | 460/520 [28:53<03:48, 3.81s/it] 89%|████████▊ | 461/520 [28:57<03:44, 3.80s/it] {'loss': 1.2284, 'grad_norm': 0.00041645757357943483, 'learning_rate': 0.0003343353159546675, 'epoch': 0.89} + 89%|████████▊ | 461/520 [28:57<03:44, 3.80s/it] 89%|████████▉ | 462/520 [29:00<03:40, 3.80s/it] {'loss': 1.3809, 'grad_norm': 0.00047129751415110334, 'learning_rate': 0.00032322065550483003, 'epoch': 0.89} + 89%|████████▉ | 462/520 [29:00<03:40, 3.80s/it] 89%|████████▉ | 463/520 [29:04<03:36, 3.80s/it] {'loss': 1.2582, 'grad_norm': 0.0005954168930769696, 'learning_rate': 0.00031228770728000454, 'epoch': 0.89} + 89%|████████▉ | 463/520 [29:04<03:36, 3.80s/it] 89%|████████▉ | 464/520 [29:08<03:32, 3.80s/it] {'loss': 1.3513, 'grad_norm': 0.0005348742392660624, 'learning_rate': 0.00030153689607045843, 'epoch': 0.89} + 89%|████████▉ | 464/520 [29:08<03:32, 3.80s/it] 89%|████████▉ | 465/520 [29:12<03:26, 3.75s/it] {'loss': 1.461, 'grad_norm': 0.0005339872090577245, 'learning_rate': 0.00029096863958968266, 'epoch': 0.89} + 89%|████████▉ | 465/520 [29:12<03:26, 3.75s/it] 90%|████████▉ | 466/520 [29:15<03:21, 3.73s/it] {'loss': 1.358, 'grad_norm': 0.0004723982294354065, 'learning_rate': 0.0002805833484581621, 'epoch': 0.9} + 90%|████████▉ | 466/520 [29:15<03:21, 3.73s/it] 90%|████████▉ | 467/520 [29:19<03:16, 3.71s/it] {'loss': 1.2416, 'grad_norm': 0.00048827383182737955, 'learning_rate': 0.0002703814261874199, 'epoch': 0.9} + 90%|████████▉ | 467/520 [29:19<03:16, 3.71s/it] 90%|█████████ | 468/520 [29:23<03:12, 3.70s/it] {'loss': 1.3087, 'grad_norm': 0.0005858972852755161, 'learning_rate': 0.0002603632691643415, 'epoch': 0.9} + 90%|█████████ | 468/520 [29:23<03:12, 3.70s/it] 90%|█████████ | 469/520 [29:26<03:08, 3.69s/it] {'loss': 1.3753, 'grad_norm': 0.0005979940564796297, 'learning_rate': 0.00025052926663577004, 'epoch': 0.9} + 90%|█████████ | 469/520 [29:26<03:08, 3.69s/it] 90%|█████████ | 470/520 [29:30<03:04, 3.69s/it] {'loss': 1.2549, 'grad_norm': 0.0006033323983337608, 'learning_rate': 0.00024087980069338822, 'epoch': 0.9} + 90%|█████████ | 470/520 [29:30<03:04, 3.69s/it] 91%|█████████ | 471/520 [29:34<03:00, 3.69s/it] {'loss': 1.3062, 'grad_norm': 0.0005992857931436495, 'learning_rate': 0.0002314152462588659, 'epoch': 0.91} + 91%|█████████ | 471/520 [29:34<03:00, 3.69s/it] 91%|█████████ | 472/520 [29:37<02:57, 3.70s/it] {'loss': 1.2842, 'grad_norm': 0.0005687628070534384, 'learning_rate': 0.00022213597106929606, 'epoch': 0.91} + 91%|█████████ | 472/520 [29:37<02:57, 3.70s/it] 91%|█████████ | 473/520 [29:41<02:53, 3.69s/it] {'loss': 1.3434, 'grad_norm': 0.0005516292859242056, 'learning_rate': 0.00021304233566290964, 'epoch': 0.91} + 91%|█████████ | 473/520 [29:41<02:53, 3.69s/it] 91%|█████████ | 474/520 [29:45<02:49, 3.68s/it] {'loss': 1.2832, 'grad_norm': 0.0004472420528261932, 'learning_rate': 0.0002041346933650612, 'epoch': 0.91} + 91%|█████████ | 474/520 [29:45<02:49, 3.68s/it] 91%|█████████▏| 475/520 [29:48<02:45, 3.68s/it] {'loss': 1.1965, 'grad_norm': 0.0004910402461082897, 'learning_rate': 0.00019541339027450257, 'epoch': 0.91} + 91%|█████████▏| 475/520 [29:48<02:45, 3.68s/it] 92%|█████████▏| 476/520 [29:52<02:41, 3.67s/it] {'loss': 1.3144, 'grad_norm': 0.0005088593536668704, 'learning_rate': 0.00018687876524993985, 'epoch': 0.92} + 92%|█████████▏| 476/520 [29:52<02:41, 3.67s/it] 92%|█████████▏| 477/520 [29:56<02:37, 3.67s/it] {'loss': 1.3231, 'grad_norm': 0.0006120404528196615, 'learning_rate': 0.0001785311498968617, 'epoch': 0.92} + 92%|█████████▏| 477/520 [29:56<02:37, 3.67s/it] 92%|█████████▏| 478/520 [29:59<02:34, 3.67s/it] {'loss': 1.2578, 'grad_norm': 0.00048402383084425286, 'learning_rate': 0.000170370868554659, 'epoch': 0.92} + 92%|█████████▏| 478/520 [29:59<02:34, 3.67s/it] 92%|█████████▏| 479/520 [30:03<02:30, 3.66s/it] {'loss': 1.2687, 'grad_norm': 0.0005259156725608081, 'learning_rate': 0.00016239823828401945, 'epoch': 0.92} + 92%|█████████▏| 479/520 [30:03<02:30, 3.66s/it] 92%|█████████▏| 480/520 [30:07<02:27, 3.69s/it] {'loss': 1.2706, 'grad_norm': 0.0005269110688556626, 'learning_rate': 0.00015461356885461075, 'epoch': 0.92} + 92%|█████████▏| 480/520 [30:07<02:27, 3.69s/it] 92%|█████████▎| 481/520 [30:10<02:23, 3.69s/it] {'loss': 1.2699, 'grad_norm': 0.0005002975120452609, 'learning_rate': 0.0001470171627330452, 'epoch': 0.93} + 92%|█████████▎| 481/520 [30:11<02:23, 3.69s/it] 93%|█████████▎| 482/520 [30:14<02:19, 3.67s/it] {'loss': 1.2888, 'grad_norm': 0.0004987022759513508, 'learning_rate': 0.0001396093150711275, 'epoch': 0.93} + 93%|█████████▎| 482/520 [30:14<02:19, 3.67s/it] 93%|█████████▎| 483/520 [30:18<02:15, 3.67s/it] {'loss': 1.3217, 'grad_norm': 0.0006583629404528187, 'learning_rate': 0.00013239031369438327, 'epoch': 0.93} + 93%|█████████▎| 483/520 [30:18<02:15, 3.67s/it] 93%|█████████▎| 484/520 [30:21<02:12, 3.67s/it] {'loss': 1.3306, 'grad_norm': 0.00048573309479940516, 'learning_rate': 0.0001253604390908819, 'epoch': 0.93} + 93%|█████████▎| 484/520 [30:21<02:12, 3.67s/it] 93%|█████████▎| 485/520 [30:25<02:08, 3.67s/it] {'loss': 1.2708, 'grad_norm': 0.0004848661312078269, 'learning_rate': 0.00011851996440033319, 'epoch': 0.93} + 93%|█████████▎| 485/520 [30:25<02:08, 3.67s/it] 93%|█████████▎| 486/520 [30:29<02:04, 3.65s/it] {'loss': 1.4076, 'grad_norm': 0.0005503720146109679, 'learning_rate': 0.00011186915540347731, 'epoch': 0.93} + 93%|█████████▎| 486/520 [30:29<02:04, 3.65s/it] 94%|█████████▎| 487/520 [30:32<02:00, 3.65s/it] {'loss': 1.2691, 'grad_norm': 0.0004965787725142296, 'learning_rate': 0.00010540827051175817, 'epoch': 0.94} + 94%|█████████▎| 487/520 [30:32<02:00, 3.65s/it] 94%|█████████▍| 488/520 [30:36<01:56, 3.65s/it] {'loss': 1.2234, 'grad_norm': 0.0005359523573172853, 'learning_rate': 9.913756075728087e-05, 'epoch': 0.94} + 94%|█████████▍| 488/520 [30:36<01:56, 3.65s/it] 94%|█████████▍| 489/520 [30:40<01:53, 3.65s/it] {'loss': 1.279, 'grad_norm': 0.0004415021120612587, 'learning_rate': 9.305726978306172e-05, 'epoch': 0.94} + 94%|█████████▍| 489/520 [30:40<01:53, 3.65s/it] 94%|█████████▍| 490/520 [30:43<01:49, 3.64s/it] {'loss': 1.3289, 'grad_norm': 0.0006076529753276604, 'learning_rate': 8.716763383355864e-05, 'epoch': 0.94} + 94%|█████████▍| 490/520 [30:43<01:49, 3.64s/it] 94%|█████████▍| 491/520 [30:47<01:47, 3.70s/it] {'loss': 1.3091, 'grad_norm': 0.0006082092000738758, 'learning_rate': 8.146888174549338e-05, 'epoch': 0.94} + 94%|█████████▍| 491/520 [30:47<01:47, 3.70s/it] 95%|█████████▍| 492/520 [30:51<01:45, 3.76s/it] {'loss': 1.4033, 'grad_norm': 0.0005121974030944733, 'learning_rate': 7.59612349389599e-05, 'epoch': 0.95} + 95%|█████████▍| 492/520 [30:51<01:45, 3.76s/it] 95%|█████████▍| 493/520 [30:55<01:42, 3.79s/it] {'loss': 1.2813, 'grad_norm': 0.0005039853309670784, 'learning_rate': 7.064490740882056e-05, 'epoch': 0.95} + 95%|█████████▍| 493/520 [30:55<01:42, 3.79s/it] 95%|█████████▌| 494/520 [30:59<01:39, 3.81s/it] {'loss': 1.3149, 'grad_norm': 0.0004643845638729322, 'learning_rate': 6.552010571639455e-05, 'epoch': 0.95} + 95%|█████████▌| 494/520 [30:59<01:39, 3.81s/it] 95%|█████████▌| 495/520 [31:03<01:35, 3.82s/it] {'loss': 1.33, 'grad_norm': 0.00053991224123967, 'learning_rate': 6.058702898142643e-05, 'epoch': 0.95} + 95%|█████████▌| 495/520 [31:03<01:35, 3.82s/it] 95%|█████████▌| 496/520 [31:06<01:32, 3.84s/it] {'loss': 1.266, 'grad_norm': 0.0005729757645760381, 'learning_rate': 5.584586887435739e-05, 'epoch': 0.95} + 95%|█████████▌| 496/520 [31:06<01:32, 3.84s/it] 96%|█████████▌| 497/520 [31:10<01:28, 3.84s/it] {'loss': 1.2241, 'grad_norm': 0.00045338575159555966, 'learning_rate': 5.129680960887006e-05, 'epoch': 0.96} + 96%|█████████▌| 497/520 [31:10<01:28, 3.84s/it] 96%|█████████▌| 498/520 [31:14<01:24, 3.84s/it] {'loss': 1.3123, 'grad_norm': 0.0006484198781336872, 'learning_rate': 4.694002793473595e-05, 'epoch': 0.96} + 96%|█████████▌| 498/520 [31:14<01:24, 3.84s/it] 96%|█████████▌| 499/520 [31:18<01:20, 3.83s/it] {'loss': 1.3809, 'grad_norm': 0.0004660130260043465, 'learning_rate': 4.277569313094809e-05, 'epoch': 0.96} + 96%|█████████▌| 499/520 [31:18<01:20, 3.83s/it] 96%|█████████▌| 500/520 [31:22<01:15, 3.77s/it] {'loss': 1.4508, 'grad_norm': 0.0006075962029060128, 'learning_rate': 3.8803966999139685e-05, 'epoch': 0.96} + 96%|█████████▌| 500/520 [31:22<01:15, 3.77s/it] 96%|█████████▋| 501/520 [31:25<01:12, 3.81s/it] {'loss': 1.2774, 'grad_norm': 0.0005583228210191791, 'learning_rate': 3.502500385730189e-05, 'epoch': 0.96} + 96%|█████████▋| 501/520 [31:26<01:12, 3.81s/it] 97%|█████████▋| 502/520 [31:29<01:09, 3.85s/it] {'loss': 1.3568, 'grad_norm': 0.0005138245490184564, 'learning_rate': 3.143895053378698e-05, 'epoch': 0.97} + 97%|█████████▋| 502/520 [31:29<01:09, 3.85s/it] 97%|█████████▋| 503/520 [31:33<01:05, 3.87s/it] {'loss': 1.2429, 'grad_norm': 0.0005313215312506365, 'learning_rate': 2.8045946361601182e-05, 'epoch': 0.97} + 97%|█████████▋| 503/520 [31:33<01:05, 3.87s/it] 97%|█████████▋| 504/520 [31:37<01:02, 3.90s/it] {'loss': 1.3358, 'grad_norm': 0.000704501096632198, 'learning_rate': 2.4846123172992953e-05, 'epoch': 0.97} + 97%|█████████▋| 504/520 [31:37<01:02, 3.90s/it] 97%|█████████▋| 505/520 [31:41<00:57, 3.85s/it] {'loss': 1.3596, 'grad_norm': 0.0006053441451827271, 'learning_rate': 2.1839605294330932e-05, 'epoch': 0.97} + 97%|█████████▋| 505/520 [31:41<00:57, 3.85s/it] 97%|█████████▋| 506/520 [31:45<00:53, 3.79s/it] {'loss': 1.2907, 'grad_norm': 0.000609397008727716, 'learning_rate': 1.9026509541272274e-05, 'epoch': 0.97} + 97%|█████████▋| 506/520 [31:45<00:53, 3.79s/it] 98%|█████████▊| 507/520 [31:48<00:48, 3.76s/it] {'loss': 1.3866, 'grad_norm': 0.0004460560516039055, 'learning_rate': 1.640694521422459e-05, 'epoch': 0.97} + 98%|█████████▊| 507/520 [31:48<00:48, 3.76s/it] 98%|█████████▊| 508/520 [31:52<00:44, 3.74s/it] {'loss': 1.4241, 'grad_norm': 0.0004863124031646003, 'learning_rate': 1.3981014094099354e-05, 'epoch': 0.98} + 98%|█████████▊| 508/520 [31:52<00:44, 3.74s/it] 98%|█████████▊| 509/520 [31:56<00:40, 3.73s/it] {'loss': 1.389, 'grad_norm': 0.0005193082268678402, 'learning_rate': 1.1748810438355628e-05, 'epoch': 0.98} + 98%|█████████▊| 509/520 [31:56<00:40, 3.73s/it] 98%|█████████▊| 510/520 [32:00<00:37, 3.72s/it] {'loss': 1.329, 'grad_norm': 0.000586161899798067, 'learning_rate': 9.710420977340761e-06, 'epoch': 0.98} + 98%|█████████▊| 510/520 [32:00<00:37, 3.72s/it] 98%|█████████▊| 511/520 [32:03<00:33, 3.70s/it] {'loss': 1.2776, 'grad_norm': 0.0005684720503537398, 'learning_rate': 7.865924910916978e-06, 'epoch': 0.98} + 98%|█████████▊| 511/520 [32:03<00:33, 3.70s/it] 98%|█████████▊| 512/520 [32:07<00:29, 3.69s/it] {'loss': 1.1775, 'grad_norm': 0.0005146026760674109, 'learning_rate': 6.215393905388278e-06, 'epoch': 0.98} + 98%|█████████▊| 512/520 [32:07<00:29, 3.69s/it] 99%|█████████▊| 513/520 [32:11<00:25, 3.70s/it] {'loss': 1.3961, 'grad_norm': 0.0005891944571663868, 'learning_rate': 4.758892090711009e-06, 'epoch': 0.99} + 99%|█████████▊| 513/520 [32:11<00:25, 3.70s/it] 99%|█████████▉| 514/520 [32:14<00:22, 3.69s/it] {'loss': 1.345, 'grad_norm': 0.0005038289306714394, 'learning_rate': 3.496476058006959e-06, 'epoch': 0.99} + 99%|█████████▉| 514/520 [32:14<00:22, 3.69s/it] 99%|█████████▉| 515/520 [32:18<00:18, 3.70s/it] {'loss': 1.4302, 'grad_norm': 0.0006006190032063123, 'learning_rate': 2.4281948573617874e-06, 'epoch': 0.99} + 99%|█████████▉| 515/520 [32:18<00:18, 3.70s/it] 99%|█████████▉| 516/520 [32:22<00:14, 3.70s/it] {'loss': 1.3402, 'grad_norm': 0.000610464199240439, 'learning_rate': 1.5540899959187727e-06, 'epoch': 0.99} + 99%|█████████▉| 516/520 [32:22<00:14, 3.70s/it] 99%|█████████▉| 517/520 [32:25<00:11, 3.68s/it] {'loss': 1.2813, 'grad_norm': 0.00045121902255326724, 'learning_rate': 8.741954362678772e-07, 'epoch': 0.99} + 99%|█████████▉| 517/520 [32:25<00:11, 3.68s/it] 100%|█████████▉| 518/520 [32:29<00:07, 3.67s/it] {'loss': 1.3175, 'grad_norm': 0.000725071087439377, 'learning_rate': 3.885375951256931e-07, 'epoch': 1.0} + 100%|█████████▉| 518/520 [32:29<00:07, 3.67s/it] 100%|█████████▉| 519/520 [32:33<00:03, 3.66s/it] {'loss': 1.2583, 'grad_norm': 0.00047529317395903224, 'learning_rate': 9.713534230904042e-08, 'epoch': 1.0} + 100%|█████████▉| 519/520 [32:33<00:03, 3.66s/it] 100%|██████████| 520/520 [32:37<00:00, 3.91s/it] {'loss': 1.2469, 'grad_norm': 0.000663764211876436, 'learning_rate': 0.0, 'epoch': 1.0} + 100%|██████████| 520/520 [32:37<00:00, 3.91s/it] {'train_runtime': 1957.5781, 'train_samples_per_second': 33.985, 'train_steps_per_second': 0.266, 'train_loss': 1.3719075583494627, 'epoch': 1.0} + 100%|██████████| 520/520 [32:37<00:00, 3.91s/it] 100%|██████████| 520/520 [32:37<00:00, 3.76s/it] +[2025-10-12 10:05:33,809] [INFO] [launch.py:348:main] Process 579482 exits successfully. +[2025-10-12 10:05:33,810] [INFO] [launch.py:348:main] Process 579483 exits successfully. +[2025-10-12 10:05:33,810] [INFO] [launch.py:348:main] Process 579486 exits successfully. +[2025-10-12 10:05:33,810] [INFO] [launch.py:348:main] Process 579481 exits successfully. +[2025-10-12 10:05:34,812] [INFO] [launch.py:348:main] Process 579484 exits successfully. +[2025-10-12 10:05:34,812] [INFO] [launch.py:348:main] Process 579485 exits successfully. +[2025-10-12 10:05:34,813] [INFO] [launch.py:348:main] Process 579487 exits successfully. +[2025-10-12 10:05:38,817] [INFO] [launch.py:348:main] Process 579480 exits successfully. +==== EXPERIMENT COMPLETED: qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_1e-2_connector-3.0_0.5_1e-2_ablation ==== +Log File: qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_1e-2_connector-3.0_0.5_1e-2_ablation_20251012_091538.log +Timestamp: 2025-10-12 10:05:41 +===================================== diff --git a/logs_oct11/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_2e-1_connector-3.0_0.5_2e-1_ablation_20251012_123546.log b/logs_oct11/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_2e-1_connector-3.0_0.5_2e-1_ablation_20251012_123546.log new file mode 100644 index 0000000000000000000000000000000000000000..05e8a8693caf9465773ad6b1854e5186eb2148fb --- /dev/null +++ b/logs_oct11/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_2e-1_connector-3.0_0.5_2e-1_ablation_20251012_123546.log @@ -0,0 +1,1109 @@ +==== STARTING EXPERIMENT: qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_2e-1_connector-3.0_0.5_2e-1_ablation ==== +Log File: qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_2e-1_connector-3.0_0.5_2e-1_ablation_20251012_123546.log +Timestamp: 2025-10-12 12:35:46 +===================================== +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 12:35:49,631] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 12:35:52,325] [WARNING] [runner.py:202:fetch_hostfile] Unable to find hostfile, will proceed with training with local resources only. +[2025-10-12 12:35:52,326] [INFO] [runner.py:568:main] cmd = /opt/conda/envs/tinyllava/bin/python3.10 -u -m deepspeed.launcher.launch --world_info=eyJsb2NhbGhvc3QiOiBbMCwgMSwgMiwgMywgNCwgNSwgNiwgN119 --master_addr=127.0.0.1 --master_port=29501 --enable_each_rank_log=None tinyllava/train/train.py --deepspeed ./scripts/zero3.json --data_path /s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json --image_folder /s3-code/ywang29/datasets/tinyllava --is_multimodal True --conv_version qwen2_base --model_name_or_path Qwen/Qwen2.5-0.5B --vision_tower google/siglip-so400m-patch14-384 --vision_tower2 --connector_type mlp2x_gelu --mm_vision_select_layer -2 --image_aspect_ratio square --attn_implementation flash_attention_2 --bf16 True --training_recipe common --tune_type_llm full --tune_type_vision_tower frozen --tune_vision_tower_from_layer 0 --tune_type_connector full --group_by_modality_length True --pretrained_model_path /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain --output_dir /nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_2e-1_connector-3.0_0.5_2e-1_ablation --num_train_epochs 1 --per_device_train_batch_size 4 --per_device_eval_batch_size 4 --gradient_accumulation_steps 4 --evaluation_strategy no --learning_rate 2e-1 --weight_decay 0. --warmup_ratio 0.03 --lr_scheduler_type cosine --logging_steps 1 --tf32 False --model_max_length 2048 --gradient_checkpointing True --dataloader_num_workers 8 --lazy_preprocess True --report_to tensorboard --tokenizer_use_fast False --run_name tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune --subnet_mode_text both --subnet_type_text None --mask_type_text soft --init_mean_text 3.0 --temperature_attn_text 0.5 --temperature_mlp_text 0.5 --backward_type_text normal --masked_layers_text all --subnet_mode_vision both --subnet_type_vision None --mask_type_vision soft --init_mean_vision 3.0 --temperature_attn_vision 0.5 --temperature_mlp_vision 0.5 --backward_type_vision normal --masked_layers_vision all --subnet_type_connector global --mask_type_connector soft --init_mean_connector 3.0 --temperature_connector 0.5 --backward_type_connector normal --mm_projector_lr 2e-1 --seed 42 --mask_model llm-connector --save_strategy steps --save_steps 50000 --save_total_limit 1 --train_data_ratio 0.1 +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 12:35:54,957] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 12:35:55,993] [INFO] [launch.py:138:main] 0 NCCL_VERSION=2.21.5 +[2025-10-12 12:35:55,993] [INFO] [launch.py:138:main] 0 NCCL_SOCKET_IFNAME=eth +[2025-10-12 12:35:55,993] [INFO] [launch.py:145:main] WORLD INFO DICT: {'localhost': [0, 1, 2, 3, 4, 5, 6, 7]} +[2025-10-12 12:35:55,993] [INFO] [launch.py:151:main] nnodes=1, num_local_procs=8, node_rank=0 +[2025-10-12 12:35:55,993] [INFO] [launch.py:162:main] global_rank_mapping=defaultdict(, {'localhost': [0, 1, 2, 3, 4, 5, 6, 7]}) +[2025-10-12 12:35:55,993] [INFO] [launch.py:163:main] dist_world_size=8 +[2025-10-12 12:35:55,993] [INFO] [launch.py:165:main] Setting CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 +[2025-10-12 12:35:55,995] [INFO] [launch.py:253:main] process 377133 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=0', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_2e-1_connector-3.0_0.5_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 12:35:55,997] [INFO] [launch.py:253:main] process 377134 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=1', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_2e-1_connector-3.0_0.5_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 12:35:56,000] [INFO] [launch.py:253:main] process 377135 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=2', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_2e-1_connector-3.0_0.5_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 12:35:56,002] [INFO] [launch.py:253:main] process 377136 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=3', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_2e-1_connector-3.0_0.5_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 12:35:56,004] [INFO] [launch.py:253:main] process 377137 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=4', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_2e-1_connector-3.0_0.5_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 12:35:56,006] [INFO] [launch.py:253:main] process 377138 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=5', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_2e-1_connector-3.0_0.5_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 12:35:56,008] [INFO] [launch.py:253:main] process 377139 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=6', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_2e-1_connector-3.0_0.5_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 12:35:56,010] [INFO] [launch.py:253:main] process 377140 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=7', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_2e-1_connector-3.0_0.5_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 12:36:02,527] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 12:36:02,594] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 12:36:02,767] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 12:36:02,880] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 12:36:02,880] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 12:36:02,913] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 12:36:02,924] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 12:36:02,924] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 12:36:02,928] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 12:36:02,992] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 12:36:03,166] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 12:36:03,166] [INFO] [comm.py:668:init_distributed] Initializing TorchBackend in DeepSpeed with backend nccl +[2025-10-12 12:36:03,275] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 12:36:03,275] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 12:36:03,308] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 12:36:03,319] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 12:36:03,319] [INFO] [comm.py:637:init_distributed] cdb=None +Apply masks for the following modules: ['llm', 'connector'] +Apply masks for the following modules: ['llm', 'connector'] +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +{'llm': {'model_name_or_path': 'Qwen/Qwen2.5-0.5B', 'cache_dir': None, 'attn_implementation': 'flash_attention_2', 'subnet_mode': 'both', 'subnet_type': 'None', 'sparsity_attn': None, 'sparsity_mlp': None, 'threshold_attn': None, 'threshold_mlp': None, 'temperature_attn': 0.5, 'temperature_mlp': 0.5, 'masked_layers': 'all', 'mask_type': 'soft', 'backward_type': 'normal'}, 'vision_tower': {'model_name_or_path': 'google/siglip-so400m-patch14-384'}, 'connector': {'connector_type': 'mlp2x_gelu', 'subnet_type': 'global', 'threshold': None, 'sparsity': None, 'temperature': 0.5, 'mask_type': 'soft', 'backward_type': 'normal'}} +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +TinyLlavaConfig { + "backward_type_connector": "normal", + "cache_dir": null, + "connector_type": "mlp2x_gelu", + "hidden_size": 896, + "ignore_index": -100, + "image_aspect_ratio": "square", + "image_token_index": -200, + "llm_model_name_or_path": "Qwen/Qwen2.5-0.5B", + "mask_model": [ + "llm", + "connector" + ], + "mask_type_connector": "soft", + "model_type": "tinyllava", + "num_queries": 128, + "num_resampler_layers": 3, + "pad_token": null, + "resampler_hidden_size": 768, + "sparsity_connector": null, + "subnet_type_connector": "global", + "temperature_connector": 0.5, + "text_config": { + "_name_or_path": "Qwen/Qwen2.5-0.5B", + "architectures": [ + "Qwen2ForCausalLM" + ], + "backward_type": "normal", + "bos_token_id": 151643, + "eos_token_id": 151643, + "hidden_size": 896, + "intermediate_size": 4864, + "mask_type": "soft", + "masked_layers": "all", + "max_position_embeddings": 32768, + "max_window_layers": 24, + "model_type": "qwen2", + "num_attention_heads": 14, + "num_hidden_layers": 24, + "num_key_value_heads": 2, + "rope_theta": 1000000.0, + "sliding_window": 32768, + "subnet_mode": "both", + "subnet_type": "None", + "temperature_attn": 0.5, + "temperature_mlp": 0.5, + "tie_word_embeddings": true, + "torch_dtype": "bfloat16", + "use_mrope": false, + "use_sliding_window": false, + "vocab_size": 151936 + }, + "threshold_connector": null, + "tokenizer_model_max_length": 2048, + "tokenizer_name_or_path": "Qwen/Qwen2.5-0.5B", + "tokenizer_padding_side": "right", + "tokenizer_use_fast": false, + "transformers_version": "4.40.1", + "tune_type_connector": "frozen", + "tune_type_llm": "frozen", + "tune_type_vision_tower": "frozen", + "tune_vision_tower_from_layer": -1, + "use_cache": false, + "vision_config": { + "hidden_act": "gelu_pytorch_tanh", + "hidden_size": 1152, + "image_size": 384, + "intermediate_size": 4304, + "layer_norm_eps": 1e-06, + "model_name_or_path": "google/siglip-so400m-patch14-384", + "model_name_or_path2": "", + "model_type": "siglip_vision_model", + "num_attention_heads": 16, + "num_hidden_layers": 27, + "patch_size": 14 + }, + "vision_feature_layer": -2, + "vision_feature_select_strategy": "patch", + "vision_hidden_size": 1152, + "vision_model_name_or_path": "google/siglip-so400m-patch14-384", + "vision_model_name_or_path2": "", + "vocab_size": 151936 +} + +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test1-worker-0:377133:377133 [0] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:377133:377133 [0] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:377133:377133 [0] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:377133:377133 [0] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:377133:377133 [0] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test1-worker-0:377133:377133 [0] NCCL INFO cudaDriverVersion 12040 +NCCL version 2.21.5+cuda12.1 +ywang29-vrdb-test1-worker-0:377134:377134 [1] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test1-worker-0:377134:377134 [1] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:377134:377134 [1] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:377135:377135 [2] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test1-worker-0:377135:377135 [2] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:377134:377134 [1] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:377134:377134 [1] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:377134:377134 [1] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test1-worker-0:377135:377135 [2] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:377137:377137 [4] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test1-worker-0:377137:377137 [4] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:377137:377137 [4] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:377135:377135 [2] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:377135:377135 [2] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:377135:377135 [2] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test1-worker-0:377137:377137 [4] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:377137:377137 [4] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:377137:377137 [4] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test1-worker-0:377133:378702 [0] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:377133:378702 [0] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:377133:378702 [0] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:377133:378702 [0] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:377133:378702 [0] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:377133:378702 [0] NCCL INFO Using network Socket +ywang29-vrdb-test1-worker-0:377134:378706 [1] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:377134:378706 [1] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:377134:378706 [1] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:377134:378706 [1] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:377134:378706 [1] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:377134:378706 [1] NCCL INFO Using network Socket +ywang29-vrdb-test1-worker-0:377137:378708 [4] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:377137:378708 [4] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:377137:378708 [4] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:377137:378708 [4] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:377137:378708 [4] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:377137:378708 [4] NCCL INFO Using network Socket +ywang29-vrdb-test1-worker-0:377135:378707 [2] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:377135:378707 [2] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:377135:378707 [2] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:377135:378707 [2] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:377135:378707 [2] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:377135:378707 [2] NCCL INFO Using network Socket +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test1-worker-0:377140:377140 [7] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test1-worker-0:377140:377140 [7] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:377140:377140 [7] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:377140:377140 [7] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:377140:377140 [7] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:377140:377140 [7] NCCL INFO NET/Plugin: Using internal network plugin. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test1-worker-0:377140:378710 [7] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:377140:378710 [7] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:377140:378710 [7] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:377140:378710 [7] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:377140:378710 [7] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:377140:378710 [7] NCCL INFO Using network Socket +ywang29-vrdb-test1-worker-0:377138:377138 [5] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test1-worker-0:377138:377138 [5] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:377138:377138 [5] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:377138:377138 [5] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:377138:377138 [5] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:377138:377138 [5] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test1-worker-0:377139:377139 [6] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test1-worker-0:377139:377139 [6] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:377139:377139 [6] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:377139:377139 [6] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:377139:377139 [6] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:377139:377139 [6] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test1-worker-0:377138:378711 [5] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:377138:378711 [5] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:377138:378711 [5] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:377138:378711 [5] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:377138:378711 [5] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:377138:378711 [5] NCCL INFO Using network Socket +ywang29-vrdb-test1-worker-0:377139:378712 [6] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:377139:378712 [6] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:377139:378712 [6] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:377139:378712 [6] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:377139:378712 [6] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:377139:378712 [6] NCCL INFO Using network Socket +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test1-worker-0:377136:377136 [3] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test1-worker-0:377136:377136 [3] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:377136:377136 [3] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:377136:377136 [3] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:377136:377136 [3] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:377136:377136 [3] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test1-worker-0:377136:378730 [3] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:377136:378730 [3] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:377136:378730 [3] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:377136:378730 [3] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:377136:378730 [3] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:377136:378730 [3] NCCL INFO Using network Socket +ywang29-vrdb-test1-worker-0:377136:378730 [3] NCCL INFO ncclCommInitRank comm 0x5572fbfb1900 rank 3 nranks 8 cudaDev 3 nvmlDev 3 busId 201d0 commId 0x531bd79cf4336d7e - Init START +ywang29-vrdb-test1-worker-0:377134:378706 [1] NCCL INFO ncclCommInitRank comm 0x55eba66cc6c0 rank 1 nranks 8 cudaDev 1 nvmlDev 1 busId 101d0 commId 0x531bd79cf4336d7e - Init START +ywang29-vrdb-test1-worker-0:377135:378707 [2] NCCL INFO ncclCommInitRank comm 0x55e614272ff0 rank 2 nranks 8 cudaDev 2 nvmlDev 2 busId 201c0 commId 0x531bd79cf4336d7e - Init START +ywang29-vrdb-test1-worker-0:377133:378702 [0] NCCL INFO ncclCommInitRank comm 0x5596a2c83f20 rank 0 nranks 8 cudaDev 0 nvmlDev 0 busId 101c0 commId 0x531bd79cf4336d7e - Init START +ywang29-vrdb-test1-worker-0:377138:378711 [5] NCCL INFO ncclCommInitRank comm 0x55c16df47420 rank 5 nranks 8 cudaDev 5 nvmlDev 5 busId 901d0 commId 0x531bd79cf4336d7e - Init START +ywang29-vrdb-test1-worker-0:377140:378710 [7] NCCL INFO ncclCommInitRank comm 0x55cd0c220340 rank 7 nranks 8 cudaDev 7 nvmlDev 7 busId a01d0 commId 0x531bd79cf4336d7e - Init START +ywang29-vrdb-test1-worker-0:377137:378708 [4] NCCL INFO ncclCommInitRank comm 0x5610c27f1020 rank 4 nranks 8 cudaDev 4 nvmlDev 4 busId 901c0 commId 0x531bd79cf4336d7e - Init START +ywang29-vrdb-test1-worker-0:377139:378712 [6] NCCL INFO ncclCommInitRank comm 0x55a39c6a1c00 rank 6 nranks 8 cudaDev 6 nvmlDev 6 busId a01c0 commId 0x531bd79cf4336d7e - Init START +ywang29-vrdb-test1-worker-0:377135:378707 [2] NCCL INFO Setting affinity for GPU 2 to ff,ffff0000,00ffffff +ywang29-vrdb-test1-worker-0:377140:378710 [7] NCCL INFO Setting affinity for GPU 7 to ffffff00,0000ffff,ff000000 +ywang29-vrdb-test1-worker-0:377140:378710 [7] NCCL INFO NVLS multicast support is not available on dev 7 +ywang29-vrdb-test1-worker-0:377135:378707 [2] NCCL INFO NVLS multicast support is not available on dev 2 +ywang29-vrdb-test1-worker-0:377134:378706 [1] NCCL INFO Setting affinity for GPU 1 to ff,ffff0000,00ffffff +ywang29-vrdb-test1-worker-0:377134:378706 [1] NCCL INFO NVLS multicast support is not available on dev 1 +ywang29-vrdb-test1-worker-0:377138:378711 [5] NCCL INFO Setting affinity for GPU 5 to ffffff00,0000ffff,ff000000 +ywang29-vrdb-test1-worker-0:377136:378730 [3] NCCL INFO Setting affinity for GPU 3 to ff,ffff0000,00ffffff +ywang29-vrdb-test1-worker-0:377138:378711 [5] NCCL INFO NVLS multicast support is not available on dev 5 +ywang29-vrdb-test1-worker-0:377136:378730 [3] NCCL INFO NVLS multicast support is not available on dev 3 +ywang29-vrdb-test1-worker-0:377133:378702 [0] NCCL INFO Setting affinity for GPU 0 to ff,ffff0000,00ffffff +ywang29-vrdb-test1-worker-0:377139:378712 [6] NCCL INFO Setting affinity for GPU 6 to ffffff00,0000ffff,ff000000 +ywang29-vrdb-test1-worker-0:377133:378702 [0] NCCL INFO NVLS multicast support is not available on dev 0 +ywang29-vrdb-test1-worker-0:377139:378712 [6] NCCL INFO NVLS multicast support is not available on dev 6 +ywang29-vrdb-test1-worker-0:377137:378708 [4] NCCL INFO Setting affinity for GPU 4 to ffffff00,0000ffff,ff000000 +ywang29-vrdb-test1-worker-0:377137:378708 [4] NCCL INFO NVLS multicast support is not available on dev 4 +ywang29-vrdb-test1-worker-0:377138:378711 [5] NCCL INFO comm 0x55c16df47420 rank 5 nRanks 8 nNodes 1 localRanks 8 localRank 5 MNNVL 0 +ywang29-vrdb-test1-worker-0:377137:378708 [4] NCCL INFO comm 0x5610c27f1020 rank 4 nRanks 8 nNodes 1 localRanks 8 localRank 4 MNNVL 0 +ywang29-vrdb-test1-worker-0:377136:378730 [3] NCCL INFO comm 0x5572fbfb1900 rank 3 nRanks 8 nNodes 1 localRanks 8 localRank 3 MNNVL 0 +ywang29-vrdb-test1-worker-0:377139:378712 [6] NCCL INFO comm 0x55a39c6a1c00 rank 6 nRanks 8 nNodes 1 localRanks 8 localRank 6 MNNVL 0 +ywang29-vrdb-test1-worker-0:377135:378707 [2] NCCL INFO comm 0x55e614272ff0 rank 2 nRanks 8 nNodes 1 localRanks 8 localRank 2 MNNVL 0 +ywang29-vrdb-test1-worker-0:377134:378706 [1] NCCL INFO comm 0x55eba66cc6c0 rank 1 nRanks 8 nNodes 1 localRanks 8 localRank 1 MNNVL 0 +ywang29-vrdb-test1-worker-0:377133:378702 [0] NCCL INFO comm 0x5596a2c83f20 rank 0 nRanks 8 nNodes 1 localRanks 8 localRank 0 MNNVL 0 +ywang29-vrdb-test1-worker-0:377140:378710 [7] NCCL INFO comm 0x55cd0c220340 rank 7 nRanks 8 nNodes 1 localRanks 8 localRank 7 MNNVL 0 +ywang29-vrdb-test1-worker-0:377137:378708 [4] NCCL INFO Trees [0] 5/-1/-1->4->3 [1] 5/-1/-1->4->3 [2] 5/-1/-1->4->3 [3] 5/-1/-1->4->3 [4] 5/-1/-1->4->3 [5] 5/-1/-1->4->3 [6] 5/-1/-1->4->3 [7] 5/-1/-1->4->3 [8] 5/-1/-1->4->3 [9] 5/-1/-1->4->3 [10] 5/-1/-1->4->3 [11] 5/-1/-1->4->3 [12] 5/-1/-1->4->3 [13] 5/-1/-1->4->3 [14] 5/-1/-1->4->3 [15] 5/-1/-1->4->3 [16] 5/-1/-1->4->3 [17] 5/-1/-1->4->3 [18] 5/-1/-1->4->3 [19] 5/-1/-1->4->3 [20] 5/-1/-1->4->3 [21] 5/-1/-1->4->3 [22] 5/-1/-1->4->3 [23] 5/-1/-1->4->3 +ywang29-vrdb-test1-worker-0:377138:378711 [5] NCCL INFO Trees [0] 6/-1/-1->5->4 [1] 6/-1/-1->5->4 [2] 6/-1/-1->5->4 [3] 6/-1/-1->5->4 [4] 6/-1/-1->5->4 [5] 6/-1/-1->5->4 [6] 6/-1/-1->5->4 [7] 6/-1/-1->5->4 [8] 6/-1/-1->5->4 [9] 6/-1/-1->5->4 [10] 6/-1/-1->5->4 [11] 6/-1/-1->5->4 [12] 6/-1/-1->5->4 [13] 6/-1/-1->5->4 [14] 6/-1/-1->5->4 [15] 6/-1/-1->5->4 [16] 6/-1/-1->5->4 [17] 6/-1/-1->5->4 [18] 6/-1/-1->5->4 [19] 6/-1/-1->5->4 [20] 6/-1/-1->5->4 [21] 6/-1/-1->5->4 [22] 6/-1/-1->5->4 [23] 6/-1/-1->5->4 +ywang29-vrdb-test1-worker-0:377137:378708 [4] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:377133:378702 [0] NCCL INFO Channel 00/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:377136:378730 [3] NCCL INFO Trees [0] 4/-1/-1->3->2 [1] 4/-1/-1->3->2 [2] 4/-1/-1->3->2 [3] 4/-1/-1->3->2 [4] 4/-1/-1->3->2 [5] 4/-1/-1->3->2 [6] 4/-1/-1->3->2 [7] 4/-1/-1->3->2 [8] 4/-1/-1->3->2 [9] 4/-1/-1->3->2 [10] 4/-1/-1->3->2 [11] 4/-1/-1->3->2 [12] 4/-1/-1->3->2 [13] 4/-1/-1->3->2 [14] 4/-1/-1->3->2 [15] 4/-1/-1->3->2 [16] 4/-1/-1->3->2 [17] 4/-1/-1->3->2 [18] 4/-1/-1->3->2 [19] 4/-1/-1->3->2 [20] 4/-1/-1->3->2 [21] 4/-1/-1->3->2 [22] 4/-1/-1->3->2 [23] 4/-1/-1->3->2 +ywang29-vrdb-test1-worker-0:377133:378702 [0] NCCL INFO Channel 01/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:377138:378711 [5] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:377133:378702 [0] NCCL INFO Channel 02/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:377136:378730 [3] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:377133:378702 [0] NCCL INFO Channel 03/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:377133:378702 [0] NCCL INFO Channel 04/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:377135:378707 [2] NCCL INFO Trees [0] 3/-1/-1->2->1 [1] 3/-1/-1->2->1 [2] 3/-1/-1->2->1 [3] 3/-1/-1->2->1 [4] 3/-1/-1->2->1 [5] 3/-1/-1->2->1 [6] 3/-1/-1->2->1 [7] 3/-1/-1->2->1 [8] 3/-1/-1->2->1 [9] 3/-1/-1->2->1 [10] 3/-1/-1->2->1 [11] 3/-1/-1->2->1 [12] 3/-1/-1->2->1 [13] 3/-1/-1->2->1 [14] 3/-1/-1->2->1 [15] 3/-1/-1->2->1 [16] 3/-1/-1->2->1 [17] 3/-1/-1->2->1 [18] 3/-1/-1->2->1 [19] 3/-1/-1->2->1 [20] 3/-1/-1->2->1 [21] 3/-1/-1->2->1 [22] 3/-1/-1->2->1 [23] 3/-1/-1->2->1 +ywang29-vrdb-test1-worker-0:377133:378702 [0] NCCL INFO Channel 05/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:377133:378702 [0] NCCL INFO Channel 06/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:377134:378706 [1] NCCL INFO Trees [0] 2/-1/-1->1->0 [1] 2/-1/-1->1->0 [2] 2/-1/-1->1->0 [3] 2/-1/-1->1->0 [4] 2/-1/-1->1->0 [5] 2/-1/-1->1->0 [6] 2/-1/-1->1->0 [7] 2/-1/-1->1->0 [8] 2/-1/-1->1->0 [9] 2/-1/-1->1->0 [10] 2/-1/-1->1->0 [11] 2/-1/-1->1->0 [12] 2/-1/-1->1->0 [13] 2/-1/-1->1->0 [14] 2/-1/-1->1->0 [15] 2/-1/-1->1->0 [16] 2/-1/-1->1->0 [17] 2/-1/-1->1->0 [18] 2/-1/-1->1->0 [19] 2/-1/-1->1->0 [20] 2/-1/-1->1->0 [21] 2/-1/-1->1->0 [22] 2/-1/-1->1->0 [23] 2/-1/-1->1->0 +ywang29-vrdb-test1-worker-0:377135:378707 [2] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:377133:378702 [0] NCCL INFO Channel 07/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:377139:378712 [6] NCCL INFO Trees [0] 7/-1/-1->6->5 [1] 7/-1/-1->6->5 [2] 7/-1/-1->6->5 [3] 7/-1/-1->6->5 [4] 7/-1/-1->6->5 [5] 7/-1/-1->6->5 [6] 7/-1/-1->6->5 [7] 7/-1/-1->6->5 [8] 7/-1/-1->6->5 [9] 7/-1/-1->6->5 [10] 7/-1/-1->6->5 [11] 7/-1/-1->6->5 [12] 7/-1/-1->6->5 [13] 7/-1/-1->6->5 [14] 7/-1/-1->6->5 [15] 7/-1/-1->6->5 [16] 7/-1/-1->6->5 [17] 7/-1/-1->6->5 [18] 7/-1/-1->6->5 [19] 7/-1/-1->6->5 [20] 7/-1/-1->6->5 [21] 7/-1/-1->6->5 [22] 7/-1/-1->6->5 [23] 7/-1/-1->6->5 +ywang29-vrdb-test1-worker-0:377133:378702 [0] NCCL INFO Channel 08/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:377133:378702 [0] NCCL INFO Channel 09/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:377134:378706 [1] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:377133:378702 [0] NCCL INFO Channel 10/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:377139:378712 [6] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:377133:378702 [0] NCCL INFO Channel 11/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:377133:378702 [0] NCCL INFO Channel 12/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:377133:378702 [0] NCCL INFO Channel 13/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:377133:378702 [0] NCCL INFO Channel 14/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:377133:378702 [0] NCCL INFO Channel 15/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:377133:378702 [0] NCCL INFO Channel 16/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:377133:378702 [0] NCCL INFO Channel 17/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:377140:378710 [7] NCCL INFO Trees [0] -1/-1/-1->7->6 [1] -1/-1/-1->7->6 [2] -1/-1/-1->7->6 [3] -1/-1/-1->7->6 [4] -1/-1/-1->7->6 [5] -1/-1/-1->7->6 [6] -1/-1/-1->7->6 [7] -1/-1/-1->7->6 [8] -1/-1/-1->7->6 [9] -1/-1/-1->7->6 [10] -1/-1/-1->7->6 [11] -1/-1/-1->7->6 [12] -1/-1/-1->7->6 [13] -1/-1/-1->7->6 [14] -1/-1/-1->7->6 [15] -1/-1/-1->7->6 [16] -1/-1/-1->7->6 [17] -1/-1/-1->7->6 [18] -1/-1/-1->7->6 [19] -1/-1/-1->7->6 [20] -1/-1/-1->7->6 [21] -1/-1/-1->7->6 [22] -1/-1/-1->7->6 [23] -1/-1/-1->7->6 +ywang29-vrdb-test1-worker-0:377133:378702 [0] NCCL INFO Channel 18/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:377133:378702 [0] NCCL INFO Channel 19/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:377133:378702 [0] NCCL INFO Channel 20/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:377133:378702 [0] NCCL INFO Channel 21/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:377140:378710 [7] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:377133:378702 [0] NCCL INFO Channel 22/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:377133:378702 [0] NCCL INFO Channel 23/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:377133:378702 [0] NCCL INFO Trees [0] 1/-1/-1->0->-1 [1] 1/-1/-1->0->-1 [2] 1/-1/-1->0->-1 [3] 1/-1/-1->0->-1 [4] 1/-1/-1->0->-1 [5] 1/-1/-1->0->-1 [6] 1/-1/-1->0->-1 [7] 1/-1/-1->0->-1 [8] 1/-1/-1->0->-1 [9] 1/-1/-1->0->-1 [10] 1/-1/-1->0->-1 [11] 1/-1/-1->0->-1 [12] 1/-1/-1->0->-1 [13] 1/-1/-1->0->-1 [14] 1/-1/-1->0->-1 [15] 1/-1/-1->0->-1 [16] 1/-1/-1->0->-1 [17] 1/-1/-1->0->-1 [18] 1/-1/-1->0->-1 [19] 1/-1/-1->0->-1 [20] 1/-1/-1->0->-1 [21] 1/-1/-1->0->-1 [22] 1/-1/-1->0->-1 [23] 1/-1/-1->0->-1 +ywang29-vrdb-test1-worker-0:377133:378702 [0] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:377138:378711 [5] NCCL INFO Channel 00/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377137:378708 [4] NCCL INFO Channel 00/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377135:378707 [2] NCCL INFO Channel 00/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377133:378702 [0] NCCL INFO Channel 00/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377136:378730 [3] NCCL INFO Channel 00/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377134:378706 [1] NCCL INFO Channel 00/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377138:378711 [5] NCCL INFO Channel 01/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377137:378708 [4] NCCL INFO Channel 01/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377135:378707 [2] NCCL INFO Channel 01/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377133:378702 [0] NCCL INFO Channel 01/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377136:378730 [3] NCCL INFO Channel 01/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377134:378706 [1] NCCL INFO Channel 01/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377138:378711 [5] NCCL INFO Channel 02/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377139:378712 [6] NCCL INFO Channel 00/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377137:378708 [4] NCCL INFO Channel 02/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377135:378707 [2] NCCL INFO Channel 02/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377133:378702 [0] NCCL INFO Channel 02/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377136:378730 [3] NCCL INFO Channel 02/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377134:378706 [1] NCCL INFO Channel 02/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377138:378711 [5] NCCL INFO Channel 03/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377137:378708 [4] NCCL INFO Channel 03/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377139:378712 [6] NCCL INFO Channel 01/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377135:378707 [2] NCCL INFO Channel 03/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377140:378710 [7] NCCL INFO Channel 00/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377133:378702 [0] NCCL INFO Channel 03/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377136:378730 [3] NCCL INFO Channel 03/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377134:378706 [1] NCCL INFO Channel 03/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377138:378711 [5] NCCL INFO Channel 04/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377137:378708 [4] NCCL INFO Channel 04/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377135:378707 [2] NCCL INFO Channel 04/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377139:378712 [6] NCCL INFO Channel 02/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377140:378710 [7] NCCL INFO Channel 01/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377133:378702 [0] NCCL INFO Channel 04/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377136:378730 [3] NCCL INFO Channel 04/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377134:378706 [1] NCCL INFO Channel 04/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377138:378711 [5] NCCL INFO Channel 05/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377137:378708 [4] NCCL INFO Channel 05/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377135:378707 [2] NCCL INFO Channel 05/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377139:378712 [6] NCCL INFO Channel 03/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377140:378710 [7] NCCL INFO Channel 02/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377133:378702 [0] NCCL INFO Channel 05/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377136:378730 [3] NCCL INFO Channel 05/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377134:378706 [1] NCCL INFO Channel 05/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377138:378711 [5] NCCL INFO Channel 06/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377137:378708 [4] NCCL INFO Channel 06/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377135:378707 [2] NCCL INFO Channel 06/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377139:378712 [6] NCCL INFO Channel 04/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377140:378710 [7] NCCL INFO Channel 03/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377133:378702 [0] NCCL INFO Channel 06/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377136:378730 [3] NCCL INFO Channel 06/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377134:378706 [1] NCCL INFO Channel 06/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377138:378711 [5] NCCL INFO Channel 07/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377137:378708 [4] NCCL INFO Channel 07/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377135:378707 [2] NCCL INFO Channel 07/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377133:378702 [0] NCCL INFO Channel 07/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377140:378710 [7] NCCL INFO Channel 04/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377139:378712 [6] NCCL INFO Channel 05/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377136:378730 [3] NCCL INFO Channel 07/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377134:378706 [1] NCCL INFO Channel 07/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377138:378711 [5] NCCL INFO Channel 08/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377137:378708 [4] NCCL INFO Channel 08/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377135:378707 [2] NCCL INFO Channel 08/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377133:378702 [0] NCCL INFO Channel 08/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377140:378710 [7] NCCL INFO Channel 05/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377136:378730 [3] NCCL INFO Channel 08/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377139:378712 [6] NCCL INFO Channel 06/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377134:378706 [1] NCCL INFO Channel 08/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377138:378711 [5] NCCL INFO Channel 09/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377137:378708 [4] NCCL INFO Channel 09/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377135:378707 [2] NCCL INFO Channel 09/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377133:378702 [0] NCCL INFO Channel 09/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377140:378710 [7] NCCL INFO Channel 06/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377136:378730 [3] NCCL INFO Channel 09/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377134:378706 [1] NCCL INFO Channel 09/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377139:378712 [6] NCCL INFO Channel 07/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377138:378711 [5] NCCL INFO Channel 10/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377137:378708 [4] NCCL INFO Channel 10/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377135:378707 [2] NCCL INFO Channel 10/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377133:378702 [0] NCCL INFO Channel 10/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377140:378710 [7] NCCL INFO Channel 07/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377134:378706 [1] NCCL INFO Channel 10/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377136:378730 [3] NCCL INFO Channel 10/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377138:378711 [5] NCCL INFO Channel 11/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377139:378712 [6] NCCL INFO Channel 08/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377137:378708 [4] NCCL INFO Channel 11/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377135:378707 [2] NCCL INFO Channel 11/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377133:378702 [0] NCCL INFO Channel 11/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377140:378710 [7] NCCL INFO Channel 08/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377134:378706 [1] NCCL INFO Channel 11/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377136:378730 [3] NCCL INFO Channel 11/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377138:378711 [5] NCCL INFO Channel 12/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377135:378707 [2] NCCL INFO Channel 12/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377139:378712 [6] NCCL INFO Channel 09/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377137:378708 [4] NCCL INFO Channel 12/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377133:378702 [0] NCCL INFO Channel 12/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377134:378706 [1] NCCL INFO Channel 12/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377140:378710 [7] NCCL INFO Channel 09/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377136:378730 [3] NCCL INFO Channel 12/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377138:378711 [5] NCCL INFO Channel 13/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377133:378702 [0] NCCL INFO Channel 13/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377135:378707 [2] NCCL INFO Channel 13/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377137:378708 [4] NCCL INFO Channel 13/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377139:378712 [6] NCCL INFO Channel 10/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377134:378706 [1] NCCL INFO Channel 13/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377140:378710 [7] NCCL INFO Channel 10/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377138:378711 [5] NCCL INFO Channel 14/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377136:378730 [3] NCCL INFO Channel 13/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377133:378702 [0] NCCL INFO Channel 14/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377135:378707 [2] NCCL INFO Channel 14/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377137:378708 [4] NCCL INFO Channel 14/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377139:378712 [6] NCCL INFO Channel 11/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377134:378706 [1] NCCL INFO Channel 14/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377140:378710 [7] NCCL INFO Channel 11/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377138:378711 [5] NCCL INFO Channel 15/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377136:378730 [3] NCCL INFO Channel 14/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377133:378702 [0] NCCL INFO Channel 15/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377135:378707 [2] NCCL INFO Channel 15/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377137:378708 [4] NCCL INFO Channel 15/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377139:378712 [6] NCCL INFO Channel 12/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377134:378706 [1] NCCL INFO Channel 15/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377138:378711 [5] NCCL INFO Channel 16/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377140:378710 [7] NCCL INFO Channel 12/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377133:378702 [0] NCCL INFO Channel 16/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377136:378730 [3] NCCL INFO Channel 15/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377135:378707 [2] NCCL INFO Channel 16/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377137:378708 [4] NCCL INFO Channel 16/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377139:378712 [6] NCCL INFO Channel 13/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377134:378706 [1] NCCL INFO Channel 16/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377138:378711 [5] NCCL INFO Channel 17/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377140:378710 [7] NCCL INFO Channel 13/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377133:378702 [0] NCCL INFO Channel 17/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377136:378730 [3] NCCL INFO Channel 16/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377135:378707 [2] NCCL INFO Channel 17/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377137:378708 [4] NCCL INFO Channel 17/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377139:378712 [6] NCCL INFO Channel 14/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377134:378706 [1] NCCL INFO Channel 17/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377138:378711 [5] NCCL INFO Channel 18/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377140:378710 [7] NCCL INFO Channel 14/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377133:378702 [0] NCCL INFO Channel 18/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377136:378730 [3] NCCL INFO Channel 17/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377135:378707 [2] NCCL INFO Channel 18/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377139:378712 [6] NCCL INFO Channel 15/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377137:378708 [4] NCCL INFO Channel 18/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377134:378706 [1] NCCL INFO Channel 18/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377138:378711 [5] NCCL INFO Channel 19/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377140:378710 [7] NCCL INFO Channel 15/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377133:378702 [0] NCCL INFO Channel 19/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377136:378730 [3] NCCL INFO Channel 18/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377135:378707 [2] NCCL INFO Channel 19/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377139:378712 [6] NCCL INFO Channel 16/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377137:378708 [4] NCCL INFO Channel 19/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377134:378706 [1] NCCL INFO Channel 19/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377138:378711 [5] NCCL INFO Channel 20/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377135:378707 [2] NCCL INFO Channel 20/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377133:378702 [0] NCCL INFO Channel 20/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377140:378710 [7] NCCL INFO Channel 16/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377136:378730 [3] NCCL INFO Channel 19/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377137:378708 [4] NCCL INFO Channel 20/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377139:378712 [6] NCCL INFO Channel 17/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377138:378711 [5] NCCL INFO Channel 21/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377134:378706 [1] NCCL INFO Channel 20/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377135:378707 [2] NCCL INFO Channel 21/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377133:378702 [0] NCCL INFO Channel 21/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377136:378730 [3] NCCL INFO Channel 20/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377140:378710 [7] NCCL INFO Channel 17/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377137:378708 [4] NCCL INFO Channel 21/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377139:378712 [6] NCCL INFO Channel 18/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377138:378711 [5] NCCL INFO Channel 22/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377134:378706 [1] NCCL INFO Channel 21/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377135:378707 [2] NCCL INFO Channel 22/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377133:378702 [0] NCCL INFO Channel 22/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377136:378730 [3] NCCL INFO Channel 21/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377140:378710 [7] NCCL INFO Channel 18/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377137:378708 [4] NCCL INFO Channel 22/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377139:378712 [6] NCCL INFO Channel 19/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377138:378711 [5] NCCL INFO Channel 23/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377134:378706 [1] NCCL INFO Channel 22/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377135:378707 [2] NCCL INFO Channel 23/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377133:378702 [0] NCCL INFO Channel 23/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377136:378730 [3] NCCL INFO Channel 22/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377140:378710 [7] NCCL INFO Channel 19/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377137:378708 [4] NCCL INFO Channel 23/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377139:378712 [6] NCCL INFO Channel 20/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377134:378706 [1] NCCL INFO Channel 23/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377136:378730 [3] NCCL INFO Channel 23/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377140:378710 [7] NCCL INFO Channel 20/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377139:378712 [6] NCCL INFO Channel 21/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377140:378710 [7] NCCL INFO Channel 21/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377139:378712 [6] NCCL INFO Channel 22/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377140:378710 [7] NCCL INFO Channel 22/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377139:378712 [6] NCCL INFO Channel 23/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377140:378710 [7] NCCL INFO Channel 23/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377135:378707 [2] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:377136:378730 [3] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:377134:378706 [1] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:377133:378702 [0] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:377137:378708 [4] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:377140:378710 [7] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:377140:378710 [7] NCCL INFO Channel 00/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377138:378711 [5] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:377139:378712 [6] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:377135:378707 [2] NCCL INFO Channel 00/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377140:378710 [7] NCCL INFO Channel 01/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377135:378707 [2] NCCL INFO Channel 01/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377140:378710 [7] NCCL INFO Channel 02/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377135:378707 [2] NCCL INFO Channel 02/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377140:378710 [7] NCCL INFO Channel 03/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377135:378707 [2] NCCL INFO Channel 03/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377140:378710 [7] NCCL INFO Channel 04/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377135:378707 [2] NCCL INFO Channel 04/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377140:378710 [7] NCCL INFO Channel 05/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377135:378707 [2] NCCL INFO Channel 05/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377135:378707 [2] NCCL INFO Channel 06/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377135:378707 [2] NCCL INFO Channel 07/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377135:378707 [2] NCCL INFO Channel 08/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377135:378707 [2] NCCL INFO Channel 09/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377135:378707 [2] NCCL INFO Channel 10/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377135:378707 [2] NCCL INFO Channel 11/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377140:378710 [7] NCCL INFO Channel 06/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377135:378707 [2] NCCL INFO Channel 12/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377140:378710 [7] NCCL INFO Channel 07/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377136:378730 [3] NCCL INFO Channel 00/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377135:378707 [2] NCCL INFO Channel 13/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377140:378710 [7] NCCL INFO Channel 08/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377136:378730 [3] NCCL INFO Channel 01/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377135:378707 [2] NCCL INFO Channel 14/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377140:378710 [7] NCCL INFO Channel 09/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377136:378730 [3] NCCL INFO Channel 02/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377135:378707 [2] NCCL INFO Channel 15/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377140:378710 [7] NCCL INFO Channel 10/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377136:378730 [3] NCCL INFO Channel 03/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377135:378707 [2] NCCL INFO Channel 16/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377136:378730 [3] NCCL INFO Channel 04/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377140:378710 [7] NCCL INFO Channel 11/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377135:378707 [2] NCCL INFO Channel 17/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377136:378730 [3] NCCL INFO Channel 05/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377140:378710 [7] NCCL INFO Channel 12/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377135:378707 [2] NCCL INFO Channel 18/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377136:378730 [3] NCCL INFO Channel 06/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377140:378710 [7] NCCL INFO Channel 13/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377135:378707 [2] NCCL INFO Channel 19/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377136:378730 [3] NCCL INFO Channel 07/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377135:378707 [2] NCCL INFO Channel 20/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377136:378730 [3] NCCL INFO Channel 08/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377137:378708 [4] NCCL INFO Channel 00/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377135:378707 [2] NCCL INFO Channel 21/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377136:378730 [3] NCCL INFO Channel 09/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377137:378708 [4] NCCL INFO Channel 01/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377135:378707 [2] NCCL INFO Channel 22/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377136:378730 [3] NCCL INFO Channel 10/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377137:378708 [4] NCCL INFO Channel 02/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377135:378707 [2] NCCL INFO Channel 23/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377136:378730 [3] NCCL INFO Channel 11/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377138:378711 [5] NCCL INFO Channel 00/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377137:378708 [4] NCCL INFO Channel 03/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377139:378712 [6] NCCL INFO Channel 00/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377136:378730 [3] NCCL INFO Channel 12/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377138:378711 [5] NCCL INFO Channel 01/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377137:378708 [4] NCCL INFO Channel 04/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377134:378706 [1] NCCL INFO Channel 00/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377139:378712 [6] NCCL INFO Channel 01/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377136:378730 [3] NCCL INFO Channel 13/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377138:378711 [5] NCCL INFO Channel 02/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377134:378706 [1] NCCL INFO Channel 01/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377137:378708 [4] NCCL INFO Channel 05/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377139:378712 [6] NCCL INFO Channel 02/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377136:378730 [3] NCCL INFO Channel 14/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377138:378711 [5] NCCL INFO Channel 03/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377134:378706 [1] NCCL INFO Channel 02/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377137:378708 [4] NCCL INFO Channel 06/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377139:378712 [6] NCCL INFO Channel 03/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377136:378730 [3] NCCL INFO Channel 15/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377138:378711 [5] NCCL INFO Channel 04/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377134:378706 [1] NCCL INFO Channel 03/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377139:378712 [6] NCCL INFO Channel 04/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377137:378708 [4] NCCL INFO Channel 07/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377136:378730 [3] NCCL INFO Channel 16/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377138:378711 [5] NCCL INFO Channel 05/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377134:378706 [1] NCCL INFO Channel 04/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377139:378712 [6] NCCL INFO Channel 05/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377137:378708 [4] NCCL INFO Channel 08/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377136:378730 [3] NCCL INFO Channel 17/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377138:378711 [5] NCCL INFO Channel 06/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377134:378706 [1] NCCL INFO Channel 05/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377140:378710 [7] NCCL INFO Channel 14/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377139:378712 [6] NCCL INFO Channel 06/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377137:378708 [4] NCCL INFO Channel 09/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377136:378730 [3] NCCL INFO Channel 18/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377138:378711 [5] NCCL INFO Channel 07/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377134:378706 [1] NCCL INFO Channel 06/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377140:378710 [7] NCCL INFO Channel 15/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377137:378708 [4] NCCL INFO Channel 10/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377139:378712 [6] NCCL INFO Channel 07/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377136:378730 [3] NCCL INFO Channel 19/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377138:378711 [5] NCCL INFO Channel 08/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377134:378706 [1] NCCL INFO Channel 07/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377140:378710 [7] NCCL INFO Channel 16/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377137:378708 [4] NCCL INFO Channel 11/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377139:378712 [6] NCCL INFO Channel 08/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377136:378730 [3] NCCL INFO Channel 20/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377138:378711 [5] NCCL INFO Channel 09/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377134:378706 [1] NCCL INFO Channel 08/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377140:378710 [7] NCCL INFO Channel 17/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377137:378708 [4] NCCL INFO Channel 12/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377139:378712 [6] NCCL INFO Channel 09/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377136:378730 [3] NCCL INFO Channel 21/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377138:378711 [5] NCCL INFO Channel 10/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377134:378706 [1] NCCL INFO Channel 09/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377140:378710 [7] NCCL INFO Channel 18/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377137:378708 [4] NCCL INFO Channel 13/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377139:378712 [6] NCCL INFO Channel 10/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377136:378730 [3] NCCL INFO Channel 22/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377138:378711 [5] NCCL INFO Channel 11/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377134:378706 [1] NCCL INFO Channel 10/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377140:378710 [7] NCCL INFO Channel 19/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377139:378712 [6] NCCL INFO Channel 11/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377137:378708 [4] NCCL INFO Channel 14/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377136:378730 [3] NCCL INFO Channel 23/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377138:378711 [5] NCCL INFO Channel 12/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377134:378706 [1] NCCL INFO Channel 11/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377140:378710 [7] NCCL INFO Channel 20/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377139:378712 [6] NCCL INFO Channel 12/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377137:378708 [4] NCCL INFO Channel 15/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377138:378711 [5] NCCL INFO Channel 13/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377134:378706 [1] NCCL INFO Channel 12/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377140:378710 [7] NCCL INFO Channel 21/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377139:378712 [6] NCCL INFO Channel 13/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377137:378708 [4] NCCL INFO Channel 16/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377138:378711 [5] NCCL INFO Channel 14/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377134:378706 [1] NCCL INFO Channel 13/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377140:378710 [7] NCCL INFO Channel 22/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377139:378712 [6] NCCL INFO Channel 14/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377137:378708 [4] NCCL INFO Channel 17/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377138:378711 [5] NCCL INFO Channel 15/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377134:378706 [1] NCCL INFO Channel 14/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377140:378710 [7] NCCL INFO Channel 23/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377139:378712 [6] NCCL INFO Channel 15/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377137:378708 [4] NCCL INFO Channel 18/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377138:378711 [5] NCCL INFO Channel 16/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377134:378706 [1] NCCL INFO Channel 15/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377139:378712 [6] NCCL INFO Channel 16/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377138:378711 [5] NCCL INFO Channel 17/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377137:378708 [4] NCCL INFO Channel 19/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377134:378706 [1] NCCL INFO Channel 16/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377139:378712 [6] NCCL INFO Channel 17/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377138:378711 [5] NCCL INFO Channel 18/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377137:378708 [4] NCCL INFO Channel 20/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377134:378706 [1] NCCL INFO Channel 17/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377139:378712 [6] NCCL INFO Channel 18/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377138:378711 [5] NCCL INFO Channel 19/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377137:378708 [4] NCCL INFO Channel 21/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377134:378706 [1] NCCL INFO Channel 18/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377139:378712 [6] NCCL INFO Channel 19/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377138:378711 [5] NCCL INFO Channel 20/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377137:378708 [4] NCCL INFO Channel 22/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377134:378706 [1] NCCL INFO Channel 19/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377139:378712 [6] NCCL INFO Channel 20/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377138:378711 [5] NCCL INFO Channel 21/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377137:378708 [4] NCCL INFO Channel 23/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377134:378706 [1] NCCL INFO Channel 20/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377139:378712 [6] NCCL INFO Channel 21/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377138:378711 [5] NCCL INFO Channel 22/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377134:378706 [1] NCCL INFO Channel 21/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377139:378712 [6] NCCL INFO Channel 22/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377138:378711 [5] NCCL INFO Channel 23/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377134:378706 [1] NCCL INFO Channel 22/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377139:378712 [6] NCCL INFO Channel 23/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377134:378706 [1] NCCL INFO Channel 23/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:377133:378702 [0] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:377133:378702 [0] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:377133:378702 [0] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:377135:378707 [2] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:377135:378707 [2] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:377135:378707 [2] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:377134:378706 [1] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:377134:378706 [1] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:377134:378706 [1] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:377136:378730 [3] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:377136:378730 [3] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:377136:378730 [3] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:377137:378708 [4] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:377137:378708 [4] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:377137:378708 [4] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:377138:378711 [5] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:377138:378711 [5] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:377138:378711 [5] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:377140:378710 [7] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:377140:378710 [7] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:377140:378710 [7] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:377139:378712 [6] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:377139:378712 [6] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:377139:378712 [6] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:377140:378710 [7] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:377137:378708 [4] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:377139:378712 [6] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:377137:378708 [4] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:377139:378712 [6] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:377137:378708 [4] NCCL INFO ncclCommInitRank comm 0x5610c27f1020 rank 4 nranks 8 cudaDev 4 nvmlDev 4 busId 901c0 commId 0x531bd79cf4336d7e - Init COMPLETE +ywang29-vrdb-test1-worker-0:377139:378712 [6] NCCL INFO ncclCommInitRank comm 0x55a39c6a1c00 rank 6 nranks 8 cudaDev 6 nvmlDev 6 busId a01c0 commId 0x531bd79cf4336d7e - Init COMPLETE +ywang29-vrdb-test1-worker-0:377140:378710 [7] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:377140:378710 [7] NCCL INFO ncclCommInitRank comm 0x55cd0c220340 rank 7 nranks 8 cudaDev 7 nvmlDev 7 busId a01d0 commId 0x531bd79cf4336d7e - Init COMPLETE +ywang29-vrdb-test1-worker-0:377135:378707 [2] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:377136:378730 [3] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:377134:378706 [1] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:377135:378707 [2] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:377136:378730 [3] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:377134:378706 [1] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:377135:378707 [2] NCCL INFO ncclCommInitRank comm 0x55e614272ff0 rank 2 nranks 8 cudaDev 2 nvmlDev 2 busId 201c0 commId 0x531bd79cf4336d7e - Init COMPLETE +ywang29-vrdb-test1-worker-0:377136:378730 [3] NCCL INFO ncclCommInitRank comm 0x5572fbfb1900 rank 3 nranks 8 cudaDev 3 nvmlDev 3 busId 201d0 commId 0x531bd79cf4336d7e - Init COMPLETE +ywang29-vrdb-test1-worker-0:377134:378706 [1] NCCL INFO ncclCommInitRank comm 0x55eba66cc6c0 rank 1 nranks 8 cudaDev 1 nvmlDev 1 busId 101d0 commId 0x531bd79cf4336d7e - Init COMPLETE +ywang29-vrdb-test1-worker-0:377138:378711 [5] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:377138:378711 [5] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:377138:378711 [5] NCCL INFO ncclCommInitRank comm 0x55c16df47420 rank 5 nranks 8 cudaDev 5 nvmlDev 5 busId 901d0 commId 0x531bd79cf4336d7e - Init COMPLETE +ywang29-vrdb-test1-worker-0:377133:378702 [0] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:377133:378702 [0] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:377133:378702 [0] NCCL INFO ncclCommInitRank comm 0x5596a2c83f20 rank 0 nranks 8 cudaDev 0 nvmlDev 0 busId 101c0 commId 0x531bd79cf4336d7e - Init COMPLETE +[2025-10-12 12:36:49,031] [INFO] [partition_parameters.py:343:__exit__] finished initializing model - num_params = 459, num_elems = 0.99B +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model + /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +[2025-10-12 12:36:50,784] [INFO] [partition_parameters.py:343:__exit__] finished initializing model - num_params = 907, num_elems = 1.42B +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +TinyLlavaForConditionalGeneration( + (language_model): Qwen2ForCausalLM( + (model): Qwen2Model( + (embed_tokens): Embedding(151936, 896) + (layers): ModuleList( + (0-23): 24 x Qwen2DecoderLayer( + (self_attn): Qwen2FlashAttention2( + (q_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=True) + (k_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=128, bias=True) + (v_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=128, bias=True) + (o_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=False) + (rotary_emb): Qwen2RotaryEmbedding() + ) + (mlp): Qwen2MLP( + (gate_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=4864, bias=False) + (up_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=4864, bias=False) + (down_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=4864, out_features=896, bias=False) + (act_fn): SiLU() + ) + (input_layernorm): Qwen2RMSNorm() + (post_attention_layernorm): Qwen2RMSNorm() + ) + ) + (norm): Qwen2RMSNorm() + ) + (lm_head): Linear(in_features=896, out_features=151936, bias=False) + ) + (vision_tower): SIGLIPVisionTower( + (_vision_tower): SiglipVisionModel( + (vision_model): SiglipVisionTransformer( + (embeddings): SiglipVisionEmbeddings( + (patch_embedding): Conv2d(3, 1152, kernel_size=(14, 14), stride=(14, 14), padding=valid) + (position_embedding): Embedding(729, 1152) + ) + (encoder): SiglipEncoder( + (layers): ModuleList( + (0-26): 27 x SiglipEncoderLayer( + (self_attn): SiglipAttention( + (k_proj): Linear(in_features=1152, out_features=1152, bias=True) + (v_proj): Linear(in_features=1152, out_features=1152, bias=True) + (q_proj): Linear(in_features=1152, out_features=1152, bias=True) + (out_proj): Linear(in_features=1152, out_features=1152, bias=True) + ) + (layer_norm1): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + (layer_norm2): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + ) + ) + ) + (post_layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (head): SiglipMultiheadAttentionPoolingHead( + (attention): MultiheadAttention( + (out_proj): NonDynamicallyQuantizableLinear(in_features=1152, out_features=1152, bias=True) + ) + (layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + ) + ) + ) + ) + (connector): MLPConnector( + (_connector): Sequential( + (0): SupermaskLinearSparsity_SoftForward_Normal(in_features=1152, out_features=896, bias=True) + (1): GELU(approximate='none') + (2): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=True) + ) + ) +) +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_towerLoading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... + +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin...Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... + +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin...[E ProcessGroupNCCL.cpp:474] [Rank 0] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=3202, OpType=_ALLGATHER_BASE, NumelIn=17016832, NumelOut=136134656, Timeout(ms)=1800000) ran for 1800286 milliseconds before timing out. + +[E ProcessGroupNCCL.cpp:474] [Rank 3] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=3202, OpType=_ALLGATHER_BASE, NumelIn=17016832, NumelOut=136134656, Timeout(ms)=1800000) ran for 1800403 milliseconds before timing out. +[E ProcessGroupNCCL.cpp:474] [Rank 6] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=3202, OpType=_ALLGATHER_BASE, NumelIn=17016832, NumelOut=136134656, Timeout(ms)=1800000) ran for 1800813 milliseconds before timing out. +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +ywang29-vrdb-test1-worker-0:377136:378736 [3] NCCL INFO [Service thread] Connection closed by localRank 3 +ywang29-vrdb-test1-worker-0:377133:378743 [0] NCCL INFO [Service thread] Connection closed by localRank 0 +ywang29-vrdb-test1-worker-0:377139:378738 [6] NCCL INFO [Service thread] Connection closed by localRank 6 +Pre-training init language_model.model.layers.0.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.0.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.0.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.0.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.0.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.0.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.0.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.mlp.down_proj.scores: Mean=3.000000 +ywang29-vrdb-test1-worker-0:377133:378251 [0] NCCL INFO comm 0x5596a2c83f20 rank 0 nranks 8 cudaDev 0 busId 101c0 - Abort COMPLETE +[E ProcessGroupNCCL.cpp:488] Some NCCL operations have failed or timed out. Due to the asynchronous nature of CUDA kernels, subsequent GPU operations might run on corrupted/incomplete data. +[E ProcessGroupNCCL.cpp:494] To avoid data inconsistency, we are taking the entire process down. +[E ProcessGroupNCCL.cpp:915] [Rank 0] NCCL watchdog thread terminated with exception: [Rank 0] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=3202, OpType=_ALLGATHER_BASE, NumelIn=17016832, NumelOut=136134656, Timeout(ms)=1800000) ran for 1800286 milliseconds before timing out. +terminate called after throwing an instance of 'std::runtime_error' + what(): [Rank 0] NCCL watchdog thread terminated with exception: [Rank 0] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=3202, OpType=_ALLGATHER_BASE, NumelIn=17016832, NumelOut=136134656, Timeout(ms)=1800000) ran for 1800286 milliseconds before timing out. +ywang29-vrdb-test1-worker-0:377136:378249 [3] NCCL INFO comm 0x5572fbfb1900 rank 3 nranks 8 cudaDev 3 busId 201d0 - Abort COMPLETE +[E ProcessGroupNCCL.cpp:488] Some NCCL operations have failed or timed out. Due to the asynchronous nature of CUDA kernels, subsequent GPU operations might run on corrupted/incomplete data. +[E ProcessGroupNCCL.cpp:494] To avoid data inconsistency, we are taking the entire process down. +[E ProcessGroupNCCL.cpp:915] [Rank 3] NCCL watchdog thread terminated with exception: [Rank 3] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=3202, OpType=_ALLGATHER_BASE, NumelIn=17016832, NumelOut=136134656, Timeout(ms)=1800000) ran for 1800403 milliseconds before timing out. +terminate called after throwing an instance of 'std::runtime_error' + what(): [Rank 3] NCCL watchdog thread terminated with exception: [Rank 3] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=3202, OpType=_ALLGATHER_BASE, NumelIn=17016832, NumelOut=136134656, Timeout(ms)=1800000) ran for 1800403 milliseconds before timing out. +ywang29-vrdb-test1-worker-0:377139:378239 [6] NCCL INFO comm 0x55a39c6a1c00 rank 6 nranks 8 cudaDev 6 busId a01c0 - Abort COMPLETE +[E ProcessGroupNCCL.cpp:488] Some NCCL operations have failed or timed out. Due to the asynchronous nature of CUDA kernels, subsequent GPU operations might run on corrupted/incomplete data. +[E ProcessGroupNCCL.cpp:494] To avoid data inconsistency, we are taking the entire process down. +[E ProcessGroupNCCL.cpp:915] [Rank 6] NCCL watchdog thread terminated with exception: [Rank 6] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=3202, OpType=_ALLGATHER_BASE, NumelIn=17016832, NumelOut=136134656, Timeout(ms)=1800000) ran for 1800813 milliseconds before timing out. +terminate called after throwing an instance of 'std::runtime_error' + what(): [Rank 6] NCCL watchdog thread terminated with exception: [Rank 6] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=3202, OpType=_ALLGATHER_BASE, NumelIn=17016832, NumelOut=136134656, Timeout(ms)=1800000) ran for 1800813 milliseconds before timing out. +[2025-10-12 13:36:58,904] [INFO] [launch.py:316:sigkill_handler] Killing subprocess 377133 +[2025-10-12 13:36:58,904] [INFO] [launch.py:316:sigkill_handler] Killing subprocess 377134 +[2025-10-12 13:36:59,241] [INFO] [launch.py:316:sigkill_handler] Killing subprocess 377135 +[2025-10-12 13:36:59,579] [INFO] [launch.py:316:sigkill_handler] Killing subprocess 377136 +[2025-10-12 13:37:00,234] [INFO] [launch.py:316:sigkill_handler] Killing subprocess 377137 +[2025-10-12 13:37:00,571] [INFO] [launch.py:316:sigkill_handler] Killing subprocess 377138 +[2025-10-12 13:37:00,948] [INFO] [launch.py:316:sigkill_handler] Killing subprocess 377139 +[2025-10-12 13:37:00,950] [INFO] [launch.py:316:sigkill_handler] Killing subprocess 377140 +[2025-10-12 13:37:01,286] [ERROR] [launch.py:322:sigkill_handler] ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=7', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_2e-1_connector-3.0_0.5_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] exits with return code = -6 +==== EXPERIMENT COMPLETED: qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_2e-1_connector-3.0_0.5_2e-1_ablation ==== +Log File: qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_2e-1_connector-3.0_0.5_2e-1_ablation_20251012_123546.log +Timestamp: 2025-10-12 13:37:02 +===================================== diff --git a/logs_oct11/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_3e-2_connector-3.0_0.5_3e-2_ablation_20251012_100541.log b/logs_oct11/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_3e-2_connector-3.0_0.5_3e-2_ablation_20251012_100541.log new file mode 100644 index 0000000000000000000000000000000000000000..7528fc7e0a38b50cb3998df2846d937440a6e5ba --- /dev/null +++ b/logs_oct11/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_3e-2_connector-3.0_0.5_3e-2_ablation_20251012_100541.log @@ -0,0 +1,2312 @@ +==== STARTING EXPERIMENT: qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_3e-2_connector-3.0_0.5_3e-2_ablation ==== +Log File: qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_3e-2_connector-3.0_0.5_3e-2_ablation_20251012_100541.log +Timestamp: 2025-10-12 10:05:41 +===================================== +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 10:05:43,998] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 10:05:46,724] [WARNING] [runner.py:202:fetch_hostfile] Unable to find hostfile, will proceed with training with local resources only. +[2025-10-12 10:05:46,726] [INFO] [runner.py:568:main] cmd = /opt/conda/envs/tinyllava/bin/python3.10 -u -m deepspeed.launcher.launch --world_info=eyJsb2NhbGhvc3QiOiBbMCwgMSwgMiwgMywgNCwgNSwgNiwgN119 --master_addr=127.0.0.1 --master_port=29501 --enable_each_rank_log=None tinyllava/train/train.py --deepspeed ./scripts/zero3.json --data_path /s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json --image_folder /s3-code/ywang29/datasets/tinyllava --is_multimodal True --conv_version qwen2_base --model_name_or_path Qwen/Qwen2.5-0.5B --vision_tower google/siglip-so400m-patch14-384 --vision_tower2 --connector_type mlp2x_gelu --mm_vision_select_layer -2 --image_aspect_ratio square --attn_implementation flash_attention_2 --bf16 True --training_recipe common --tune_type_llm full --tune_type_vision_tower frozen --tune_vision_tower_from_layer 0 --tune_type_connector full --group_by_modality_length True --pretrained_model_path /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain --output_dir /nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_3e-2_connector-3.0_0.5_3e-2_ablation --num_train_epochs 1 --per_device_train_batch_size 4 --per_device_eval_batch_size 4 --gradient_accumulation_steps 4 --evaluation_strategy no --learning_rate 3e-2 --weight_decay 0. --warmup_ratio 0.03 --lr_scheduler_type cosine --logging_steps 1 --tf32 False --model_max_length 2048 --gradient_checkpointing True --dataloader_num_workers 8 --lazy_preprocess True --report_to tensorboard --tokenizer_use_fast False --run_name tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune --subnet_mode_text both --subnet_type_text None --mask_type_text soft --init_mean_text 3.0 --temperature_attn_text 0.5 --temperature_mlp_text 0.5 --backward_type_text normal --masked_layers_text all --subnet_mode_vision both --subnet_type_vision None --mask_type_vision soft --init_mean_vision 3.0 --temperature_attn_vision 0.5 --temperature_mlp_vision 0.5 --backward_type_vision normal --masked_layers_vision all --subnet_type_connector global --mask_type_connector soft --init_mean_connector 3.0 --temperature_connector 0.5 --backward_type_connector normal --mm_projector_lr 3e-2 --seed 42 --mask_model llm-connector --save_strategy steps --save_steps 50000 --save_total_limit 1 --train_data_ratio 0.1 +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 10:05:49,284] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 10:05:50,395] [INFO] [launch.py:138:main] 0 NCCL_VERSION=2.21.5 +[2025-10-12 10:05:50,395] [INFO] [launch.py:138:main] 0 NCCL_SOCKET_IFNAME=eth +[2025-10-12 10:05:50,395] [INFO] [launch.py:145:main] WORLD INFO DICT: {'localhost': [0, 1, 2, 3, 4, 5, 6, 7]} +[2025-10-12 10:05:50,395] [INFO] [launch.py:151:main] nnodes=1, num_local_procs=8, node_rank=0 +[2025-10-12 10:05:50,395] [INFO] [launch.py:162:main] global_rank_mapping=defaultdict(, {'localhost': [0, 1, 2, 3, 4, 5, 6, 7]}) +[2025-10-12 10:05:50,395] [INFO] [launch.py:163:main] dist_world_size=8 +[2025-10-12 10:05:50,395] [INFO] [launch.py:165:main] Setting CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 +[2025-10-12 10:05:50,398] [INFO] [launch.py:253:main] process 684538 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=0', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_3e-2_connector-3.0_0.5_3e-2_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '3e-2', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '3e-2', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 10:05:50,400] [INFO] [launch.py:253:main] process 684539 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=1', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_3e-2_connector-3.0_0.5_3e-2_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '3e-2', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '3e-2', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 10:05:50,402] [INFO] [launch.py:253:main] process 684540 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=2', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_3e-2_connector-3.0_0.5_3e-2_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '3e-2', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '3e-2', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 10:05:50,403] [INFO] [launch.py:253:main] process 684541 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=3', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_3e-2_connector-3.0_0.5_3e-2_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '3e-2', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '3e-2', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 10:05:50,405] [INFO] [launch.py:253:main] process 684542 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=4', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_3e-2_connector-3.0_0.5_3e-2_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '3e-2', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '3e-2', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 10:05:50,407] [INFO] [launch.py:253:main] process 684543 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=5', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_3e-2_connector-3.0_0.5_3e-2_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '3e-2', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '3e-2', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 10:05:50,409] [INFO] [launch.py:253:main] process 684544 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=6', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_3e-2_connector-3.0_0.5_3e-2_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '3e-2', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '3e-2', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 10:05:50,411] [INFO] [launch.py:253:main] process 684545 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=7', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_3e-2_connector-3.0_0.5_3e-2_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '3e-2', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '3e-2', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 10:05:57,103] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 10:05:57,349] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 10:05:57,391] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 10:05:57,413] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 10:05:57,435] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 10:05:57,446] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 10:05:57,446] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 10:05:57,449] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 10:05:57,506] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 10:05:57,754] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 10:05:57,793] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 10:05:57,794] [INFO] [comm.py:668:init_distributed] Initializing TorchBackend in DeepSpeed with backend nccl +[2025-10-12 10:05:57,814] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 10:05:57,831] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 10:05:57,843] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 10:05:57,843] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 10:05:57,849] [INFO] [comm.py:637:init_distributed] cdb=None +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +{'llm': {'model_name_or_path': 'Qwen/Qwen2.5-0.5B', 'cache_dir': None, 'attn_implementation': 'flash_attention_2', 'subnet_mode': 'both', 'subnet_type': 'None', 'sparsity_attn': None, 'sparsity_mlp': None, 'threshold_attn': None, 'threshold_mlp': None, 'temperature_attn': 0.5, 'temperature_mlp': 0.5, 'masked_layers': 'all', 'mask_type': 'soft', 'backward_type': 'normal'}, 'vision_tower': {'model_name_or_path': 'google/siglip-so400m-patch14-384'}, 'connector': {'connector_type': 'mlp2x_gelu', 'subnet_type': 'global', 'threshold': None, 'sparsity': None, 'temperature': 0.5, 'mask_type': 'soft', 'backward_type': 'normal'}} +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +TinyLlavaConfig { + "backward_type_connector": "normal", + "cache_dir": null, + "connector_type": "mlp2x_gelu", + "hidden_size": 896, + "ignore_index": -100, + "image_aspect_ratio": "square", + "image_token_index": -200, + "llm_model_name_or_path": "Qwen/Qwen2.5-0.5B", + "mask_model": [ + "llm", + "connector" + ], + "mask_type_connector": "soft", + "model_type": "tinyllava", + "num_queries": 128, + "num_resampler_layers": 3, + "pad_token": null, + "resampler_hidden_size": 768, + "sparsity_connector": null, + "subnet_type_connector": "global", + "temperature_connector": 0.5, + "text_config": { + "_name_or_path": "Qwen/Qwen2.5-0.5B", + "architectures": [ + "Qwen2ForCausalLM" + ], + "backward_type": "normal", + "bos_token_id": 151643, + "eos_token_id": 151643, + "hidden_size": 896, + "intermediate_size": 4864, + "mask_type": "soft", + "masked_layers": "all", + "max_position_embeddings": 32768, + "max_window_layers": 24, + "model_type": "qwen2", + "num_attention_heads": 14, + "num_hidden_layers": 24, + "num_key_value_heads": 2, + "rope_theta": 1000000.0, + "sliding_window": 32768, + "subnet_mode": "both", + "subnet_type": "None", + "temperature_attn": 0.5, + "temperature_mlp": 0.5, + "tie_word_embeddings": true, + "torch_dtype": "bfloat16", + "use_mrope": false, + "use_sliding_window": false, + "vocab_size": 151936 + }, + "threshold_connector": null, + "tokenizer_model_max_length": 2048, + "tokenizer_name_or_path": "Qwen/Qwen2.5-0.5B", + "tokenizer_padding_side": "right", + "tokenizer_use_fast": false, + "transformers_version": "4.40.1", + "tune_type_connector": "frozen", + "tune_type_llm": "frozen", + "tune_type_vision_tower": "frozen", + "tune_vision_tower_from_layer": -1, + "use_cache": false, + "vision_config": { + "hidden_act": "gelu_pytorch_tanh", + "hidden_size": 1152, + "image_size": 384, + "intermediate_size": 4304, + "layer_norm_eps": 1e-06, + "model_name_or_path": "google/siglip-so400m-patch14-384", + "model_name_or_path2": "", + "model_type": "siglip_vision_model", + "num_attention_heads": 16, + "num_hidden_layers": 27, + "patch_size": 14 + }, + "vision_feature_layer": -2, + "vision_feature_select_strategy": "patch", + "vision_hidden_size": 1152, + "vision_model_name_or_path": "google/siglip-so400m-patch14-384", + "vision_model_name_or_path2": "", + "vocab_size": 151936 +} + +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test2-worker-0:684538:684538 [0] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:684538:684538 [0] NCCL INFO Bootstrap : Using eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:684538:684538 [0] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test2-worker-0:684538:684538 [0] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test2-worker-0:684538:684538 [0] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test2-worker-0:684538:684538 [0] NCCL INFO cudaDriverVersion 12040 +NCCL version 2.21.5+cuda12.1 +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test2-worker-0:684544:684544 [6] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test2-worker-0:684544:684544 [6] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:684544:684544 [6] NCCL INFO Bootstrap : Using eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:684543:684543 [5] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test2-worker-0:684543:684543 [5] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:684544:684544 [6] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test2-worker-0:684544:684544 [6] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test2-worker-0:684544:684544 [6] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test2-worker-0:684543:684543 [5] NCCL INFO Bootstrap : Using eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:684543:684543 [5] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test2-worker-0:684543:684543 [5] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test2-worker-0:684543:684543 [5] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test2-worker-0:684541:684541 [3] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test2-worker-0:684541:684541 [3] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:684541:684541 [3] NCCL INFO Bootstrap : Using eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:684541:684541 [3] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test2-worker-0:684541:684541 [3] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test2-worker-0:684541:684541 [3] NCCL INFO NET/Plugin: Using internal network plugin. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test2-worker-0:684539:684539 [1] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test2-worker-0:684539:684539 [1] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:684539:684539 [1] NCCL INFO Bootstrap : Using eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:684539:684539 [1] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test2-worker-0:684539:684539 [1] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test2-worker-0:684539:684539 [1] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test2-worker-0:684538:686127 [0] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:684538:686127 [0] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test2-worker-0:684538:686127 [0] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:684538:686127 [0] NCCL INFO NET/Socket : Using [0]eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:684538:686127 [0] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test2-worker-0:684538:686127 [0] NCCL INFO Using network Socket +ywang29-vrdb-test2-worker-0:684540:684540 [2] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test2-worker-0:684540:684540 [2] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:684540:684540 [2] NCCL INFO Bootstrap : Using eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:684540:684540 [2] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test2-worker-0:684540:684540 [2] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test2-worker-0:684540:684540 [2] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test2-worker-0:684543:686129 [5] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:684544:686128 [6] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:684543:686129 [5] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test2-worker-0:684544:686128 [6] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test2-worker-0:684544:686128 [6] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:684543:686129 [5] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:684543:686129 [5] NCCL INFO NET/Socket : Using [0]eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:684544:686128 [6] NCCL INFO NET/Socket : Using [0]eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:684543:686129 [5] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test2-worker-0:684544:686128 [6] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test2-worker-0:684544:686128 [6] NCCL INFO Using network Socket +ywang29-vrdb-test2-worker-0:684543:686129 [5] NCCL INFO Using network Socket +ywang29-vrdb-test2-worker-0:684541:686130 [3] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:684541:686130 [3] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test2-worker-0:684541:686130 [3] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:684541:686130 [3] NCCL INFO NET/Socket : Using [0]eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:684541:686130 [3] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test2-worker-0:684541:686130 [3] NCCL INFO Using network Socket +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test2-worker-0:684542:684542 [4] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test2-worker-0:684542:684542 [4] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:684542:684542 [4] NCCL INFO Bootstrap : Using eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:684542:684542 [4] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test2-worker-0:684542:684542 [4] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test2-worker-0:684542:684542 [4] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test2-worker-0:684539:686147 [1] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:684539:686147 [1] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test2-worker-0:684539:686147 [1] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:684539:686147 [1] NCCL INFO NET/Socket : Using [0]eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:684539:686147 [1] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test2-worker-0:684539:686147 [1] NCCL INFO Using network Socket +ywang29-vrdb-test2-worker-0:684540:686148 [2] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:684540:686148 [2] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test2-worker-0:684540:686148 [2] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:684540:686148 [2] NCCL INFO NET/Socket : Using [0]eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:684540:686148 [2] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test2-worker-0:684540:686148 [2] NCCL INFO Using network Socket +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test2-worker-0:684545:684545 [7] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test2-worker-0:684545:684545 [7] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:684545:684545 [7] NCCL INFO Bootstrap : Using eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:684545:684545 [7] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test2-worker-0:684545:684545 [7] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test2-worker-0:684545:684545 [7] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test2-worker-0:684542:686149 [4] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:684542:686149 [4] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test2-worker-0:684542:686149 [4] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:684542:686149 [4] NCCL INFO NET/Socket : Using [0]eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:684542:686149 [4] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test2-worker-0:684542:686149 [4] NCCL INFO Using network Socket +ywang29-vrdb-test2-worker-0:684545:686150 [7] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:684545:686150 [7] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test2-worker-0:684545:686150 [7] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:684545:686150 [7] NCCL INFO NET/Socket : Using [0]eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:684545:686150 [7] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test2-worker-0:684545:686150 [7] NCCL INFO Using network Socket +ywang29-vrdb-test2-worker-0:684539:686147 [1] NCCL INFO ncclCommInitRank comm 0x555d80a6c920 rank 1 nranks 8 cudaDev 1 nvmlDev 1 busId 101d0 commId 0x4743d905a7e968d0 - Init START +ywang29-vrdb-test2-worker-0:684545:686150 [7] NCCL INFO ncclCommInitRank comm 0x55ac6433d190 rank 7 nranks 8 cudaDev 7 nvmlDev 7 busId a01d0 commId 0x4743d905a7e968d0 - Init START +ywang29-vrdb-test2-worker-0:684544:686128 [6] NCCL INFO ncclCommInitRank comm 0x55eb15243410 rank 6 nranks 8 cudaDev 6 nvmlDev 6 busId a01c0 commId 0x4743d905a7e968d0 - Init START +ywang29-vrdb-test2-worker-0:684543:686129 [5] NCCL INFO ncclCommInitRank comm 0x55f482687f60 rank 5 nranks 8 cudaDev 5 nvmlDev 5 busId 901d0 commId 0x4743d905a7e968d0 - Init START +ywang29-vrdb-test2-worker-0:684540:686148 [2] NCCL INFO ncclCommInitRank comm 0x55cd7aba0240 rank 2 nranks 8 cudaDev 2 nvmlDev 2 busId 201c0 commId 0x4743d905a7e968d0 - Init START +ywang29-vrdb-test2-worker-0:684538:686127 [0] NCCL INFO ncclCommInitRank comm 0x563ee01f2300 rank 0 nranks 8 cudaDev 0 nvmlDev 0 busId 101c0 commId 0x4743d905a7e968d0 - Init START +ywang29-vrdb-test2-worker-0:684541:686130 [3] NCCL INFO ncclCommInitRank comm 0x5574dcd75f70 rank 3 nranks 8 cudaDev 3 nvmlDev 3 busId 201d0 commId 0x4743d905a7e968d0 - Init START +ywang29-vrdb-test2-worker-0:684542:686149 [4] NCCL INFO ncclCommInitRank comm 0x555c97c3dea0 rank 4 nranks 8 cudaDev 4 nvmlDev 4 busId 901c0 commId 0x4743d905a7e968d0 - Init START +ywang29-vrdb-test2-worker-0:684538:686127 [0] NCCL INFO Setting affinity for GPU 0 to ff,ffff0000,00ffffff +ywang29-vrdb-test2-worker-0:684538:686127 [0] NCCL INFO NVLS multicast support is not available on dev 0 +ywang29-vrdb-test2-worker-0:684545:686150 [7] NCCL INFO Setting affinity for GPU 7 to ffffff00,0000ffff,ff000000 +ywang29-vrdb-test2-worker-0:684545:686150 [7] NCCL INFO NVLS multicast support is not available on dev 7 +ywang29-vrdb-test2-worker-0:684543:686129 [5] NCCL INFO Setting affinity for GPU 5 to ffffff00,0000ffff,ff000000 +ywang29-vrdb-test2-worker-0:684541:686130 [3] NCCL INFO Setting affinity for GPU 3 to ff,ffff0000,00ffffff +ywang29-vrdb-test2-worker-0:684541:686130 [3] NCCL INFO NVLS multicast support is not available on dev 3 +ywang29-vrdb-test2-worker-0:684543:686129 [5] NCCL INFO NVLS multicast support is not available on dev 5 +ywang29-vrdb-test2-worker-0:684544:686128 [6] NCCL INFO Setting affinity for GPU 6 to ffffff00,0000ffff,ff000000 +ywang29-vrdb-test2-worker-0:684544:686128 [6] NCCL INFO NVLS multicast support is not available on dev 6 +ywang29-vrdb-test2-worker-0:684540:686148 [2] NCCL INFO Setting affinity for GPU 2 to ff,ffff0000,00ffffff +ywang29-vrdb-test2-worker-0:684540:686148 [2] NCCL INFO NVLS multicast support is not available on dev 2 +ywang29-vrdb-test2-worker-0:684539:686147 [1] NCCL INFO Setting affinity for GPU 1 to ff,ffff0000,00ffffff +ywang29-vrdb-test2-worker-0:684539:686147 [1] NCCL INFO NVLS multicast support is not available on dev 1 +ywang29-vrdb-test2-worker-0:684542:686149 [4] NCCL INFO Setting affinity for GPU 4 to ffffff00,0000ffff,ff000000 +ywang29-vrdb-test2-worker-0:684542:686149 [4] NCCL INFO NVLS multicast support is not available on dev 4 +ywang29-vrdb-test2-worker-0:684544:686128 [6] NCCL INFO comm 0x55eb15243410 rank 6 nRanks 8 nNodes 1 localRanks 8 localRank 6 MNNVL 0 +ywang29-vrdb-test2-worker-0:684543:686129 [5] NCCL INFO comm 0x55f482687f60 rank 5 nRanks 8 nNodes 1 localRanks 8 localRank 5 MNNVL 0 +ywang29-vrdb-test2-worker-0:684540:686148 [2] NCCL INFO comm 0x55cd7aba0240 rank 2 nRanks 8 nNodes 1 localRanks 8 localRank 2 MNNVL 0 +ywang29-vrdb-test2-worker-0:684542:686149 [4] NCCL INFO comm 0x555c97c3dea0 rank 4 nRanks 8 nNodes 1 localRanks 8 localRank 4 MNNVL 0 +ywang29-vrdb-test2-worker-0:684539:686147 [1] NCCL INFO comm 0x555d80a6c920 rank 1 nRanks 8 nNodes 1 localRanks 8 localRank 1 MNNVL 0 +ywang29-vrdb-test2-worker-0:684541:686130 [3] NCCL INFO comm 0x5574dcd75f70 rank 3 nRanks 8 nNodes 1 localRanks 8 localRank 3 MNNVL 0 +ywang29-vrdb-test2-worker-0:684538:686127 [0] NCCL INFO comm 0x563ee01f2300 rank 0 nRanks 8 nNodes 1 localRanks 8 localRank 0 MNNVL 0 +ywang29-vrdb-test2-worker-0:684545:686150 [7] NCCL INFO comm 0x55ac6433d190 rank 7 nRanks 8 nNodes 1 localRanks 8 localRank 7 MNNVL 0 +ywang29-vrdb-test2-worker-0:684538:686127 [0] NCCL INFO Channel 00/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:684538:686127 [0] NCCL INFO Channel 01/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:684538:686127 [0] NCCL INFO Channel 02/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:684544:686128 [6] NCCL INFO Trees [0] 7/-1/-1->6->5 [1] 7/-1/-1->6->5 [2] 7/-1/-1->6->5 [3] 7/-1/-1->6->5 [4] 7/-1/-1->6->5 [5] 7/-1/-1->6->5 [6] 7/-1/-1->6->5 [7] 7/-1/-1->6->5 [8] 7/-1/-1->6->5 [9] 7/-1/-1->6->5 [10] 7/-1/-1->6->5 [11] 7/-1/-1->6->5 [12] 7/-1/-1->6->5 [13] 7/-1/-1->6->5 [14] 7/-1/-1->6->5 [15] 7/-1/-1->6->5 [16] 7/-1/-1->6->5 [17] 7/-1/-1->6->5 [18] 7/-1/-1->6->5 [19] 7/-1/-1->6->5 [20] 7/-1/-1->6->5 [21] 7/-1/-1->6->5 [22] 7/-1/-1->6->5 [23] 7/-1/-1->6->5 +ywang29-vrdb-test2-worker-0:684538:686127 [0] NCCL INFO Channel 03/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:684538:686127 [0] NCCL INFO Channel 04/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:684543:686129 [5] NCCL INFO Trees [0] 6/-1/-1->5->4 [1] 6/-1/-1->5->4 [2] 6/-1/-1->5->4 [3] 6/-1/-1->5->4 [4] 6/-1/-1->5->4 [5] 6/-1/-1->5->4 [6] 6/-1/-1->5->4 [7] 6/-1/-1->5->4 [8] 6/-1/-1->5->4 [9] 6/-1/-1->5->4 [10] 6/-1/-1->5->4 [11] 6/-1/-1->5->4 [12] 6/-1/-1->5->4 [13] 6/-1/-1->5->4 [14] 6/-1/-1->5->4 [15] 6/-1/-1->5->4 [16] 6/-1/-1->5->4 [17] 6/-1/-1->5->4 [18] 6/-1/-1->5->4 [19] 6/-1/-1->5->4 [20] 6/-1/-1->5->4 [21] 6/-1/-1->5->4 [22] 6/-1/-1->5->4 [23] 6/-1/-1->5->4 +ywang29-vrdb-test2-worker-0:684540:686148 [2] NCCL INFO Trees [0] 3/-1/-1->2->1 [1] 3/-1/-1->2->1 [2] 3/-1/-1->2->1 [3] 3/-1/-1->2->1 [4] 3/-1/-1->2->1 [5] 3/-1/-1->2->1 [6] 3/-1/-1->2->1 [7] 3/-1/-1->2->1 [8] 3/-1/-1->2->1 [9] 3/-1/-1->2->1 [10] 3/-1/-1->2->1 [11] 3/-1/-1->2->1 [12] 3/-1/-1->2->1 [13] 3/-1/-1->2->1 [14] 3/-1/-1->2->1 [15] 3/-1/-1->2->1 [16] 3/-1/-1->2->1 [17] 3/-1/-1->2->1 [18] 3/-1/-1->2->1 [19] 3/-1/-1->2->1 [20] 3/-1/-1->2->1 [21] 3/-1/-1->2->1 [22] 3/-1/-1->2->1 [23] 3/-1/-1->2->1 +ywang29-vrdb-test2-worker-0:684538:686127 [0] NCCL INFO Channel 05/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:684544:686128 [6] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:684538:686127 [0] NCCL INFO Channel 06/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:684540:686148 [2] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:684538:686127 [0] NCCL INFO Channel 07/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:684543:686129 [5] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:684538:686127 [0] NCCL INFO Channel 08/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:684538:686127 [0] NCCL INFO Channel 09/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:684538:686127 [0] NCCL INFO Channel 10/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:684538:686127 [0] NCCL INFO Channel 11/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:684542:686149 [4] NCCL INFO Trees [0] 5/-1/-1->4->3 [1] 5/-1/-1->4->3 [2] 5/-1/-1->4->3 [3] 5/-1/-1->4->3 [4] 5/-1/-1->4->3 [5] 5/-1/-1->4->3 [6] 5/-1/-1->4->3 [7] 5/-1/-1->4->3 [8] 5/-1/-1->4->3 [9] 5/-1/-1->4->3 [10] 5/-1/-1->4->3 [11] 5/-1/-1->4->3 [12] 5/-1/-1->4->3 [13] 5/-1/-1->4->3 [14] 5/-1/-1->4->3 [15] 5/-1/-1->4->3 [16] 5/-1/-1->4->3 [17] 5/-1/-1->4->3 [18] 5/-1/-1->4->3 [19] 5/-1/-1->4->3 [20] 5/-1/-1->4->3 [21] 5/-1/-1->4->3 [22] 5/-1/-1->4->3 [23] 5/-1/-1->4->3 +ywang29-vrdb-test2-worker-0:684538:686127 [0] NCCL INFO Channel 12/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:684538:686127 [0] NCCL INFO Channel 13/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:684538:686127 [0] NCCL INFO Channel 14/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:684538:686127 [0] NCCL INFO Channel 15/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:684542:686149 [4] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:684538:686127 [0] NCCL INFO Channel 16/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:684539:686147 [1] NCCL INFO Trees [0] 2/-1/-1->1->0 [1] 2/-1/-1->1->0 [2] 2/-1/-1->1->0 [3] 2/-1/-1->1->0 [4] 2/-1/-1->1->0 [5] 2/-1/-1->1->0 [6] 2/-1/-1->1->0 [7] 2/-1/-1->1->0 [8] 2/-1/-1->1->0 [9] 2/-1/-1->1->0 [10] 2/-1/-1->1->0 [11] 2/-1/-1->1->0 [12] 2/-1/-1->1->0 [13] 2/-1/-1->1->0 [14] 2/-1/-1->1->0 [15] 2/-1/-1->1->0 [16] 2/-1/-1->1->0 [17] 2/-1/-1->1->0 [18] 2/-1/-1->1->0 [19] 2/-1/-1->1->0 [20] 2/-1/-1->1->0 [21] 2/-1/-1->1->0 [22] 2/-1/-1->1->0 [23] 2/-1/-1->1->0 +ywang29-vrdb-test2-worker-0:684538:686127 [0] NCCL INFO Channel 17/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:684541:686130 [3] NCCL INFO Trees [0] 4/-1/-1->3->2 [1] 4/-1/-1->3->2 [2] 4/-1/-1->3->2 [3] 4/-1/-1->3->2 [4] 4/-1/-1->3->2 [5] 4/-1/-1->3->2 [6] 4/-1/-1->3->2 [7] 4/-1/-1->3->2 [8] 4/-1/-1->3->2 [9] 4/-1/-1->3->2 [10] 4/-1/-1->3->2 [11] 4/-1/-1->3->2 [12] 4/-1/-1->3->2 [13] 4/-1/-1->3->2 [14] 4/-1/-1->3->2 [15] 4/-1/-1->3->2 [16] 4/-1/-1->3->2 [17] 4/-1/-1->3->2 [18] 4/-1/-1->3->2 [19] 4/-1/-1->3->2 [20] 4/-1/-1->3->2 [21] 4/-1/-1->3->2 [22] 4/-1/-1->3->2 [23] 4/-1/-1->3->2 +ywang29-vrdb-test2-worker-0:684538:686127 [0] NCCL INFO Channel 18/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:684538:686127 [0] NCCL INFO Channel 19/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:684538:686127 [0] NCCL INFO Channel 20/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:684545:686150 [7] NCCL INFO Trees [0] -1/-1/-1->7->6 [1] -1/-1/-1->7->6 [2] -1/-1/-1->7->6 [3] -1/-1/-1->7->6 [4] -1/-1/-1->7->6 [5] -1/-1/-1->7->6 [6] -1/-1/-1->7->6 [7] -1/-1/-1->7->6 [8] -1/-1/-1->7->6 [9] -1/-1/-1->7->6 [10] -1/-1/-1->7->6 [11] -1/-1/-1->7->6 [12] -1/-1/-1->7->6 [13] -1/-1/-1->7->6 [14] -1/-1/-1->7->6 [15] -1/-1/-1->7->6 [16] -1/-1/-1->7->6 [17] -1/-1/-1->7->6 [18] -1/-1/-1->7->6 [19] -1/-1/-1->7->6 [20] -1/-1/-1->7->6 [21] -1/-1/-1->7->6 [22] -1/-1/-1->7->6 [23] -1/-1/-1->7->6 +ywang29-vrdb-test2-worker-0:684539:686147 [1] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:684538:686127 [0] NCCL INFO Channel 21/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:684541:686130 [3] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:684538:686127 [0] NCCL INFO Channel 22/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:684538:686127 [0] NCCL INFO Channel 23/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:684545:686150 [7] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:684538:686127 [0] NCCL INFO Trees [0] 1/-1/-1->0->-1 [1] 1/-1/-1->0->-1 [2] 1/-1/-1->0->-1 [3] 1/-1/-1->0->-1 [4] 1/-1/-1->0->-1 [5] 1/-1/-1->0->-1 [6] 1/-1/-1->0->-1 [7] 1/-1/-1->0->-1 [8] 1/-1/-1->0->-1 [9] 1/-1/-1->0->-1 [10] 1/-1/-1->0->-1 [11] 1/-1/-1->0->-1 [12] 1/-1/-1->0->-1 [13] 1/-1/-1->0->-1 [14] 1/-1/-1->0->-1 [15] 1/-1/-1->0->-1 [16] 1/-1/-1->0->-1 [17] 1/-1/-1->0->-1 [18] 1/-1/-1->0->-1 [19] 1/-1/-1->0->-1 [20] 1/-1/-1->0->-1 [21] 1/-1/-1->0->-1 [22] 1/-1/-1->0->-1 [23] 1/-1/-1->0->-1 +ywang29-vrdb-test2-worker-0:684538:686127 [0] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:684544:686128 [6] NCCL INFO Channel 00/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:686129 [5] NCCL INFO Channel 00/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:686128 [6] NCCL INFO Channel 01/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684538:686127 [0] NCCL INFO Channel 00/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:686129 [5] NCCL INFO Channel 01/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:686128 [6] NCCL INFO Channel 02/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684538:686127 [0] NCCL INFO Channel 01/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:686129 [5] NCCL INFO Channel 02/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:686149 [4] NCCL INFO Channel 00/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:686128 [6] NCCL INFO Channel 03/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:686130 [3] NCCL INFO Channel 00/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:686150 [7] NCCL INFO Channel 00/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684538:686127 [0] NCCL INFO Channel 02/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:686147 [1] NCCL INFO Channel 00/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:686129 [5] NCCL INFO Channel 03/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:686149 [4] NCCL INFO Channel 01/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:686128 [6] NCCL INFO Channel 04/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:686130 [3] NCCL INFO Channel 01/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684538:686127 [0] NCCL INFO Channel 03/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:686150 [7] NCCL INFO Channel 01/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:686129 [5] NCCL INFO Channel 04/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:686147 [1] NCCL INFO Channel 01/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:686149 [4] NCCL INFO Channel 02/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:686128 [6] NCCL INFO Channel 05/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:686130 [3] NCCL INFO Channel 02/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684538:686127 [0] NCCL INFO Channel 04/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:686150 [7] NCCL INFO Channel 02/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:686129 [5] NCCL INFO Channel 05/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:686147 [1] NCCL INFO Channel 02/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:686128 [6] NCCL INFO Channel 06/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:686149 [4] NCCL INFO Channel 03/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:686130 [3] NCCL INFO Channel 03/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684538:686127 [0] NCCL INFO Channel 05/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:686150 [7] NCCL INFO Channel 03/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:686129 [5] NCCL INFO Channel 06/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:686147 [1] NCCL INFO Channel 03/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:686128 [6] NCCL INFO Channel 07/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:686149 [4] NCCL INFO Channel 04/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684538:686127 [0] NCCL INFO Channel 06/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:686130 [3] NCCL INFO Channel 04/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:686150 [7] NCCL INFO Channel 04/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:686129 [5] NCCL INFO Channel 07/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:686128 [6] NCCL INFO Channel 08/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:686147 [1] NCCL INFO Channel 04/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:686149 [4] NCCL INFO Channel 05/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:686148 [2] NCCL INFO Channel 00/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684538:686127 [0] NCCL INFO Channel 07/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:686130 [3] NCCL INFO Channel 05/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:686150 [7] NCCL INFO Channel 05/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:686129 [5] NCCL INFO Channel 08/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:686128 [6] NCCL INFO Channel 09/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:686147 [1] NCCL INFO Channel 05/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:686148 [2] NCCL INFO Channel 01/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684538:686127 [0] NCCL INFO Channel 08/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:686149 [4] NCCL INFO Channel 06/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:686130 [3] NCCL INFO Channel 06/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:686128 [6] NCCL INFO Channel 10/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:686150 [7] NCCL INFO Channel 06/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:686129 [5] NCCL INFO Channel 09/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:686148 [2] NCCL INFO Channel 02/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684538:686127 [0] NCCL INFO Channel 09/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:686147 [1] NCCL INFO Channel 06/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:686149 [4] NCCL INFO Channel 07/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:686130 [3] NCCL INFO Channel 07/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:686128 [6] NCCL INFO Channel 11/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:686150 [7] NCCL INFO Channel 07/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:686129 [5] NCCL INFO Channel 10/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:686148 [2] NCCL INFO Channel 03/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684538:686127 [0] NCCL INFO Channel 10/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:686147 [1] NCCL INFO Channel 07/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:686149 [4] NCCL INFO Channel 08/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:686130 [3] NCCL INFO Channel 08/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:686128 [6] NCCL INFO Channel 12/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:686150 [7] NCCL INFO Channel 08/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:686129 [5] NCCL INFO Channel 11/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:686148 [2] NCCL INFO Channel 04/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684538:686127 [0] NCCL INFO Channel 11/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:686147 [1] NCCL INFO Channel 08/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:686149 [4] NCCL INFO Channel 09/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:686130 [3] NCCL INFO Channel 09/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:686128 [6] NCCL INFO Channel 13/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:686129 [5] NCCL INFO Channel 12/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:686150 [7] NCCL INFO Channel 09/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:686148 [2] NCCL INFO Channel 05/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684538:686127 [0] NCCL INFO Channel 12/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:686147 [1] NCCL INFO Channel 09/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:686149 [4] NCCL INFO Channel 10/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:686129 [5] NCCL INFO Channel 13/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:686128 [6] NCCL INFO Channel 14/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:686130 [3] NCCL INFO Channel 10/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:686150 [7] NCCL INFO Channel 10/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:686148 [2] NCCL INFO Channel 06/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684538:686127 [0] NCCL INFO Channel 13/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:686147 [1] NCCL INFO Channel 10/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:686149 [4] NCCL INFO Channel 11/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:686129 [5] NCCL INFO Channel 14/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:686128 [6] NCCL INFO Channel 15/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:686130 [3] NCCL INFO Channel 11/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684538:686127 [0] NCCL INFO Channel 14/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:686148 [2] NCCL INFO Channel 07/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:686150 [7] NCCL INFO Channel 11/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:686147 [1] NCCL INFO Channel 11/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:686128 [6] NCCL INFO Channel 16/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:686149 [4] NCCL INFO Channel 12/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:686129 [5] NCCL INFO Channel 15/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:686130 [3] NCCL INFO Channel 12/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684538:686127 [0] NCCL INFO Channel 15/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:686148 [2] NCCL INFO Channel 08/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:686150 [7] NCCL INFO Channel 12/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:686129 [5] NCCL INFO Channel 16/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:686147 [1] NCCL INFO Channel 12/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:686128 [6] NCCL INFO Channel 17/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:686149 [4] NCCL INFO Channel 13/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:686130 [3] NCCL INFO Channel 13/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684538:686127 [0] NCCL INFO Channel 16/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:686148 [2] NCCL INFO Channel 09/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:686150 [7] NCCL INFO Channel 13/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:686129 [5] NCCL INFO Channel 17/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:686147 [1] NCCL INFO Channel 13/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:686128 [6] NCCL INFO Channel 18/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:686149 [4] NCCL INFO Channel 14/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684538:686127 [0] NCCL INFO Channel 17/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:686130 [3] NCCL INFO Channel 14/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:686148 [2] NCCL INFO Channel 10/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:686150 [7] NCCL INFO Channel 14/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:686129 [5] NCCL INFO Channel 18/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:686128 [6] NCCL INFO Channel 19/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:686147 [1] NCCL INFO Channel 14/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:686149 [4] NCCL INFO Channel 15/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684538:686127 [0] NCCL INFO Channel 18/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:686130 [3] NCCL INFO Channel 15/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:686148 [2] NCCL INFO Channel 11/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:686150 [7] NCCL INFO Channel 15/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:686129 [5] NCCL INFO Channel 19/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:686128 [6] NCCL INFO Channel 20/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:686147 [1] NCCL INFO Channel 15/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:686149 [4] NCCL INFO Channel 16/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684538:686127 [0] NCCL INFO Channel 19/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:686130 [3] NCCL INFO Channel 16/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:686148 [2] NCCL INFO Channel 12/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:686150 [7] NCCL INFO Channel 16/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:686129 [5] NCCL INFO Channel 20/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:686128 [6] NCCL INFO Channel 21/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:686147 [1] NCCL INFO Channel 16/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:686149 [4] NCCL INFO Channel 17/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:686130 [3] NCCL INFO Channel 17/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684538:686127 [0] NCCL INFO Channel 20/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:686148 [2] NCCL INFO Channel 13/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:686150 [7] NCCL INFO Channel 17/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:686129 [5] NCCL INFO Channel 21/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:686128 [6] NCCL INFO Channel 22/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:686147 [1] NCCL INFO Channel 17/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:686149 [4] NCCL INFO Channel 18/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:686130 [3] NCCL INFO Channel 18/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684538:686127 [0] NCCL INFO Channel 21/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:686148 [2] NCCL INFO Channel 14/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:686129 [5] NCCL INFO Channel 22/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:686150 [7] NCCL INFO Channel 18/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:686128 [6] NCCL INFO Channel 23/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:686147 [1] NCCL INFO Channel 18/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:686149 [4] NCCL INFO Channel 19/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:686130 [3] NCCL INFO Channel 19/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684538:686127 [0] NCCL INFO Channel 22/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:686148 [2] NCCL INFO Channel 15/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:686129 [5] NCCL INFO Channel 23/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:686150 [7] NCCL INFO Channel 19/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:686147 [1] NCCL INFO Channel 19/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:686149 [4] NCCL INFO Channel 20/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684538:686127 [0] NCCL INFO Channel 23/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:686130 [3] NCCL INFO Channel 20/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:686148 [2] NCCL INFO Channel 16/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:686150 [7] NCCL INFO Channel 20/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:686147 [1] NCCL INFO Channel 20/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:686149 [4] NCCL INFO Channel 21/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:686148 [2] NCCL INFO Channel 17/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:686130 [3] NCCL INFO Channel 21/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:686150 [7] NCCL INFO Channel 21/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:686147 [1] NCCL INFO Channel 21/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:686148 [2] NCCL INFO Channel 18/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:686149 [4] NCCL INFO Channel 22/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:686130 [3] NCCL INFO Channel 22/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:686150 [7] NCCL INFO Channel 22/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:686147 [1] NCCL INFO Channel 22/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:686148 [2] NCCL INFO Channel 19/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:686149 [4] NCCL INFO Channel 23/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:686130 [3] NCCL INFO Channel 23/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:686150 [7] NCCL INFO Channel 23/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:686148 [2] NCCL INFO Channel 20/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:686147 [1] NCCL INFO Channel 23/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:686148 [2] NCCL INFO Channel 21/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:686148 [2] NCCL INFO Channel 22/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:686148 [2] NCCL INFO Channel 23/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:686148 [2] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:684539:686147 [1] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:684541:686130 [3] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:684538:686127 [0] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:684542:686149 [4] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:684545:686150 [7] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:684545:686150 [7] NCCL INFO Channel 00/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:686129 [5] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:684544:686128 [6] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:684545:686150 [7] NCCL INFO Channel 01/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:686148 [2] NCCL INFO Channel 00/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:686150 [7] NCCL INFO Channel 02/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:686148 [2] NCCL INFO Channel 01/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:686150 [7] NCCL INFO Channel 03/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:686148 [2] NCCL INFO Channel 02/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:686150 [7] NCCL INFO Channel 04/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:686148 [2] NCCL INFO Channel 03/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:686150 [7] NCCL INFO Channel 05/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:686148 [2] NCCL INFO Channel 04/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:686150 [7] NCCL INFO Channel 06/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:686148 [2] NCCL INFO Channel 05/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:686150 [7] NCCL INFO Channel 07/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:686150 [7] NCCL INFO Channel 08/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:686147 [1] NCCL INFO Channel 00/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:686150 [7] NCCL INFO Channel 09/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:686147 [1] NCCL INFO Channel 01/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:686150 [7] NCCL INFO Channel 10/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:686147 [1] NCCL INFO Channel 02/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:686150 [7] NCCL INFO Channel 11/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:686147 [1] NCCL INFO Channel 03/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:686150 [7] NCCL INFO Channel 12/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:686147 [1] NCCL INFO Channel 04/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:686148 [2] NCCL INFO Channel 06/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:686147 [1] NCCL INFO Channel 05/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:686148 [2] NCCL INFO Channel 07/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:686147 [1] NCCL INFO Channel 06/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:686148 [2] NCCL INFO Channel 08/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:686147 [1] NCCL INFO Channel 07/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:686148 [2] NCCL INFO Channel 09/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:686147 [1] NCCL INFO Channel 08/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:686148 [2] NCCL INFO Channel 10/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:686147 [1] NCCL INFO Channel 09/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:686148 [2] NCCL INFO Channel 11/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:686147 [1] NCCL INFO Channel 10/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:686148 [2] NCCL INFO Channel 12/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:686147 [1] NCCL INFO Channel 11/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:686150 [7] NCCL INFO Channel 13/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:686148 [2] NCCL INFO Channel 13/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:686147 [1] NCCL INFO Channel 12/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:686150 [7] NCCL INFO Channel 14/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:686148 [2] NCCL INFO Channel 14/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:686147 [1] NCCL INFO Channel 13/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:686150 [7] NCCL INFO Channel 15/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:686148 [2] NCCL INFO Channel 15/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:686147 [1] NCCL INFO Channel 14/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:686150 [7] NCCL INFO Channel 16/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:686148 [2] NCCL INFO Channel 16/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:686147 [1] NCCL INFO Channel 15/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:686128 [6] NCCL INFO Channel 00/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:686130 [3] NCCL INFO Channel 00/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:686150 [7] NCCL INFO Channel 17/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:686148 [2] NCCL INFO Channel 17/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:686147 [1] NCCL INFO Channel 16/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:686128 [6] NCCL INFO Channel 01/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:686150 [7] NCCL INFO Channel 18/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:686130 [3] NCCL INFO Channel 01/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:686148 [2] NCCL INFO Channel 18/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:686147 [1] NCCL INFO Channel 17/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:686128 [6] NCCL INFO Channel 02/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:686150 [7] NCCL INFO Channel 19/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:686130 [3] NCCL INFO Channel 02/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:686148 [2] NCCL INFO Channel 19/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:686147 [1] NCCL INFO Channel 18/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:686128 [6] NCCL INFO Channel 03/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:686150 [7] NCCL INFO Channel 20/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:686130 [3] NCCL INFO Channel 03/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:686147 [1] NCCL INFO Channel 19/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:686148 [2] NCCL INFO Channel 20/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:686129 [5] NCCL INFO Channel 00/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:686128 [6] NCCL INFO Channel 04/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:686150 [7] NCCL INFO Channel 21/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:686130 [3] NCCL INFO Channel 04/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:686147 [1] NCCL INFO Channel 20/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:686148 [2] NCCL INFO Channel 21/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:686128 [6] NCCL INFO Channel 05/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:686129 [5] NCCL INFO Channel 01/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:686150 [7] NCCL INFO Channel 22/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:686130 [3] NCCL INFO Channel 05/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:686147 [1] NCCL INFO Channel 21/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:686148 [2] NCCL INFO Channel 22/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:686128 [6] NCCL INFO Channel 06/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:686129 [5] NCCL INFO Channel 02/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:686150 [7] NCCL INFO Channel 23/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:686130 [3] NCCL INFO Channel 06/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:686147 [1] NCCL INFO Channel 22/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:686148 [2] NCCL INFO Channel 23/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:686128 [6] NCCL INFO Channel 07/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:686129 [5] NCCL INFO Channel 03/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:686130 [3] NCCL INFO Channel 07/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:686147 [1] NCCL INFO Channel 23/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:686128 [6] NCCL INFO Channel 08/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:686129 [5] NCCL INFO Channel 04/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:686130 [3] NCCL INFO Channel 08/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:686128 [6] NCCL INFO Channel 09/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:686130 [3] NCCL INFO Channel 09/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:686129 [5] NCCL INFO Channel 05/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:686128 [6] NCCL INFO Channel 10/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:686130 [3] NCCL INFO Channel 10/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:686129 [5] NCCL INFO Channel 06/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:686128 [6] NCCL INFO Channel 11/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:686130 [3] NCCL INFO Channel 11/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:686128 [6] NCCL INFO Channel 12/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:686129 [5] NCCL INFO Channel 07/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:686130 [3] NCCL INFO Channel 12/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:686128 [6] NCCL INFO Channel 13/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:686129 [5] NCCL INFO Channel 08/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:686130 [3] NCCL INFO Channel 13/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:686128 [6] NCCL INFO Channel 14/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:686129 [5] NCCL INFO Channel 09/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:686130 [3] NCCL INFO Channel 14/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:686128 [6] NCCL INFO Channel 15/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:686130 [3] NCCL INFO Channel 15/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:686129 [5] NCCL INFO Channel 10/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:686130 [3] NCCL INFO Channel 16/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:686130 [3] NCCL INFO Channel 17/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:686130 [3] NCCL INFO Channel 18/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:686129 [5] NCCL INFO Channel 11/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:686128 [6] NCCL INFO Channel 16/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:686130 [3] NCCL INFO Channel 19/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:686129 [5] NCCL INFO Channel 12/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:686128 [6] NCCL INFO Channel 17/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:686130 [3] NCCL INFO Channel 20/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:686129 [5] NCCL INFO Channel 13/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:686128 [6] NCCL INFO Channel 18/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:686130 [3] NCCL INFO Channel 21/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:686128 [6] NCCL INFO Channel 19/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:686149 [4] NCCL INFO Channel 00/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:686129 [5] NCCL INFO Channel 14/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:686130 [3] NCCL INFO Channel 22/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:686129 [5] NCCL INFO Channel 15/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:686130 [3] NCCL INFO Channel 23/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:686128 [6] NCCL INFO Channel 20/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:686128 [6] NCCL INFO Channel 21/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:686128 [6] NCCL INFO Channel 22/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:686128 [6] NCCL INFO Channel 23/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:686129 [5] NCCL INFO Channel 16/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:686149 [4] NCCL INFO Channel 01/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:686129 [5] NCCL INFO Channel 17/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:686149 [4] NCCL INFO Channel 02/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:686149 [4] NCCL INFO Channel 03/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:686129 [5] NCCL INFO Channel 18/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:686149 [4] NCCL INFO Channel 04/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:686129 [5] NCCL INFO Channel 19/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:686149 [4] NCCL INFO Channel 05/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:686129 [5] NCCL INFO Channel 20/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:686149 [4] NCCL INFO Channel 06/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:686149 [4] NCCL INFO Channel 07/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:686129 [5] NCCL INFO Channel 21/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:686149 [4] NCCL INFO Channel 08/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:686129 [5] NCCL INFO Channel 22/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:686149 [4] NCCL INFO Channel 09/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:686129 [5] NCCL INFO Channel 23/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:686149 [4] NCCL INFO Channel 10/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:686149 [4] NCCL INFO Channel 11/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:686149 [4] NCCL INFO Channel 12/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:686149 [4] NCCL INFO Channel 13/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:686149 [4] NCCL INFO Channel 14/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:686149 [4] NCCL INFO Channel 15/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:686149 [4] NCCL INFO Channel 16/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:686149 [4] NCCL INFO Channel 17/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:686149 [4] NCCL INFO Channel 18/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:686149 [4] NCCL INFO Channel 19/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:686149 [4] NCCL INFO Channel 20/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:686149 [4] NCCL INFO Channel 21/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:686149 [4] NCCL INFO Channel 22/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:686149 [4] NCCL INFO Channel 23/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684538:686127 [0] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:684538:686127 [0] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:684538:686127 [0] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:684539:686147 [1] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:684539:686147 [1] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:684539:686147 [1] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:684545:686150 [7] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:684545:686150 [7] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:684545:686150 [7] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:684540:686148 [2] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:684540:686148 [2] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:684540:686148 [2] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:684541:686130 [3] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:684541:686130 [3] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:684541:686130 [3] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:684542:686149 [4] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:684542:686149 [4] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:684542:686149 [4] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:684544:686128 [6] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:684543:686129 [5] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:684544:686128 [6] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:684544:686128 [6] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:684543:686129 [5] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:684543:686129 [5] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:684543:686129 [5] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test2-worker-0:684543:686129 [5] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test2-worker-0:684543:686129 [5] NCCL INFO ncclCommInitRank comm 0x55f482687f60 rank 5 nranks 8 cudaDev 5 nvmlDev 5 busId 901d0 commId 0x4743d905a7e968d0 - Init COMPLETE +ywang29-vrdb-test2-worker-0:684545:686150 [7] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test2-worker-0:684545:686150 [7] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test2-worker-0:684545:686150 [7] NCCL INFO ncclCommInitRank comm 0x55ac6433d190 rank 7 nranks 8 cudaDev 7 nvmlDev 7 busId a01d0 commId 0x4743d905a7e968d0 - Init COMPLETE +ywang29-vrdb-test2-worker-0:684541:686130 [3] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test2-worker-0:684541:686130 [3] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test2-worker-0:684541:686130 [3] NCCL INFO ncclCommInitRank comm 0x5574dcd75f70 rank 3 nranks 8 cudaDev 3 nvmlDev 3 busId 201d0 commId 0x4743d905a7e968d0 - Init COMPLETE +ywang29-vrdb-test2-worker-0:684539:686147 [1] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test2-worker-0:684539:686147 [1] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test2-worker-0:684539:686147 [1] NCCL INFO ncclCommInitRank comm 0x555d80a6c920 rank 1 nranks 8 cudaDev 1 nvmlDev 1 busId 101d0 commId 0x4743d905a7e968d0 - Init COMPLETE +ywang29-vrdb-test2-worker-0:684540:686148 [2] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test2-worker-0:684540:686148 [2] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test2-worker-0:684540:686148 [2] NCCL INFO ncclCommInitRank comm 0x55cd7aba0240 rank 2 nranks 8 cudaDev 2 nvmlDev 2 busId 201c0 commId 0x4743d905a7e968d0 - Init COMPLETE +ywang29-vrdb-test2-worker-0:684544:686128 [6] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test2-worker-0:684544:686128 [6] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test2-worker-0:684544:686128 [6] NCCL INFO ncclCommInitRank comm 0x55eb15243410 rank 6 nranks 8 cudaDev 6 nvmlDev 6 busId a01c0 commId 0x4743d905a7e968d0 - Init COMPLETE +ywang29-vrdb-test2-worker-0:684542:686149 [4] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test2-worker-0:684542:686149 [4] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test2-worker-0:684538:686127 [0] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test2-worker-0:684538:686127 [0] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test2-worker-0:684542:686149 [4] NCCL INFO ncclCommInitRank comm 0x555c97c3dea0 rank 4 nranks 8 cudaDev 4 nvmlDev 4 busId 901c0 commId 0x4743d905a7e968d0 - Init COMPLETE +ywang29-vrdb-test2-worker-0:684538:686127 [0] NCCL INFO ncclCommInitRank comm 0x563ee01f2300 rank 0 nranks 8 cudaDev 0 nvmlDev 0 busId 101c0 commId 0x4743d905a7e968d0 - Init COMPLETE +[2025-10-12 10:06:45,191] [INFO] [partition_parameters.py:343:__exit__] finished initializing model - num_params = 459, num_elems = 0.99B +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +[2025-10-12 10:33:29,764] [INFO] [partition_parameters.py:343:__exit__] finished initializing model - num_params = 907, num_elems = 1.42B +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +TinyLlavaForConditionalGeneration( + (language_model): Qwen2ForCausalLM( + (model): Qwen2Model( + (embed_tokens): Embedding(151936, 896) + (layers): ModuleList( + (0-23): 24 x Qwen2DecoderLayer( + (self_attn): Qwen2FlashAttention2( + (q_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=True) + (k_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=128, bias=True) + (v_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=128, bias=True) + (o_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=False) + (rotary_emb): Qwen2RotaryEmbedding() + ) + (mlp): Qwen2MLP( + (gate_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=4864, bias=False) + (up_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=4864, bias=False) + (down_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=4864, out_features=896, bias=False) + (act_fn): SiLU() + ) + (input_layernorm): Qwen2RMSNorm() + (post_attention_layernorm): Qwen2RMSNorm() + ) + ) + (norm): Qwen2RMSNorm() + ) + (lm_head): Linear(in_features=896, out_features=151936, bias=False) + ) + (vision_tower): SIGLIPVisionTower( + (_vision_tower): SiglipVisionModel( + (vision_model): SiglipVisionTransformer( + (embeddings): SiglipVisionEmbeddings( + (patch_embedding): Conv2d(3, 1152, kernel_size=(14, 14), stride=(14, 14), padding=valid) + (position_embedding): Embedding(729, 1152) + ) + (encoder): SiglipEncoder( + (layers): ModuleList( + (0-26): 27 x SiglipEncoderLayer( + (self_attn): SiglipAttention( + (k_proj): Linear(in_features=1152, out_features=1152, bias=True) + (v_proj): Linear(in_features=1152, out_features=1152, bias=True) + (q_proj): Linear(in_features=1152, out_features=1152, bias=True) + (out_proj): Linear(in_features=1152, out_features=1152, bias=True) + ) + (layer_norm1): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + (layer_norm2): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + ) + ) + ) + (post_layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (head): SiglipMultiheadAttentionPoolingHead( + (attention): MultiheadAttention( + (out_proj): NonDynamicallyQuantizableLinear(in_features=1152, out_features=1152, bias=True) + ) + (layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + ) + ) + ) + ) + (connector): MLPConnector( + (_connector): Sequential( + (0): SupermaskLinearSparsity_SoftForward_Normal(in_features=1152, out_features=896, bias=True) + (1): GELU(approximate='none') + (2): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=True) + ) + ) +) +Pre-training init language_model.model.layers.0.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.0.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.0.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.0.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.0.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.0.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.0.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.mlp.down_proj.scores: Mean=3.000000 +Pre-training init connector._connector.0.scores: Mean=3.000005 +Pre-training init connector._connector.2.scores: Mean=2.999970 +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +2025-10-12 10:33:48,207 | INFO: Total Parameters: 1283756736, Total Trainable Parameters: 359661568 +2025-10-12 10:33:48,212 | INFO: Trainable Parameters: +language_model.model.layers.0.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.0.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.0.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.0.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.0.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.0.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.0.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.1.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.1.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.1.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.1.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.1.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.1.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.1.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.2.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.2.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.2.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.2.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.2.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.2.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.2.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.3.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.3.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.3.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.3.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.3.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.3.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.3.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.4.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.4.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.4.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.4.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.4.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.4.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.4.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.5.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.5.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.5.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.5.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.5.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.5.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.5.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.6.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.6.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.6.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.6.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.6.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.6.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.6.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.7.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.7.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.7.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.7.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.7.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.7.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.7.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.8.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.8.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.8.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.8.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.8.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.8.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.8.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.9.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.9.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.9.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.9.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.9.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.9.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.9.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.10.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.10.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.10.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.10.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.10.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.10.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.10.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.11.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.11.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.11.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.11.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.11.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.11.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.11.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.12.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.12.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.12.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.12.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.12.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.12.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.12.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.13.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.13.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.13.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.13.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.13.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.13.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.13.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.14.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.14.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.14.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.14.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.14.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.14.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.14.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.15.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.15.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.15.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.15.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.15.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.15.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.15.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.16.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.16.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.16.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.16.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.16.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.16.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.16.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.17.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.17.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.17.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.17.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.17.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.17.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.17.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.18.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.18.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.18.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.18.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.18.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.18.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.18.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.19.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.19.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.19.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.19.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.19.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.19.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.19.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.20.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.20.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.20.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.20.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.20.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.20.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.20.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.21.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.21.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.21.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.21.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.21.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.21.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.21.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.22.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.22.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.22.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.22.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.22.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.22.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.22.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.23.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.23.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.23.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.23.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.23.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.23.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.23.mlp.down_proj.scores: 4358144 parameters +connector._connector.0.scores: 1032192 parameters +connector._connector.2.scores: 802816 parameters +Parameter Offload: Total persistent parameters: 486464 in 403 params + 0%| | 0/520 [00:004->3 [1] 5/-1/-1->4->3 [2] 5/-1/-1->4->3 [3] 5/-1/-1->4->3 [4] 5/-1/-1->4->3 [5] 5/-1/-1->4->3 [6] 5/-1/-1->4->3 [7] 5/-1/-1->4->3 [8] 5/-1/-1->4->3 [9] 5/-1/-1->4->3 [10] 5/-1/-1->4->3 [11] 5/-1/-1->4->3 [12] 5/-1/-1->4->3 [13] 5/-1/-1->4->3 [14] 5/-1/-1->4->3 [15] 5/-1/-1->4->3 [16] 5/-1/-1->4->3 [17] 5/-1/-1->4->3 [18] 5/-1/-1->4->3 [19] 5/-1/-1->4->3 [20] 5/-1/-1->4->3 [21] 5/-1/-1->4->3 [22] 5/-1/-1->4->3 [23] 5/-1/-1->4->3 +ywang29-vrdb-test2-worker-0:684538:691610 [0] NCCL INFO Channel 02/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:684543:691611 [5] NCCL INFO Trees [0] 6/-1/-1->5->4 [1] 6/-1/-1->5->4 [2] 6/-1/-1->5->4 [3] 6/-1/-1->5->4 [4] 6/-1/-1->5->4 [5] 6/-1/-1->5->4 [6] 6/-1/-1->5->4 [7] 6/-1/-1->5->4 [8] 6/-1/-1->5->4 [9] 6/-1/-1->5->4 [10] 6/-1/-1->5->4 [11] 6/-1/-1->5->4 [12] 6/-1/-1->5->4 [13] 6/-1/-1->5->4 [14] 6/-1/-1->5->4 [15] 6/-1/-1->5->4 [16] 6/-1/-1->5->4 [17] 6/-1/-1->5->4 [18] 6/-1/-1->5->4 [19] 6/-1/-1->5->4 [20] 6/-1/-1->5->4 [21] 6/-1/-1->5->4 [22] 6/-1/-1->5->4 [23] 6/-1/-1->5->4 +ywang29-vrdb-test2-worker-0:684542:691615 [4] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:684538:691610 [0] NCCL INFO Channel 03/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:684543:691611 [5] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:684538:691610 [0] NCCL INFO Channel 04/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:684541:691617 [3] NCCL INFO Trees [0] 4/-1/-1->3->2 [1] 4/-1/-1->3->2 [2] 4/-1/-1->3->2 [3] 4/-1/-1->3->2 [4] 4/-1/-1->3->2 [5] 4/-1/-1->3->2 [6] 4/-1/-1->3->2 [7] 4/-1/-1->3->2 [8] 4/-1/-1->3->2 [9] 4/-1/-1->3->2 [10] 4/-1/-1->3->2 [11] 4/-1/-1->3->2 [12] 4/-1/-1->3->2 [13] 4/-1/-1->3->2 [14] 4/-1/-1->3->2 [15] 4/-1/-1->3->2 [16] 4/-1/-1->3->2 [17] 4/-1/-1->3->2 [18] 4/-1/-1->3->2 [19] 4/-1/-1->3->2 [20] 4/-1/-1->3->2 [21] 4/-1/-1->3->2 [22] 4/-1/-1->3->2 [23] 4/-1/-1->3->2 +ywang29-vrdb-test2-worker-0:684538:691610 [0] NCCL INFO Channel 05/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:684540:691612 [2] NCCL INFO Trees [0] 3/-1/-1->2->1 [1] 3/-1/-1->2->1 [2] 3/-1/-1->2->1 [3] 3/-1/-1->2->1 [4] 3/-1/-1->2->1 [5] 3/-1/-1->2->1 [6] 3/-1/-1->2->1 [7] 3/-1/-1->2->1 [8] 3/-1/-1->2->1 [9] 3/-1/-1->2->1 [10] 3/-1/-1->2->1 [11] 3/-1/-1->2->1 [12] 3/-1/-1->2->1 [13] 3/-1/-1->2->1 [14] 3/-1/-1->2->1 [15] 3/-1/-1->2->1 [16] 3/-1/-1->2->1 [17] 3/-1/-1->2->1 [18] 3/-1/-1->2->1 [19] 3/-1/-1->2->1 [20] 3/-1/-1->2->1 [21] 3/-1/-1->2->1 [22] 3/-1/-1->2->1 [23] 3/-1/-1->2->1 +ywang29-vrdb-test2-worker-0:684541:691617 [3] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:684539:691613 [1] NCCL INFO Trees [0] 2/-1/-1->1->0 [1] 2/-1/-1->1->0 [2] 2/-1/-1->1->0 [3] 2/-1/-1->1->0 [4] 2/-1/-1->1->0 [5] 2/-1/-1->1->0 [6] 2/-1/-1->1->0 [7] 2/-1/-1->1->0 [8] 2/-1/-1->1->0 [9] 2/-1/-1->1->0 [10] 2/-1/-1->1->0 [11] 2/-1/-1->1->0 [12] 2/-1/-1->1->0 [13] 2/-1/-1->1->0 [14] 2/-1/-1->1->0 [15] 2/-1/-1->1->0 [16] 2/-1/-1->1->0 [17] 2/-1/-1->1->0 [18] 2/-1/-1->1->0 [19] 2/-1/-1->1->0 [20] 2/-1/-1->1->0 [21] 2/-1/-1->1->0 [22] 2/-1/-1->1->0 [23] 2/-1/-1->1->0 +ywang29-vrdb-test2-worker-0:684538:691610 [0] NCCL INFO Channel 06/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:684540:691612 [2] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:684539:691613 [1] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:684538:691610 [0] NCCL INFO Channel 07/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:684538:691610 [0] NCCL INFO Channel 08/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:684538:691610 [0] NCCL INFO Channel 09/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:684538:691610 [0] NCCL INFO Channel 10/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:684538:691610 [0] NCCL INFO Channel 11/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:684538:691610 [0] NCCL INFO Channel 12/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:684538:691610 [0] NCCL INFO Channel 13/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:684538:691610 [0] NCCL INFO Channel 14/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:684544:691616 [6] NCCL INFO Trees [0] 7/-1/-1->6->5 [1] 7/-1/-1->6->5 [2] 7/-1/-1->6->5 [3] 7/-1/-1->6->5 [4] 7/-1/-1->6->5 [5] 7/-1/-1->6->5 [6] 7/-1/-1->6->5 [7] 7/-1/-1->6->5 [8] 7/-1/-1->6->5 [9] 7/-1/-1->6->5 [10] 7/-1/-1->6->5 [11] 7/-1/-1->6->5 [12] 7/-1/-1->6->5 [13] 7/-1/-1->6->5 [14] 7/-1/-1->6->5 [15] 7/-1/-1->6->5 [16] 7/-1/-1->6->5 [17] 7/-1/-1->6->5 [18] 7/-1/-1->6->5 [19] 7/-1/-1->6->5 [20] 7/-1/-1->6->5 [21] 7/-1/-1->6->5 [22] 7/-1/-1->6->5 [23] 7/-1/-1->6->5 +ywang29-vrdb-test2-worker-0:684538:691610 [0] NCCL INFO Channel 15/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:684544:691616 [6] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:684538:691610 [0] NCCL INFO Channel 16/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:684538:691610 [0] NCCL INFO Channel 17/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:684538:691610 [0] NCCL INFO Channel 18/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:684545:691614 [7] NCCL INFO Trees [0] -1/-1/-1->7->6 [1] -1/-1/-1->7->6 [2] -1/-1/-1->7->6 [3] -1/-1/-1->7->6 [4] -1/-1/-1->7->6 [5] -1/-1/-1->7->6 [6] -1/-1/-1->7->6 [7] -1/-1/-1->7->6 [8] -1/-1/-1->7->6 [9] -1/-1/-1->7->6 [10] -1/-1/-1->7->6 [11] -1/-1/-1->7->6 [12] -1/-1/-1->7->6 [13] -1/-1/-1->7->6 [14] -1/-1/-1->7->6 [15] -1/-1/-1->7->6 [16] -1/-1/-1->7->6 [17] -1/-1/-1->7->6 [18] -1/-1/-1->7->6 [19] -1/-1/-1->7->6 [20] -1/-1/-1->7->6 [21] -1/-1/-1->7->6 [22] -1/-1/-1->7->6 [23] -1/-1/-1->7->6 +ywang29-vrdb-test2-worker-0:684538:691610 [0] NCCL INFO Channel 19/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:684545:691614 [7] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:684538:691610 [0] NCCL INFO Channel 20/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:684538:691610 [0] NCCL INFO Channel 21/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:684538:691610 [0] NCCL INFO Channel 22/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:684538:691610 [0] NCCL INFO Channel 23/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:684538:691610 [0] NCCL INFO Trees [0] 1/-1/-1->0->-1 [1] 1/-1/-1->0->-1 [2] 1/-1/-1->0->-1 [3] 1/-1/-1->0->-1 [4] 1/-1/-1->0->-1 [5] 1/-1/-1->0->-1 [6] 1/-1/-1->0->-1 [7] 1/-1/-1->0->-1 [8] 1/-1/-1->0->-1 [9] 1/-1/-1->0->-1 [10] 1/-1/-1->0->-1 [11] 1/-1/-1->0->-1 [12] 1/-1/-1->0->-1 [13] 1/-1/-1->0->-1 [14] 1/-1/-1->0->-1 [15] 1/-1/-1->0->-1 [16] 1/-1/-1->0->-1 [17] 1/-1/-1->0->-1 [18] 1/-1/-1->0->-1 [19] 1/-1/-1->0->-1 [20] 1/-1/-1->0->-1 [21] 1/-1/-1->0->-1 [22] 1/-1/-1->0->-1 [23] 1/-1/-1->0->-1 +ywang29-vrdb-test2-worker-0:684538:691610 [0] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:684540:691612 [2] NCCL INFO Channel 00/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:691612 [2] NCCL INFO Channel 01/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:691612 [2] NCCL INFO Channel 02/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:691612 [2] NCCL INFO Channel 03/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:691612 [2] NCCL INFO Channel 04/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:691612 [2] NCCL INFO Channel 05/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:691612 [2] NCCL INFO Channel 06/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:691611 [5] NCCL INFO Channel 00/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:691613 [1] NCCL INFO Channel 00/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:691612 [2] NCCL INFO Channel 07/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:691611 [5] NCCL INFO Channel 01/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684538:691610 [0] NCCL INFO Channel 00/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:691613 [1] NCCL INFO Channel 01/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:691612 [2] NCCL INFO Channel 08/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:691611 [5] NCCL INFO Channel 02/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684538:691610 [0] NCCL INFO Channel 01/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:691617 [3] NCCL INFO Channel 00/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:691613 [1] NCCL INFO Channel 02/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:691612 [2] NCCL INFO Channel 09/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:691611 [5] NCCL INFO Channel 03/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684538:691610 [0] NCCL INFO Channel 02/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:691617 [3] NCCL INFO Channel 01/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:691613 [1] NCCL INFO Channel 03/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:691612 [2] NCCL INFO Channel 10/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:691611 [5] NCCL INFO Channel 04/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684538:691610 [0] NCCL INFO Channel 03/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:691617 [3] NCCL INFO Channel 02/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:691613 [1] NCCL INFO Channel 04/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:691612 [2] NCCL INFO Channel 11/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:691616 [6] NCCL INFO Channel 00/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:691611 [5] NCCL INFO Channel 05/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684538:691610 [0] NCCL INFO Channel 04/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:691617 [3] NCCL INFO Channel 03/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:691614 [7] NCCL INFO Channel 00/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:691613 [1] NCCL INFO Channel 05/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:691612 [2] NCCL INFO Channel 12/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:691616 [6] NCCL INFO Channel 01/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:691611 [5] NCCL INFO Channel 06/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684538:691610 [0] NCCL INFO Channel 05/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:691617 [3] NCCL INFO Channel 04/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:691613 [1] NCCL INFO Channel 06/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:691612 [2] NCCL INFO Channel 13/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:691614 [7] NCCL INFO Channel 01/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:691616 [6] NCCL INFO Channel 02/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:691611 [5] NCCL INFO Channel 07/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684538:691610 [0] NCCL INFO Channel 06/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:691617 [3] NCCL INFO Channel 05/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:691613 [1] NCCL INFO Channel 07/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:691612 [2] NCCL INFO Channel 14/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:691614 [7] NCCL INFO Channel 02/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:691616 [6] NCCL INFO Channel 03/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:691611 [5] NCCL INFO Channel 08/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684538:691610 [0] NCCL INFO Channel 07/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:691617 [3] NCCL INFO Channel 06/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:691613 [1] NCCL INFO Channel 08/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:691612 [2] NCCL INFO Channel 15/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:691614 [7] NCCL INFO Channel 03/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:691616 [6] NCCL INFO Channel 04/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:691611 [5] NCCL INFO Channel 09/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684538:691610 [0] NCCL INFO Channel 08/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:691617 [3] NCCL INFO Channel 07/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:691613 [1] NCCL INFO Channel 09/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:691612 [2] NCCL INFO Channel 16/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:691614 [7] NCCL INFO Channel 04/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:691616 [6] NCCL INFO Channel 05/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:691611 [5] NCCL INFO Channel 10/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684538:691610 [0] NCCL INFO Channel 09/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:691617 [3] NCCL INFO Channel 08/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:691613 [1] NCCL INFO Channel 10/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:691612 [2] NCCL INFO Channel 17/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:691614 [7] NCCL INFO Channel 05/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:691611 [5] NCCL INFO Channel 11/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:691616 [6] NCCL INFO Channel 06/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684538:691610 [0] NCCL INFO Channel 10/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:691617 [3] NCCL INFO Channel 09/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:691613 [1] NCCL INFO Channel 11/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:691612 [2] NCCL INFO Channel 18/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:691614 [7] NCCL INFO Channel 06/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:691611 [5] NCCL INFO Channel 12/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:691616 [6] NCCL INFO Channel 07/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684538:691610 [0] NCCL INFO Channel 11/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:691617 [3] NCCL INFO Channel 10/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:691613 [1] NCCL INFO Channel 12/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:691612 [2] NCCL INFO Channel 19/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:691614 [7] NCCL INFO Channel 07/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:691611 [5] NCCL INFO Channel 13/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684538:691610 [0] NCCL INFO Channel 12/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:691616 [6] NCCL INFO Channel 08/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:691617 [3] NCCL INFO Channel 11/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:691613 [1] NCCL INFO Channel 13/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:691612 [2] NCCL INFO Channel 20/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:691614 [7] NCCL INFO Channel 08/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:691611 [5] NCCL INFO Channel 14/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684538:691610 [0] NCCL INFO Channel 13/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:691616 [6] NCCL INFO Channel 09/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:691613 [1] NCCL INFO Channel 14/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:691612 [2] NCCL INFO Channel 21/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:691617 [3] NCCL INFO Channel 12/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:691614 [7] NCCL INFO Channel 09/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:691611 [5] NCCL INFO Channel 15/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684538:691610 [0] NCCL INFO Channel 14/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:691616 [6] NCCL INFO Channel 10/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:691613 [1] NCCL INFO Channel 15/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:691612 [2] NCCL INFO Channel 22/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:691617 [3] NCCL INFO Channel 13/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:691611 [5] NCCL INFO Channel 16/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:691614 [7] NCCL INFO Channel 10/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684538:691610 [0] NCCL INFO Channel 15/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:691616 [6] NCCL INFO Channel 11/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:691613 [1] NCCL INFO Channel 16/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:691612 [2] NCCL INFO Channel 23/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:691617 [3] NCCL INFO Channel 14/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:691611 [5] NCCL INFO Channel 17/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684538:691610 [0] NCCL INFO Channel 16/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:691614 [7] NCCL INFO Channel 11/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:691616 [6] NCCL INFO Channel 12/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:691613 [1] NCCL INFO Channel 17/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:691617 [3] NCCL INFO Channel 15/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:691611 [5] NCCL INFO Channel 18/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684538:691610 [0] NCCL INFO Channel 17/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:691614 [7] NCCL INFO Channel 12/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:691616 [6] NCCL INFO Channel 13/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:691613 [1] NCCL INFO Channel 18/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:691617 [3] NCCL INFO Channel 16/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:691611 [5] NCCL INFO Channel 19/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684538:691610 [0] NCCL INFO Channel 18/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:691614 [7] NCCL INFO Channel 13/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:691616 [6] NCCL INFO Channel 14/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:691613 [1] NCCL INFO Channel 19/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:691617 [3] NCCL INFO Channel 17/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:691611 [5] NCCL INFO Channel 20/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684538:691610 [0] NCCL INFO Channel 19/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:691614 [7] NCCL INFO Channel 14/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:691616 [6] NCCL INFO Channel 15/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:691613 [1] NCCL INFO Channel 20/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:691617 [3] NCCL INFO Channel 18/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:691611 [5] NCCL INFO Channel 21/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684538:691610 [0] NCCL INFO Channel 20/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:691614 [7] NCCL INFO Channel 15/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:691616 [6] NCCL INFO Channel 16/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:691617 [3] NCCL INFO Channel 19/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:691611 [5] NCCL INFO Channel 22/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684538:691610 [0] NCCL INFO Channel 21/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:691613 [1] NCCL INFO Channel 21/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:691614 [7] NCCL INFO Channel 16/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:691616 [6] NCCL INFO Channel 17/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:691617 [3] NCCL INFO Channel 20/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:691611 [5] NCCL INFO Channel 23/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684538:691610 [0] NCCL INFO Channel 22/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:691613 [1] NCCL INFO Channel 22/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:691614 [7] NCCL INFO Channel 17/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:691616 [6] NCCL INFO Channel 18/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:691617 [3] NCCL INFO Channel 21/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684538:691610 [0] NCCL INFO Channel 23/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:691613 [1] NCCL INFO Channel 23/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:691614 [7] NCCL INFO Channel 18/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:691616 [6] NCCL INFO Channel 19/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:691617 [3] NCCL INFO Channel 22/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:691614 [7] NCCL INFO Channel 19/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:691616 [6] NCCL INFO Channel 20/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:691617 [3] NCCL INFO Channel 23/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:691615 [4] NCCL INFO Channel 00/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:691616 [6] NCCL INFO Channel 21/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:691614 [7] NCCL INFO Channel 20/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:691615 [4] NCCL INFO Channel 01/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:691616 [6] NCCL INFO Channel 22/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:691615 [4] NCCL INFO Channel 02/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:691614 [7] NCCL INFO Channel 21/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:691616 [6] NCCL INFO Channel 23/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:691615 [4] NCCL INFO Channel 03/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:691614 [7] NCCL INFO Channel 22/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:691615 [4] NCCL INFO Channel 04/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:691614 [7] NCCL INFO Channel 23/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:691615 [4] NCCL INFO Channel 05/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:691615 [4] NCCL INFO Channel 06/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:691615 [4] NCCL INFO Channel 07/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:691615 [4] NCCL INFO Channel 08/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:691615 [4] NCCL INFO Channel 09/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:691615 [4] NCCL INFO Channel 10/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:691615 [4] NCCL INFO Channel 11/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:691615 [4] NCCL INFO Channel 12/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:691615 [4] NCCL INFO Channel 13/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:691615 [4] NCCL INFO Channel 14/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:691615 [4] NCCL INFO Channel 15/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:691615 [4] NCCL INFO Channel 16/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:691615 [4] NCCL INFO Channel 17/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:691615 [4] NCCL INFO Channel 18/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:691615 [4] NCCL INFO Channel 19/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:691615 [4] NCCL INFO Channel 20/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:691615 [4] NCCL INFO Channel 21/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:691615 [4] NCCL INFO Channel 22/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:691615 [4] NCCL INFO Channel 23/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:691613 [1] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:684538:691610 [0] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:684540:691612 [2] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:684541:691617 [3] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:684545:691614 [7] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:684545:691614 [7] NCCL INFO Channel 00/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:691614 [7] NCCL INFO Channel 01/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:691614 [7] NCCL INFO Channel 02/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:691614 [7] NCCL INFO Channel 03/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:691614 [7] NCCL INFO Channel 04/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:691613 [1] NCCL INFO Channel 00/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:691614 [7] NCCL INFO Channel 05/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:691613 [1] NCCL INFO Channel 01/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:691614 [7] NCCL INFO Channel 06/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:691613 [1] NCCL INFO Channel 02/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:691616 [6] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:684542:691615 [4] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:684543:691611 [5] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:684545:691614 [7] NCCL INFO Channel 07/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:691613 [1] NCCL INFO Channel 03/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:691613 [1] NCCL INFO Channel 04/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:691614 [7] NCCL INFO Channel 08/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:691614 [7] NCCL INFO Channel 09/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:691613 [1] NCCL INFO Channel 05/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:691614 [7] NCCL INFO Channel 10/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:691613 [1] NCCL INFO Channel 06/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:691614 [7] NCCL INFO Channel 11/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:691613 [1] NCCL INFO Channel 07/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:691614 [7] NCCL INFO Channel 12/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:691613 [1] NCCL INFO Channel 08/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:691614 [7] NCCL INFO Channel 13/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:691613 [1] NCCL INFO Channel 09/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:691614 [7] NCCL INFO Channel 14/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:691613 [1] NCCL INFO Channel 10/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:691614 [7] NCCL INFO Channel 15/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:691613 [1] NCCL INFO Channel 11/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:691613 [1] NCCL INFO Channel 12/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:691613 [1] NCCL INFO Channel 13/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:691613 [1] NCCL INFO Channel 14/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:691613 [1] NCCL INFO Channel 15/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:691617 [3] NCCL INFO Channel 00/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:691613 [1] NCCL INFO Channel 16/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:691617 [3] NCCL INFO Channel 01/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:691614 [7] NCCL INFO Channel 16/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:691613 [1] NCCL INFO Channel 17/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:691617 [3] NCCL INFO Channel 02/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:691612 [2] NCCL INFO Channel 00/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:691614 [7] NCCL INFO Channel 17/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:691613 [1] NCCL INFO Channel 18/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:691617 [3] NCCL INFO Channel 03/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:691612 [2] NCCL INFO Channel 01/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:691614 [7] NCCL INFO Channel 18/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:691613 [1] NCCL INFO Channel 19/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:691617 [3] NCCL INFO Channel 04/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:691612 [2] NCCL INFO Channel 02/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:691614 [7] NCCL INFO Channel 19/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:691613 [1] NCCL INFO Channel 20/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:691617 [3] NCCL INFO Channel 05/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:691612 [2] NCCL INFO Channel 03/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:691614 [7] NCCL INFO Channel 20/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:691613 [1] NCCL INFO Channel 21/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:691617 [3] NCCL INFO Channel 06/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:691612 [2] NCCL INFO Channel 04/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:691614 [7] NCCL INFO Channel 21/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:691613 [1] NCCL INFO Channel 22/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:691617 [3] NCCL INFO Channel 07/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:691612 [2] NCCL INFO Channel 05/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:691614 [7] NCCL INFO Channel 22/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684539:691613 [1] NCCL INFO Channel 23/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:691617 [3] NCCL INFO Channel 08/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:691612 [2] NCCL INFO Channel 06/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684545:691614 [7] NCCL INFO Channel 23/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:691617 [3] NCCL INFO Channel 09/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:691612 [2] NCCL INFO Channel 07/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:691617 [3] NCCL INFO Channel 10/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:691612 [2] NCCL INFO Channel 08/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:691617 [3] NCCL INFO Channel 11/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:691615 [4] NCCL INFO Channel 00/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:691616 [6] NCCL INFO Channel 00/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:691615 [4] NCCL INFO Channel 01/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:691612 [2] NCCL INFO Channel 09/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:691611 [5] NCCL INFO Channel 00/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:691617 [3] NCCL INFO Channel 12/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:691616 [6] NCCL INFO Channel 01/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:691615 [4] NCCL INFO Channel 02/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:691612 [2] NCCL INFO Channel 10/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:691611 [5] NCCL INFO Channel 01/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:691617 [3] NCCL INFO Channel 13/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:691616 [6] NCCL INFO Channel 02/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:691612 [2] NCCL INFO Channel 11/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:691611 [5] NCCL INFO Channel 02/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:691617 [3] NCCL INFO Channel 14/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:691612 [2] NCCL INFO Channel 12/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:691617 [3] NCCL INFO Channel 15/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:691611 [5] NCCL INFO Channel 03/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:691612 [2] NCCL INFO Channel 13/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:691616 [6] NCCL INFO Channel 03/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:691617 [3] NCCL INFO Channel 16/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:691615 [4] NCCL INFO Channel 03/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:691611 [5] NCCL INFO Channel 04/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:691612 [2] NCCL INFO Channel 14/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:691616 [6] NCCL INFO Channel 04/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:691617 [3] NCCL INFO Channel 17/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:691611 [5] NCCL INFO Channel 05/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:691612 [2] NCCL INFO Channel 15/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:691615 [4] NCCL INFO Channel 04/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:691616 [6] NCCL INFO Channel 05/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:691611 [5] NCCL INFO Channel 06/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:691612 [2] NCCL INFO Channel 16/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:691615 [4] NCCL INFO Channel 05/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:691616 [6] NCCL INFO Channel 06/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:691611 [5] NCCL INFO Channel 07/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:691612 [2] NCCL INFO Channel 17/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:691615 [4] NCCL INFO Channel 06/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:691616 [6] NCCL INFO Channel 07/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:691611 [5] NCCL INFO Channel 08/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:691612 [2] NCCL INFO Channel 18/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:691615 [4] NCCL INFO Channel 07/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:691616 [6] NCCL INFO Channel 08/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:691611 [5] NCCL INFO Channel 09/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:691612 [2] NCCL INFO Channel 19/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:691617 [3] NCCL INFO Channel 18/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:691611 [5] NCCL INFO Channel 10/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:691612 [2] NCCL INFO Channel 20/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:691617 [3] NCCL INFO Channel 19/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:691611 [5] NCCL INFO Channel 11/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:691612 [2] NCCL INFO Channel 21/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:691617 [3] NCCL INFO Channel 20/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:691611 [5] NCCL INFO Channel 12/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:691616 [6] NCCL INFO Channel 09/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:691612 [2] NCCL INFO Channel 22/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:691617 [3] NCCL INFO Channel 21/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:691611 [5] NCCL INFO Channel 13/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:691616 [6] NCCL INFO Channel 10/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684540:691612 [2] NCCL INFO Channel 23/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:691617 [3] NCCL INFO Channel 22/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:691615 [4] NCCL INFO Channel 08/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:691611 [5] NCCL INFO Channel 14/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:691611 [5] NCCL INFO Channel 15/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:691611 [5] NCCL INFO Channel 16/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:691611 [5] NCCL INFO Channel 17/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:691611 [5] NCCL INFO Channel 18/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:691611 [5] NCCL INFO Channel 19/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:691611 [5] NCCL INFO Channel 20/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:691611 [5] NCCL INFO Channel 21/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:691611 [5] NCCL INFO Channel 22/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684543:691611 [5] NCCL INFO Channel 23/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:691616 [6] NCCL INFO Channel 11/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684541:691617 [3] NCCL INFO Channel 23/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:691615 [4] NCCL INFO Channel 09/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:691616 [6] NCCL INFO Channel 12/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:691615 [4] NCCL INFO Channel 10/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:691616 [6] NCCL INFO Channel 13/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:691615 [4] NCCL INFO Channel 11/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:691616 [6] NCCL INFO Channel 14/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:691615 [4] NCCL INFO Channel 12/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:691616 [6] NCCL INFO Channel 15/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:691615 [4] NCCL INFO Channel 13/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:691616 [6] NCCL INFO Channel 16/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:691615 [4] NCCL INFO Channel 14/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:691616 [6] NCCL INFO Channel 17/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:691615 [4] NCCL INFO Channel 15/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:691616 [6] NCCL INFO Channel 18/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:691615 [4] NCCL INFO Channel 16/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:691616 [6] NCCL INFO Channel 19/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:691615 [4] NCCL INFO Channel 17/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:691616 [6] NCCL INFO Channel 20/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:691615 [4] NCCL INFO Channel 18/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:691616 [6] NCCL INFO Channel 21/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:691615 [4] NCCL INFO Channel 19/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:691615 [4] NCCL INFO Channel 20/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:691615 [4] NCCL INFO Channel 21/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:691615 [4] NCCL INFO Channel 22/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684542:691615 [4] NCCL INFO Channel 23/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:691616 [6] NCCL INFO Channel 22/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684544:691616 [6] NCCL INFO Channel 23/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:684538:691610 [0] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:684538:691610 [0] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:684538:691610 [0] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:684539:691613 [1] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:684539:691613 [1] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:684539:691613 [1] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:684540:691612 [2] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:684540:691612 [2] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:684540:691612 [2] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:684541:691617 [3] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:684541:691617 [3] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:684541:691617 [3] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:684542:691615 [4] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:684542:691615 [4] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:684542:691615 [4] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:684543:691611 [5] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:684543:691611 [5] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:684543:691611 [5] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:684545:691614 [7] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:684545:691614 [7] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:684545:691614 [7] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:684544:691616 [6] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:684544:691616 [6] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:684544:691616 [6] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:684545:691614 [7] NCCL INFO ncclCommInitRank comm 0x7ff4d406a3a0 rank 7 nranks 8 cudaDev 7 nvmlDev 7 busId a01d0 commId 0xac02844318e11ae - Init COMPLETE +ywang29-vrdb-test2-worker-0:684541:691617 [3] NCCL INFO ncclCommInitRank comm 0x7fbd3006a9c0 rank 3 nranks 8 cudaDev 3 nvmlDev 3 busId 201d0 commId 0xac02844318e11ae - Init COMPLETE +ywang29-vrdb-test2-worker-0:684543:691611 [5] NCCL INFO ncclCommInitRank comm 0x7f557806b490 rank 5 nranks 8 cudaDev 5 nvmlDev 5 busId 901d0 commId 0xac02844318e11ae - Init COMPLETE +ywang29-vrdb-test2-worker-0:684539:691613 [1] NCCL INFO ncclCommInitRank comm 0x7f653806b050 rank 1 nranks 8 cudaDev 1 nvmlDev 1 busId 101d0 commId 0xac02844318e11ae - Init COMPLETE +ywang29-vrdb-test2-worker-0:684538:691610 [0] NCCL INFO ncclCommInitRank comm 0x7f98d806b5c0 rank 0 nranks 8 cudaDev 0 nvmlDev 0 busId 101c0 commId 0xac02844318e11ae - Init COMPLETE +ywang29-vrdb-test2-worker-0:684542:691615 [4] NCCL INFO ncclCommInitRank comm 0x7f79c406b1f0 rank 4 nranks 8 cudaDev 4 nvmlDev 4 busId 901c0 commId 0xac02844318e11ae - Init COMPLETE +ywang29-vrdb-test2-worker-0:684540:691612 [2] NCCL INFO ncclCommInitRank comm 0x7f706c06a8a0 rank 2 nranks 8 cudaDev 2 nvmlDev 2 busId 201c0 commId 0xac02844318e11ae - Init COMPLETE +ywang29-vrdb-test2-worker-0:684544:691616 [6] NCCL INFO ncclCommInitRank comm 0x7fea5806b600 rank 6 nranks 8 cudaDev 6 nvmlDev 6 busId a01c0 commId 0xac02844318e11ae - Init COMPLETE + 0%| | 1/520 [00:14<2:04:13, 14.36s/it] {'loss': 2.0453, 'grad_norm': 0.004835594380436832, 'learning_rate': 0.001875, 'epoch': 0.0} + 0%| | 1/520 [00:14<2:04:13, 14.36s/it] 0%| | 2/520 [00:17<1:09:31, 8.05s/it] {'loss': 2.0549, 'grad_norm': 0.0052485741057041, 'learning_rate': 0.00375, 'epoch': 0.0} + 0%| | 2/520 [00:18<1:09:31, 8.05s/it] 1%| | 3/520 [00:21<52:23, 6.08s/it] {'loss': 2.1899, 'grad_norm': 0.006004686724119165, 'learning_rate': 0.005625, 'epoch': 0.01} + 1%| | 3/520 [00:21<52:23, 6.08s/it] 1%| | 4/520 [00:25<44:03, 5.12s/it] {'loss': 2.0656, 'grad_norm': 0.004963434437933776, 'learning_rate': 0.0075, 'epoch': 0.01} + 1%| | 4/520 [00:25<44:03, 5.12s/it] 1%| | 5/520 [00:29<39:22, 4.59s/it] {'loss': 2.2333, 'grad_norm': 0.005481988044809933, 'learning_rate': 0.009375, 'epoch': 0.01} + 1%| | 5/520 [00:29<39:22, 4.59s/it] 1%| | 6/520 [00:32<36:32, 4.26s/it] {'loss': 1.6754, 'grad_norm': 0.0028032746225227427, 'learning_rate': 0.01125, 'epoch': 0.01} + 1%| | 6/520 [00:32<36:32, 4.26s/it] 1%|▏ | 7/520 [00:36<34:40, 4.05s/it] {'loss': 2.0776, 'grad_norm': 0.005415242279065533, 'learning_rate': 0.013125, 'epoch': 0.01} + 1%|▏ | 7/520 [00:36<34:40, 4.05s/it] 2%|▏ | 8/520 [00:40<35:13, 4.13s/it] {'loss': 2.0541, 'grad_norm': 0.004572123228535455, 'learning_rate': 0.015, 'epoch': 0.02} + 2%|▏ | 8/520 [00:40<35:13, 4.13s/it] 2%|▏ | 9/520 [00:44<33:52, 3.98s/it] {'loss': 2.19, 'grad_norm': 0.005017654283622185, 'learning_rate': 0.016875, 'epoch': 0.02} + 2%|▏ | 9/520 [00:44<33:52, 3.98s/it] 2%|▏ | 10/520 [00:47<32:55, 3.87s/it] {'loss': 2.0841, 'grad_norm': 0.005625821069469326, 'learning_rate': 0.01875, 'epoch': 0.02} + 2%|▏ | 10/520 [00:47<32:55, 3.87s/it] 2%|▏ | 11/520 [00:51<32:40, 3.85s/it] {'loss': 2.0582, 'grad_norm': 0.00495668376479793, 'learning_rate': 0.020624999999999998, 'epoch': 0.02} + 2%|▏ | 11/520 [00:51<32:40, 3.85s/it] 2%|▏ | 12/520 [00:55<32:12, 3.80s/it] {'loss': 1.5579, 'grad_norm': 0.0019327303519781425, 'learning_rate': 0.0225, 'epoch': 0.02} + 2%|▏ | 12/520 [00:55<32:12, 3.80s/it][2025-10-12 10:34:52,864] [WARNING] [stage3.py:2069:step] 1 pytorch allocator cache flushes since last step. this happens when there is high memory pressure and is detrimental to performance. if this is happening frequently consider adjusting settings to reduce memory consumption. If you are unable to make the cache flushes go away consider adding get_accelerator().empty_cache() calls in your training loop to ensure that all ranks flush their caches at the same time + 2%|▎ | 13/520 [00:59<33:31, 3.97s/it] {'loss': 1.6319, 'grad_norm': 0.0017294238090760235, 'learning_rate': 0.024375, 'epoch': 0.03} + 2%|▎ | 13/520 [00:59<33:31, 3.97s/it] 3%|▎ | 14/520 [01:03<32:51, 3.90s/it] {'loss': 1.598, 'grad_norm': 0.0009241708973226582, 'learning_rate': 0.02625, 'epoch': 0.03} + 3%|▎ | 14/520 [01:03<32:51, 3.90s/it] 3%|▎ | 15/520 [01:07<32:21, 3.84s/it] {'loss': 1.484, 'grad_norm': 0.0005662976315291759, 'learning_rate': 0.028124999999999997, 'epoch': 0.03} + 3%|▎ | 15/520 [01:07<32:21, 3.84s/it] 3%|▎ | 16/520 [01:10<31:49, 3.79s/it] {'loss': 1.4514, 'grad_norm': 0.000554599920918518, 'learning_rate': 0.03, 'epoch': 0.03} + 3%|▎ | 16/520 [01:10<31:49, 3.79s/it] 3%|▎ | 17/520 [01:14<31:32, 3.76s/it] {'loss': 1.6173, 'grad_norm': 0.0006302062772187279, 'learning_rate': 0.02999970859397307, 'epoch': 0.03} + 3%|▎ | 17/520 [01:14<31:32, 3.76s/it] 3%|▎ | 18/520 [01:18<31:14, 3.73s/it] {'loss': 1.4772, 'grad_norm': 0.000551258971477743, 'learning_rate': 0.02999883438721462, 'epoch': 0.03} + 3%|▎ | 18/520 [01:18<31:14, 3.73s/it] 4%|▎ | 19/520 [01:21<30:59, 3.71s/it] {'loss': 1.4469, 'grad_norm': 0.000481835682509408, 'learning_rate': 0.029997377413691195, 'epoch': 0.04} + 4%|▎ | 19/520 [01:21<30:59, 3.71s/it] 4%|▍ | 20/520 [01:25<30:47, 3.69s/it] {'loss': 1.4394, 'grad_norm': 0.0004965409518334734, 'learning_rate': 0.029995337730012242, 'epoch': 0.04} + 4%|▍ | 20/520 [01:25<30:47, 3.69s/it] 4%|▍ | 21/520 [01:29<30:36, 3.68s/it] {'loss': 1.438, 'grad_norm': 0.0004579723872077416, 'learning_rate': 0.02999271541542791, 'epoch': 0.04} + 4%|▍ | 21/520 [01:29<30:36, 3.68s/it] 4%|▍ | 22/520 [01:32<30:25, 3.66s/it] {'loss': 1.5614, 'grad_norm': 0.0004278067213599666, 'learning_rate': 0.02998951057182598, 'epoch': 0.04} + 4%|▍ | 22/520 [01:32<30:25, 3.66s/it] 4%|▍ | 23/520 [01:36<30:18, 3.66s/it] {'loss': 1.497, 'grad_norm': 0.00038908704171535603, 'learning_rate': 0.029985723323727866, 'epoch': 0.04} + 4%|▍ | 23/520 [01:36<30:18, 3.66s/it] 5%|▍ | 24/520 [01:40<30:09, 3.65s/it] {'loss': 1.4125, 'grad_norm': 0.00039297356391049606, 'learning_rate': 0.029981353818283832, 'epoch': 0.05} + 5%|▍ | 24/520 [01:40<30:09, 3.65s/it] 5%|▍ | 25/520 [01:43<30:07, 3.65s/it] {'loss': 1.5433, 'grad_norm': 0.0004936519216780346, 'learning_rate': 0.02997640222526725, 'epoch': 0.05} + 5%|▍ | 25/520 [01:43<30:07, 3.65s/it] 5%|▌ | 26/520 [01:47<30:03, 3.65s/it] {'loss': 1.4415, 'grad_norm': 0.00034871121672394107, 'learning_rate': 0.029970868737067977, 'epoch': 0.05} + 5%|▌ | 26/520 [01:47<30:03, 3.65s/it] 5%|▌ | 27/520 [01:50<29:55, 3.64s/it] {'loss': 1.38, 'grad_norm': 0.0003647226840342541, 'learning_rate': 0.029964753568684933, 'epoch': 0.05} + 5%|▌ | 27/520 [01:50<29:55, 3.64s/it] 5%|▌ | 28/520 [01:54<29:52, 3.64s/it] {'loss': 1.4192, 'grad_norm': 0.00039637869643329343, 'learning_rate': 0.0299580569577177, 'epoch': 0.05} + 5%|▌ | 28/520 [01:54<29:52, 3.64s/it] 6%|▌ | 29/520 [01:58<29:56, 3.66s/it] {'loss': 1.4192, 'grad_norm': 0.0003694360893471985, 'learning_rate': 0.029950779164357327, 'epoch': 0.06} + 6%|▌ | 29/520 [01:58<29:56, 3.66s/it] 6%|▌ | 30/520 [02:01<29:52, 3.66s/it] {'loss': 1.475, 'grad_norm': 0.00032249200651695225, 'learning_rate': 0.029942920471376182, 'epoch': 0.06} + 6%|▌ | 30/520 [02:01<29:52, 3.66s/it] 6%|▌ | 31/520 [02:05<29:50, 3.66s/it] {'loss': 1.3852, 'grad_norm': 0.00033919634997431976, 'learning_rate': 0.029934481184117005, 'epoch': 0.06} + 6%|▌ | 31/520 [02:05<29:50, 3.66s/it] 6%|▌ | 32/520 [02:09<29:48, 3.66s/it] {'loss': 1.2639, 'grad_norm': 0.0002867671988599108, 'learning_rate': 0.029925461630481022, 'epoch': 0.06} + 6%|▌ | 32/520 [02:09<29:48, 3.66s/it] 6%|▋ | 33/520 [02:12<29:39, 3.65s/it] {'loss': 1.404, 'grad_norm': 0.000404423319503682, 'learning_rate': 0.029915862160915196, 'epoch': 0.06} + 6%|▋ | 33/520 [02:12<29:39, 3.65s/it] 7%|▋ | 34/520 [02:16<29:39, 3.66s/it] {'loss': 1.3893, 'grad_norm': 0.0003835931219638484, 'learning_rate': 0.02990568314839864, 'epoch': 0.07} + 7%|▋ | 34/520 [02:16<29:39, 3.66s/it] 7%|▋ | 35/520 [02:20<29:28, 3.65s/it] {'loss': 1.3922, 'grad_norm': 0.000399838959202088, 'learning_rate': 0.02989492498842809, 'epoch': 0.07} + 7%|▋ | 35/520 [02:20<29:28, 3.65s/it] 7%|▋ | 36/520 [02:23<29:21, 3.64s/it] {'loss': 1.4802, 'grad_norm': 0.00031430042258179215, 'learning_rate': 0.029883588099002583, 'epoch': 0.07} + 7%|▋ | 36/520 [02:23<29:21, 3.64s/it] 7%|▋ | 37/520 [02:27<29:21, 3.65s/it] {'loss': 1.4293, 'grad_norm': 0.0002830065557818753, 'learning_rate': 0.029871672920607155, 'epoch': 0.07} + 7%|▋ | 37/520 [02:27<29:21, 3.65s/it] 7%|▋ | 38/520 [02:31<29:17, 3.65s/it] {'loss': 1.54, 'grad_norm': 0.00031552975897229516, 'learning_rate': 0.02985917991619579, 'epoch': 0.07} + 7%|▋ | 38/520 [02:31<29:17, 3.65s/it] 8%|▊ | 39/520 [02:34<29:16, 3.65s/it] {'loss': 1.4361, 'grad_norm': 0.0003796008335081465, 'learning_rate': 0.02984610957117339, 'epoch': 0.07} + 8%|▊ | 39/520 [02:34<29:16, 3.65s/it] 8%|▊ | 40/520 [02:38<29:23, 3.67s/it] {'loss': 1.4404, 'grad_norm': 0.0003154352238096016, 'learning_rate': 0.029832462393376925, 'epoch': 0.08} + 8%|▊ | 40/520 [02:38<29:23, 3.67s/it] 8%|▊ | 41/520 [02:42<29:26, 3.69s/it] {'loss': 1.4257, 'grad_norm': 0.0003549376283614547, 'learning_rate': 0.029818238913055718, 'epoch': 0.08} + 8%|▊ | 41/520 [02:42<29:26, 3.69s/it] 8%|▊ | 42/520 [02:45<29:18, 3.68s/it] {'loss': 1.4198, 'grad_norm': 0.00044877656572188195, 'learning_rate': 0.029803439682850815, 'epoch': 0.08} + 8%|▊ | 42/520 [02:45<29:18, 3.68s/it] 8%|▊ | 43/520 [02:49<29:13, 3.68s/it] {'loss': 1.3271, 'grad_norm': 0.00030504492743563017, 'learning_rate': 0.029788065277773536, 'epoch': 0.08} + 8%|▊ | 43/520 [02:49<29:13, 3.68s/it] 8%|▊ | 44/520 [02:53<29:09, 3.67s/it] {'loss': 1.4237, 'grad_norm': 0.00032128880279373386, 'learning_rate': 0.029772116295183122, 'epoch': 0.08} + 8%|▊ | 44/520 [02:53<29:09, 3.67s/it] 9%|▊ | 45/520 [02:56<29:06, 3.68s/it] {'loss': 1.4392, 'grad_norm': 0.0003550985722736327, 'learning_rate': 0.029755593354763517, 'epoch': 0.09} + 9%|▊ | 45/520 [02:56<29:06, 3.68s/it] 9%|▉ | 46/520 [03:00<29:02, 3.68s/it] {'loss': 1.4506, 'grad_norm': 0.000302421933791612, 'learning_rate': 0.029738497098499326, 'epoch': 0.09} + 9%|▉ | 46/520 [03:00<29:02, 3.68s/it] 9%|▉ | 47/520 [03:04<28:59, 3.68s/it] {'loss': 1.3994, 'grad_norm': 0.0003874062961979314, 'learning_rate': 0.029720828190650816, 'epoch': 0.09} + 9%|▉ | 47/520 [03:04<28:59, 3.68s/it] 9%|▉ | 48/520 [03:07<28:51, 3.67s/it] {'loss': 1.4327, 'grad_norm': 0.0004699034488223386, 'learning_rate': 0.029702587317728155, 'epoch': 0.09} + 9%|▉ | 48/520 [03:07<28:51, 3.67s/it] 9%|▉ | 49/520 [03:11<28:44, 3.66s/it] {'loss': 1.4404, 'grad_norm': 0.000353953551602808, 'learning_rate': 0.029683775188464726, 'epoch': 0.09} + 9%|▉ | 49/520 [03:11<28:44, 3.66s/it] 10%|▉ | 50/520 [03:15<28:38, 3.66s/it] {'loss': 1.4408, 'grad_norm': 0.0003718196051818948, 'learning_rate': 0.02966439253378957, 'epoch': 0.1} + 10%|▉ | 50/520 [03:15<28:38, 3.66s/it] 10%|▉ | 51/520 [03:18<28:38, 3.66s/it] {'loss': 1.3699, 'grad_norm': 0.00043129414758246424, 'learning_rate': 0.029644440106798998, 'epoch': 0.1} + 10%|▉ | 51/520 [03:18<28:38, 3.66s/it] 10%|█ | 52/520 [03:22<28:34, 3.66s/it] {'loss': 1.4933, 'grad_norm': 0.00040005464385684024, 'learning_rate': 0.029623918682727353, 'epoch': 0.1} + 10%|█ | 52/520 [03:22<28:34, 3.66s/it] 10%|█ | 53/520 [03:26<28:27, 3.66s/it] {'loss': 1.4636, 'grad_norm': 0.00036149514973362153, 'learning_rate': 0.029602829058916848, 'epoch': 0.1} + 10%|█ | 53/520 [03:26<28:27, 3.66s/it] 10%|█ | 54/520 [03:29<28:20, 3.65s/it] {'loss': 1.4046, 'grad_norm': 0.0003700474590101797, 'learning_rate': 0.029581172054786616, 'epoch': 0.1} + 10%|█ | 54/520 [03:29<28:20, 3.65s/it] 11%|█ | 55/520 [03:33<28:22, 3.66s/it] {'loss': 1.3552, 'grad_norm': 0.00042232326103588356, 'learning_rate': 0.029558948511800864, 'epoch': 0.11} + 11%|█ | 55/520 [03:33<28:22, 3.66s/it] 11%|█ | 56/520 [03:37<28:15, 3.65s/it] {'loss': 1.4796, 'grad_norm': 0.0003963055846747318, 'learning_rate': 0.029536159293436167, 'epoch': 0.11} + 11%|█ | 56/520 [03:37<28:15, 3.65s/it] 11%|█ | 57/520 [03:41<28:44, 3.72s/it] {'loss': 1.3718, 'grad_norm': 0.00045797006353181445, 'learning_rate': 0.029512805285147943, 'epoch': 0.11} + 11%|█ | 57/520 [03:41<28:44, 3.72s/it] 11%|█ | 58/520 [03:44<29:07, 3.78s/it] {'loss': 1.4799, 'grad_norm': 0.0003412595822135265, 'learning_rate': 0.02948888739433602, 'epoch': 0.11} + 11%|█ | 58/520 [03:44<29:07, 3.78s/it] 11%|█▏ | 59/520 [03:48<29:18, 3.82s/it] {'loss': 1.2956, 'grad_norm': 0.00034995612026358615, 'learning_rate': 0.029464406550309415, 'epoch': 0.11} + 11%|█▏ | 59/520 [03:48<29:18, 3.82s/it] 12%|█▏ | 60/520 [03:52<29:09, 3.80s/it] {'loss': 1.4011, 'grad_norm': 0.0003796939964566601, 'learning_rate': 0.02943936370425018, 'epoch': 0.12} + 12%|█▏ | 60/520 [03:52<29:09, 3.80s/it] 12%|█▏ | 61/520 [03:56<29:06, 3.81s/it] {'loss': 1.3609, 'grad_norm': 0.0003981686904519445, 'learning_rate': 0.029413759829176496, 'epoch': 0.12} + 12%|█▏ | 61/520 [03:56<29:06, 3.81s/it] 12%|█▏ | 62/520 [04:00<29:10, 3.82s/it] {'loss': 1.3772, 'grad_norm': 0.0003934051943784505, 'learning_rate': 0.029387595919904815, 'epoch': 0.12} + 12%|█▏ | 62/520 [04:00<29:10, 3.82s/it] 12%|█▏ | 63/520 [04:04<29:09, 3.83s/it] {'loss': 1.3883, 'grad_norm': 0.00039257072842810274, 'learning_rate': 0.02936087299301127, 'epoch': 0.12} + 12%|█▏ | 63/520 [04:04<29:09, 3.83s/it] 12%|█▏ | 64/520 [04:08<29:10, 3.84s/it] {'loss': 1.4074, 'grad_norm': 0.0003956278297975135, 'learning_rate': 0.02933359208679211, 'epoch': 0.12} + 12%|█▏ | 64/520 [04:08<29:10, 3.84s/it] 12%|█▎ | 65/520 [04:11<29:11, 3.85s/it] {'loss': 1.4154, 'grad_norm': 0.0004793450622583016, 'learning_rate': 0.0293057542612234, 'epoch': 0.12} + 12%|█▎ | 65/520 [04:11<29:11, 3.85s/it] 13%|█▎ | 66/520 [04:15<29:10, 3.85s/it] {'loss': 1.348, 'grad_norm': 0.00039842346245967897, 'learning_rate': 0.029277360597919837, 'epoch': 0.13} + 13%|█▎ | 66/520 [04:15<29:10, 3.85s/it] 13%|█▎ | 67/520 [04:19<29:03, 3.85s/it] {'loss': 1.2772, 'grad_norm': 0.0004264007654619666, 'learning_rate': 0.029248412200092687, 'epoch': 0.13} + 13%|█▎ | 67/520 [04:19<29:03, 3.85s/it] 13%|█▎ | 68/520 [04:23<29:02, 3.86s/it] {'loss': 1.3433, 'grad_norm': 0.00042616199302678606, 'learning_rate': 0.029218910192506976, 'epoch': 0.13} + 13%|█▎ | 68/520 [04:23<29:02, 3.86s/it] 13%|█▎ | 69/520 [04:27<28:56, 3.85s/it] {'loss': 1.3274, 'grad_norm': 0.0005140154012321072, 'learning_rate': 0.02918885572143774, 'epoch': 0.13} + 13%|█▎ | 69/520 [04:27<28:56, 3.85s/it] 13%|█▎ | 70/520 [04:31<28:30, 3.80s/it] {'loss': 1.3393, 'grad_norm': 0.00042946307621232995, 'learning_rate': 0.02915824995462551, 'epoch': 0.13} + 13%|█▎ | 70/520 [04:31<28:30, 3.80s/it] 14%|█▎ | 71/520 [04:34<28:08, 3.76s/it] {'loss': 1.2909, 'grad_norm': 0.00041021789457535763, 'learning_rate': 0.02912709408123095, 'epoch': 0.14} + 14%|█▎ | 71/520 [04:34<28:08, 3.76s/it] 14%|█▍ | 72/520 [04:38<27:48, 3.72s/it] {'loss': 1.4416, 'grad_norm': 0.00043509469304480223, 'learning_rate': 0.029095389311788625, 'epoch': 0.14} + 14%|█▍ | 72/520 [04:38<27:48, 3.72s/it] 14%|█▍ | 73/520 [04:41<27:35, 3.70s/it] {'loss': 1.2546, 'grad_norm': 0.0004919769859420356, 'learning_rate': 0.02906313687815999, 'epoch': 0.14} + 14%|█▍ | 73/520 [04:41<27:35, 3.70s/it] 14%|█▍ | 74/520 [04:45<27:26, 3.69s/it] {'loss': 1.3585, 'grad_norm': 0.0004890207343314258, 'learning_rate': 0.02903033803348551, 'epoch': 0.14} + 14%|█▍ | 74/520 [04:45<27:26, 3.69s/it] 14%|█▍ | 75/520 [04:49<27:16, 3.68s/it] {'loss': 1.2856, 'grad_norm': 0.00044465278677225864, 'learning_rate': 0.028996994052135995, 'epoch': 0.14} + 14%|█▍ | 75/520 [04:49<27:16, 3.68s/it] 15%|█▍ | 76/520 [04:52<27:09, 3.67s/it] {'loss': 1.3817, 'grad_norm': 0.0004221120307845972, 'learning_rate': 0.028963106229663062, 'epoch': 0.15} + 15%|█▍ | 76/520 [04:52<27:09, 3.67s/it] 15%|█▍ | 77/520 [04:56<27:03, 3.67s/it] {'loss': 1.2247, 'grad_norm': 0.0005325491764362268, 'learning_rate': 0.028928675882748802, 'epoch': 0.15} + 15%|█▍ | 77/520 [04:56<27:03, 3.67s/it] 15%|█▌ | 78/520 [05:00<26:55, 3.66s/it] {'loss': 1.3125, 'grad_norm': 0.0005325798487854726, 'learning_rate': 0.028893704349154628, 'epoch': 0.15} + 15%|█▌ | 78/520 [05:00<26:55, 3.66s/it] 15%|█▌ | 79/520 [05:03<26:54, 3.66s/it] {'loss': 1.2973, 'grad_norm': 0.0005428254076706863, 'learning_rate': 0.0288581929876693, 'epoch': 0.15} + 15%|█▌ | 79/520 [05:03<26:54, 3.66s/it] 15%|█▌ | 80/520 [05:07<26:48, 3.66s/it] {'loss': 1.3603, 'grad_norm': 0.0004872295806094779, 'learning_rate': 0.028822143178056113, 'epoch': 0.15} + 15%|█▌ | 80/520 [05:07<26:48, 3.66s/it] 16%|█▌ | 81/520 [05:11<26:46, 3.66s/it] {'loss': 1.45, 'grad_norm': 0.0006749968517338096, 'learning_rate': 0.028785556320999307, 'epoch': 0.16} + 16%|█▌ | 81/520 [05:11<26:46, 3.66s/it] 16%|█▌ | 82/520 [05:14<26:46, 3.67s/it] {'loss': 1.3711, 'grad_norm': 0.0005080441431273979, 'learning_rate': 0.02874843383804964, 'epoch': 0.16} + 16%|█▌ | 82/520 [05:14<26:46, 3.67s/it] 16%|█▌ | 83/520 [05:18<26:39, 3.66s/it] {'loss': 1.3656, 'grad_norm': 0.0005281294450091858, 'learning_rate': 0.028710777171569148, 'epoch': 0.16} + 16%|█▌ | 83/520 [05:18<26:39, 3.66s/it] 16%|█▌ | 84/520 [05:22<26:39, 3.67s/it] {'loss': 1.3952, 'grad_norm': 0.0005942562259370714, 'learning_rate': 0.0286725877846751, 'epoch': 0.16} + 16%|█▌ | 84/520 [05:22<26:39, 3.67s/it] 16%|█▋ | 85/520 [05:25<26:38, 3.68s/it] {'loss': 1.4433, 'grad_norm': 0.0005009813411895455, 'learning_rate': 0.028633867161183164, 'epoch': 0.16} + 16%|█▋ | 85/520 [05:25<26:38, 3.68s/it] 17%|█▋ | 86/520 [05:29<26:35, 3.68s/it] {'loss': 1.4142, 'grad_norm': 0.000501438509980385, 'learning_rate': 0.028594616805549748, 'epoch': 0.17} + 17%|█▋ | 86/520 [05:29<26:35, 3.68s/it] 17%|█▋ | 87/520 [05:33<26:32, 3.68s/it] {'loss': 1.3222, 'grad_norm': 0.000544057612197893, 'learning_rate': 0.02855483824281355, 'epoch': 0.17} + 17%|█▋ | 87/520 [05:33<26:32, 3.68s/it] 17%|█▋ | 88/520 [05:36<26:35, 3.69s/it] {'loss': 1.2279, 'grad_norm': 0.00045857561946968336, 'learning_rate': 0.028514533018536285, 'epoch': 0.17} + 17%|█▋ | 88/520 [05:37<26:35, 3.69s/it] 17%|█▋ | 89/520 [05:40<26:27, 3.68s/it] {'loss': 1.378, 'grad_norm': 0.0006604545244930115, 'learning_rate': 0.02847370269874266, 'epoch': 0.17} + 17%|█▋ | 89/520 [05:40<26:27, 3.68s/it] 17%|█▋ | 90/520 [05:44<26:20, 3.67s/it] {'loss': 1.3138, 'grad_norm': 0.000546561476390485, 'learning_rate': 0.028432348869859514, 'epoch': 0.17} + 17%|█▋ | 90/520 [05:44<26:20, 3.67s/it] 18%|█▊ | 91/520 [05:47<26:12, 3.67s/it] {'loss': 1.3665, 'grad_norm': 0.0005323360366279856, 'learning_rate': 0.02839047313865417, 'epoch': 0.17} + 18%|█▊ | 91/520 [05:47<26:12, 3.67s/it] 18%|█▊ | 92/520 [05:51<26:08, 3.66s/it] {'loss': 1.304, 'grad_norm': 0.0005566549748558746, 'learning_rate': 0.02834807713217203, 'epoch': 0.18} + 18%|█▊ | 92/520 [05:51<26:08, 3.66s/it] 18%|█▊ | 93/520 [05:55<26:00, 3.65s/it] {'loss': 1.3303, 'grad_norm': 0.000653768373236387, 'learning_rate': 0.028305162497673324, 'epoch': 0.18} + 18%|█▊ | 93/520 [05:55<26:00, 3.65s/it] 18%|█▊ | 94/520 [05:58<25:53, 3.65s/it] {'loss': 1.404, 'grad_norm': 0.0006769354327703279, 'learning_rate': 0.028261730902569147, 'epoch': 0.18} + 18%|█▊ | 94/520 [05:58<25:53, 3.65s/it] 18%|█▊ | 95/520 [06:02<25:49, 3.65s/it] {'loss': 1.3242, 'grad_norm': 0.0006897663877048986, 'learning_rate': 0.028217784034356635, 'epoch': 0.18} + 18%|█▊ | 95/520 [06:02<25:49, 3.65s/it] 18%|█▊ | 96/520 [06:06<25:46, 3.65s/it] {'loss': 1.3258, 'grad_norm': 0.0005850255783332522, 'learning_rate': 0.028173323600553427, 'epoch': 0.18} + 18%|█▊ | 96/520 [06:06<25:46, 3.65s/it] 19%|█▊ | 97/520 [06:09<25:43, 3.65s/it] {'loss': 1.3068, 'grad_norm': 0.0007037481632079662, 'learning_rate': 0.028128351328631306, 'epoch': 0.19} + 19%|█▊ | 97/520 [06:09<25:43, 3.65s/it] 19%|█▉ | 98/520 [06:13<25:42, 3.66s/it] {'loss': 1.2959, 'grad_norm': 0.0005157541493376634, 'learning_rate': 0.02808286896594908, 'epoch': 0.19} + 19%|█▉ | 98/520 [06:13<25:42, 3.66s/it] 19%|█▉ | 99/520 [06:17<25:37, 3.65s/it] {'loss': 1.291, 'grad_norm': 0.0006827166617448436, 'learning_rate': 0.028036878279684697, 'epoch': 0.19} + 19%|█▉ | 99/520 [06:17<25:37, 3.65s/it] 19%|█▉ | 100/520 [06:20<25:30, 3.64s/it] {'loss': 1.2719, 'grad_norm': 0.0005587536348274973, 'learning_rate': 0.027990381056766583, 'epoch': 0.19} + 19%|█▉ | 100/520 [06:20<25:30, 3.64s/it] 19%|█▉ | 101/520 [06:24<25:30, 3.65s/it] {'loss': 1.3216, 'grad_norm': 0.0006629347887962675, 'learning_rate': 0.027943379103804195, 'epoch': 0.19} + 19%|█▉ | 101/520 [06:24<25:30, 3.65s/it] 20%|█▉ | 102/520 [06:28<25:24, 3.65s/it] {'loss': 1.3282, 'grad_norm': 0.0006528670745697654, 'learning_rate': 0.027895874247017855, 'epoch': 0.2} + 20%|█▉ | 102/520 [06:28<25:24, 3.65s/it] 20%|█▉ | 103/520 [06:31<25:23, 3.65s/it] {'loss': 1.2628, 'grad_norm': 0.0006140983464057081, 'learning_rate': 0.027847868332167774, 'epoch': 0.2} + 20%|█▉ | 103/520 [06:31<25:23, 3.65s/it] 20%|██ | 104/520 [06:35<25:23, 3.66s/it] {'loss': 1.3402, 'grad_norm': 0.0007079043668166379, 'learning_rate': 0.027799363224482333, 'epoch': 0.2} + 20%|██ | 104/520 [06:35<25:23, 3.66s/it] 20%|██ | 105/520 [06:39<25:18, 3.66s/it] {'loss': 1.3218, 'grad_norm': 0.0006243768976182147, 'learning_rate': 0.027750360808585634, 'epoch': 0.2} + 20%|██ | 105/520 [06:39<25:18, 3.66s/it] 20%|██ | 106/520 [06:42<25:17, 3.66s/it] {'loss': 1.2857, 'grad_norm': 0.000648565721423432, 'learning_rate': 0.027700862988424262, 'epoch': 0.2} + 20%|██ | 106/520 [06:42<25:17, 3.66s/it] 21%|██ | 107/520 [06:46<25:13, 3.67s/it] {'loss': 1.271, 'grad_norm': 0.0007202879371890406, 'learning_rate': 0.027650871687193285, 'epoch': 0.21} + 21%|██ | 107/520 [06:46<25:13, 3.67s/it] 21%|██ | 108/520 [06:50<25:08, 3.66s/it] {'loss': 1.2831, 'grad_norm': 0.000766395640514277, 'learning_rate': 0.027600388847261573, 'epoch': 0.21} + 21%|██ | 108/520 [06:50<25:08, 3.66s/it] 21%|██ | 109/520 [06:53<25:09, 3.67s/it] {'loss': 1.2369, 'grad_norm': 0.0006777536749361198, 'learning_rate': 0.027549416430096296, 'epoch': 0.21} + 21%|██ | 109/520 [06:53<25:09, 3.67s/it] 21%|██ | 110/520 [06:57<25:05, 3.67s/it] {'loss': 1.4437, 'grad_norm': 0.0006841481761176572, 'learning_rate': 0.02749795641618673, 'epoch': 0.21} + 21%|██ | 110/520 [06:57<25:05, 3.67s/it] 21%|██▏ | 111/520 [07:01<24:59, 3.67s/it] {'loss': 1.4486, 'grad_norm': 0.0006683235485019258, 'learning_rate': 0.02744601080496731, 'epoch': 0.21} + 21%|██▏ | 111/520 [07:01<24:59, 3.67s/it] 22%|██▏ | 112/520 [07:04<24:56, 3.67s/it] {'loss': 1.3248, 'grad_norm': 0.0006492036770391092, 'learning_rate': 0.027393581614739922, 'epoch': 0.22} + 22%|██▏ | 112/520 [07:04<24:56, 3.67s/it] 22%|██▏ | 113/520 [07:08<24:53, 3.67s/it] {'loss': 1.2392, 'grad_norm': 0.0006602486755048801, 'learning_rate': 0.027340670882595498, 'epoch': 0.22} + 22%|██▏ | 113/520 [07:08<24:53, 3.67s/it] 22%|██▏ | 114/520 [07:12<24:52, 3.68s/it] {'loss': 1.3373, 'grad_norm': 0.0006311746108924108, 'learning_rate': 0.02728728066433488, 'epoch': 0.22} + 22%|██▏ | 114/520 [07:12<24:52, 3.68s/it] 22%|██▏ | 115/520 [07:15<24:51, 3.68s/it] {'loss': 1.4047, 'grad_norm': 0.0007021620052228481, 'learning_rate': 0.027233413034388903, 'epoch': 0.22} + 22%|██▏ | 115/520 [07:15<24:51, 3.68s/it] 22%|██▏ | 116/520 [07:19<24:50, 3.69s/it] {'loss': 1.4249, 'grad_norm': 0.000625885660106651, 'learning_rate': 0.02717907008573785, 'epoch': 0.22} + 22%|██▏ | 116/520 [07:19<24:50, 3.69s/it] 22%|██▎ | 117/520 [07:23<24:46, 3.69s/it] {'loss': 1.3882, 'grad_norm': 0.0007747109503155006, 'learning_rate': 0.02712425392983008, 'epoch': 0.23} + 22%|██▎ | 117/520 [07:23<24:46, 3.69s/it] 23%|██▎ | 118/520 [07:26<24:38, 3.68s/it] {'loss': 1.328, 'grad_norm': 0.0006778689234536668, 'learning_rate': 0.027068966696500024, 'epoch': 0.23} + 23%|██▎ | 118/520 [07:26<24:38, 3.68s/it] 23%|██▎ | 119/520 [07:30<24:35, 3.68s/it] {'loss': 1.2751, 'grad_norm': 0.000780429026507387, 'learning_rate': 0.02701321053388542, 'epoch': 0.23} + 23%|██▎ | 119/520 [07:30<24:35, 3.68s/it] 23%|██▎ | 120/520 [07:34<24:44, 3.71s/it] {'loss': 1.2933, 'grad_norm': 0.0008654211510387514, 'learning_rate': 0.026956987608343835, 'epoch': 0.23} + 23%|██▎ | 120/520 [07:34<24:44, 3.71s/it] 23%|██▎ | 121/520 [07:38<24:52, 3.74s/it] {'loss': 1.3287, 'grad_norm': 0.0007196070417299428, 'learning_rate': 0.026900300104368526, 'epoch': 0.23} + 23%|██▎ | 121/520 [07:38<24:52, 3.74s/it] 23%|██▎ | 122/520 [07:42<25:03, 3.78s/it] {'loss': 1.2584, 'grad_norm': 0.0006708638900504419, 'learning_rate': 0.02684315022450353, 'epoch': 0.23} + 23%|██▎ | 122/520 [07:42<25:03, 3.78s/it] 24%|██▎ | 123/520 [07:45<25:05, 3.79s/it] {'loss': 1.3106, 'grad_norm': 0.0006560170946630828, 'learning_rate': 0.026785540189258106, 'epoch': 0.24} + 24%|██▎ | 123/520 [07:45<25:05, 3.79s/it] 24%|██▍ | 124/520 [07:49<25:04, 3.80s/it] {'loss': 1.2975, 'grad_norm': 0.0007779716467823844, 'learning_rate': 0.026727472237020447, 'epoch': 0.24} + 24%|██▍ | 124/520 [07:49<25:04, 3.80s/it] 24%|██▍ | 125/520 [07:53<25:09, 3.82s/it] {'loss': 1.2917, 'grad_norm': 0.00071955098214077, 'learning_rate': 0.026668948623970724, 'epoch': 0.24} + 24%|██▍ | 125/520 [07:53<25:09, 3.82s/it] 24%|██▍ | 126/520 [07:58<26:25, 4.02s/it] {'loss': 1.242, 'grad_norm': 0.0005929945852311654, 'learning_rate': 0.026609971623993412, 'epoch': 0.24} + 24%|██▍ | 126/520 [07:58<26:25, 4.02s/it] 24%|██▍ | 127/520 [08:01<25:46, 3.94s/it] {'loss': 1.2983, 'grad_norm': 0.0008430613584785757, 'learning_rate': 0.026550543528588943, 'epoch': 0.24} + 24%|██▍ | 127/520 [08:01<25:46, 3.94s/it] 25%|██▍ | 128/520 [08:05<25:07, 3.84s/it] {'loss': 1.3135, 'grad_norm': 0.0007097623827274476, 'learning_rate': 0.026490666646784667, 'epoch': 0.25} + 25%|██▍ | 128/520 [08:05<25:07, 3.84s/it] 25%|██▍ | 129/520 [08:09<24:38, 3.78s/it] {'loss': 1.2935, 'grad_norm': 0.0006729630190410913, 'learning_rate': 0.02643034330504516, 'epoch': 0.25} + 25%|██▍ | 129/520 [08:09<24:38, 3.78s/it] 25%|██▌ | 130/520 [08:12<24:33, 3.78s/it] {'loss': 1.3074, 'grad_norm': 0.0006699415787602603, 'learning_rate': 0.026369575847181793, 'epoch': 0.25} + 25%|██▌ | 130/520 [08:12<24:33, 3.78s/it] 25%|██▌ | 131/520 [08:16<24:37, 3.80s/it] {'loss': 1.2063, 'grad_norm': 0.0006138767925995708, 'learning_rate': 0.026308366634261695, 'epoch': 0.25} + 25%|██▌ | 131/520 [08:16<24:37, 3.80s/it] 25%|██▌ | 132/520 [08:20<24:36, 3.80s/it] {'loss': 1.3619, 'grad_norm': 0.0008009230967978993, 'learning_rate': 0.02624671804451601, 'epoch': 0.25} + 25%|██▌ | 132/520 [08:20<24:36, 3.80s/it] 26%|██▌ | 133/520 [08:24<24:36, 3.82s/it] {'loss': 1.2789, 'grad_norm': 0.000857170425054599, 'learning_rate': 0.026184632473247482, 'epoch': 0.26} + 26%|██▌ | 133/520 [08:24<24:36, 3.82s/it] 26%|██▌ | 134/520 [08:28<24:33, 3.82s/it] {'loss': 1.3457, 'grad_norm': 0.0007385486623634334, 'learning_rate': 0.0261221123327374, 'epoch': 0.26} + 26%|██▌ | 134/520 [08:28<24:33, 3.82s/it] 26%|██▌ | 135/520 [08:31<24:29, 3.82s/it] {'loss': 1.3988, 'grad_norm': 0.0007194279055569345, 'learning_rate': 0.026059160052151857, 'epoch': 0.26} + 26%|██▌ | 135/520 [08:31<24:29, 3.82s/it] 26%|██▌ | 136/520 [08:35<24:25, 3.82s/it] {'loss': 1.3644, 'grad_norm': 0.0007438541592708173, 'learning_rate': 0.025995778077447393, 'epoch': 0.26} + 26%|██▌ | 136/520 [08:35<24:25, 3.82s/it] 26%|██▋ | 137/520 [08:39<24:25, 3.83s/it] {'loss': 1.2828, 'grad_norm': 0.0008940401264137671, 'learning_rate': 0.025931968871275926, 'epoch': 0.26} + 26%|██▋ | 137/520 [08:39<24:25, 3.83s/it] 27%|██▋ | 138/520 [08:43<24:20, 3.82s/it] {'loss': 1.2836, 'grad_norm': 0.0007045898468181, 'learning_rate': 0.02586773491288909, 'epoch': 0.27} + 27%|██▋ | 138/520 [08:43<24:20, 3.82s/it] 27%|██▋ | 139/520 [08:47<24:17, 3.83s/it] {'loss': 1.1269, 'grad_norm': 0.0007400132663673564, 'learning_rate': 0.0258030786980419, 'epoch': 0.27} + 27%|██▋ | 139/520 [08:47<24:17, 3.83s/it] 27%|██▋ | 140/520 [08:50<23:52, 3.77s/it] {'loss': 1.2654, 'grad_norm': 0.0007127153966784771, 'learning_rate': 0.025738002738895777, 'epoch': 0.27} + 27%|██▋ | 140/520 [08:50<23:52, 3.77s/it] 27%|██▋ | 141/520 [08:54<23:33, 3.73s/it] {'loss': 1.3644, 'grad_norm': 0.0006830944188338759, 'learning_rate': 0.02567250956392095, 'epoch': 0.27} + 27%|██▋ | 141/520 [08:54<23:33, 3.73s/it] 27%|██▋ | 142/520 [08:58<23:21, 3.71s/it] {'loss': 1.2757, 'grad_norm': 0.0006627871006710643, 'learning_rate': 0.02560660171779821, 'epoch': 0.27} + 27%|██▋ | 142/520 [08:58<23:21, 3.71s/it] 28%|██▊ | 143/520 [09:01<23:10, 3.69s/it] {'loss': 1.2999, 'grad_norm': 0.0008630744635515056, 'learning_rate': 0.025540281761320036, 'epoch': 0.28} + 28%|██▊ | 143/520 [09:01<23:10, 3.69s/it] 28%|██▊ | 144/520 [09:05<23:01, 3.68s/it] {'loss': 1.2967, 'grad_norm': 0.0008450835960211104, 'learning_rate': 0.02547355227129109, 'epoch': 0.28} + 28%|██▊ | 144/520 [09:05<23:01, 3.68s/it] 28%|██▊ | 145/520 [09:09<22:54, 3.66s/it] {'loss': 1.2116, 'grad_norm': 0.0007695565669394956, 'learning_rate': 0.025406415840428125, 'epoch': 0.28} + 28%|██▊ | 145/520 [09:09<22:54, 3.66s/it] 28%|██▊ | 146/520 [09:12<22:47, 3.66s/it] {'loss': 1.3251, 'grad_norm': 0.0007585666664902014, 'learning_rate': 0.025338875077259203, 'epoch': 0.28} + 28%|██▊ | 146/520 [09:12<22:47, 3.66s/it] 28%|██▊ | 147/520 [09:16<22:41, 3.65s/it] {'loss': 1.2596, 'grad_norm': 0.0008833940177645538, 'learning_rate': 0.025270932606022402, 'epoch': 0.28} + 28%|██▊ | 147/520 [09:16<22:41, 3.65s/it] 28%|██▊ | 148/520 [09:20<22:38, 3.65s/it] {'loss': 1.2904, 'grad_norm': 0.0007675233849834222, 'learning_rate': 0.02520259106656379, 'epoch': 0.28} + 28%|██▊ | 148/520 [09:20<22:38, 3.65s/it] 29%|██▊ | 149/520 [09:23<22:37, 3.66s/it] {'loss': 1.2142, 'grad_norm': 0.0007455119202169978, 'learning_rate': 0.025133853114234904, 'epoch': 0.29} + 29%|██▊ | 149/520 [09:23<22:37, 3.66s/it] 29%|██▉ | 150/520 [09:27<22:32, 3.66s/it] {'loss': 1.4306, 'grad_norm': 0.0008597090116355185, 'learning_rate': 0.02506472141978955, 'epoch': 0.29} + 29%|██▉ | 150/520 [09:27<22:32, 3.66s/it] 29%|██▉ | 151/520 [09:30<22:23, 3.64s/it] {'loss': 1.2768, 'grad_norm': 0.0007932696790208784, 'learning_rate': 0.024995198669280058, 'epoch': 0.29} + 29%|██▉ | 151/520 [09:30<22:23, 3.64s/it] 29%|██▉ | 152/520 [09:34<22:18, 3.64s/it] {'loss': 1.2414, 'grad_norm': 0.0008585683005392958, 'learning_rate': 0.024925287563952892, 'epoch': 0.29} + 29%|██▉ | 152/520 [09:34<22:18, 3.64s/it] 29%|██▉ | 153/520 [09:38<22:13, 3.63s/it] {'loss': 1.2665, 'grad_norm': 0.0007985780768509842, 'learning_rate': 0.024854990820143707, 'epoch': 0.29} + 29%|██▉ | 153/520 [09:38<22:13, 3.63s/it] 30%|██▉ | 154/520 [09:41<22:10, 3.64s/it] {'loss': 1.3509, 'grad_norm': 0.0007530812943933175, 'learning_rate': 0.024784311169171815, 'epoch': 0.3} + 30%|██▉ | 154/520 [09:41<22:10, 3.64s/it] 30%|██▉ | 155/520 [09:45<22:08, 3.64s/it] {'loss': 1.2722, 'grad_norm': 0.000884638636242677, 'learning_rate': 0.024713251357234057, 'epoch': 0.3} + 30%|██▉ | 155/520 [09:45<22:08, 3.64s/it] 30%|███ | 156/520 [09:49<22:03, 3.64s/it] {'loss': 1.297, 'grad_norm': 0.0009683781049774049, 'learning_rate': 0.024641814145298088, 'epoch': 0.3} + 30%|███ | 156/520 [09:49<22:03, 3.64s/it] 30%|███ | 157/520 [09:52<22:02, 3.64s/it] {'loss': 1.3233, 'grad_norm': 0.0007496413907054461, 'learning_rate': 0.02457000230899513, 'epoch': 0.3} + 30%|███ | 157/520 [09:52<22:02, 3.64s/it] 30%|███ | 158/520 [09:56<21:54, 3.63s/it] {'loss': 1.2793, 'grad_norm': 0.0008363843335815012, 'learning_rate': 0.0244978186385121, 'epoch': 0.3} + 30%|███ | 158/520 [09:56<21:54, 3.63s/it] 31%|███ | 159/520 [10:00<21:48, 3.63s/it] {'loss': 1.334, 'grad_norm': 0.0008109655470152422, 'learning_rate': 0.02442526593848321, 'epoch': 0.31} + 31%|███ | 159/520 [10:00<21:48, 3.63s/it] 31%|███ | 160/520 [10:03<21:47, 3.63s/it] {'loss': 1.3282, 'grad_norm': 0.0008054986786546615, 'learning_rate': 0.024352347027881003, 'epoch': 0.31} + 31%|███ | 160/520 [10:03<21:47, 3.63s/it] 31%|███ | 161/520 [10:07<21:46, 3.64s/it] {'loss': 1.3064, 'grad_norm': 0.0008444420019589041, 'learning_rate': 0.024279064739906823, 'epoch': 0.31} + 31%|███ | 161/520 [10:07<21:46, 3.64s/it] 31%|███ | 162/520 [10:10<21:44, 3.64s/it] {'loss': 1.257, 'grad_norm': 0.0008245364860828269, 'learning_rate': 0.024205421921880712, 'epoch': 0.31} + 31%|███ | 162/520 [10:10<21:44, 3.64s/it] 31%|███▏ | 163/520 [10:14<21:37, 3.64s/it] {'loss': 1.2157, 'grad_norm': 0.0009922190258395587, 'learning_rate': 0.02413142143513081, 'epoch': 0.31} + 31%|███▏ | 163/520 [10:14<21:37, 3.64s/it] 32%|███▏ | 164/520 [10:18<21:35, 3.64s/it] {'loss': 1.17, 'grad_norm': 0.0008488050272326251, 'learning_rate': 0.02405706615488216, 'epoch': 0.32} + 32%|███▏ | 164/520 [10:18<21:35, 3.64s/it] 32%|███▏ | 165/520 [10:21<21:32, 3.64s/it] {'loss': 1.3408, 'grad_norm': 0.0008259492939655172, 'learning_rate': 0.023982358970145005, 'epoch': 0.32} + 32%|███▏ | 165/520 [10:21<21:32, 3.64s/it] 32%|███▏ | 166/520 [10:25<21:26, 3.63s/it] {'loss': 1.2901, 'grad_norm': 0.0009143467437389575, 'learning_rate': 0.02390730278360252, 'epoch': 0.32} + 32%|███▏ | 166/520 [10:25<21:26, 3.63s/it] 32%|███▏ | 167/520 [10:29<21:24, 3.64s/it] {'loss': 1.2854, 'grad_norm': 0.0008220348040744097, 'learning_rate': 0.02383190051149807, 'epoch': 0.32} + 32%|███▏ | 167/520 [10:29<21:24, 3.64s/it] 32%|███▏ | 168/520 [10:32<21:23, 3.65s/it] {'loss': 1.2235, 'grad_norm': 0.0007839477643849899, 'learning_rate': 0.02375615508352185, 'epoch': 0.32} + 32%|███▏ | 168/520 [10:32<21:23, 3.65s/it] 32%|███▎ | 169/520 [10:36<21:21, 3.65s/it] {'loss': 1.2989, 'grad_norm': 0.0008023151934483476, 'learning_rate': 0.02368006944269709, 'epoch': 0.33} + 32%|███▎ | 169/520 [10:36<21:21, 3.65s/it] 33%|███▎ | 170/520 [10:40<21:20, 3.66s/it] {'loss': 1.2118, 'grad_norm': 0.0006989785786115103, 'learning_rate': 0.02360364654526569, 'epoch': 0.33} + 33%|███▎ | 170/520 [10:40<21:20, 3.66s/it] 33%|███▎ | 171/520 [10:43<21:19, 3.67s/it] {'loss': 1.2489, 'grad_norm': 0.0009187978654139596, 'learning_rate': 0.023526889360573386, 'epoch': 0.33} + 33%|███▎ | 171/520 [10:43<21:19, 3.67s/it] 33%|███▎ | 172/520 [10:47<21:18, 3.67s/it] {'loss': 1.3128, 'grad_norm': 0.000813772942755099, 'learning_rate': 0.02344980087095433, 'epoch': 0.33} + 33%|███▎ | 172/520 [10:47<21:18, 3.67s/it] 33%|███▎ | 173/520 [10:51<21:12, 3.67s/it] {'loss': 1.2446, 'grad_norm': 0.0008262589690076697, 'learning_rate': 0.023372384071615262, 'epoch': 0.33} + 33%|███▎ | 173/520 [10:51<21:12, 3.67s/it] 33%|███▎ | 174/520 [10:54<21:06, 3.66s/it] {'loss': 1.2996, 'grad_norm': 0.0008642753367094732, 'learning_rate': 0.02329464197051909, 'epoch': 0.33} + 33%|███▎ | 174/520 [10:54<21:06, 3.66s/it] 34%|███▎ | 175/520 [10:58<21:09, 3.68s/it] {'loss': 1.2102, 'grad_norm': 0.0008250039117610039, 'learning_rate': 0.023216577588268072, 'epoch': 0.34} + 34%|███▎ | 175/520 [10:58<21:09, 3.68s/it] 34%|███▍ | 176/520 [11:02<21:06, 3.68s/it] {'loss': 1.2695, 'grad_norm': 0.0007781260116089651, 'learning_rate': 0.02313819395798639, 'epoch': 0.34} + 34%|███▍ | 176/520 [11:02<21:06, 3.68s/it] 34%|███▍ | 177/520 [11:05<21:02, 3.68s/it] {'loss': 1.171, 'grad_norm': 0.0008151131196888738, 'learning_rate': 0.02305949412520236, 'epoch': 0.34} + 34%|███▍ | 177/520 [11:05<21:02, 3.68s/it] 34%|███▍ | 178/520 [11:09<20:56, 3.67s/it] {'loss': 1.2759, 'grad_norm': 0.0009246719598748152, 'learning_rate': 0.022980481147730047, 'epoch': 0.34} + 34%|███▍ | 178/520 [11:09<20:56, 3.67s/it] 34%|███▍ | 179/520 [11:13<20:52, 3.67s/it] {'loss': 1.3549, 'grad_norm': 0.0008042667962826618, 'learning_rate': 0.02290115809555051, 'epoch': 0.34} + 34%|███▍ | 179/520 [11:13<20:52, 3.67s/it] 35%|███▍ | 180/520 [11:16<20:45, 3.66s/it] {'loss': 1.2721, 'grad_norm': 0.0008760983307098861, 'learning_rate': 0.022821528050692468, 'epoch': 0.35} + 35%|███▍ | 180/520 [11:16<20:45, 3.66s/it] 35%|███▍ | 181/520 [11:20<20:40, 3.66s/it] {'loss': 1.2492, 'grad_norm': 0.0007566313775123798, 'learning_rate': 0.022741594107112597, 'epoch': 0.35} + 35%|███▍ | 181/520 [11:20<20:40, 3.66s/it] 35%|███▌ | 182/520 [11:24<20:38, 3.66s/it] {'loss': 1.264, 'grad_norm': 0.0008993199681273479, 'learning_rate': 0.022661359370575285, 'epoch': 0.35} + 35%|███▌ | 182/520 [11:24<20:38, 3.66s/it] 35%|███▌ | 183/520 [11:27<20:33, 3.66s/it] {'loss': 1.2759, 'grad_norm': 0.0008186312351071321, 'learning_rate': 0.022580826958531964, 'epoch': 0.35} + 35%|███▌ | 183/520 [11:27<20:33, 3.66s/it] 35%|███▌ | 184/520 [11:31<20:29, 3.66s/it] {'loss': 1.2477, 'grad_norm': 0.0010100870756264417, 'learning_rate': 0.0225, 'epoch': 0.35} + 35%|███▌ | 184/520 [11:31<20:29, 3.66s/it] 36%|███▌ | 185/520 [11:35<20:27, 3.67s/it] {'loss': 1.3327, 'grad_norm': 0.0008446850915422703, 'learning_rate': 0.02241888163544111, 'epoch': 0.36} + 36%|███▌ | 185/520 [11:35<20:27, 3.67s/it] 36%|███▌ | 186/520 [11:38<20:25, 3.67s/it] {'loss': 1.2631, 'grad_norm': 0.0009315730168197859, 'learning_rate': 0.02233747501663934, 'epoch': 0.36} + 36%|███▌ | 186/520 [11:38<20:25, 3.67s/it] 36%|███▌ | 187/520 [11:42<20:21, 3.67s/it] {'loss': 1.2515, 'grad_norm': 0.0010128939416518337, 'learning_rate': 0.022255783306578597, 'epoch': 0.36} + 36%|███▌ | 187/520 [11:42<20:21, 3.67s/it] 36%|███▌ | 188/520 [11:46<20:16, 3.66s/it] {'loss': 1.3482, 'grad_norm': 0.0009073029620618887, 'learning_rate': 0.022173809679319773, 'epoch': 0.36} + 36%|███▌ | 188/520 [11:46<20:16, 3.66s/it] 36%|███▋ | 189/520 [11:49<20:10, 3.66s/it] {'loss': 1.3329, 'grad_norm': 0.0008103915788697835, 'learning_rate': 0.022091557319877407, 'epoch': 0.36} + 36%|███▋ | 189/520 [11:49<20:10, 3.66s/it] 37%|███▋ | 190/520 [11:53<20:11, 3.67s/it] {'loss': 1.2519, 'grad_norm': 0.0009711083674249614, 'learning_rate': 0.02200902942409593, 'epoch': 0.37} + 37%|███▋ | 190/520 [11:53<20:11, 3.67s/it] 37%|███▋ | 191/520 [11:57<20:17, 3.70s/it] {'loss': 1.225, 'grad_norm': 0.000938907665447702, 'learning_rate': 0.02192622919852551, 'epoch': 0.37} + 37%|███▋ | 191/520 [11:57<20:17, 3.70s/it] 37%|███▋ | 192/520 [12:01<20:28, 3.75s/it] {'loss': 1.2922, 'grad_norm': 0.0008784363260931927, 'learning_rate': 0.021843159860297445, 'epoch': 0.37} + 37%|███▋ | 192/520 [12:01<20:28, 3.75s/it] 37%|███▋ | 193/520 [12:05<20:33, 3.77s/it] {'loss': 1.2284, 'grad_norm': 0.0009497974054842567, 'learning_rate': 0.02175982463699918, 'epoch': 0.37} + 37%|███▋ | 193/520 [12:05<20:33, 3.77s/it] 37%|███▋ | 194/520 [12:08<20:24, 3.75s/it] {'loss': 1.119, 'grad_norm': 0.0008273750079332058, 'learning_rate': 0.021676226766548883, 'epoch': 0.37} + 37%|███▋ | 194/520 [12:08<20:24, 3.75s/it] 38%|███▊ | 195/520 [12:12<20:10, 3.72s/it] {'loss': 1.3005, 'grad_norm': 0.0008374957518066912, 'learning_rate': 0.021592369497069674, 'epoch': 0.38} + 38%|███▊ | 195/520 [12:12<20:10, 3.72s/it] 38%|███▊ | 196/520 [12:16<19:57, 3.70s/it] {'loss': 1.3066, 'grad_norm': 0.0010315803768696018, 'learning_rate': 0.02150825608676337, 'epoch': 0.38} + 38%|███▊ | 196/520 [12:16<19:57, 3.70s/it] 38%|███▊ | 197/520 [12:19<19:48, 3.68s/it] {'loss': 1.2419, 'grad_norm': 0.0009523246174179482, 'learning_rate': 0.02142388980378394, 'epoch': 0.38} + 38%|███▊ | 197/520 [12:19<19:48, 3.68s/it] 38%|███▊ | 198/520 [12:23<19:42, 3.67s/it] {'loss': 1.324, 'grad_norm': 0.001021414651065932, 'learning_rate': 0.021339273926110493, 'epoch': 0.38} + 38%|███▊ | 198/520 [12:23<19:42, 3.67s/it] 38%|███▊ | 199/520 [12:26<19:36, 3.67s/it] {'loss': 1.2269, 'grad_norm': 0.0009357834699004217, 'learning_rate': 0.021254411741419923, 'epoch': 0.38} + 38%|███▊ | 199/520 [12:26<19:36, 3.67s/it] 38%|███▊ | 200/520 [12:30<19:32, 3.66s/it] {'loss': 1.1841, 'grad_norm': 0.0008964465918059075, 'learning_rate': 0.021169306546959175, 'epoch': 0.38} + 38%|███▊ | 200/520 [12:30<19:32, 3.66s/it] 39%|███▊ | 201/520 [12:34<19:29, 3.67s/it] {'loss': 1.1959, 'grad_norm': 0.0008113761922487723, 'learning_rate': 0.021083961649417126, 'epoch': 0.39} + 39%|███▊ | 201/520 [12:34<19:29, 3.67s/it] 39%|███▉ | 202/520 [12:37<19:26, 3.67s/it] {'loss': 1.2561, 'grad_norm': 0.0009589694587939786, 'learning_rate': 0.02099838036479611, 'epoch': 0.39} + 39%|███▉ | 202/520 [12:37<19:26, 3.67s/it] 39%|███▉ | 203/520 [12:41<19:22, 3.67s/it] {'loss': 1.2837, 'grad_norm': 0.0009768267269237082, 'learning_rate': 0.02091256601828309, 'epoch': 0.39} + 39%|███▉ | 203/520 [12:41<19:22, 3.67s/it] 39%|███▉ | 204/520 [12:45<19:18, 3.67s/it] {'loss': 1.3012, 'grad_norm': 0.0009600825228655656, 'learning_rate': 0.02082652194412042, 'epoch': 0.39} + 39%|███▉ | 204/520 [12:45<19:18, 3.67s/it] 39%|███▉ | 205/520 [12:48<19:16, 3.67s/it] {'loss': 1.1959, 'grad_norm': 0.0008754248592212839, 'learning_rate': 0.020740251485476348, 'epoch': 0.39} + 39%|███▉ | 205/520 [12:48<19:16, 3.67s/it] 40%|███▉ | 206/520 [12:52<19:11, 3.67s/it] {'loss': 1.3117, 'grad_norm': 0.0008985222115809133, 'learning_rate': 0.020653757994315076, 'epoch': 0.4} + 40%|███▉ | 206/520 [12:52<19:11, 3.67s/it] 40%|███▉ | 207/520 [12:56<19:09, 3.67s/it] {'loss': 1.1556, 'grad_norm': 0.0007625674254939507, 'learning_rate': 0.02056704483126657, 'epoch': 0.4} + 40%|███▉ | 207/520 [12:56<19:09, 3.67s/it] 40%|████ | 208/520 [12:59<19:02, 3.66s/it] {'loss': 1.3115, 'grad_norm': 0.0010188224664782952, 'learning_rate': 0.02048011536549593, 'epoch': 0.4} + 40%|████ | 208/520 [12:59<19:02, 3.66s/it] 40%|████ | 209/520 [13:03<19:04, 3.68s/it] {'loss': 1.229, 'grad_norm': 0.0008829060933269436, 'learning_rate': 0.020392972974572512, 'epoch': 0.4} + 40%|████ | 209/520 [13:03<19:04, 3.68s/it] 40%|████ | 210/520 [13:07<18:59, 3.68s/it] {'loss': 1.2978, 'grad_norm': 0.0009463447548183801, 'learning_rate': 0.02030562104433872, 'epoch': 0.4} + 40%|████ | 210/520 [13:07<18:59, 3.68s/it] 41%|████ | 211/520 [13:10<18:54, 3.67s/it] {'loss': 1.2896, 'grad_norm': 0.0008598625731582578, 'learning_rate': 0.020218062968778407, 'epoch': 0.41} + 41%|████ | 211/520 [13:10<18:54, 3.67s/it] 41%|████ | 212/520 [13:14<18:49, 3.67s/it] {'loss': 1.302, 'grad_norm': 0.0008959193726149853, 'learning_rate': 0.020130302149885034, 'epoch': 0.41} + 41%|████ | 212/520 [13:14<18:49, 3.67s/it] 41%|████ | 213/520 [13:18<18:44, 3.66s/it] {'loss': 1.255, 'grad_norm': 0.0010050967015773112, 'learning_rate': 0.020042341997529464, 'epoch': 0.41} + 41%|████ | 213/520 [13:18<18:44, 3.66s/it] 41%|████ | 214/520 [13:21<18:41, 3.67s/it] {'loss': 1.2587, 'grad_norm': 0.0009878914441678005, 'learning_rate': 0.019954185929327507, 'epoch': 0.41} + 41%|████ | 214/520 [13:21<18:41, 3.67s/it] 41%|████▏ | 215/520 [13:25<18:37, 3.66s/it] {'loss': 1.1324, 'grad_norm': 0.0009328977789096166, 'learning_rate': 0.01986583737050711, 'epoch': 0.41} + 41%|████▏ | 215/520 [13:25<18:37, 3.66s/it] 42%|████▏ | 216/520 [13:29<18:33, 3.66s/it] {'loss': 1.1659, 'grad_norm': 0.0009379149352242593, 'learning_rate': 0.019777299753775266, 'epoch': 0.42} + 42%|████▏ | 216/520 [13:29<18:33, 3.66s/it] 42%|████▏ | 217/520 [13:32<18:29, 3.66s/it] {'loss': 1.2933, 'grad_norm': 0.0009782138375184456, 'learning_rate': 0.019688576519184667, 'epoch': 0.42} + 42%|████▏ | 217/520 [13:32<18:29, 3.66s/it] 42%|████▏ | 218/520 [13:36<18:37, 3.70s/it] {'loss': 1.27, 'grad_norm': 0.0010059022448080607, 'learning_rate': 0.019599671114000013, 'epoch': 0.42} + 42%|████▏ | 218/520 [13:36<18:37, 3.70s/it] 42%|████▏ | 219/520 [13:40<18:44, 3.74s/it] {'loss': 1.2943, 'grad_norm': 0.0009004866146098169, 'learning_rate': 0.019510586992564093, 'epoch': 0.42} + 42%|████▏ | 219/520 [13:40<18:44, 3.74s/it] 42%|████▏ | 220/520 [13:44<18:41, 3.74s/it] {'loss': 1.1726, 'grad_norm': 0.0008658525804900313, 'learning_rate': 0.019421327616163563, 'epoch': 0.42} + 42%|████▏ | 220/520 [13:44<18:41, 3.74s/it] 42%|████▎ | 221/520 [13:47<18:30, 3.71s/it] {'loss': 1.2849, 'grad_norm': 0.0009881235488127652, 'learning_rate': 0.01933189645289445, 'epoch': 0.42} + 42%|████▎ | 221/520 [13:47<18:30, 3.71s/it] 43%|████▎ | 222/520 [13:51<18:36, 3.75s/it] {'loss': 1.2148, 'grad_norm': 0.0009693171374883983, 'learning_rate': 0.019242296977527412, 'epoch': 0.43} + 43%|████▎ | 222/520 [13:51<18:36, 3.75s/it] 43%|████▎ | 223/520 [13:55<18:28, 3.73s/it] {'loss': 1.2329, 'grad_norm': 0.0009313291509963202, 'learning_rate': 0.01915253267137274, 'epoch': 0.43} + 43%|████▎ | 223/520 [13:55<18:28, 3.73s/it] 43%|████▎ | 224/520 [13:59<18:20, 3.72s/it] {'loss': 1.205, 'grad_norm': 0.0008193076065644014, 'learning_rate': 0.019062607022145076, 'epoch': 0.43} + 43%|████▎ | 224/520 [13:59<18:20, 3.72s/it] 43%|████▎ | 225/520 [14:02<18:10, 3.70s/it] {'loss': 1.2243, 'grad_norm': 0.0009763351141909621, 'learning_rate': 0.018972523523827906, 'epoch': 0.43} + 43%|████▎ | 225/520 [14:02<18:10, 3.70s/it] 43%|████▎ | 226/520 [14:06<18:03, 3.68s/it] {'loss': 1.3401, 'grad_norm': 0.0009440369033874206, 'learning_rate': 0.01888228567653781, 'epoch': 0.43} + 43%|████▎ | 226/520 [14:06<18:03, 3.68s/it] 44%|████▎ | 227/520 [14:10<17:56, 3.68s/it] {'loss': 1.3077, 'grad_norm': 0.00094683339487539, 'learning_rate': 0.01879189698638846, 'epoch': 0.44} + 44%|████▎ | 227/520 [14:10<17:56, 3.68s/it] 44%|████▍ | 228/520 [14:13<17:52, 3.67s/it] {'loss': 1.2594, 'grad_norm': 0.0008958683098356264, 'learning_rate': 0.0187013609653544, 'epoch': 0.44} + 44%|████▍ | 228/520 [14:13<17:52, 3.67s/it] 44%|████▍ | 229/520 [14:17<17:52, 3.68s/it] {'loss': 1.2822, 'grad_norm': 0.000882776738226201, 'learning_rate': 0.018610681131134597, 'epoch': 0.44} + 44%|████▍ | 229/520 [14:17<17:52, 3.68s/it] 44%|████▍ | 230/520 [14:21<17:49, 3.69s/it] {'loss': 1.1804, 'grad_norm': 0.0009150633973949253, 'learning_rate': 0.01851986100701573, 'epoch': 0.44} + 44%|████▍ | 230/520 [14:21<17:49, 3.69s/it] 44%|████▍ | 231/520 [14:24<17:42, 3.67s/it] {'loss': 1.2527, 'grad_norm': 0.0008665995516684892, 'learning_rate': 0.018428904121735343, 'epoch': 0.44} + 44%|████▍ | 231/520 [14:24<17:42, 3.67s/it] 45%|████▍ | 232/520 [14:28<17:37, 3.67s/it] {'loss': 1.3053, 'grad_norm': 0.0009044624638966961, 'learning_rate': 0.018337814009344713, 'epoch': 0.45} + 45%|████▍ | 232/520 [14:28<17:37, 3.67s/it] 45%|████▍ | 233/520 [14:32<17:32, 3.67s/it] {'loss': 1.2078, 'grad_norm': 0.0010070583158172494, 'learning_rate': 0.018246594209071543, 'epoch': 0.45} + 45%|████▍ | 233/520 [14:32<17:32, 3.67s/it] 45%|████▌ | 234/520 [14:35<17:29, 3.67s/it] {'loss': 1.1999, 'grad_norm': 0.0010019114177457252, 'learning_rate': 0.018155248265182437, 'epoch': 0.45} + 45%|████▌ | 234/520 [14:35<17:29, 3.67s/it] 45%|████▌ | 235/520 [14:39<17:29, 3.68s/it] {'loss': 1.2622, 'grad_norm': 0.0009920837917784268, 'learning_rate': 0.018063779726845204, 'epoch': 0.45} + 45%|████▌ | 235/520 [14:39<17:29, 3.68s/it] 45%|████▌ | 236/520 [14:43<17:24, 3.68s/it] {'loss': 1.3177, 'grad_norm': 0.0009604553229069513, 'learning_rate': 0.017972192147990965, 'epoch': 0.45} + 45%|████▌ | 236/520 [14:43<17:24, 3.68s/it] 46%|████▌ | 237/520 [14:46<17:21, 3.68s/it] {'loss': 1.3144, 'grad_norm': 0.0009623488714060682, 'learning_rate': 0.017880489087176044, 'epoch': 0.46} + 46%|████▌ | 237/520 [14:46<17:21, 3.68s/it] 46%|████▌ | 238/520 [14:50<17:14, 3.67s/it] {'loss': 1.2569, 'grad_norm': 0.001015740279692093, 'learning_rate': 0.01778867410744372, 'epoch': 0.46} + 46%|████▌ | 238/520 [14:50<17:14, 3.67s/it] 46%|████▌ | 239/520 [14:54<17:12, 3.67s/it] {'loss': 1.3104, 'grad_norm': 0.0009564795575148513, 'learning_rate': 0.01769675077618579, 'epoch': 0.46} + 46%|████▌ | 239/520 [14:54<17:12, 3.67s/it] 46%|████▌ | 240/520 [14:57<17:09, 3.68s/it] {'loss': 1.153, 'grad_norm': 0.0009168435327977884, 'learning_rate': 0.017604722665003958, 'epoch': 0.46} + 46%|████▌ | 240/520 [14:57<17:09, 3.68s/it] 46%|████▋ | 241/520 [15:01<17:04, 3.67s/it] {'loss': 1.2335, 'grad_norm': 0.001000072613410532, 'learning_rate': 0.017512593349571046, 'epoch': 0.46} + 46%|████▋ | 241/520 [15:01<17:04, 3.67s/it] 47%|████▋ | 242/520 [15:05<17:05, 3.69s/it] {'loss': 1.2269, 'grad_norm': 0.0008904451222971796, 'learning_rate': 0.0174203664094921, 'epoch': 0.47} + 47%|████▋ | 242/520 [15:05<17:05, 3.69s/it] 47%|████▋ | 243/520 [15:09<17:01, 3.69s/it] {'loss': 1.2339, 'grad_norm': 0.0009423638940394828, 'learning_rate': 0.01732804542816527, 'epoch': 0.47} + 47%|████▋ | 243/520 [15:09<17:01, 3.69s/it] 47%|████▋ | 244/520 [15:12<17:03, 3.71s/it] {'loss': 1.3099, 'grad_norm': 0.0009343442653766102, 'learning_rate': 0.017235633992642614, 'epoch': 0.47} + 47%|████▋ | 244/520 [15:12<17:03, 3.71s/it] 47%|████▋ | 245/520 [15:16<16:57, 3.70s/it] {'loss': 1.2153, 'grad_norm': 0.0009982015855854725, 'learning_rate': 0.0171431356934907, 'epoch': 0.47} + 47%|████▋ | 245/520 [15:16<16:57, 3.70s/it] 47%|████▋ | 246/520 [15:20<16:50, 3.69s/it] {'loss': 1.2914, 'grad_norm': 0.0009455863146618818, 'learning_rate': 0.0170505541246511, 'epoch': 0.47} + 47%|████▋ | 246/520 [15:20<16:50, 3.69s/it] 48%|████▊ | 247/520 [15:23<16:44, 3.68s/it] {'loss': 1.3814, 'grad_norm': 0.0009730871437207335, 'learning_rate': 0.016957892883300775, 'epoch': 0.47} + 48%|████▊ | 247/520 [15:23<16:44, 3.68s/it] 48%|████▊ | 248/520 [15:27<16:38, 3.67s/it] {'loss': 1.1966, 'grad_norm': 0.0009958342547350097, 'learning_rate': 0.01686515556971228, 'epoch': 0.48} + 48%|████▊ | 248/520 [15:27<16:38, 3.67s/it] 48%|████▊ | 249/520 [15:31<16:31, 3.66s/it] {'loss': 1.2923, 'grad_norm': 0.0009589674455226537, 'learning_rate': 0.016772345787113892, 'epoch': 0.48} + 48%|████▊ | 249/520 [15:31<16:31, 3.66s/it] 48%|████▊ | 250/520 [15:34<16:30, 3.67s/it] {'loss': 1.2515, 'grad_norm': 0.0010743761923630796, 'learning_rate': 0.016679467141549616, 'epoch': 0.48} + 48%|████▊ | 250/520 [15:34<16:30, 3.67s/it] 48%|████▊ | 251/520 [15:38<16:29, 3.68s/it] {'loss': 1.3097, 'grad_norm': 0.0009284950589982723, 'learning_rate': 0.01658652324173907, 'epoch': 0.48} + 48%|████▊ | 251/520 [15:38<16:29, 3.68s/it] 48%|████▊ | 252/520 [15:42<16:24, 3.67s/it] {'loss': 1.2048, 'grad_norm': 0.0009053021117386091, 'learning_rate': 0.01649351769893725, 'epoch': 0.48} + 48%|████▊ | 252/520 [15:42<16:24, 3.67s/it] 49%|████▊ | 253/520 [15:45<16:19, 3.67s/it] {'loss': 1.2809, 'grad_norm': 0.0010446974230909909, 'learning_rate': 0.016400454126794257, 'epoch': 0.49} + 49%|████▊ | 253/520 [15:45<16:19, 3.67s/it] 49%|████▉ | 254/520 [15:49<16:12, 3.66s/it] {'loss': 1.2325, 'grad_norm': 0.0009267425085548351, 'learning_rate': 0.016307336141214875, 'epoch': 0.49} + 49%|████▉ | 254/520 [15:49<16:12, 3.66s/it] 49%|████▉ | 255/520 [15:53<16:20, 3.70s/it] {'loss': 1.2374, 'grad_norm': 0.0010570158392624977, 'learning_rate': 0.01621416736021805, 'epoch': 0.49} + 49%|████▉ | 255/520 [15:53<16:20, 3.70s/it] 49%|████▉ | 256/520 [15:57<16:26, 3.74s/it] {'loss': 1.3068, 'grad_norm': 0.001053583476002272, 'learning_rate': 0.016120951403796366, 'epoch': 0.49} + 49%|████▉ | 256/520 [15:57<16:26, 3.74s/it] 49%|████▉ | 257/520 [16:00<16:31, 3.77s/it] {'loss': 1.2825, 'grad_norm': 0.0009976607902031844, 'learning_rate': 0.01602769189377535, 'epoch': 0.49} + 49%|████▉ | 257/520 [16:00<16:31, 3.77s/it] 50%|████▉ | 258/520 [16:04<16:20, 3.74s/it] {'loss': 1.2767, 'grad_norm': 0.0008439246112002426, 'learning_rate': 0.015934392453672782, 'epoch': 0.5} + 50%|████▉ | 258/520 [16:04<16:20, 3.74s/it] 50%|████▉ | 259/520 [16:08<16:09, 3.72s/it] {'loss': 1.346, 'grad_norm': 0.0010827734740306793, 'learning_rate': 0.015841056708557875, 'epoch': 0.5} + 50%|████▉ | 259/520 [16:08<16:09, 3.72s/it] 50%|█████ | 260/520 [16:11<16:01, 3.70s/it] {'loss': 1.2484, 'grad_norm': 0.0007617148035356368, 'learning_rate': 0.015747688284910456, 'epoch': 0.5} + 50%|█████ | 260/520 [16:11<16:01, 3.70s/it] 50%|█████ | 261/520 [16:15<16:11, 3.75s/it] {'loss': 1.1971, 'grad_norm': 0.0009399068026278555, 'learning_rate': 0.015654290810480042, 'epoch': 0.5} + 50%|█████ | 261/520 [16:15<16:11, 3.75s/it] 50%|█████ | 262/520 [16:19<16:08, 3.75s/it] {'loss': 1.2257, 'grad_norm': 0.0010336754231396027, 'learning_rate': 0.015560867914144887, 'epoch': 0.5} + 50%|█████ | 262/520 [16:19<16:08, 3.75s/it] 51%|█████ | 263/520 [16:23<15:56, 3.72s/it] {'loss': 1.2261, 'grad_norm': 0.001027925304057171, 'learning_rate': 0.015467423225770997, 'epoch': 0.51} + 51%|█████ | 263/520 [16:23<15:56, 3.72s/it] 51%|█████ | 264/520 [16:26<15:49, 3.71s/it] {'loss': 1.3024, 'grad_norm': 0.0009190007977762193, 'learning_rate': 0.015373960376071093, 'epoch': 0.51} + 51%|█████ | 264/520 [16:26<15:49, 3.71s/it] 51%|█████ | 265/520 [16:30<15:42, 3.70s/it] {'loss': 1.248, 'grad_norm': 0.0011315669728947813, 'learning_rate': 0.015280482996463533, 'epoch': 0.51} + 51%|█████ | 265/520 [16:30<15:42, 3.70s/it] 51%|█████ | 266/520 [16:34<15:37, 3.69s/it] {'loss': 1.1217, 'grad_norm': 0.0009014352708325167, 'learning_rate': 0.015186994718931226, 'epoch': 0.51} + 51%|█████ | 266/520 [16:34<15:37, 3.69s/it] 51%|█████▏ | 267/520 [16:37<15:30, 3.68s/it] {'loss': 1.2313, 'grad_norm': 0.0009701420729030409, 'learning_rate': 0.015093499175880503, 'epoch': 0.51} + 51%|█████▏ | 267/520 [16:37<15:30, 3.68s/it] 52%|█████▏ | 268/520 [16:41<15:25, 3.67s/it] {'loss': 1.3237, 'grad_norm': 0.0009676296963726292, 'learning_rate': 0.015, 'epoch': 0.52} + 52%|█████▏ | 268/520 [16:41<15:25, 3.67s/it] 52%|█████▏ | 269/520 [16:45<15:23, 3.68s/it] {'loss': 1.3348, 'grad_norm': 0.0010186702073570022, 'learning_rate': 0.014906500824119496, 'epoch': 0.52} + 52%|█████▏ | 269/520 [16:45<15:23, 3.68s/it] 52%|█████▏ | 270/520 [16:48<15:19, 3.68s/it] {'loss': 1.1583, 'grad_norm': 0.0009397945481725943, 'learning_rate': 0.014813005281068775, 'epoch': 0.52} + 52%|█████▏ | 270/520 [16:48<15:19, 3.68s/it] 52%|█████▏ | 271/520 [16:52<15:15, 3.67s/it] {'loss': 1.3087, 'grad_norm': 0.0010375661447304932, 'learning_rate': 0.01471951700353647, 'epoch': 0.52} + 52%|█████▏ | 271/520 [16:52<15:15, 3.67s/it] 52%|█████▏ | 272/520 [16:56<15:13, 3.68s/it] {'loss': 1.1855, 'grad_norm': 0.0009523310972348567, 'learning_rate': 0.014626039623928909, 'epoch': 0.52} + 52%|█████▏ | 272/520 [16:56<15:13, 3.68s/it] 52%|█████▎ | 273/520 [16:59<15:08, 3.68s/it] {'loss': 1.2685, 'grad_norm': 0.0009451348199455199, 'learning_rate': 0.014532576774229006, 'epoch': 0.53} + 52%|█████▎ | 273/520 [16:59<15:08, 3.68s/it] 53%|█████▎ | 274/520 [17:03<15:05, 3.68s/it] {'loss': 1.3149, 'grad_norm': 0.0011451579555945798, 'learning_rate': 0.014439132085855117, 'epoch': 0.53} + 53%|█████▎ | 274/520 [17:03<15:05, 3.68s/it] 53%|█████▎ | 275/520 [17:07<14:59, 3.67s/it] {'loss': 1.2418, 'grad_norm': 0.0010434109433560006, 'learning_rate': 0.01434570918951996, 'epoch': 0.53} + 53%|█████▎ | 275/520 [17:07<14:59, 3.67s/it] 53%|█████▎ | 276/520 [17:10<14:57, 3.68s/it] {'loss': 1.2949, 'grad_norm': 0.0010792687163951383, 'learning_rate': 0.01425231171508954, 'epoch': 0.53} + 53%|█████▎ | 276/520 [17:10<14:57, 3.68s/it] 53%|█████▎ | 277/520 [17:14<14:55, 3.68s/it] {'loss': 1.2887, 'grad_norm': 0.0009218021367609773, 'learning_rate': 0.01415894329144212, 'epoch': 0.53} + 53%|█████▎ | 277/520 [17:14<14:55, 3.68s/it] 53%|█████▎ | 278/520 [17:18<14:52, 3.69s/it] {'loss': 1.2061, 'grad_norm': 0.0009637677952362289, 'learning_rate': 0.014065607546327219, 'epoch': 0.53} + 53%|█████▎ | 278/520 [17:18<14:52, 3.69s/it] 54%|█████▎ | 279/520 [17:21<14:49, 3.69s/it] {'loss': 1.1756, 'grad_norm': 0.0010375299584695116, 'learning_rate': 0.01397230810622465, 'epoch': 0.54} + 54%|█████▎ | 279/520 [17:21<14:49, 3.69s/it] 54%|█████▍ | 280/520 [17:25<14:42, 3.68s/it] {'loss': 1.2364, 'grad_norm': 0.0010680646057579693, 'learning_rate': 0.013879048596203637, 'epoch': 0.54} + 54%|█████▍ | 280/520 [17:25<14:42, 3.68s/it] 54%|█████▍ | 281/520 [17:29<14:37, 3.67s/it] {'loss': 1.3319, 'grad_norm': 0.0010671278103569026, 'learning_rate': 0.01378583263978195, 'epoch': 0.54} + 54%|█████▍ | 281/520 [17:29<14:37, 3.67s/it] 54%|█████▍ | 282/520 [17:32<14:32, 3.67s/it] {'loss': 1.2162, 'grad_norm': 0.0010128561081606377, 'learning_rate': 0.013692663858785125, 'epoch': 0.54} + 54%|█████▍ | 282/520 [17:32<14:32, 3.67s/it] 54%|█████▍ | 283/520 [17:36<14:28, 3.66s/it] {'loss': 1.3348, 'grad_norm': 0.001052500907248367, 'learning_rate': 0.013599545873205741, 'epoch': 0.54} + 54%|█████▍ | 283/520 [17:36<14:28, 3.66s/it] 55%|█████▍ | 284/520 [17:40<14:24, 3.66s/it] {'loss': 1.1838, 'grad_norm': 0.001027665163059778, 'learning_rate': 0.013506482301062751, 'epoch': 0.55} + 55%|█████▍ | 284/520 [17:40<14:24, 3.66s/it] 55%|█████▍ | 285/520 [17:43<14:23, 3.67s/it] {'loss': 1.2384, 'grad_norm': 0.0010155599875438562, 'learning_rate': 0.013413476758260934, 'epoch': 0.55} + 55%|█████▍ | 285/520 [17:43<14:23, 3.67s/it] 55%|█████▌ | 286/520 [17:47<14:31, 3.72s/it] {'loss': 1.1204, 'grad_norm': 0.0010390637609906774, 'learning_rate': 0.013320532858450381, 'epoch': 0.55} + 55%|█████▌ | 286/520 [17:47<14:31, 3.72s/it] 55%|█████▌ | 287/520 [17:51<14:35, 3.76s/it] {'loss': 1.3429, 'grad_norm': 0.0009943076853294412, 'learning_rate': 0.013227654212886108, 'epoch': 0.55} + 55%|█████▌ | 287/520 [17:51<14:35, 3.76s/it] 55%|█████▌ | 288/520 [17:55<14:37, 3.78s/it] {'loss': 1.3686, 'grad_norm': 0.0009033528978388391, 'learning_rate': 0.013134844430287725, 'epoch': 0.55} + 55%|█████▌ | 288/520 [17:55<14:37, 3.78s/it] 56%|█████▌ | 289/520 [17:59<14:37, 3.80s/it] {'loss': 1.2662, 'grad_norm': 0.0009919472736108186, 'learning_rate': 0.013042107116699229, 'epoch': 0.56} + 56%|█████▌ | 289/520 [17:59<14:37, 3.80s/it] 56%|█████▌ | 290/520 [18:03<14:34, 3.80s/it] {'loss': 1.1791, 'grad_norm': 0.0009183591391537831, 'learning_rate': 0.012949445875348901, 'epoch': 0.56} + 56%|█████▌ | 290/520 [18:03<14:34, 3.80s/it] 56%|█████▌ | 291/520 [18:06<14:31, 3.81s/it] {'loss': 1.209, 'grad_norm': 0.001014513819604561, 'learning_rate': 0.012856864306509301, 'epoch': 0.56} + 56%|█████▌ | 291/520 [18:06<14:31, 3.81s/it] 56%|█████▌ | 292/520 [18:10<14:27, 3.80s/it] {'loss': 1.2754, 'grad_norm': 0.0010277743264729272, 'learning_rate': 0.012764366007357381, 'epoch': 0.56} + 56%|█████▌ | 292/520 [18:10<14:27, 3.80s/it] 56%|█████▋ | 293/520 [18:14<14:26, 3.82s/it] {'loss': 1.2237, 'grad_norm': 0.0010847137698505602, 'learning_rate': 0.012671954571834725, 'epoch': 0.56} + 56%|█████▋ | 293/520 [18:14<14:26, 3.82s/it] 57%|█████▋ | 294/520 [18:18<14:28, 3.84s/it] {'loss': 1.2475, 'grad_norm': 0.00109583859780748, 'learning_rate': 0.0125796335905079, 'epoch': 0.57} + 57%|█████▋ | 294/520 [18:18<14:28, 3.84s/it] 57%|█████▋ | 295/520 [18:22<14:25, 3.85s/it] {'loss': 1.1992, 'grad_norm': 0.0008639071730013199, 'learning_rate': 0.012487406650428955, 'epoch': 0.57} + 57%|█████▋ | 295/520 [18:22<14:25, 3.85s/it] 57%|█████▋ | 296/520 [18:26<14:22, 3.85s/it] {'loss': 1.1948, 'grad_norm': 0.0010487248170156604, 'learning_rate': 0.012395277334996044, 'epoch': 0.57} + 57%|█████▋ | 296/520 [18:26<14:22, 3.85s/it] 57%|█████▋ | 297/520 [18:30<14:19, 3.85s/it] {'loss': 1.3298, 'grad_norm': 0.0010979976215577064, 'learning_rate': 0.012303249223814209, 'epoch': 0.57} + 57%|█████▋ | 297/520 [18:30<14:19, 3.85s/it] 57%|█████▋ | 298/520 [18:33<14:15, 3.85s/it] {'loss': 1.2791, 'grad_norm': 0.0009144195670493575, 'learning_rate': 0.012211325892556281, 'epoch': 0.57} + 57%|█████▋ | 298/520 [18:33<14:15, 3.85s/it] 57%|█████▊ | 299/520 [18:37<14:12, 3.86s/it] {'loss': 1.2617, 'grad_norm': 0.0009611333143505522, 'learning_rate': 0.012119510912823959, 'epoch': 0.57} + 57%|█████▊ | 299/520 [18:37<14:12, 3.86s/it] 58%|█████▊ | 300/520 [18:41<14:10, 3.87s/it] {'loss': 1.3266, 'grad_norm': 0.0010110199456112878, 'learning_rate': 0.01202780785200904, 'epoch': 0.58} + 58%|█████▊ | 300/520 [18:41<14:10, 3.87s/it] 58%|█████▊ | 301/520 [18:45<14:09, 3.88s/it] {'loss': 1.3274, 'grad_norm': 0.0010207852372854417, 'learning_rate': 0.011936220273154797, 'epoch': 0.58} + 58%|█████▊ | 301/520 [18:45<14:09, 3.88s/it] 58%|█████▊ | 302/520 [18:49<14:05, 3.88s/it] {'loss': 1.2544, 'grad_norm': 0.0009583056815779603, 'learning_rate': 0.011844751734817565, 'epoch': 0.58} + 58%|█████▊ | 302/520 [18:49<14:05, 3.88s/it] 58%|█████▊ | 303/520 [18:53<14:01, 3.88s/it] {'loss': 1.2392, 'grad_norm': 0.0011193722005337443, 'learning_rate': 0.011753405790928456, 'epoch': 0.58} + 58%|█████▊ | 303/520 [18:53<14:01, 3.88s/it] 58%|█████▊ | 304/520 [18:57<14:15, 3.96s/it] {'loss': 1.1759, 'grad_norm': 0.00102077501348105, 'learning_rate': 0.011662185990655284, 'epoch': 0.58} + 58%|█████▊ | 304/520 [18:57<14:15, 3.96s/it] 59%|█████▊ | 305/520 [19:01<14:05, 3.93s/it] {'loss': 1.3405, 'grad_norm': 0.0010905912367058102, 'learning_rate': 0.011571095878264659, 'epoch': 0.59} + 59%|█████▊ | 305/520 [19:01<14:05, 3.93s/it] 59%|█████▉ | 306/520 [19:05<13:49, 3.87s/it] {'loss': 1.2884, 'grad_norm': 0.0010170239132675932, 'learning_rate': 0.011480138992984276, 'epoch': 0.59} + 59%|█████▉ | 306/520 [19:05<13:49, 3.87s/it] 59%|█████▉ | 307/520 [19:08<13:30, 3.80s/it] {'loss': 1.224, 'grad_norm': 0.0009654142268637024, 'learning_rate': 0.011389318868865407, 'epoch': 0.59} + 59%|█████▉ | 307/520 [19:08<13:30, 3.80s/it] 59%|█████▉ | 308/520 [19:12<13:20, 3.78s/it] {'loss': 1.3497, 'grad_norm': 0.0009237520083747198, 'learning_rate': 0.011298639034645595, 'epoch': 0.59} + 59%|█████▉ | 308/520 [19:12<13:20, 3.78s/it] 59%|█████▉ | 309/520 [19:16<13:10, 3.75s/it] {'loss': 1.2388, 'grad_norm': 0.0009407987040018813, 'learning_rate': 0.011208103013611536, 'epoch': 0.59} + 59%|█████▉ | 309/520 [19:16<13:10, 3.75s/it] 60%|█████▉ | 310/520 [19:19<13:02, 3.73s/it] {'loss': 1.2122, 'grad_norm': 0.0009981516715314406, 'learning_rate': 0.011117714323462187, 'epoch': 0.6} + 60%|█████▉ | 310/520 [19:19<13:02, 3.73s/it] 60%|█████▉ | 311/520 [19:23<12:54, 3.71s/it] {'loss': 1.2213, 'grad_norm': 0.001056675115375442, 'learning_rate': 0.011027476476172091, 'epoch': 0.6} + 60%|█████▉ | 311/520 [19:23<12:54, 3.71s/it] 60%|██████ | 312/520 [19:27<12:50, 3.70s/it] {'loss': 1.1914, 'grad_norm': 0.001060459843698656, 'learning_rate': 0.010937392977854924, 'epoch': 0.6} + 60%|██████ | 312/520 [19:27<12:50, 3.70s/it] 60%|██████ | 313/520 [19:30<12:45, 3.70s/it] {'loss': 1.162, 'grad_norm': 0.0009262636819004037, 'learning_rate': 0.01084746732862726, 'epoch': 0.6} + 60%|██████ | 313/520 [19:30<12:45, 3.70s/it] 60%|██████ | 314/520 [19:34<13:08, 3.83s/it] {'loss': 1.2148, 'grad_norm': 0.001002569984896748, 'learning_rate': 0.010757703022472587, 'epoch': 0.6} + 60%|██████ | 314/520 [19:34<13:08, 3.83s/it] 61%|██████ | 315/520 [19:38<13:07, 3.84s/it] {'loss': 1.2238, 'grad_norm': 0.001067081781267858, 'learning_rate': 0.010668103547105554, 'epoch': 0.61} + 61%|██████ | 315/520 [19:38<13:07, 3.84s/it] 61%|██████ | 316/520 [19:43<13:35, 4.00s/it] {'loss': 1.2011, 'grad_norm': 0.0010830851512692937, 'learning_rate': 0.010578672383836436, 'epoch': 0.61} + 61%|██████ | 316/520 [19:43<13:35, 4.00s/it] 61%|██████ | 317/520 [19:47<13:23, 3.96s/it] {'loss': 1.1929, 'grad_norm': 0.0009396370021223601, 'learning_rate': 0.010489413007435904, 'epoch': 0.61} + 61%|██████ | 317/520 [19:47<13:23, 3.96s/it] 61%|██████ | 318/520 [19:50<13:15, 3.94s/it] {'loss': 1.3126, 'grad_norm': 0.0010667020529355543, 'learning_rate': 0.010400328885999987, 'epoch': 0.61} + 61%|██████ | 318/520 [19:50<13:15, 3.94s/it] 61%|██████▏ | 319/520 [19:55<13:29, 4.03s/it] {'loss': 1.1868, 'grad_norm': 0.0009162772411876583, 'learning_rate': 0.010311423480815334, 'epoch': 0.61} + 61%|██████▏ | 319/520 [19:55<13:29, 4.03s/it] 62%|██████▏ | 320/520 [19:59<13:15, 3.98s/it] {'loss': 1.1224, 'grad_norm': 0.0010407504891551272, 'learning_rate': 0.010222700246224735, 'epoch': 0.62} + 62%|██████▏ | 320/520 [19:59<13:15, 3.98s/it] 62%|██████▏ | 321/520 [20:02<13:04, 3.94s/it] {'loss': 1.3347, 'grad_norm': 0.0009863494297320972, 'learning_rate': 0.010134162629492896, 'epoch': 0.62} + 62%|██████▏ | 321/520 [20:02<13:04, 3.94s/it] 62%|██████▏ | 322/520 [20:06<12:54, 3.91s/it] {'loss': 1.1207, 'grad_norm': 0.0009718939124940094, 'learning_rate': 0.010045814070672498, 'epoch': 0.62} + 62%|██████▏ | 322/520 [20:06<12:54, 3.91s/it] 62%|██████▏ | 323/520 [20:10<12:47, 3.90s/it] {'loss': 1.2051, 'grad_norm': 0.0010018919830707006, 'learning_rate': 0.00995765800247054, 'epoch': 0.62} + 62%|██████▏ | 323/520 [20:10<12:47, 3.90s/it] 62%|██████▏ | 324/520 [20:14<12:40, 3.88s/it] {'loss': 1.2752, 'grad_norm': 0.0010422507148508027, 'learning_rate': 0.00986969785011497, 'epoch': 0.62} + 62%|██████▏ | 324/520 [20:14<12:40, 3.88s/it] 62%|██████▎ | 325/520 [20:18<12:35, 3.87s/it] {'loss': 1.2632, 'grad_norm': 0.0010705675970890405, 'learning_rate': 0.00978193703122159, 'epoch': 0.62} + 62%|██████▎ | 325/520 [20:18<12:35, 3.87s/it] 63%|██████▎ | 326/520 [20:22<12:28, 3.86s/it] {'loss': 1.2691, 'grad_norm': 0.001062906352300604, 'learning_rate': 0.00969437895566128, 'epoch': 0.63} + 63%|██████▎ | 326/520 [20:22<12:28, 3.86s/it] 63%|██████▎ | 327/520 [20:25<12:22, 3.85s/it] {'loss': 1.2303, 'grad_norm': 0.0010300650352991895, 'learning_rate': 0.009607027025427487, 'epoch': 0.63} + 63%|██████▎ | 327/520 [20:25<12:22, 3.85s/it] 63%|██████▎ | 328/520 [20:29<12:18, 3.85s/it] {'loss': 1.3096, 'grad_norm': 0.0010642960255325693, 'learning_rate': 0.009519884634504074, 'epoch': 0.63} + 63%|██████▎ | 328/520 [20:29<12:18, 3.85s/it] 63%|██████▎ | 329/520 [20:33<12:15, 3.85s/it] {'loss': 1.1907, 'grad_norm': 0.000927023418039672, 'learning_rate': 0.009432955168733432, 'epoch': 0.63} + 63%|██████▎ | 329/520 [20:33<12:15, 3.85s/it] 63%|██████▎ | 330/520 [20:37<12:12, 3.85s/it] {'loss': 1.2763, 'grad_norm': 0.0010010215853973733, 'learning_rate': 0.00934624200568492, 'epoch': 0.63} + 63%|██████▎ | 330/520 [20:37<12:12, 3.85s/it] 64%|██████▎ | 331/520 [20:41<12:06, 3.84s/it] {'loss': 1.2266, 'grad_norm': 0.0010772127134450783, 'learning_rate': 0.009259748514523655, 'epoch': 0.64} + 64%|██████▎ | 331/520 [20:41<12:06, 3.84s/it] 64%|██████▍ | 332/520 [20:45<12:02, 3.84s/it] {'loss': 1.2523, 'grad_norm': 0.0009112429505284896, 'learning_rate': 0.00917347805587958, 'epoch': 0.64} + 64%|██████▍ | 332/520 [20:45<12:02, 3.84s/it] 64%|██████▍ | 333/520 [20:49<11:57, 3.84s/it] {'loss': 1.3554, 'grad_norm': 0.001079187110601605, 'learning_rate': 0.009087433981716912, 'epoch': 0.64} + 64%|██████▍ | 333/520 [20:49<11:57, 3.84s/it] 64%|██████▍ | 334/520 [20:52<11:56, 3.85s/it] {'loss': 1.2827, 'grad_norm': 0.0010902944309576273, 'learning_rate': 0.009001619635203888, 'epoch': 0.64} + 64%|██████▍ | 334/520 [20:52<11:56, 3.85s/it] 64%|██████▍ | 335/520 [20:56<11:50, 3.84s/it] {'loss': 1.2753, 'grad_norm': 0.0009775126003367498, 'learning_rate': 0.008916038350582876, 'epoch': 0.64} + 64%|██████▍ | 335/520 [20:56<11:50, 3.84s/it] 65%|██████▍ | 336/520 [21:00<11:45, 3.83s/it] {'loss': 1.1884, 'grad_norm': 0.0011790475354697112, 'learning_rate': 0.00883069345304083, 'epoch': 0.65} + 65%|██████▍ | 336/520 [21:00<11:45, 3.83s/it] 65%|██████▍ | 337/520 [21:04<11:41, 3.83s/it] {'loss': 1.1782, 'grad_norm': 0.0010939840237006054, 'learning_rate': 0.008745588258580083, 'epoch': 0.65} + 65%|██████▍ | 337/520 [21:04<11:41, 3.83s/it] 65%|██████▌ | 338/520 [21:08<11:38, 3.84s/it] {'loss': 1.2853, 'grad_norm': 0.0010482298090016223, 'learning_rate': 0.008660726073889511, 'epoch': 0.65} + 65%|██████▌ | 338/520 [21:08<11:38, 3.84s/it] 65%|██████▌ | 339/520 [21:12<11:34, 3.84s/it] {'loss': 1.2233, 'grad_norm': 0.001094255689375324, 'learning_rate': 0.008576110196216057, 'epoch': 0.65} + 65%|██████▌ | 339/520 [21:12<11:34, 3.84s/it] 65%|██████▌ | 340/520 [21:15<11:31, 3.84s/it] {'loss': 1.2087, 'grad_norm': 0.0010360714816863763, 'learning_rate': 0.008491743913236628, 'epoch': 0.65} + 65%|██████▌ | 340/520 [21:15<11:31, 3.84s/it] 66%|██████▌ | 341/520 [21:19<11:27, 3.84s/it] {'loss': 1.248, 'grad_norm': 0.0010860696654090994, 'learning_rate': 0.008407630502930323, 'epoch': 0.66} + 66%|██████▌ | 341/520 [21:19<11:27, 3.84s/it] 66%|██████▌ | 342/520 [21:23<11:22, 3.83s/it] {'loss': 1.2488, 'grad_norm': 0.0011366358070904828, 'learning_rate': 0.008323773233451114, 'epoch': 0.66} + 66%|██████▌ | 342/520 [21:23<11:22, 3.83s/it] 66%|██████▌ | 343/520 [21:27<11:17, 3.83s/it] {'loss': 1.1745, 'grad_norm': 0.0008214189194782609, 'learning_rate': 0.00824017536300082, 'epoch': 0.66} + 66%|██████▌ | 343/520 [21:27<11:17, 3.83s/it] 66%|██████▌ | 344/520 [21:31<11:14, 3.83s/it] {'loss': 1.1964, 'grad_norm': 0.0010010879087531078, 'learning_rate': 0.008156840139702554, 'epoch': 0.66} + 66%|██████▌ | 344/520 [21:31<11:14, 3.83s/it] 66%|██████▋ | 345/520 [21:35<11:12, 3.84s/it] {'loss': 1.3007, 'grad_norm': 0.0010787562520281585, 'learning_rate': 0.008073770801474494, 'epoch': 0.66} + 66%|██████▋ | 345/520 [21:35<11:12, 3.84s/it] 67%|██████▋ | 346/520 [21:38<11:10, 3.86s/it] {'loss': 1.2052, 'grad_norm': 0.0009769458785245228, 'learning_rate': 0.00799097057590407, 'epoch': 0.67} + 67%|██████▋ | 346/520 [21:38<11:10, 3.86s/it] 67%|██████▋ | 347/520 [21:42<11:11, 3.88s/it] {'loss': 1.2151, 'grad_norm': 0.0009387815105766709, 'learning_rate': 0.007908442680122597, 'epoch': 0.67} + 67%|██████▋ | 347/520 [21:42<11:11, 3.88s/it]Token indices sequence length is longer than the specified maximum sequence length for this model (2778 > 2048). Running this sequence through the model will result in indexing errors + 67%|██████▋ | 348/520 [21:46<11:07, 3.88s/it] {'loss': 1.1747, 'grad_norm': 0.0012367626311997981, 'learning_rate': 0.00782619032068023, 'epoch': 0.67} + 67%|██████▋ | 348/520 [21:46<11:07, 3.88s/it] 67%|██████▋ | 349/520 [21:50<11:05, 3.89s/it] {'loss': 1.2047, 'grad_norm': 0.001105405813482122, 'learning_rate': 0.007744216693421403, 'epoch': 0.67} + 67%|██████▋ | 349/520 [21:50<11:05, 3.89s/it] 67%|██████▋ | 350/520 [21:54<10:59, 3.88s/it] {'loss': 1.2515, 'grad_norm': 0.0010856967729312196, 'learning_rate': 0.007662524983360665, 'epoch': 0.67} + 67%|██████▋ | 350/520 [21:54<10:59, 3.88s/it] 68%|██████▊ | 351/520 [21:58<10:46, 3.82s/it] {'loss': 1.1638, 'grad_norm': 0.0010158028638521475, 'learning_rate': 0.007581118364558889, 'epoch': 0.68} + 68%|██████▊ | 351/520 [21:58<10:46, 3.82s/it] 68%|██████▊ | 352/520 [22:01<10:34, 3.77s/it] {'loss': 1.2796, 'grad_norm': 0.0009684680605428732, 'learning_rate': 0.007500000000000003, 'epoch': 0.68} + 68%|██████▊ | 352/520 [22:01<10:34, 3.77s/it] 68%|██████▊ | 353/520 [22:05<10:25, 3.75s/it] {'loss': 1.179, 'grad_norm': 0.0008367084503974196, 'learning_rate': 0.007419173041468043, 'epoch': 0.68} + 68%|██████▊ | 353/520 [22:05<10:25, 3.75s/it] 68%|██████▊ | 354/520 [22:09<10:16, 3.71s/it] {'loss': 1.2735, 'grad_norm': 0.0009233511031684456, 'learning_rate': 0.007338640629424719, 'epoch': 0.68} + 68%|██████▊ | 354/520 [22:09<10:16, 3.71s/it] 68%|██████▊ | 355/520 [22:12<10:10, 3.70s/it] {'loss': 1.2356, 'grad_norm': 0.0010678297407360228, 'learning_rate': 0.0072584058928873985, 'epoch': 0.68} + 68%|██████▊ | 355/520 [22:12<10:10, 3.70s/it] 68%|██████▊ | 356/520 [22:16<10:04, 3.68s/it] {'loss': 1.238, 'grad_norm': 0.0010922618005099612, 'learning_rate': 0.00717847194930753, 'epoch': 0.68} + 68%|██████▊ | 356/520 [22:16<10:04, 3.68s/it] 69%|██████▊ | 357/520 [22:20<09:58, 3.67s/it] {'loss': 1.268, 'grad_norm': 0.0009883049567465082, 'learning_rate': 0.007098841904449488, 'epoch': 0.69} + 69%|██████▊ | 357/520 [22:20<09:58, 3.67s/it] 69%|██████▉ | 358/520 [22:23<09:53, 3.67s/it] {'loss': 1.1974, 'grad_norm': 0.00103546495114353, 'learning_rate': 0.007019518852269953, 'epoch': 0.69} + 69%|██████▉ | 358/520 [22:23<09:53, 3.67s/it] 69%|██████▉ | 359/520 [22:27<09:49, 3.66s/it] {'loss': 1.217, 'grad_norm': 0.0010530594295878183, 'learning_rate': 0.00694050587479764, 'epoch': 0.69} + 69%|██████▉ | 359/520 [22:27<09:49, 3.66s/it] 69%|██████▉ | 360/520 [22:31<09:46, 3.66s/it] {'loss': 1.2204, 'grad_norm': 0.00098805588996493, 'learning_rate': 0.00686180604201361, 'epoch': 0.69} + 69%|██████▉ | 360/520 [22:31<09:46, 3.66s/it] 69%|██████▉ | 361/520 [22:34<09:42, 3.66s/it] {'loss': 1.2426, 'grad_norm': 0.0009052712931978537, 'learning_rate': 0.006783422411731931, 'epoch': 0.69} + 69%|██████▉ | 361/520 [22:34<09:42, 3.66s/it] 70%|██████▉ | 362/520 [22:38<09:37, 3.66s/it] {'loss': 1.2361, 'grad_norm': 0.001125804406594741, 'learning_rate': 0.006705358029480908, 'epoch': 0.7} + 70%|██████▉ | 362/520 [22:38<09:37, 3.66s/it] 70%|██████▉ | 363/520 [22:42<09:33, 3.65s/it] {'loss': 1.2722, 'grad_norm': 0.0009885093118214615, 'learning_rate': 0.006627615928384743, 'epoch': 0.7} + 70%|██████▉ | 363/520 [22:42<09:33, 3.65s/it] 70%|███████ | 364/520 [22:45<09:30, 3.66s/it] {'loss': 1.246, 'grad_norm': 0.0009706221143329806, 'learning_rate': 0.006550199129045669, 'epoch': 0.7} + 70%|███████ | 364/520 [22:45<09:30, 3.66s/it] 70%|███████ | 365/520 [22:49<09:25, 3.65s/it] {'loss': 1.3186, 'grad_norm': 0.0010374131999102382, 'learning_rate': 0.0064731106394266165, 'epoch': 0.7} + 70%|███████ | 365/520 [22:49<09:25, 3.65s/it] 70%|███████ | 366/520 [22:53<09:22, 3.66s/it] {'loss': 1.3011, 'grad_norm': 0.0009909853966016489, 'learning_rate': 0.006396353454734312, 'epoch': 0.7} + 70%|███████ | 366/520 [22:53<09:22, 3.66s/it] 71%|███████ | 367/520 [22:56<09:18, 3.65s/it] {'loss': 1.2915, 'grad_norm': 0.001089601077916453, 'learning_rate': 0.0063199305573029135, 'epoch': 0.71} + 71%|███████ | 367/520 [22:56<09:18, 3.65s/it] 71%|███████ | 368/520 [23:00<09:14, 3.65s/it] {'loss': 1.1344, 'grad_norm': 0.001075298088789813, 'learning_rate': 0.006243844916478156, 'epoch': 0.71} + 71%|███████ | 368/520 [23:00<09:14, 3.65s/it] 71%|███████ | 369/520 [23:04<09:10, 3.65s/it] {'loss': 1.2162, 'grad_norm': 0.0009110591860983578, 'learning_rate': 0.00616809948850193, 'epoch': 0.71} + 71%|███████ | 369/520 [23:04<09:10, 3.65s/it] 71%|███████ | 370/520 [23:07<09:06, 3.64s/it] {'loss': 1.2005, 'grad_norm': 0.0009980599271397195, 'learning_rate': 0.006092697216397477, 'epoch': 0.71} + 71%|███████ | 370/520 [23:07<09:06, 3.64s/it] 71%|███████▏ | 371/520 [23:11<09:03, 3.64s/it] {'loss': 1.1857, 'grad_norm': 0.0011235359988677662, 'learning_rate': 0.006017641029854996, 'epoch': 0.71} + 71%|███████▏ | 371/520 [23:11<09:03, 3.64s/it] 72%|███████▏ | 372/520 [23:14<09:01, 3.66s/it] {'loss': 1.2697, 'grad_norm': 0.0008705419600067514, 'learning_rate': 0.005942933845117836, 'epoch': 0.72} + 72%|███████▏ | 372/520 [23:14<09:01, 3.66s/it] 72%|███████▏ | 373/520 [23:18<08:57, 3.66s/it] {'loss': 1.1701, 'grad_norm': 0.0010756409786377276, 'learning_rate': 0.00586857856486919, 'epoch': 0.72} + 72%|███████▏ | 373/520 [23:18<08:57, 3.66s/it] 72%|███████▏ | 374/520 [23:22<08:53, 3.66s/it] {'loss': 1.2788, 'grad_norm': 0.0010900885228881212, 'learning_rate': 0.005794578078119291, 'epoch': 0.72} + 72%|███████▏ | 374/520 [23:22<08:53, 3.66s/it] 72%|███████▏ | 375/520 [23:25<08:50, 3.66s/it] {'loss': 1.2015, 'grad_norm': 0.0010775693033181972, 'learning_rate': 0.005720935260093177, 'epoch': 0.72} + 72%|███████▏ | 375/520 [23:25<08:50, 3.66s/it] 72%|███████▏ | 376/520 [23:29<08:46, 3.66s/it] {'loss': 1.3074, 'grad_norm': 0.0009779802322198471, 'learning_rate': 0.005647652972118998, 'epoch': 0.72} + 72%|███████▏ | 376/520 [23:29<08:46, 3.66s/it] 72%|███████▎ | 377/520 [23:33<08:43, 3.66s/it] {'loss': 1.2331, 'grad_norm': 0.001121843133088533, 'learning_rate': 0.00557473406151679, 'epoch': 0.72} + 72%|███████▎ | 377/520 [23:33<08:43, 3.66s/it] 73%|███████▎ | 378/520 [23:36<08:38, 3.65s/it] {'loss': 1.3024, 'grad_norm': 0.0009541063709216088, 'learning_rate': 0.005502181361487904, 'epoch': 0.73} + 73%|███████▎ | 378/520 [23:36<08:38, 3.65s/it] 73%|███████▎ | 379/520 [23:40<08:35, 3.65s/it] {'loss': 1.2593, 'grad_norm': 0.0009673333650461929, 'learning_rate': 0.005429997691004873, 'epoch': 0.73} + 73%|███████▎ | 379/520 [23:40<08:35, 3.65s/it] 73%|███████▎ | 380/520 [23:44<08:30, 3.65s/it] {'loss': 1.2589, 'grad_norm': 0.0009935553329097693, 'learning_rate': 0.00535818585470191, 'epoch': 0.73} + 73%|███████▎ | 380/520 [23:44<08:30, 3.65s/it] 73%|███████▎ | 381/520 [23:47<08:28, 3.66s/it] {'loss': 1.2794, 'grad_norm': 0.0009895920984991158, 'learning_rate': 0.0052867486427659455, 'epoch': 0.73} + 73%|███████▎ | 381/520 [23:47<08:28, 3.66s/it] 73%|███████▎ | 382/520 [23:51<08:25, 3.67s/it] {'loss': 1.2361, 'grad_norm': 0.0009139993902204973, 'learning_rate': 0.005215688830828187, 'epoch': 0.73} + 73%|███████▎ | 382/520 [23:51<08:25, 3.67s/it] 74%|███████▎ | 383/520 [23:55<08:22, 3.67s/it] {'loss': 1.1212, 'grad_norm': 0.0011982018243678232, 'learning_rate': 0.005145009179856295, 'epoch': 0.74} + 74%|███████▎ | 383/520 [23:55<08:22, 3.67s/it] 74%|███████▍ | 384/520 [23:58<08:18, 3.67s/it] {'loss': 1.2361, 'grad_norm': 0.0008671458722054691, 'learning_rate': 0.005074712436047112, 'epoch': 0.74} + 74%|███████▍ | 384/520 [23:58<08:18, 3.67s/it] 74%|███████▍ | 385/520 [24:02<08:18, 3.69s/it] {'loss': 1.2711, 'grad_norm': 0.0009844947067758798, 'learning_rate': 0.005004801330719941, 'epoch': 0.74} + 74%|███████▍ | 385/520 [24:02<08:18, 3.69s/it] 74%|███████▍ | 386/520 [24:06<08:14, 3.69s/it] {'loss': 1.2135, 'grad_norm': 0.0009063626388284338, 'learning_rate': 0.00493527858021045, 'epoch': 0.74} + 74%|███████▍ | 386/520 [49:58<08:14, 3.69s/it] 74%|███████▍ | 387/520 [50:02<17:20:31, 469.41s/it] {'loss': 1.273, 'grad_norm': 0.0009559317911186838, 'learning_rate': 0.004866146885765096, 'epoch': 0.74} + 74%|███████▍ | 387/520 [50:02<17:20:31, 469.41s/it] 75%|███████▍ | 388/520 [50:06<12:05:20, 329.70s/it] {'loss': 1.1807, 'grad_norm': 0.0010136521165996532, 'learning_rate': 0.004797408933436206, 'epoch': 0.75} + 75%|███████▍ | 388/520 [50:06<12:05:20, 329.70s/it] 75%|███████▍ | 389/520 [50:09<8:26:21, 231.92s/it] {'loss': 1.2331, 'grad_norm': 0.0013087750037018172, 'learning_rate': 0.004729067393977597, 'epoch': 0.75} + 75%|███████▍ | 389/520 [50:09<8:26:21, 231.92s/it] 75%|███████▌ | 390/520 [50:13<5:54:16, 163.51s/it] {'loss': 1.2902, 'grad_norm': 0.0009673446150658872, 'learning_rate': 0.004661124922740794, 'epoch': 0.75} + 75%|███████▌ | 390/520 [50:13<5:54:16, 163.51s/it] 75%|███████▌ | 391/520 [50:17<4:08:31, 115.59s/it] {'loss': 1.3449, 'grad_norm': 0.0010094960604910895, 'learning_rate': 0.004593584159571875, 'epoch': 0.75} + 75%|███████▌ | 391/520 [50:17<4:08:31, 115.59s/it] 75%|███████▌ | 392/520 [50:21<2:55:05, 82.08s/it] {'loss': 1.1809, 'grad_norm': 0.0009918992744579315, 'learning_rate': 0.004526447728708909, 'epoch': 0.75} + 75%|███████▌ | 392/520 [50:21<2:55:05, 82.08s/it] 76%|███████▌ | 393/520 [50:25<2:03:59, 58.58s/it] {'loss': 1.1429, 'grad_norm': 0.000876804742224855, 'learning_rate': 0.004459718238679963, 'epoch': 0.76} + 76%|███████▌ | 393/520 [50:25<2:03:59, 58.58s/it] 76%|███████▌ | 394/520 [50:29<1:28:32, 42.16s/it] {'loss': 1.2562, 'grad_norm': 0.0010767160154079027, 'learning_rate': 0.004393398282201788, 'epoch': 0.76} + 76%|███████▌ | 394/520 [50:29<1:28:32, 42.16s/it] 76%|███████▌ | 395/520 [50:32<1:03:55, 30.69s/it] {'loss': 1.2239, 'grad_norm': 0.001112070976784576, 'learning_rate': 0.004327490436079051, 'epoch': 0.76} + 76%|███████▌ | 395/520 [50:32<1:03:55, 30.69s/it] 76%|███████▌ | 396/520 [50:36<46:46, 22.63s/it] {'loss': 1.285, 'grad_norm': 0.001073405027457537, 'learning_rate': 0.004261997261104223, 'epoch': 0.76} + 76%|███████▌ | 396/520 [50:36<46:46, 22.63s/it] 76%|███████▋ | 397/520 [50:40<34:45, 16.96s/it] {'loss': 1.2575, 'grad_norm': 0.0009981483127297726, 'learning_rate': 0.004196921301958104, 'epoch': 0.76} + 76%|███████▋ | 397/520 [50:40<34:45, 16.96s/it] 77%|███████▋ | 398/520 [50:44<26:22, 12.97s/it] {'loss': 1.256, 'grad_norm': 0.001109389282866114, 'learning_rate': 0.00413226508711091, 'epoch': 0.77} + 77%|███████▋ | 398/520 [50:44<26:22, 12.97s/it] 77%|███████▋ | 399/520 [50:47<20:32, 10.19s/it] {'loss': 1.1683, 'grad_norm': 0.0009665282414647108, 'learning_rate': 0.0040680311287240745, 'epoch': 0.77} + 77%|███████▋ | 399/520 [50:47<20:32, 10.19s/it] 77%|███████▋ | 400/520 [50:51<16:29, 8.24s/it] {'loss': 1.2096, 'grad_norm': 0.0009165812794406251, 'learning_rate': 0.004004221922552608, 'epoch': 0.77} + 77%|███████▋ | 400/520 [50:51<16:29, 8.24s/it] 77%|███████▋ | 401/520 [50:55<13:39, 6.89s/it] {'loss': 1.0984, 'grad_norm': 0.0011466026593103702, 'learning_rate': 0.00394083994784814, 'epoch': 0.77} + 77%|███████▋ | 401/520 [50:55<13:39, 6.89s/it] 77%|███████▋ | 402/520 [50:58<11:39, 5.92s/it] {'loss': 1.2296, 'grad_norm': 0.0010683545319274705, 'learning_rate': 0.003877887667262599, 'epoch': 0.77} + 77%|███████▋ | 402/520 [50:58<11:39, 5.92s/it] 78%|███████▊ | 403/520 [51:02<10:17, 5.28s/it] {'loss': 1.2518, 'grad_norm': 0.0011266644688464975, 'learning_rate': 0.003815367526752516, 'epoch': 0.78} + 78%|███████▊ | 403/520 [51:02<10:17, 5.28s/it] 78%|███████▊ | 404/520 [51:06<09:21, 4.84s/it] {'loss': 1.1737, 'grad_norm': 0.0012637446295185799, 'learning_rate': 0.003753281955483985, 'epoch': 0.78} + 78%|███████▊ | 404/520 [51:06<09:21, 4.84s/it] 78%|███████▊ | 405/520 [51:10<08:43, 4.56s/it] {'loss': 1.1888, 'grad_norm': 0.0009817903375638549, 'learning_rate': 0.0036916333657383026, 'epoch': 0.78} + 78%|███████▊ | 405/520 [51:10<08:43, 4.56s/it] 78%|███████▊ | 406/520 [51:14<08:14, 4.34s/it] {'loss': 1.1238, 'grad_norm': 0.0012206373442656736, 'learning_rate': 0.0036304241528182027, 'epoch': 0.78} + 78%|███████▊ | 406/520 [51:14<08:14, 4.34s/it] 78%|███████▊ | 407/520 [51:18<07:52, 4.18s/it] {'loss': 1.3204, 'grad_norm': 0.0010577014896870423, 'learning_rate': 0.003569656694954838, 'epoch': 0.78} + 78%|███████▊ | 407/520 [51:18<07:52, 4.18s/it] 78%|███████▊ | 408/520 [51:21<07:36, 4.08s/it] {'loss': 1.2459, 'grad_norm': 0.001173312899703609, 'learning_rate': 0.0035093333532153314, 'epoch': 0.78} + 78%|███████▊ | 408/520 [51:21<07:36, 4.08s/it] 79%|███████▊ | 409/520 [51:25<07:23, 4.00s/it] {'loss': 1.3752, 'grad_norm': 0.0011227868701239418, 'learning_rate': 0.0034494564714110582, 'epoch': 0.79} + 79%|███████▊ | 409/520 [51:25<07:23, 4.00s/it] 79%|███████▉ | 410/520 [51:29<07:15, 3.96s/it] {'loss': 1.1039, 'grad_norm': 0.0010917209366866486, 'learning_rate': 0.003390028376006589, 'epoch': 0.79} + 79%|███████▉ | 410/520 [51:29<07:15, 3.96s/it] 79%|███████▉ | 411/520 [51:33<07:06, 3.92s/it] {'loss': 1.3412, 'grad_norm': 0.0011003014571270484, 'learning_rate': 0.0033310513760292787, 'epoch': 0.79} + 79%|███████▉ | 411/520 [51:33<07:06, 3.92s/it] 79%|███████▉ | 412/520 [51:37<07:01, 3.90s/it] {'loss': 1.2553, 'grad_norm': 0.0009949102286188954, 'learning_rate': 0.0032725277629795527, 'epoch': 0.79} + 79%|███████▉ | 412/520 [51:37<07:01, 3.90s/it] 79%|███████▉ | 413/520 [51:41<06:53, 3.87s/it] {'loss': 1.2165, 'grad_norm': 0.0009744867427823784, 'learning_rate': 0.0032144598107418975, 'epoch': 0.79} + 79%|███████▉ | 413/520 [51:41<06:53, 3.87s/it] 80%|███████▉ | 414/520 [51:44<06:50, 3.87s/it] {'loss': 1.0158, 'grad_norm': 0.0008841239811144257, 'learning_rate': 0.00315684977549647, 'epoch': 0.8} + 80%|███████▉ | 414/520 [51:44<06:50, 3.87s/it] 80%|███████▉ | 415/520 [51:48<06:45, 3.86s/it] {'loss': 1.236, 'grad_norm': 0.0010093177226886206, 'learning_rate': 0.003099699895631474, 'epoch': 0.8} + 80%|███████▉ | 415/520 [51:48<06:45, 3.86s/it] 80%|████████ | 416/520 [51:52<06:41, 3.86s/it] {'loss': 1.1431, 'grad_norm': 0.0011779675160632677, 'learning_rate': 0.0030430123916561675, 'epoch': 0.8} + 80%|████████ | 416/520 [51:52<06:41, 3.86s/it] 80%|████████ | 417/520 [51:56<06:35, 3.84s/it] {'loss': 1.2949, 'grad_norm': 0.0010077842602906617, 'learning_rate': 0.002986789466114582, 'epoch': 0.8} + 80%|████████ | 417/520 [51:56<06:35, 3.84s/it] 80%|████████ | 418/520 [52:00<06:32, 3.85s/it] {'loss': 1.2939, 'grad_norm': 0.0009692174614131077, 'learning_rate': 0.0029310333034999747, 'epoch': 0.8} + 80%|████████ | 418/520 [52:00<06:32, 3.85s/it] 81%|████████ | 419/520 [52:04<06:25, 3.82s/it] {'loss': 1.2985, 'grad_norm': 0.00113036625538842, 'learning_rate': 0.0028757460701699217, 'epoch': 0.81} + 81%|████████ | 419/520 [52:04<06:25, 3.82s/it] 81%|████████ | 420/520 [52:07<06:20, 3.81s/it] {'loss': 1.1825, 'grad_norm': 0.0011158456741614292, 'learning_rate': 0.0028209299142621523, 'epoch': 0.81} + 81%|████████ | 420/520 [52:07<06:20, 3.81s/it] 81%|████████ | 421/520 [52:11<06:13, 3.77s/it] {'loss': 1.1193, 'grad_norm': 0.001124687013168201, 'learning_rate': 0.0027665869656110973, 'epoch': 0.81} + 81%|████████ | 421/520 [52:11<06:13, 3.77s/it] 81%|████████ | 422/520 [52:15<06:06, 3.74s/it] {'loss': 1.2478, 'grad_norm': 0.0011170641031315151, 'learning_rate': 0.0027127193356651213, 'epoch': 0.81} + 81%|████████ | 422/520 [52:15<06:06, 3.74s/it] 81%|████████▏ | 423/520 [52:18<06:00, 3.72s/it] {'loss': 1.2146, 'grad_norm': 0.0011569058165777485, 'learning_rate': 0.0026593291174044995, 'epoch': 0.81} + 81%|████████▏ | 423/520 [52:18<06:00, 3.72s/it] 82%|████████▏ | 424/520 [52:22<05:57, 3.72s/it] {'loss': 1.281, 'grad_norm': 0.0008887611284222406, 'learning_rate': 0.0026064183852600796, 'epoch': 0.82} + 82%|████████▏ | 424/520 [52:22<05:57, 3.72s/it] 82%|████████▏ | 425/520 [52:26<05:50, 3.69s/it] {'loss': 1.2132, 'grad_norm': 0.0010143542659560053, 'learning_rate': 0.0025539891950326875, 'epoch': 0.82} + 82%|████████▏ | 425/520 [52:26<05:50, 3.69s/it] 82%|████████▏ | 426/520 [52:29<05:46, 3.69s/it] {'loss': 1.2717, 'grad_norm': 0.0014135446524518012, 'learning_rate': 0.0025020435838132675, 'epoch': 0.82} + 82%|████████▏ | 426/520 [52:29<05:46, 3.69s/it] 82%|████████▏ | 427/520 [52:33<05:41, 3.67s/it] {'loss': 1.1455, 'grad_norm': 0.0010394490698979507, 'learning_rate': 0.0024505835699037038, 'epoch': 0.82} + 82%|████████▏ | 427/520 [52:33<05:41, 3.67s/it] 82%|████████▏ | 428/520 [52:37<05:36, 3.66s/it] {'loss': 1.1503, 'grad_norm': 0.0011256411802948272, 'learning_rate': 0.002399611152738429, 'epoch': 0.82} + 82%|████████▏ | 428/520 [52:37<05:36, 3.66s/it] 82%|████████▎ | 429/520 [52:40<05:32, 3.65s/it] {'loss': 1.2595, 'grad_norm': 0.0010498983185541467, 'learning_rate': 0.0023491283128067174, 'epoch': 0.82} + 82%|████████▎ | 429/520 [52:40<05:32, 3.65s/it]Token indices sequence length is longer than the specified maximum sequence length for this model (2076 > 2048). Running this sequence through the model will result in indexing errors + 83%|████████▎ | 430/520 [52:44<05:28, 3.65s/it] {'loss': 1.2519, 'grad_norm': 0.0010166042443828269, 'learning_rate': 0.002299137011575738, 'epoch': 0.83} + 83%|████████▎ | 430/520 [52:44<05:28, 3.65s/it] 83%|████████▎ | 431/520 [52:48<05:25, 3.65s/it] {'loss': 1.1743, 'grad_norm': 0.0009819572811085497, 'learning_rate': 0.002249639191414363, 'epoch': 0.83} + 83%|████████▎ | 431/520 [52:48<05:25, 3.65s/it] 83%|████████▎ | 432/520 [52:51<05:21, 3.66s/it] {'loss': 1.1643, 'grad_norm': 0.0010699043078423193, 'learning_rate': 0.002200636775517666, 'epoch': 0.83} + 83%|████████▎ | 432/520 [52:51<05:21, 3.66s/it] 83%|████████▎ | 433/520 [52:55<05:17, 3.65s/it] {'loss': 1.2927, 'grad_norm': 0.0010491307014260248, 'learning_rate': 0.00215213166783223, 'epoch': 0.83} + 83%|████████▎ | 433/520 [52:55<05:17, 3.65s/it] 83%|████████▎ | 434/520 [52:59<05:15, 3.67s/it] {'loss': 1.0591, 'grad_norm': 0.0010867835229959406, 'learning_rate': 0.0021041257529821455, 'epoch': 0.83} + 83%|████████▎ | 434/520 [52:59<05:15, 3.67s/it] 84%|████████▎ | 435/520 [53:02<05:10, 3.66s/it] {'loss': 1.3328, 'grad_norm': 0.0011506013535946295, 'learning_rate': 0.002056620896195804, 'epoch': 0.84} + 84%|████████▎ | 435/520 [53:02<05:10, 3.66s/it] 84%|████████▍ | 436/520 [53:06<05:05, 3.64s/it] {'loss': 1.1331, 'grad_norm': 0.00113085914044143, 'learning_rate': 0.0020096189432334192, 'epoch': 0.84} + 84%|████████▍ | 436/520 [53:06<05:05, 3.64s/it] 84%|████████▍ | 437/520 [53:10<05:01, 3.64s/it] {'loss': 1.3475, 'grad_norm': 0.001060993442542984, 'learning_rate': 0.001963121720315304, 'epoch': 0.84} + 84%|████████▍ | 437/520 [53:10<05:01, 3.64s/it] 84%|████████▍ | 438/520 [53:13<04:58, 3.64s/it] {'loss': 1.1637, 'grad_norm': 0.0011305833458570634, 'learning_rate': 0.00191713103405092, 'epoch': 0.84} + 84%|████████▍ | 438/520 [53:13<04:58, 3.64s/it] 84%|████████▍ | 439/520 [53:17<04:55, 3.64s/it] {'loss': 1.1637, 'grad_norm': 0.0008217239529941164, 'learning_rate': 0.0018716486713686946, 'epoch': 0.84} + 84%|████████▍ | 439/520 [53:17<04:55, 3.64s/it] 85%|████████▍ | 440/520 [53:20<04:51, 3.65s/it] {'loss': 1.2114, 'grad_norm': 0.0010795486347380011, 'learning_rate': 0.00182667639944657, 'epoch': 0.85} + 85%|████████▍ | 440/520 [53:20<04:51, 3.65s/it] 85%|████████▍ | 441/520 [53:24<04:48, 3.66s/it] {'loss': 1.1726, 'grad_norm': 0.000995059907088892, 'learning_rate': 0.0017822159656433639, 'epoch': 0.85} + 85%|████████▍ | 441/520 [53:24<04:48, 3.66s/it] 85%|████████▌ | 442/520 [53:28<04:45, 3.65s/it] {'loss': 1.2692, 'grad_norm': 0.0011453804554356834, 'learning_rate': 0.001738269097430855, 'epoch': 0.85} + 85%|████████▌ | 442/520 [53:28<04:45, 3.65s/it] 85%|████████▌ | 443/520 [53:31<04:41, 3.65s/it] {'loss': 1.2711, 'grad_norm': 0.0010026585058462041, 'learning_rate': 0.0016948375023266742, 'epoch': 0.85} + 85%|████████▌ | 443/520 [53:31<04:41, 3.65s/it] 85%|████████▌ | 444/520 [53:35<04:38, 3.66s/it] {'loss': 1.2331, 'grad_norm': 0.0009375430907612804, 'learning_rate': 0.0016519228678279717, 'epoch': 0.85} + 85%|████████▌ | 444/520 [53:35<04:38, 3.66s/it] 86%|████████▌ | 445/520 [53:39<04:33, 3.65s/it] {'loss': 1.1563, 'grad_norm': 0.000982293480444834, 'learning_rate': 0.0016095268613458302, 'epoch': 0.86} + 86%|████████▌ | 445/520 [53:39<04:33, 3.65s/it] 86%|████████▌ | 446/520 [53:42<04:30, 3.66s/it] {'loss': 1.2581, 'grad_norm': 0.0008959759400525094, 'learning_rate': 0.001567651130140486, 'epoch': 0.86} + 86%|████████▌ | 446/520 [53:42<04:30, 3.66s/it] 86%|████████▌ | 447/520 [53:46<04:27, 3.66s/it] {'loss': 1.2314, 'grad_norm': 0.0010104701065706651, 'learning_rate': 0.0015262973012573394, 'epoch': 0.86} + 86%|████████▌ | 447/520 [53:46<04:27, 3.66s/it] 86%|████████▌ | 448/520 [53:50<04:23, 3.66s/it] {'loss': 1.2319, 'grad_norm': 0.001138736672416326, 'learning_rate': 0.0014854669814637145, 'epoch': 0.86} + 86%|████████▌ | 448/520 [53:50<04:23, 3.66s/it] 86%|████████▋ | 449/520 [53:53<04:19, 3.66s/it] {'loss': 1.2222, 'grad_norm': 0.0010047709401080453, 'learning_rate': 0.0014451617571864528, 'epoch': 0.86} + 86%|████████▋ | 449/520 [53:53<04:19, 3.66s/it] 87%|████████▋ | 450/520 [53:57<04:16, 3.66s/it] {'loss': 1.2624, 'grad_norm': 0.0010672755399010912, 'learning_rate': 0.0014053831944502509, 'epoch': 0.87} + 87%|████████▋ | 450/520 [53:57<04:16, 3.66s/it] 87%|████████▋ | 451/520 [54:01<04:12, 3.66s/it] {'loss': 1.257, 'grad_norm': 0.0010621773059563155, 'learning_rate': 0.0013661328388168358, 'epoch': 0.87} + 87%|████████▋ | 451/520 [54:01<04:12, 3.66s/it] 87%|████████▋ | 452/520 [54:04<04:09, 3.67s/it] {'loss': 1.2628, 'grad_norm': 0.0009321076446993207, 'learning_rate': 0.001327412215324903, 'epoch': 0.87} + 87%|████████▋ | 452/520 [54:04<04:09, 3.67s/it] 87%|████████▋ | 453/520 [54:08<04:05, 3.67s/it] {'loss': 1.2499, 'grad_norm': 0.0010082743164604282, 'learning_rate': 0.0012892228284308548, 'epoch': 0.87} + 87%|████████▋ | 453/520 [54:08<04:05, 3.67s/it] 87%|████████▋ | 454/520 [54:12<04:02, 3.68s/it] {'loss': 1.1692, 'grad_norm': 0.0010259252676219283, 'learning_rate': 0.001251566161950357, 'epoch': 0.87} + 87%|████████▋ | 454/520 [54:12<04:02, 3.68s/it] 88%|████████▊ | 455/520 [54:15<03:59, 3.69s/it] {'loss': 1.3062, 'grad_norm': 0.0010531909491636102, 'learning_rate': 0.00121444367900069, 'epoch': 0.88} + 88%|████████▊ | 455/520 [54:15<03:59, 3.69s/it] 88%|████████▊ | 456/520 [54:19<03:55, 3.68s/it] {'loss': 1.254, 'grad_norm': 0.0010926569476875944, 'learning_rate': 0.001177856821943884, 'epoch': 0.88} + 88%|████████▊ | 456/520 [54:19<03:55, 3.68s/it] 88%|████████▊ | 457/520 [54:23<03:52, 3.69s/it] {'loss': 1.0987, 'grad_norm': 0.000923909039787047, 'learning_rate': 0.0011418070123306989, 'epoch': 0.88} + 88%|████████▊ | 457/520 [54:23<03:52, 3.69s/it] 88%|████████▊ | 458/520 [54:27<03:48, 3.69s/it] {'loss': 1.3653, 'grad_norm': 0.001102461075984995, 'learning_rate': 0.0011062956508453702, 'epoch': 0.88} + 88%|████████▊ | 458/520 [54:27<03:48, 3.69s/it] 88%|████████▊ | 459/520 [54:30<03:45, 3.70s/it] {'loss': 1.2853, 'grad_norm': 0.001053106070307233, 'learning_rate': 0.0010713241172511966, 'epoch': 0.88} + 88%|████████▊ | 459/520 [54:30<03:45, 3.70s/it] 88%|████████▊ | 460/520 [54:34<03:41, 3.69s/it] {'loss': 1.186, 'grad_norm': 0.0011162256579475946, 'learning_rate': 0.001036893770336938, 'epoch': 0.88} + 88%|████████▊ | 460/520 [54:34<03:41, 3.69s/it] 89%|████████▊ | 461/520 [54:38<03:37, 3.68s/it] {'loss': 1.1853, 'grad_norm': 0.0007247565269524276, 'learning_rate': 0.0010030059478640023, 'epoch': 0.89} + 89%|████████▊ | 461/520 [54:38<03:37, 3.68s/it] 89%|████████▉ | 462/520 [54:41<03:32, 3.67s/it] {'loss': 1.3045, 'grad_norm': 0.0009548242324750761, 'learning_rate': 0.0009696619665144901, 'epoch': 0.89} + 89%|████████▉ | 462/520 [54:41<03:32, 3.67s/it] 89%|████████▉ | 463/520 [54:45<03:28, 3.66s/it] {'loss': 1.1673, 'grad_norm': 0.0011080691389801914, 'learning_rate': 0.0009368631218400136, 'epoch': 0.89} + 89%|████████▉ | 463/520 [54:45<03:28, 3.66s/it] 89%|████████▉ | 464/520 [54:49<03:24, 3.66s/it] {'loss': 1.2759, 'grad_norm': 0.0010458710590456996, 'learning_rate': 0.0009046106882113752, 'epoch': 0.89} + 89%|████████▉ | 464/520 [54:49<03:24, 3.66s/it] 89%|████████▉ | 465/520 [54:52<03:21, 3.66s/it] {'loss': 1.3795, 'grad_norm': 0.001047449223986602, 'learning_rate': 0.0008729059187690479, 'epoch': 0.89} + 89%|████████▉ | 465/520 [54:52<03:21, 3.66s/it] 90%|████████▉ | 466/520 [54:56<03:17, 3.66s/it] {'loss': 1.2788, 'grad_norm': 0.0009713359620464108, 'learning_rate': 0.0008417500453744863, 'epoch': 0.9} + 90%|████████▉ | 466/520 [54:56<03:17, 3.66s/it] 90%|████████▉ | 467/520 [54:59<03:13, 3.65s/it] {'loss': 1.1854, 'grad_norm': 0.0009409468856748312, 'learning_rate': 0.0008111442785622596, 'epoch': 0.9} + 90%|████████▉ | 467/520 [54:59<03:13, 3.65s/it] 90%|█████████ | 468/520 [55:03<03:09, 3.65s/it] {'loss': 1.2369, 'grad_norm': 0.001188198417640022, 'learning_rate': 0.0007810898074930245, 'epoch': 0.9} + 90%|█████████ | 468/520 [55:03<03:09, 3.65s/it] 90%|█████████ | 469/520 [55:07<03:08, 3.69s/it] {'loss': 1.304, 'grad_norm': 0.0010865893402428825, 'learning_rate': 0.00075158779990731, 'epoch': 0.9} + 90%|█████████ | 469/520 [55:07<03:08, 3.69s/it] 90%|█████████ | 470/520 [55:11<03:06, 3.74s/it] {'loss': 1.1768, 'grad_norm': 0.0009655923783474294, 'learning_rate': 0.0007226394020801646, 'epoch': 0.9} + 90%|█████████ | 470/520 [55:11<03:06, 3.74s/it] 91%|█████████ | 471/520 [55:15<03:05, 3.78s/it] {'loss': 1.221, 'grad_norm': 0.0011154587362388238, 'learning_rate': 0.0006942457387765976, 'epoch': 0.91} + 91%|█████████ | 471/520 [55:15<03:05, 3.78s/it] 91%|█████████ | 472/520 [55:18<03:00, 3.77s/it] {'loss': 1.1883, 'grad_norm': 0.0010517747763035754, 'learning_rate': 0.0006664079132078882, 'epoch': 0.91} + 91%|█████████ | 472/520 [55:18<03:00, 3.77s/it] 91%|█████████ | 473/520 [55:22<02:55, 3.73s/it] {'loss': 1.2586, 'grad_norm': 0.0011036779064981673, 'learning_rate': 0.0006391270069887289, 'epoch': 0.91} + 91%|█████████ | 473/520 [55:22<02:55, 3.73s/it] 91%|█████████ | 474/520 [55:26<02:50, 3.70s/it] {'loss': 1.215, 'grad_norm': 0.000913114109292765, 'learning_rate': 0.0006124040800951835, 'epoch': 0.91} + 91%|█████████ | 474/520 [55:26<02:50, 3.70s/it] 91%|█████████▏| 475/520 [55:29<02:45, 3.68s/it] {'loss': 1.1373, 'grad_norm': 0.0009489257208069559, 'learning_rate': 0.0005862401708235076, 'epoch': 0.91} + 91%|█████████▏| 475/520 [55:29<02:45, 3.68s/it] 92%|█████████▏| 476/520 [55:33<02:41, 3.67s/it] {'loss': 1.233, 'grad_norm': 0.001113423609058057, 'learning_rate': 0.0005606362957498195, 'epoch': 0.92} + 92%|█████████▏| 476/520 [55:33<02:41, 3.67s/it] 92%|█████████▏| 477/520 [55:37<02:37, 3.66s/it] {'loss': 1.2319, 'grad_norm': 0.0011770277433286117, 'learning_rate': 0.000535593449690585, 'epoch': 0.92} + 92%|█████████▏| 477/520 [55:37<02:37, 3.66s/it] 92%|█████████▏| 478/520 [55:40<02:33, 3.65s/it] {'loss': 1.1702, 'grad_norm': 0.0009920618705267464, 'learning_rate': 0.000511112605663977, 'epoch': 0.92} + 92%|█████████▏| 478/520 [55:40<02:33, 3.65s/it] 92%|█████████▏| 479/520 [55:44<02:30, 3.66s/it] {'loss': 1.1997, 'grad_norm': 0.001050898479132669, 'learning_rate': 0.00048719471485205834, 'epoch': 0.92} + 92%|█████████▏| 479/520 [55:44<02:30, 3.66s/it] 92%|█████████▏| 480/520 [55:48<02:26, 3.66s/it] {'loss': 1.2059, 'grad_norm': 0.0009409677533893671, 'learning_rate': 0.0004638407065638322, 'epoch': 0.92} + 92%|█████████▏| 480/520 [55:48<02:26, 3.66s/it] 92%|█████████▎| 481/520 [55:51<02:23, 3.67s/it] {'loss': 1.2011, 'grad_norm': 0.0009147096361355962, 'learning_rate': 0.0004410514881991356, 'epoch': 0.93} + 92%|█████████▎| 481/520 [55:51<02:23, 3.67s/it] 93%|█████████▎| 482/520 [55:55<02:19, 3.68s/it] {'loss': 1.2236, 'grad_norm': 0.0009085434009664605, 'learning_rate': 0.0004188279452133825, 'epoch': 0.93} + 93%|█████████▎| 482/520 [55:55<02:19, 3.68s/it] 93%|█████████▎| 483/520 [55:59<02:16, 3.68s/it] {'loss': 1.2363, 'grad_norm': 0.0010449924911193995, 'learning_rate': 0.00039717094108314974, 'epoch': 0.93} + 93%|█████████▎| 483/520 [55:59<02:16, 3.68s/it] 93%|█████████▎| 484/520 [56:02<02:12, 3.69s/it] {'loss': 1.2502, 'grad_norm': 0.0010320382157586467, 'learning_rate': 0.0003760813172726457, 'epoch': 0.93} + 93%|█████████▎| 484/520 [56:02<02:12, 3.69s/it] 93%|█████████▎| 485/520 [56:06<02:09, 3.70s/it] {'loss': 1.1992, 'grad_norm': 0.0010207329815395956, 'learning_rate': 0.0003555598932009996, 'epoch': 0.93} + 93%|█████████▎| 485/520 [56:06<02:09, 3.70s/it] 93%|█████████▎| 486/520 [56:10<02:05, 3.69s/it] {'loss': 1.3241, 'grad_norm': 0.0010407544508563025, 'learning_rate': 0.0003356074662104319, 'epoch': 0.93} + 93%|█████████▎| 486/520 [56:10<02:05, 3.69s/it] 94%|█████████▎| 487/520 [56:13<02:02, 3.72s/it] {'loss': 1.189, 'grad_norm': 0.0010057248173658048, 'learning_rate': 0.00031622481153527446, 'epoch': 0.94} + 94%|█████████▎| 487/520 [56:13<02:02, 3.72s/it] 94%|█████████▍| 488/520 [56:17<01:58, 3.70s/it] {'loss': 1.139, 'grad_norm': 0.001065657798306652, 'learning_rate': 0.0002974126822718426, 'epoch': 0.94} + 94%|█████████▍| 488/520 [56:17<01:58, 3.70s/it] 94%|█████████▍| 489/520 [56:21<01:54, 3.69s/it] {'loss': 1.2191, 'grad_norm': 0.0008675048284257239, 'learning_rate': 0.00027917180934918515, 'epoch': 0.94} + 94%|█████████▍| 489/520 [56:21<01:54, 3.69s/it] 94%|█████████▍| 490/520 [56:24<01:50, 3.68s/it] {'loss': 1.245, 'grad_norm': 0.0010713788613794274, 'learning_rate': 0.0002615029015006759, 'epoch': 0.94} + 94%|█████████▍| 490/520 [56:24<01:50, 3.68s/it] 94%|█████████▍| 491/520 [56:28<01:46, 3.67s/it] {'loss': 1.2166, 'grad_norm': 0.0011323563492149635, 'learning_rate': 0.00024440664523648015, 'epoch': 0.94} + 94%|█████████▍| 491/520 [56:28<01:46, 3.67s/it] 95%|█████████▍| 492/520 [56:32<01:42, 3.67s/it] {'loss': 1.3213, 'grad_norm': 0.0010637999171094765, 'learning_rate': 0.00022788370481687968, 'epoch': 0.95} + 95%|█████████▍| 492/520 [56:32<01:42, 3.67s/it] 95%|█████████▍| 493/520 [56:35<01:39, 3.67s/it] {'loss': 1.22, 'grad_norm': 0.0010280214146055325, 'learning_rate': 0.0002119347222264617, 'epoch': 0.95} + 95%|█████████▍| 493/520 [56:35<01:39, 3.67s/it] 95%|█████████▌| 494/520 [56:39<01:35, 3.66s/it] {'loss': 1.2436, 'grad_norm': 0.0009596659861884596, 'learning_rate': 0.00019656031714918365, 'epoch': 0.95} + 95%|█████████▌| 494/520 [56:39<01:35, 3.66s/it] 95%|█████████▌| 495/520 [56:43<01:31, 3.67s/it] {'loss': 1.2386, 'grad_norm': 0.001076369629006069, 'learning_rate': 0.0001817610869442793, 'epoch': 0.95} + 95%|█████████▌| 495/520 [56:43<01:31, 3.67s/it] 95%|█████████▌| 496/520 [56:47<01:28, 3.69s/it] {'loss': 1.18, 'grad_norm': 0.0011041579644478021, 'learning_rate': 0.00016753760662307216, 'epoch': 0.95} + 95%|█████████▌| 496/520 [56:47<01:28, 3.69s/it] 96%|█████████▌| 497/520 [56:50<01:24, 3.68s/it] {'loss': 1.1568, 'grad_norm': 0.0008853007745122089, 'learning_rate': 0.00015389042882661019, 'epoch': 0.96} + 96%|█████████▌| 497/520 [56:50<01:24, 3.68s/it] 96%|█████████▌| 498/520 [56:54<01:20, 3.67s/it] {'loss': 1.2275, 'grad_norm': 0.0010664323101820638, 'learning_rate': 0.00014082008380420785, 'epoch': 0.96} + 96%|█████████▌| 498/520 [56:54<01:20, 3.67s/it] 96%|█████████▌| 499/520 [56:58<01:17, 3.68s/it] {'loss': 1.3005, 'grad_norm': 0.0009494016242162053, 'learning_rate': 0.00012832707939284426, 'epoch': 0.96} + 96%|█████████▌| 499/520 [56:58<01:17, 3.68s/it] 96%|█████████▌| 500/520 [57:01<01:13, 3.67s/it] {'loss': 1.3547, 'grad_norm': 0.0012074876749991975, 'learning_rate': 0.00011641190099741904, 'epoch': 0.96} + 96%|█████████▌| 500/520 [57:01<01:13, 3.67s/it] 96%|█████████▋| 501/520 [57:05<01:09, 3.68s/it] {'loss': 1.2005, 'grad_norm': 0.001031010500192741, 'learning_rate': 0.00010507501157190568, 'epoch': 0.96} + 96%|█████████▋| 501/520 [57:05<01:09, 3.68s/it] 97%|█████████▋| 502/520 [57:09<01:06, 3.68s/it] {'loss': 1.2621, 'grad_norm': 0.0010076135151260248, 'learning_rate': 9.431685160136093e-05, 'epoch': 0.97} + 97%|█████████▋| 502/520 [57:09<01:06, 3.68s/it] 97%|█████████▋| 503/520 [57:12<01:02, 3.70s/it] {'loss': 1.1761, 'grad_norm': 0.0009832609463530772, 'learning_rate': 8.413783908480354e-05, 'epoch': 0.97} + 97%|█████████▋| 503/520 [57:12<01:02, 3.70s/it] 97%|█████████▋| 504/520 [57:16<00:59, 3.69s/it] {'loss': 1.2524, 'grad_norm': 0.0012063765894004196, 'learning_rate': 7.453836951897885e-05, 'epoch': 0.97} + 97%|█████████▋| 504/520 [57:16<00:59, 3.69s/it] 97%|█████████▋| 505/520 [57:20<00:55, 3.68s/it] {'loss': 1.274, 'grad_norm': 0.001064295717239553, 'learning_rate': 6.551881588299279e-05, 'epoch': 0.97} + 97%|█████████▋| 505/520 [57:20<00:55, 3.68s/it] 97%|█████████▋| 506/520 [57:23<00:51, 3.67s/it] {'loss': 1.21, 'grad_norm': 0.0011329322074716895, 'learning_rate': 5.707952862381682e-05, 'epoch': 0.97} + 97%|█████████▋| 506/520 [57:23<00:51, 3.67s/it] 98%|█████████▊| 507/520 [57:27<00:47, 3.67s/it] {'loss': 1.3228, 'grad_norm': 0.0008834324669361336, 'learning_rate': 4.9220835642673764e-05, 'epoch': 0.97} + 98%|█████████▊| 507/520 [57:27<00:47, 3.67s/it] 98%|█████████▊| 508/520 [57:31<00:44, 3.67s/it] {'loss': 1.3458, 'grad_norm': 0.0010565365781888543, 'learning_rate': 4.194304228229806e-05, 'epoch': 0.98} + 98%|█████████▊| 508/520 [57:31<00:44, 3.67s/it] 98%|█████████▊| 509/520 [57:34<00:40, 3.72s/it] {'loss': 1.3047, 'grad_norm': 0.0010108007675160534, 'learning_rate': 3.524643131506688e-05, 'epoch': 0.98} + 98%|█████████▊| 509/520 [57:34<00:40, 3.72s/it] 98%|█████████▊| 510/520 [57:38<00:37, 3.77s/it] {'loss': 1.2506, 'grad_norm': 0.001041465330259901, 'learning_rate': 2.913126293202228e-05, 'epoch': 0.98} + 98%|█████████▊| 510/520 [57:38<00:37, 3.77s/it] 98%|█████████▊| 511/520 [57:42<00:33, 3.78s/it] {'loss': 1.2064, 'grad_norm': 0.001036437075080947, 'learning_rate': 2.3597774732750932e-05, 'epoch': 0.98} + 98%|█████████▊| 511/520 [57:42<00:33, 3.78s/it] 98%|█████████▊| 512/520 [57:46<00:30, 3.76s/it] {'loss': 1.1019, 'grad_norm': 0.001034088864042137, 'learning_rate': 1.864618171616483e-05, 'epoch': 0.98} + 98%|█████████▊| 512/520 [57:46<00:30, 3.76s/it] 99%|█████████▊| 513/520 [57:50<00:26, 3.73s/it] {'loss': 1.3066, 'grad_norm': 0.0011771675051074752, 'learning_rate': 1.4276676272133025e-05, 'epoch': 0.99} + 99%|█████████▊| 513/520 [57:50<00:26, 3.73s/it] 99%|█████████▉| 514/520 [57:53<00:22, 3.72s/it] {'loss': 1.2692, 'grad_norm': 0.000974260184700523, 'learning_rate': 1.0489428174020876e-05, 'epoch': 0.99} + 99%|█████████▉| 514/520 [57:53<00:22, 3.72s/it] 99%|█████████▉| 515/520 [57:57<00:18, 3.69s/it] {'loss': 1.3363, 'grad_norm': 0.0013031240298195105, 'learning_rate': 7.2845845720853615e-06, 'epoch': 0.99} + 99%|█████████▉| 515/520 [57:57<00:18, 3.69s/it] 99%|█████████▉| 516/520 [58:00<00:14, 3.67s/it] {'loss': 1.2435, 'grad_norm': 0.0010645522517159448, 'learning_rate': 4.662269987756317e-06, 'epoch': 0.99} + 99%|█████████▉| 516/520 [58:00<00:14, 3.67s/it] 99%|█████████▉| 517/520 [58:04<00:10, 3.64s/it] {'loss': 1.2154, 'grad_norm': 0.0009212822390667412, 'learning_rate': 2.6225863088036314e-06, 'epoch': 0.99} + 99%|█████████▉| 517/520 [58:04<00:10, 3.64s/it] 100%|█████████▉| 518/520 [58:08<00:07, 3.61s/it] {'loss': 1.2387, 'grad_norm': 0.0011106171193100349, 'learning_rate': 1.1656127853770792e-06, 'epoch': 1.0} + 100%|█████████▉| 518/520 [58:08<00:07, 3.61s/it] 100%|█████████▉| 519/520 [58:11<00:03, 3.59s/it] {'loss': 1.1994, 'grad_norm': 0.000987146651646325, 'learning_rate': 2.9140602692712126e-07, 'epoch': 1.0} + 100%|█████████▉| 519/520 [58:11<00:03, 3.59s/it] 100%|██████████| 520/520 [58:16<00:00, 3.86s/it] {'loss': 1.1822, 'grad_norm': 0.0009083504240305505, 'learning_rate': 0.0, 'epoch': 1.0} + 100%|██████████| 520/520 [58:16<00:00, 3.86s/it] {'train_runtime': 3496.1476, 'train_samples_per_second': 19.029, 'train_steps_per_second': 0.149, 'train_loss': 1.2936651511834218, 'epoch': 1.0} + 100%|██████████| 520/520 [58:16<00:00, 3.86s/it] 100%|██████████| 520/520 [58:16<00:00, 6.72s/it] +[2025-10-12 11:32:14,885] [INFO] [launch.py:348:main] Process 684544 exits successfully. +[2025-10-12 11:32:15,887] [INFO] [launch.py:348:main] Process 684545 exits successfully. +[2025-10-12 11:32:15,887] [INFO] [launch.py:348:main] Process 684540 exits successfully. +[2025-10-12 11:32:15,887] [INFO] [launch.py:348:main] Process 684542 exits successfully. +[2025-10-12 11:32:15,888] [INFO] [launch.py:348:main] Process 684539 exits successfully. +[2025-10-12 11:32:16,889] [INFO] [launch.py:348:main] Process 684543 exits successfully. +[2025-10-12 11:32:16,889] [INFO] [launch.py:348:main] Process 684541 exits successfully. +[2025-10-12 11:32:19,893] [INFO] [launch.py:348:main] Process 684538 exits successfully. +==== EXPERIMENT COMPLETED: qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_3e-2_connector-3.0_0.5_3e-2_ablation ==== +Log File: qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_3e-2_connector-3.0_0.5_3e-2_ablation_20251012_100541.log +Timestamp: 2025-10-12 11:32:22 +===================================== diff --git a/logs_oct11/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_5e-1_connector-3.0_0.5_5e-1_ablation_20251012_052751.log b/logs_oct11/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_5e-1_connector-3.0_0.5_5e-1_ablation_20251012_052751.log new file mode 100644 index 0000000000000000000000000000000000000000..14b649b1c1c4904ceef3ac020e6c8c113619e002 --- /dev/null +++ b/logs_oct11/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_5e-1_connector-3.0_0.5_5e-1_ablation_20251012_052751.log @@ -0,0 +1,2314 @@ +==== STARTING EXPERIMENT: qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_5e-1_connector-3.0_0.5_5e-1_ablation ==== +Log File: qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_5e-1_connector-3.0_0.5_5e-1_ablation_20251012_052751.log +Timestamp: 2025-10-12 05:27:51 +===================================== +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +The cache for model files in Transformers v4.22.0 has been updated. Migrating your old cache. This is a one-time only operation. You can interrupt this and resume the migration later on by calling `transformers.utils.move_cache()`. + 0it [00:00, ?it/s] 0it [00:00, ?it/s] +[2025-10-12 05:27:53,843] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 05:27:57,239] [WARNING] [runner.py:202:fetch_hostfile] Unable to find hostfile, will proceed with training with local resources only. +[2025-10-12 05:27:57,240] [INFO] [runner.py:568:main] cmd = /opt/conda/envs/tinyllava/bin/python3.10 -u -m deepspeed.launcher.launch --world_info=eyJsb2NhbGhvc3QiOiBbMCwgMSwgMiwgMywgNCwgNSwgNiwgN119 --master_addr=127.0.0.1 --master_port=29501 --enable_each_rank_log=None tinyllava/train/train.py --deepspeed ./scripts/zero3.json --data_path /s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json --image_folder /s3-code/ywang29/datasets/tinyllava --is_multimodal True --conv_version qwen2_base --model_name_or_path Qwen/Qwen2.5-0.5B --vision_tower google/siglip-so400m-patch14-384 --vision_tower2 --connector_type mlp2x_gelu --mm_vision_select_layer -2 --image_aspect_ratio square --attn_implementation flash_attention_2 --bf16 True --training_recipe common --tune_type_llm full --tune_type_vision_tower frozen --tune_vision_tower_from_layer 0 --tune_type_connector full --group_by_modality_length True --pretrained_model_path /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain --output_dir /nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_5e-1_connector-3.0_0.5_5e-1_ablation --num_train_epochs 1 --per_device_train_batch_size 4 --per_device_eval_batch_size 4 --gradient_accumulation_steps 4 --evaluation_strategy no --learning_rate 5e-1 --weight_decay 0. --warmup_ratio 0.03 --lr_scheduler_type cosine --logging_steps 1 --tf32 False --model_max_length 2048 --gradient_checkpointing True --dataloader_num_workers 8 --lazy_preprocess True --report_to tensorboard --tokenizer_use_fast False --run_name tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune --subnet_mode_text both --subnet_type_text None --mask_type_text soft --init_mean_text 3.0 --temperature_attn_text 0.5 --temperature_mlp_text 0.5 --backward_type_text normal --masked_layers_text all --subnet_mode_vision both --subnet_type_vision None --mask_type_vision soft --init_mean_vision 3.0 --temperature_attn_vision 0.5 --temperature_mlp_vision 0.5 --backward_type_vision normal --masked_layers_vision all --subnet_type_connector global --mask_type_connector soft --init_mean_connector 3.0 --temperature_connector 0.5 --backward_type_connector normal --mm_projector_lr 5e-1 --seed 42 --mask_model llm-connector --save_strategy steps --save_steps 50000 --save_total_limit 1 --train_data_ratio 0.1 +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 05:27:59,887] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 05:28:00,994] [INFO] [launch.py:138:main] 0 NCCL_VERSION=2.21.5 +[2025-10-12 05:28:00,994] [INFO] [launch.py:138:main] 0 NCCL_SOCKET_IFNAME=eth +[2025-10-12 05:28:00,994] [INFO] [launch.py:145:main] WORLD INFO DICT: {'localhost': [0, 1, 2, 3, 4, 5, 6, 7]} +[2025-10-12 05:28:00,994] [INFO] [launch.py:151:main] nnodes=1, num_local_procs=8, node_rank=0 +[2025-10-12 05:28:00,994] [INFO] [launch.py:162:main] global_rank_mapping=defaultdict(, {'localhost': [0, 1, 2, 3, 4, 5, 6, 7]}) +[2025-10-12 05:28:00,994] [INFO] [launch.py:163:main] dist_world_size=8 +[2025-10-12 05:28:00,994] [INFO] [launch.py:165:main] Setting CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 +[2025-10-12 05:28:00,996] [INFO] [launch.py:253:main] process 496287 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=0', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_5e-1_connector-3.0_0.5_5e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '5e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '5e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 05:28:00,998] [INFO] [launch.py:253:main] process 496288 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=1', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_5e-1_connector-3.0_0.5_5e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '5e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '5e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 05:28:01,000] [INFO] [launch.py:253:main] process 496289 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=2', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_5e-1_connector-3.0_0.5_5e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '5e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '5e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 05:28:01,002] [INFO] [launch.py:253:main] process 496290 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=3', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_5e-1_connector-3.0_0.5_5e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '5e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '5e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 05:28:01,004] [INFO] [launch.py:253:main] process 496291 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=4', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_5e-1_connector-3.0_0.5_5e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '5e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '5e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 05:28:01,006] [INFO] [launch.py:253:main] process 496292 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=5', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_5e-1_connector-3.0_0.5_5e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '5e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '5e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 05:28:01,008] [INFO] [launch.py:253:main] process 496293 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=6', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_5e-1_connector-3.0_0.5_5e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '5e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '5e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 05:28:01,010] [INFO] [launch.py:253:main] process 496294 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=7', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_5e-1_connector-3.0_0.5_5e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '5e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '5e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 05:28:07,995] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 05:28:07,995] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 05:28:08,026] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 05:28:08,028] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 05:28:08,031] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 05:28:08,035] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 05:28:08,051] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 05:28:08,055] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 05:28:08,589] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 05:28:08,589] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 05:28:08,589] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 05:28:08,589] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 05:28:08,589] [INFO] [comm.py:668:init_distributed] Initializing TorchBackend in DeepSpeed with backend nccl +[2025-10-12 05:28:08,589] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 05:28:08,589] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 05:28:08,589] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 05:28:08,589] [INFO] [comm.py:637:init_distributed] cdb=None +Apply masks for the following modules: ['llm', 'connector'] +{'llm': {'model_name_or_path': 'Qwen/Qwen2.5-0.5B', 'cache_dir': None, 'attn_implementation': 'flash_attention_2', 'subnet_mode': 'both', 'subnet_type': 'None', 'sparsity_attn': None, 'sparsity_mlp': None, 'threshold_attn': None, 'threshold_mlp': None, 'temperature_attn': 0.5, 'temperature_mlp': 0.5, 'masked_layers': 'all', 'mask_type': 'soft', 'backward_type': 'normal'}, 'vision_tower': {'model_name_or_path': 'google/siglip-so400m-patch14-384'}, 'connector': {'connector_type': 'mlp2x_gelu', 'subnet_type': 'global', 'threshold': None, 'sparsity': None, 'temperature': 0.5, 'mask_type': 'soft', 'backward_type': 'normal'}} +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector']Apply masks for the following modules: + ['llm', 'connector'] +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +TinyLlavaConfig { + "backward_type_connector": "normal", + "cache_dir": null, + "connector_type": "mlp2x_gelu", + "hidden_size": 896, + "ignore_index": -100, + "image_aspect_ratio": "square", + "image_token_index": -200, + "llm_model_name_or_path": "Qwen/Qwen2.5-0.5B", + "mask_model": [ + "llm", + "connector" + ], + "mask_type_connector": "soft", + "model_type": "tinyllava", + "num_queries": 128, + "num_resampler_layers": 3, + "pad_token": null, + "resampler_hidden_size": 768, + "sparsity_connector": null, + "subnet_type_connector": "global", + "temperature_connector": 0.5, + "text_config": { + "_name_or_path": "Qwen/Qwen2.5-0.5B", + "architectures": [ + "Qwen2ForCausalLM" + ], + "backward_type": "normal", + "bos_token_id": 151643, + "eos_token_id": 151643, + "hidden_size": 896, + "intermediate_size": 4864, + "mask_type": "soft", + "masked_layers": "all", + "max_position_embeddings": 32768, + "max_window_layers": 24, + "model_type": "qwen2", + "num_attention_heads": 14, + "num_hidden_layers": 24, + "num_key_value_heads": 2, + "rope_theta": 1000000.0, + "sliding_window": 32768, + "subnet_mode": "both", + "subnet_type": "None", + "temperature_attn": 0.5, + "temperature_mlp": 0.5, + "tie_word_embeddings": true, + "torch_dtype": "bfloat16", + "use_mrope": false, + "use_sliding_window": false, + "vocab_size": 151936 + }, + "threshold_connector": null, + "tokenizer_model_max_length": 2048, + "tokenizer_name_or_path": "Qwen/Qwen2.5-0.5B", + "tokenizer_padding_side": "right", + "tokenizer_use_fast": false, + "transformers_version": "4.40.1", + "tune_type_connector": "frozen", + "tune_type_llm": "frozen", + "tune_type_vision_tower": "frozen", + "tune_vision_tower_from_layer": -1, + "use_cache": false, + "vision_config": { + "hidden_act": "gelu_pytorch_tanh", + "hidden_size": 1152, + "image_size": 384, + "intermediate_size": 4304, + "layer_norm_eps": 1e-06, + "model_name_or_path": "google/siglip-so400m-patch14-384", + "model_name_or_path2": "", + "model_type": "siglip_vision_model", + "num_attention_heads": 16, + "num_hidden_layers": 27, + "patch_size": 14 + }, + "vision_feature_layer": -2, + "vision_feature_select_strategy": "patch", + "vision_hidden_size": 1152, + "vision_model_name_or_path": "google/siglip-so400m-patch14-384", + "vision_model_name_or_path2": "", + "vocab_size": 151936 +} + +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test2-worker-0:496287:496287 [0] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:496287:496287 [0] NCCL INFO Bootstrap : Using eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:496287:496287 [0] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test2-worker-0:496287:496287 [0] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test2-worker-0:496287:496287 [0] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test2-worker-0:496287:496287 [0] NCCL INFO cudaDriverVersion 12040 +NCCL version 2.21.5+cuda12.1 +ywang29-vrdb-test2-worker-0:496287:497904 [0] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:496287:497904 [0] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test2-worker-0:496287:497904 [0] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:496287:497904 [0] NCCL INFO NET/Socket : Using [0]eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:496287:497904 [0] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test2-worker-0:496287:497904 [0] NCCL INFO Using network Socket +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test2-worker-0:496288:496288 [1] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test2-worker-0:496288:496288 [1] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:496288:496288 [1] NCCL INFO Bootstrap : Using eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:496288:496288 [1] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test2-worker-0:496288:496288 [1] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test2-worker-0:496288:496288 [1] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test2-worker-0:496289:496289 [2] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test2-worker-0:496289:496289 [2] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:496289:496289 [2] NCCL INFO Bootstrap : Using eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:496289:496289 [2] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test2-worker-0:496289:496289 [2] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test2-worker-0:496289:496289 [2] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test2-worker-0:496290:496290 [3] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test2-worker-0:496290:496290 [3] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:496290:496290 [3] NCCL INFO Bootstrap : Using eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:496290:496290 [3] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test2-worker-0:496290:496290 [3] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test2-worker-0:496290:496290 [3] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test2-worker-0:496294:496294 [7] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test2-worker-0:496294:496294 [7] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:496294:496294 [7] NCCL INFO Bootstrap : Using eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:496294:496294 [7] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test2-worker-0:496294:496294 [7] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test2-worker-0:496294:496294 [7] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test2-worker-0:496292:496292 [5] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test2-worker-0:496292:496292 [5] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:496292:496292 [5] NCCL INFO Bootstrap : Using eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:496292:496292 [5] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test2-worker-0:496292:496292 [5] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test2-worker-0:496292:496292 [5] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test2-worker-0:496293:496293 [6] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test2-worker-0:496293:496293 [6] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:496291:496291 [4] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test2-worker-0:496293:496293 [6] NCCL INFO Bootstrap : Using eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:496291:496291 [4] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:496291:496291 [4] NCCL INFO Bootstrap : Using eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:496293:496293 [6] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test2-worker-0:496293:496293 [6] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test2-worker-0:496293:496293 [6] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test2-worker-0:496291:496291 [4] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test2-worker-0:496291:496291 [4] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test2-worker-0:496291:496291 [4] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test2-worker-0:496288:497922 [1] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:496288:497922 [1] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test2-worker-0:496288:497922 [1] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:496288:497922 [1] NCCL INFO NET/Socket : Using [0]eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:496288:497922 [1] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test2-worker-0:496288:497922 [1] NCCL INFO Using network Socket +ywang29-vrdb-test2-worker-0:496289:497923 [2] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:496289:497923 [2] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test2-worker-0:496289:497923 [2] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:496289:497923 [2] NCCL INFO NET/Socket : Using [0]eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:496289:497923 [2] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test2-worker-0:496289:497923 [2] NCCL INFO Using network Socket +ywang29-vrdb-test2-worker-0:496294:497925 [7] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:496294:497925 [7] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test2-worker-0:496294:497925 [7] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:496294:497925 [7] NCCL INFO NET/Socket : Using [0]eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:496294:497925 [7] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test2-worker-0:496294:497925 [7] NCCL INFO Using network Socket +ywang29-vrdb-test2-worker-0:496292:497926 [5] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:496292:497926 [5] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test2-worker-0:496292:497926 [5] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:496292:497926 [5] NCCL INFO NET/Socket : Using [0]eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:496292:497926 [5] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test2-worker-0:496292:497926 [5] NCCL INFO Using network Socket +ywang29-vrdb-test2-worker-0:496290:497924 [3] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:496290:497924 [3] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test2-worker-0:496290:497924 [3] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:496290:497924 [3] NCCL INFO NET/Socket : Using [0]eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:496290:497924 [3] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test2-worker-0:496290:497924 [3] NCCL INFO Using network Socket +ywang29-vrdb-test2-worker-0:496291:497928 [4] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:496291:497928 [4] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test2-worker-0:496291:497928 [4] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:496291:497928 [4] NCCL INFO NET/Socket : Using [0]eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:496291:497928 [4] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test2-worker-0:496291:497928 [4] NCCL INFO Using network Socket +ywang29-vrdb-test2-worker-0:496293:497927 [6] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:496293:497927 [6] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test2-worker-0:496293:497927 [6] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:496293:497927 [6] NCCL INFO NET/Socket : Using [0]eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:496293:497927 [6] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test2-worker-0:496293:497927 [6] NCCL INFO Using network Socket +ywang29-vrdb-test2-worker-0:496289:497923 [2] NCCL INFO ncclCommInitRank comm 0x560c47d89c60 rank 2 nranks 8 cudaDev 2 nvmlDev 2 busId 201c0 commId 0xc46ad9a90899c95d - Init START +ywang29-vrdb-test2-worker-0:496288:497922 [1] NCCL INFO ncclCommInitRank comm 0x55725d7ccfd0 rank 1 nranks 8 cudaDev 1 nvmlDev 1 busId 101d0 commId 0xc46ad9a90899c95d - Init START +ywang29-vrdb-test2-worker-0:496287:497904 [0] NCCL INFO ncclCommInitRank comm 0x56307caafc10 rank 0 nranks 8 cudaDev 0 nvmlDev 0 busId 101c0 commId 0xc46ad9a90899c95d - Init START +ywang29-vrdb-test2-worker-0:496292:497926 [5] NCCL INFO ncclCommInitRank comm 0x5569a50a3a40 rank 5 nranks 8 cudaDev 5 nvmlDev 5 busId 901d0 commId 0xc46ad9a90899c95d - Init START +ywang29-vrdb-test2-worker-0:496291:497928 [4] NCCL INFO ncclCommInitRank comm 0x5557a5b850c0 rank 4 nranks 8 cudaDev 4 nvmlDev 4 busId 901c0 commId 0xc46ad9a90899c95d - Init START +ywang29-vrdb-test2-worker-0:496293:497927 [6] NCCL INFO ncclCommInitRank comm 0x55d51f8bdf40 rank 6 nranks 8 cudaDev 6 nvmlDev 6 busId a01c0 commId 0xc46ad9a90899c95d - Init START +ywang29-vrdb-test2-worker-0:496290:497924 [3] NCCL INFO ncclCommInitRank comm 0x55d9aee1cb30 rank 3 nranks 8 cudaDev 3 nvmlDev 3 busId 201d0 commId 0xc46ad9a90899c95d - Init START +ywang29-vrdb-test2-worker-0:496294:497925 [7] NCCL INFO ncclCommInitRank comm 0x56246aa1d250 rank 7 nranks 8 cudaDev 7 nvmlDev 7 busId a01d0 commId 0xc46ad9a90899c95d - Init START +ywang29-vrdb-test2-worker-0:496291:497928 [4] NCCL INFO Setting affinity for GPU 4 to ffffff00,0000ffff,ff000000 +ywang29-vrdb-test2-worker-0:496291:497928 [4] NCCL INFO NVLS multicast support is not available on dev 4 +ywang29-vrdb-test2-worker-0:496288:497922 [1] NCCL INFO Setting affinity for GPU 1 to ff,ffff0000,00ffffff +ywang29-vrdb-test2-worker-0:496288:497922 [1] NCCL INFO NVLS multicast support is not available on dev 1 +ywang29-vrdb-test2-worker-0:496293:497927 [6] NCCL INFO Setting affinity for GPU 6 to ffffff00,0000ffff,ff000000 +ywang29-vrdb-test2-worker-0:496293:497927 [6] NCCL INFO NVLS multicast support is not available on dev 6 +ywang29-vrdb-test2-worker-0:496287:497904 [0] NCCL INFO Setting affinity for GPU 0 to ff,ffff0000,00ffffff +ywang29-vrdb-test2-worker-0:496287:497904 [0] NCCL INFO NVLS multicast support is not available on dev 0 +ywang29-vrdb-test2-worker-0:496292:497926 [5] NCCL INFO Setting affinity for GPU 5 to ffffff00,0000ffff,ff000000 +ywang29-vrdb-test2-worker-0:496290:497924 [3] NCCL INFO Setting affinity for GPU 3 to ff,ffff0000,00ffffff +ywang29-vrdb-test2-worker-0:496290:497924 [3] NCCL INFO NVLS multicast support is not available on dev 3 +ywang29-vrdb-test2-worker-0:496292:497926 [5] NCCL INFO NVLS multicast support is not available on dev 5 +ywang29-vrdb-test2-worker-0:496294:497925 [7] NCCL INFO Setting affinity for GPU 7 to ffffff00,0000ffff,ff000000 +ywang29-vrdb-test2-worker-0:496289:497923 [2] NCCL INFO Setting affinity for GPU 2 to ff,ffff0000,00ffffff +ywang29-vrdb-test2-worker-0:496289:497923 [2] NCCL INFO NVLS multicast support is not available on dev 2 +ywang29-vrdb-test2-worker-0:496294:497925 [7] NCCL INFO NVLS multicast support is not available on dev 7 +ywang29-vrdb-test2-worker-0:496294:497925 [7] NCCL INFO comm 0x56246aa1d250 rank 7 nRanks 8 nNodes 1 localRanks 8 localRank 7 MNNVL 0 +ywang29-vrdb-test2-worker-0:496290:497924 [3] NCCL INFO comm 0x55d9aee1cb30 rank 3 nRanks 8 nNodes 1 localRanks 8 localRank 3 MNNVL 0 +ywang29-vrdb-test2-worker-0:496293:497927 [6] NCCL INFO comm 0x55d51f8bdf40 rank 6 nRanks 8 nNodes 1 localRanks 8 localRank 6 MNNVL 0 +ywang29-vrdb-test2-worker-0:496289:497923 [2] NCCL INFO comm 0x560c47d89c60 rank 2 nRanks 8 nNodes 1 localRanks 8 localRank 2 MNNVL 0 +ywang29-vrdb-test2-worker-0:496287:497904 [0] NCCL INFO comm 0x56307caafc10 rank 0 nRanks 8 nNodes 1 localRanks 8 localRank 0 MNNVL 0 +ywang29-vrdb-test2-worker-0:496292:497926 [5] NCCL INFO comm 0x5569a50a3a40 rank 5 nRanks 8 nNodes 1 localRanks 8 localRank 5 MNNVL 0 +ywang29-vrdb-test2-worker-0:496288:497922 [1] NCCL INFO comm 0x55725d7ccfd0 rank 1 nRanks 8 nNodes 1 localRanks 8 localRank 1 MNNVL 0 +ywang29-vrdb-test2-worker-0:496291:497928 [4] NCCL INFO comm 0x5557a5b850c0 rank 4 nRanks 8 nNodes 1 localRanks 8 localRank 4 MNNVL 0 +ywang29-vrdb-test2-worker-0:496287:497904 [0] NCCL INFO Channel 00/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:496287:497904 [0] NCCL INFO Channel 01/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:496294:497925 [7] NCCL INFO Trees [0] -1/-1/-1->7->6 [1] -1/-1/-1->7->6 [2] -1/-1/-1->7->6 [3] -1/-1/-1->7->6 [4] -1/-1/-1->7->6 [5] -1/-1/-1->7->6 [6] -1/-1/-1->7->6 [7] -1/-1/-1->7->6 [8] -1/-1/-1->7->6 [9] -1/-1/-1->7->6 [10] -1/-1/-1->7->6 [11] -1/-1/-1->7->6 [12] -1/-1/-1->7->6 [13] -1/-1/-1->7->6 [14] -1/-1/-1->7->6 [15] -1/-1/-1->7->6 [16] -1/-1/-1->7->6 [17] -1/-1/-1->7->6 [18] -1/-1/-1->7->6 [19] -1/-1/-1->7->6 [20] -1/-1/-1->7->6 [21] -1/-1/-1->7->6 [22] -1/-1/-1->7->6 [23] -1/-1/-1->7->6 +ywang29-vrdb-test2-worker-0:496287:497904 [0] NCCL INFO Channel 02/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:496287:497904 [0] NCCL INFO Channel 03/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:496290:497924 [3] NCCL INFO Trees [0] 4/-1/-1->3->2 [1] 4/-1/-1->3->2 [2] 4/-1/-1->3->2 [3] 4/-1/-1->3->2 [4] 4/-1/-1->3->2 [5] 4/-1/-1->3->2 [6] 4/-1/-1->3->2 [7] 4/-1/-1->3->2 [8] 4/-1/-1->3->2 [9] 4/-1/-1->3->2 [10] 4/-1/-1->3->2 [11] 4/-1/-1->3->2 [12] 4/-1/-1->3->2 [13] 4/-1/-1->3->2 [14] 4/-1/-1->3->2 [15] 4/-1/-1->3->2 [16] 4/-1/-1->3->2 [17] 4/-1/-1->3->2 [18] 4/-1/-1->3->2 [19] 4/-1/-1->3->2 [20] 4/-1/-1->3->2 [21] 4/-1/-1->3->2 [22] 4/-1/-1->3->2 [23] 4/-1/-1->3->2 +ywang29-vrdb-test2-worker-0:496294:497925 [7] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:496293:497927 [6] NCCL INFO Trees [0] 7/-1/-1->6->5 [1] 7/-1/-1->6->5 [2] 7/-1/-1->6->5 [3] 7/-1/-1->6->5 [4] 7/-1/-1->6->5 [5] 7/-1/-1->6->5 [6] 7/-1/-1->6->5 [7] 7/-1/-1->6->5 [8] 7/-1/-1->6->5 [9] 7/-1/-1->6->5 [10] 7/-1/-1->6->5 [11] 7/-1/-1->6->5 [12] 7/-1/-1->6->5 [13] 7/-1/-1->6->5 [14] 7/-1/-1->6->5 [15] 7/-1/-1->6->5 [16] 7/-1/-1->6->5 [17] 7/-1/-1->6->5 [18] 7/-1/-1->6->5 [19] 7/-1/-1->6->5 [20] 7/-1/-1->6->5 [21] 7/-1/-1->6->5 [22] 7/-1/-1->6->5 [23] 7/-1/-1->6->5 +ywang29-vrdb-test2-worker-0:496289:497923 [2] NCCL INFO Trees [0] 3/-1/-1->2->1 [1] 3/-1/-1->2->1 [2] 3/-1/-1->2->1 [3] 3/-1/-1->2->1 [4] 3/-1/-1->2->1 [5] 3/-1/-1->2->1 [6] 3/-1/-1->2->1 [7] 3/-1/-1->2->1 [8] 3/-1/-1->2->1 [9] 3/-1/-1->2->1 [10] 3/-1/-1->2->1 [11] 3/-1/-1->2->1 [12] 3/-1/-1->2->1 [13] 3/-1/-1->2->1 [14] 3/-1/-1->2->1 [15] 3/-1/-1->2->1 [16] 3/-1/-1->2->1 [17] 3/-1/-1->2->1 [18] 3/-1/-1->2->1 [19] 3/-1/-1->2->1 [20] 3/-1/-1->2->1 [21] 3/-1/-1->2->1 [22] 3/-1/-1->2->1 [23] 3/-1/-1->2->1 +ywang29-vrdb-test2-worker-0:496287:497904 [0] NCCL INFO Channel 04/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:496290:497924 [3] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:496288:497922 [1] NCCL INFO Trees [0] 2/-1/-1->1->0 [1] 2/-1/-1->1->0 [2] 2/-1/-1->1->0 [3] 2/-1/-1->1->0 [4] 2/-1/-1->1->0 [5] 2/-1/-1->1->0 [6] 2/-1/-1->1->0 [7] 2/-1/-1->1->0 [8] 2/-1/-1->1->0 [9] 2/-1/-1->1->0 [10] 2/-1/-1->1->0 [11] 2/-1/-1->1->0 [12] 2/-1/-1->1->0 [13] 2/-1/-1->1->0 [14] 2/-1/-1->1->0 [15] 2/-1/-1->1->0 [16] 2/-1/-1->1->0 [17] 2/-1/-1->1->0 [18] 2/-1/-1->1->0 [19] 2/-1/-1->1->0 [20] 2/-1/-1->1->0 [21] 2/-1/-1->1->0 [22] 2/-1/-1->1->0 [23] 2/-1/-1->1->0 +ywang29-vrdb-test2-worker-0:496293:497927 [6] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:496287:497904 [0] NCCL INFO Channel 05/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:496289:497923 [2] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:496287:497904 [0] NCCL INFO Channel 06/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:496288:497922 [1] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:496287:497904 [0] NCCL INFO Channel 07/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:496287:497904 [0] NCCL INFO Channel 08/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:496287:497904 [0] NCCL INFO Channel 09/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:496287:497904 [0] NCCL INFO Channel 10/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:496287:497904 [0] NCCL INFO Channel 11/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:496287:497904 [0] NCCL INFO Channel 12/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:496287:497904 [0] NCCL INFO Channel 13/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:496292:497926 [5] NCCL INFO Trees [0] 6/-1/-1->5->4 [1] 6/-1/-1->5->4 [2] 6/-1/-1->5->4 [3] 6/-1/-1->5->4 [4] 6/-1/-1->5->4 [5] 6/-1/-1->5->4 [6] 6/-1/-1->5->4 [7] 6/-1/-1->5->4 [8] 6/-1/-1->5->4 [9] 6/-1/-1->5->4 [10] 6/-1/-1->5->4 [11] 6/-1/-1->5->4 [12] 6/-1/-1->5->4 [13] 6/-1/-1->5->4 [14] 6/-1/-1->5->4 [15] 6/-1/-1->5->4 [16] 6/-1/-1->5->4 [17] 6/-1/-1->5->4 [18] 6/-1/-1->5->4 [19] 6/-1/-1->5->4 [20] 6/-1/-1->5->4 [21] 6/-1/-1->5->4 [22] 6/-1/-1->5->4 [23] 6/-1/-1->5->4 +ywang29-vrdb-test2-worker-0:496287:497904 [0] NCCL INFO Channel 14/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:496287:497904 [0] NCCL INFO Channel 15/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:496287:497904 [0] NCCL INFO Channel 16/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:496291:497928 [4] NCCL INFO Trees [0] 5/-1/-1->4->3 [1] 5/-1/-1->4->3 [2] 5/-1/-1->4->3 [3] 5/-1/-1->4->3 [4] 5/-1/-1->4->3 [5] 5/-1/-1->4->3 [6] 5/-1/-1->4->3 [7] 5/-1/-1->4->3 [8] 5/-1/-1->4->3 [9] 5/-1/-1->4->3 [10] 5/-1/-1->4->3 [11] 5/-1/-1->4->3 [12] 5/-1/-1->4->3 [13] 5/-1/-1->4->3 [14] 5/-1/-1->4->3 [15] 5/-1/-1->4->3 [16] 5/-1/-1->4->3 [17] 5/-1/-1->4->3 [18] 5/-1/-1->4->3 [19] 5/-1/-1->4->3 [20] 5/-1/-1->4->3 [21] 5/-1/-1->4->3 [22] 5/-1/-1->4->3 [23] 5/-1/-1->4->3 +ywang29-vrdb-test2-worker-0:496292:497926 [5] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:496287:497904 [0] NCCL INFO Channel 17/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:496287:497904 [0] NCCL INFO Channel 18/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:496287:497904 [0] NCCL INFO Channel 19/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:496287:497904 [0] NCCL INFO Channel 20/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:496287:497904 [0] NCCL INFO Channel 21/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:496291:497928 [4] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:496287:497904 [0] NCCL INFO Channel 22/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:496287:497904 [0] NCCL INFO Channel 23/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:496287:497904 [0] NCCL INFO Trees [0] 1/-1/-1->0->-1 [1] 1/-1/-1->0->-1 [2] 1/-1/-1->0->-1 [3] 1/-1/-1->0->-1 [4] 1/-1/-1->0->-1 [5] 1/-1/-1->0->-1 [6] 1/-1/-1->0->-1 [7] 1/-1/-1->0->-1 [8] 1/-1/-1->0->-1 [9] 1/-1/-1->0->-1 [10] 1/-1/-1->0->-1 [11] 1/-1/-1->0->-1 [12] 1/-1/-1->0->-1 [13] 1/-1/-1->0->-1 [14] 1/-1/-1->0->-1 [15] 1/-1/-1->0->-1 [16] 1/-1/-1->0->-1 [17] 1/-1/-1->0->-1 [18] 1/-1/-1->0->-1 [19] 1/-1/-1->0->-1 [20] 1/-1/-1->0->-1 [21] 1/-1/-1->0->-1 [22] 1/-1/-1->0->-1 [23] 1/-1/-1->0->-1 +ywang29-vrdb-test2-worker-0:496287:497904 [0] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:496293:497927 [6] NCCL INFO Channel 00/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:497927 [6] NCCL INFO Channel 01/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:497923 [2] NCCL INFO Channel 00/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:497924 [3] NCCL INFO Channel 00/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:497925 [7] NCCL INFO Channel 00/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:497927 [6] NCCL INFO Channel 02/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:497922 [1] NCCL INFO Channel 00/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496287:497904 [0] NCCL INFO Channel 00/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:497923 [2] NCCL INFO Channel 01/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:497924 [3] NCCL INFO Channel 01/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:497925 [7] NCCL INFO Channel 01/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:497927 [6] NCCL INFO Channel 03/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:497922 [1] NCCL INFO Channel 01/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496287:497904 [0] NCCL INFO Channel 01/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:497923 [2] NCCL INFO Channel 02/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:497924 [3] NCCL INFO Channel 02/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:497928 [4] NCCL INFO Channel 00/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:497926 [5] NCCL INFO Channel 00/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:497925 [7] NCCL INFO Channel 02/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:497927 [6] NCCL INFO Channel 04/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:497922 [1] NCCL INFO Channel 02/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496287:497904 [0] NCCL INFO Channel 02/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:497923 [2] NCCL INFO Channel 03/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:497924 [3] NCCL INFO Channel 03/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:497928 [4] NCCL INFO Channel 01/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:497926 [5] NCCL INFO Channel 01/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:497925 [7] NCCL INFO Channel 03/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:497927 [6] NCCL INFO Channel 05/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:497922 [1] NCCL INFO Channel 03/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496287:497904 [0] NCCL INFO Channel 03/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:497923 [2] NCCL INFO Channel 04/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:497924 [3] NCCL INFO Channel 04/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:497928 [4] NCCL INFO Channel 02/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:497925 [7] NCCL INFO Channel 04/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:497926 [5] NCCL INFO Channel 02/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:497927 [6] NCCL INFO Channel 06/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:497922 [1] NCCL INFO Channel 04/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496287:497904 [0] NCCL INFO Channel 04/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:497923 [2] NCCL INFO Channel 05/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:497924 [3] NCCL INFO Channel 05/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:497928 [4] NCCL INFO Channel 03/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:497927 [6] NCCL INFO Channel 07/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:497925 [7] NCCL INFO Channel 05/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:497926 [5] NCCL INFO Channel 03/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:497922 [1] NCCL INFO Channel 05/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496287:497904 [0] NCCL INFO Channel 05/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:497923 [2] NCCL INFO Channel 06/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:497924 [3] NCCL INFO Channel 06/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:497928 [4] NCCL INFO Channel 04/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:497927 [6] NCCL INFO Channel 08/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:497922 [1] NCCL INFO Channel 06/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:497925 [7] NCCL INFO Channel 06/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:497926 [5] NCCL INFO Channel 04/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496287:497904 [0] NCCL INFO Channel 06/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:497923 [2] NCCL INFO Channel 07/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:497924 [3] NCCL INFO Channel 07/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:497927 [6] NCCL INFO Channel 09/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:497928 [4] NCCL INFO Channel 05/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:497922 [1] NCCL INFO Channel 07/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:497925 [7] NCCL INFO Channel 07/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:497926 [5] NCCL INFO Channel 05/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496287:497904 [0] NCCL INFO Channel 07/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:497923 [2] NCCL INFO Channel 08/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:497924 [3] NCCL INFO Channel 08/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:497927 [6] NCCL INFO Channel 10/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:497928 [4] NCCL INFO Channel 06/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:497922 [1] NCCL INFO Channel 08/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:497925 [7] NCCL INFO Channel 08/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:497926 [5] NCCL INFO Channel 06/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496287:497904 [0] NCCL INFO Channel 08/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:497923 [2] NCCL INFO Channel 09/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:497924 [3] NCCL INFO Channel 09/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:497927 [6] NCCL INFO Channel 11/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:497922 [1] NCCL INFO Channel 09/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:497928 [4] NCCL INFO Channel 07/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:497925 [7] NCCL INFO Channel 09/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496287:497904 [0] NCCL INFO Channel 09/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:497926 [5] NCCL INFO Channel 07/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:497923 [2] NCCL INFO Channel 10/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:497924 [3] NCCL INFO Channel 10/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:497927 [6] NCCL INFO Channel 12/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:497922 [1] NCCL INFO Channel 10/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:497928 [4] NCCL INFO Channel 08/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:497925 [7] NCCL INFO Channel 10/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496287:497904 [0] NCCL INFO Channel 10/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:497923 [2] NCCL INFO Channel 11/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:497924 [3] NCCL INFO Channel 11/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:497926 [5] NCCL INFO Channel 08/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:497927 [6] NCCL INFO Channel 13/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:497922 [1] NCCL INFO Channel 11/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:497925 [7] NCCL INFO Channel 11/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:497928 [4] NCCL INFO Channel 09/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496287:497904 [0] NCCL INFO Channel 11/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:497923 [2] NCCL INFO Channel 12/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:497924 [3] NCCL INFO Channel 12/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:497926 [5] NCCL INFO Channel 09/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:497927 [6] NCCL INFO Channel 14/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:497922 [1] NCCL INFO Channel 12/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:497925 [7] NCCL INFO Channel 12/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:497928 [4] NCCL INFO Channel 10/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496287:497904 [0] NCCL INFO Channel 12/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:497923 [2] NCCL INFO Channel 13/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:497924 [3] NCCL INFO Channel 13/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:497926 [5] NCCL INFO Channel 10/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:497927 [6] NCCL INFO Channel 15/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:497922 [1] NCCL INFO Channel 13/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:497925 [7] NCCL INFO Channel 13/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:497928 [4] NCCL INFO Channel 11/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496287:497904 [0] NCCL INFO Channel 13/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:497923 [2] NCCL INFO Channel 14/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:497924 [3] NCCL INFO Channel 14/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:497926 [5] NCCL INFO Channel 11/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:497927 [6] NCCL INFO Channel 16/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:497922 [1] NCCL INFO Channel 14/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:497925 [7] NCCL INFO Channel 14/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:497928 [4] NCCL INFO Channel 12/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496287:497904 [0] NCCL INFO Channel 14/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:497923 [2] NCCL INFO Channel 15/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:497924 [3] NCCL INFO Channel 15/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:497926 [5] NCCL INFO Channel 12/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:497927 [6] NCCL INFO Channel 17/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:497922 [1] NCCL INFO Channel 15/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:497925 [7] NCCL INFO Channel 15/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:497928 [4] NCCL INFO Channel 13/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496287:497904 [0] NCCL INFO Channel 15/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:497923 [2] NCCL INFO Channel 16/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:497924 [3] NCCL INFO Channel 16/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:497926 [5] NCCL INFO Channel 13/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:497922 [1] NCCL INFO Channel 16/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:497927 [6] NCCL INFO Channel 18/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:497925 [7] NCCL INFO Channel 16/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:497928 [4] NCCL INFO Channel 14/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496287:497904 [0] NCCL INFO Channel 16/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:497923 [2] NCCL INFO Channel 17/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:497924 [3] NCCL INFO Channel 17/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:497926 [5] NCCL INFO Channel 14/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:497922 [1] NCCL INFO Channel 17/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:497927 [6] NCCL INFO Channel 19/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:497925 [7] NCCL INFO Channel 17/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:497928 [4] NCCL INFO Channel 15/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496287:497904 [0] NCCL INFO Channel 17/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:497923 [2] NCCL INFO Channel 18/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:497924 [3] NCCL INFO Channel 18/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:497926 [5] NCCL INFO Channel 15/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:497922 [1] NCCL INFO Channel 18/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:497927 [6] NCCL INFO Channel 20/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:497925 [7] NCCL INFO Channel 18/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:497928 [4] NCCL INFO Channel 16/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:497923 [2] NCCL INFO Channel 19/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496287:497904 [0] NCCL INFO Channel 18/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:497924 [3] NCCL INFO Channel 19/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:497926 [5] NCCL INFO Channel 16/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:497922 [1] NCCL INFO Channel 19/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:497927 [6] NCCL INFO Channel 21/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:497925 [7] NCCL INFO Channel 19/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:497928 [4] NCCL INFO Channel 17/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:497923 [2] NCCL INFO Channel 20/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496287:497904 [0] NCCL INFO Channel 19/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:497924 [3] NCCL INFO Channel 20/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:497922 [1] NCCL INFO Channel 20/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:497926 [5] NCCL INFO Channel 17/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:497927 [6] NCCL INFO Channel 22/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:497925 [7] NCCL INFO Channel 20/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:497923 [2] NCCL INFO Channel 21/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:497928 [4] NCCL INFO Channel 18/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496287:497904 [0] NCCL INFO Channel 20/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:497924 [3] NCCL INFO Channel 21/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:497922 [1] NCCL INFO Channel 21/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:497927 [6] NCCL INFO Channel 23/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:497926 [5] NCCL INFO Channel 18/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:497925 [7] NCCL INFO Channel 21/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:497923 [2] NCCL INFO Channel 22/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:497928 [4] NCCL INFO Channel 19/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496287:497904 [0] NCCL INFO Channel 21/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:497924 [3] NCCL INFO Channel 22/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:497922 [1] NCCL INFO Channel 22/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:497926 [5] NCCL INFO Channel 19/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:497925 [7] NCCL INFO Channel 22/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:497923 [2] NCCL INFO Channel 23/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:497928 [4] NCCL INFO Channel 20/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496287:497904 [0] NCCL INFO Channel 22/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:497924 [3] NCCL INFO Channel 23/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:497922 [1] NCCL INFO Channel 23/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:497925 [7] NCCL INFO Channel 23/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:497926 [5] NCCL INFO Channel 20/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496287:497904 [0] NCCL INFO Channel 23/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:497928 [4] NCCL INFO Channel 21/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:497926 [5] NCCL INFO Channel 21/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:497928 [4] NCCL INFO Channel 22/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:497926 [5] NCCL INFO Channel 22/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:497926 [5] NCCL INFO Channel 23/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:497928 [4] NCCL INFO Channel 23/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:497923 [2] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:496290:497924 [3] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:496288:497922 [1] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:496287:497904 [0] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:496291:497928 [4] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:496289:497923 [2] NCCL INFO Channel 00/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:497923 [2] NCCL INFO Channel 01/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:497923 [2] NCCL INFO Channel 02/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:497923 [2] NCCL INFO Channel 03/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:497923 [2] NCCL INFO Channel 04/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:497923 [2] NCCL INFO Channel 05/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:497924 [3] NCCL INFO Channel 00/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:497923 [2] NCCL INFO Channel 06/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:497924 [3] NCCL INFO Channel 01/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:497923 [2] NCCL INFO Channel 07/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:497923 [2] NCCL INFO Channel 08/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:497923 [2] NCCL INFO Channel 09/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:497926 [5] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:496293:497927 [6] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:496289:497923 [2] NCCL INFO Channel 10/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:497923 [2] NCCL INFO Channel 11/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:497923 [2] NCCL INFO Channel 12/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:497923 [2] NCCL INFO Channel 13/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:497924 [3] NCCL INFO Channel 02/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:497925 [7] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:496294:497925 [7] NCCL INFO Channel 00/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:497923 [2] NCCL INFO Channel 14/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:497924 [3] NCCL INFO Channel 03/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:497925 [7] NCCL INFO Channel 01/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:497923 [2] NCCL INFO Channel 15/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:497924 [3] NCCL INFO Channel 04/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:497925 [7] NCCL INFO Channel 02/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:497923 [2] NCCL INFO Channel 16/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:497924 [3] NCCL INFO Channel 05/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:497925 [7] NCCL INFO Channel 03/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:497922 [1] NCCL INFO Channel 00/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:497923 [2] NCCL INFO Channel 17/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:497924 [3] NCCL INFO Channel 06/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:497925 [7] NCCL INFO Channel 04/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:497922 [1] NCCL INFO Channel 01/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:497923 [2] NCCL INFO Channel 18/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:497928 [4] NCCL INFO Channel 00/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:497924 [3] NCCL INFO Channel 07/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:497925 [7] NCCL INFO Channel 05/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:497922 [1] NCCL INFO Channel 02/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:497923 [2] NCCL INFO Channel 19/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:497928 [4] NCCL INFO Channel 01/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:497924 [3] NCCL INFO Channel 08/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:497925 [7] NCCL INFO Channel 06/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:497922 [1] NCCL INFO Channel 03/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:497923 [2] NCCL INFO Channel 20/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:497924 [3] NCCL INFO Channel 09/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:497922 [1] NCCL INFO Channel 04/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:497923 [2] NCCL INFO Channel 21/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:497924 [3] NCCL INFO Channel 10/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:497923 [2] NCCL INFO Channel 22/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:497922 [1] NCCL INFO Channel 05/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:497924 [3] NCCL INFO Channel 11/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:497923 [2] NCCL INFO Channel 23/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:497922 [1] NCCL INFO Channel 06/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:497924 [3] NCCL INFO Channel 12/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:497922 [1] NCCL INFO Channel 07/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:497924 [3] NCCL INFO Channel 13/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:497922 [1] NCCL INFO Channel 08/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:497924 [3] NCCL INFO Channel 14/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:497922 [1] NCCL INFO Channel 09/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:497924 [3] NCCL INFO Channel 15/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:497922 [1] NCCL INFO Channel 10/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:497924 [3] NCCL INFO Channel 16/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:497922 [1] NCCL INFO Channel 11/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:497924 [3] NCCL INFO Channel 17/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:497922 [1] NCCL INFO Channel 12/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:497924 [3] NCCL INFO Channel 18/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:497922 [1] NCCL INFO Channel 13/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:497924 [3] NCCL INFO Channel 19/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:497922 [1] NCCL INFO Channel 14/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:497924 [3] NCCL INFO Channel 20/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:497922 [1] NCCL INFO Channel 15/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:497924 [3] NCCL INFO Channel 21/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:497922 [1] NCCL INFO Channel 16/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:497924 [3] NCCL INFO Channel 22/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:497927 [6] NCCL INFO Channel 00/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:497922 [1] NCCL INFO Channel 17/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:497924 [3] NCCL INFO Channel 23/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:497927 [6] NCCL INFO Channel 01/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:497922 [1] NCCL INFO Channel 18/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:497927 [6] NCCL INFO Channel 02/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:497928 [4] NCCL INFO Channel 02/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:497922 [1] NCCL INFO Channel 19/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:497927 [6] NCCL INFO Channel 03/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:497928 [4] NCCL INFO Channel 03/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:497922 [1] NCCL INFO Channel 20/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:497927 [6] NCCL INFO Channel 04/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:497928 [4] NCCL INFO Channel 04/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:497922 [1] NCCL INFO Channel 21/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:497927 [6] NCCL INFO Channel 05/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:497928 [4] NCCL INFO Channel 05/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:497926 [5] NCCL INFO Channel 00/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:497922 [1] NCCL INFO Channel 22/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:497927 [6] NCCL INFO Channel 06/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:497928 [4] NCCL INFO Channel 06/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:497926 [5] NCCL INFO Channel 01/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:497922 [1] NCCL INFO Channel 23/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:497927 [6] NCCL INFO Channel 07/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:497928 [4] NCCL INFO Channel 07/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:497926 [5] NCCL INFO Channel 02/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:497927 [6] NCCL INFO Channel 08/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:497928 [4] NCCL INFO Channel 08/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:497926 [5] NCCL INFO Channel 03/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:497927 [6] NCCL INFO Channel 09/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:497928 [4] NCCL INFO Channel 09/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:497926 [5] NCCL INFO Channel 04/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:497928 [4] NCCL INFO Channel 10/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:497927 [6] NCCL INFO Channel 10/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:497926 [5] NCCL INFO Channel 05/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:497925 [7] NCCL INFO Channel 07/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:497925 [7] NCCL INFO Channel 08/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:497925 [7] NCCL INFO Channel 09/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:497928 [4] NCCL INFO Channel 11/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:497928 [4] NCCL INFO Channel 12/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:497927 [6] NCCL INFO Channel 11/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:497925 [7] NCCL INFO Channel 10/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:497928 [4] NCCL INFO Channel 13/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:497927 [6] NCCL INFO Channel 12/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:497925 [7] NCCL INFO Channel 11/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:497926 [5] NCCL INFO Channel 06/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:497928 [4] NCCL INFO Channel 14/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:497926 [5] NCCL INFO Channel 07/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:497925 [7] NCCL INFO Channel 12/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:497927 [6] NCCL INFO Channel 13/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:497928 [4] NCCL INFO Channel 15/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:497926 [5] NCCL INFO Channel 08/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:497927 [6] NCCL INFO Channel 14/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:497928 [4] NCCL INFO Channel 16/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:497926 [5] NCCL INFO Channel 09/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:497927 [6] NCCL INFO Channel 15/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:497925 [7] NCCL INFO Channel 13/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:497928 [4] NCCL INFO Channel 17/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:497928 [4] NCCL INFO Channel 18/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:497928 [4] NCCL INFO Channel 19/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:497926 [5] NCCL INFO Channel 10/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:497927 [6] NCCL INFO Channel 16/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:497925 [7] NCCL INFO Channel 14/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:497928 [4] NCCL INFO Channel 20/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:497926 [5] NCCL INFO Channel 11/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:497927 [6] NCCL INFO Channel 17/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:497925 [7] NCCL INFO Channel 15/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:497928 [4] NCCL INFO Channel 21/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:497926 [5] NCCL INFO Channel 12/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:497927 [6] NCCL INFO Channel 18/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:497925 [7] NCCL INFO Channel 16/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:497928 [4] NCCL INFO Channel 22/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:497926 [5] NCCL INFO Channel 13/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:497927 [6] NCCL INFO Channel 19/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:497925 [7] NCCL INFO Channel 17/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:497928 [4] NCCL INFO Channel 23/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:497926 [5] NCCL INFO Channel 14/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:497927 [6] NCCL INFO Channel 20/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:497925 [7] NCCL INFO Channel 18/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:497926 [5] NCCL INFO Channel 15/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:497927 [6] NCCL INFO Channel 21/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:497925 [7] NCCL INFO Channel 19/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:497927 [6] NCCL INFO Channel 22/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:497925 [7] NCCL INFO Channel 20/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:497926 [5] NCCL INFO Channel 16/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:497925 [7] NCCL INFO Channel 21/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:497927 [6] NCCL INFO Channel 23/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:497926 [5] NCCL INFO Channel 17/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:497925 [7] NCCL INFO Channel 22/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:497926 [5] NCCL INFO Channel 18/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:497925 [7] NCCL INFO Channel 23/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:497926 [5] NCCL INFO Channel 19/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:497926 [5] NCCL INFO Channel 20/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:497926 [5] NCCL INFO Channel 21/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:497926 [5] NCCL INFO Channel 22/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:497926 [5] NCCL INFO Channel 23/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496287:497904 [0] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:496287:497904 [0] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:496287:497904 [0] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:496288:497922 [1] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:496288:497922 [1] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:496288:497922 [1] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:496289:497923 [2] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:496289:497923 [2] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:496289:497923 [2] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:496290:497924 [3] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:496290:497924 [3] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:496290:497924 [3] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:496294:497925 [7] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:496294:497925 [7] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:496294:497925 [7] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:496291:497928 [4] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:496291:497928 [4] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:496291:497928 [4] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:496293:497927 [6] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:496292:497926 [5] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:496293:497927 [6] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:496293:497927 [6] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:496292:497926 [5] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:496292:497926 [5] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:496294:497925 [7] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test2-worker-0:496290:497924 [3] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test2-worker-0:496294:497925 [7] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test2-worker-0:496290:497924 [3] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test2-worker-0:496294:497925 [7] NCCL INFO ncclCommInitRank comm 0x56246aa1d250 rank 7 nranks 8 cudaDev 7 nvmlDev 7 busId a01d0 commId 0xc46ad9a90899c95d - Init COMPLETE +ywang29-vrdb-test2-worker-0:496290:497924 [3] NCCL INFO ncclCommInitRank comm 0x55d9aee1cb30 rank 3 nranks 8 cudaDev 3 nvmlDev 3 busId 201d0 commId 0xc46ad9a90899c95d - Init COMPLETE +ywang29-vrdb-test2-worker-0:496288:497922 [1] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test2-worker-0:496288:497922 [1] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test2-worker-0:496288:497922 [1] NCCL INFO ncclCommInitRank comm 0x55725d7ccfd0 rank 1 nranks 8 cudaDev 1 nvmlDev 1 busId 101d0 commId 0xc46ad9a90899c95d - Init COMPLETE +ywang29-vrdb-test2-worker-0:496292:497926 [5] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test2-worker-0:496292:497926 [5] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test2-worker-0:496292:497926 [5] NCCL INFO ncclCommInitRank comm 0x5569a50a3a40 rank 5 nranks 8 cudaDev 5 nvmlDev 5 busId 901d0 commId 0xc46ad9a90899c95d - Init COMPLETE +ywang29-vrdb-test2-worker-0:496289:497923 [2] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test2-worker-0:496289:497923 [2] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test2-worker-0:496289:497923 [2] NCCL INFO ncclCommInitRank comm 0x560c47d89c60 rank 2 nranks 8 cudaDev 2 nvmlDev 2 busId 201c0 commId 0xc46ad9a90899c95d - Init COMPLETE +ywang29-vrdb-test2-worker-0:496287:497904 [0] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test2-worker-0:496291:497928 [4] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test2-worker-0:496291:497928 [4] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test2-worker-0:496287:497904 [0] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test2-worker-0:496291:497928 [4] NCCL INFO ncclCommInitRank comm 0x5557a5b850c0 rank 4 nranks 8 cudaDev 4 nvmlDev 4 busId 901c0 commId 0xc46ad9a90899c95d - Init COMPLETE +ywang29-vrdb-test2-worker-0:496293:497927 [6] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test2-worker-0:496287:497904 [0] NCCL INFO ncclCommInitRank comm 0x56307caafc10 rank 0 nranks 8 cudaDev 0 nvmlDev 0 busId 101c0 commId 0xc46ad9a90899c95d - Init COMPLETE +ywang29-vrdb-test2-worker-0:496293:497927 [6] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test2-worker-0:496293:497927 [6] NCCL INFO ncclCommInitRank comm 0x55d51f8bdf40 rank 6 nranks 8 cudaDev 6 nvmlDev 6 busId a01c0 commId 0xc46ad9a90899c95d - Init COMPLETE +[2025-10-12 05:28:58,441] [INFO] [partition_parameters.py:343:__exit__] finished initializing model - num_params = 459, num_elems = 0.99B +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +[2025-10-12 05:29:01,984] [INFO] [partition_parameters.py:343:__exit__] finished initializing model - num_params = 907, num_elems = 1.42B +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +TinyLlavaForConditionalGeneration( + (language_model): Qwen2ForCausalLM( + (model): Qwen2Model( + (embed_tokens): Embedding(151936, 896) + (layers): ModuleList( + (0-23): 24 x Qwen2DecoderLayer( + (self_attn): Qwen2FlashAttention2( + (q_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=True) + (k_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=128, bias=True) + (v_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=128, bias=True) + (o_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=False) + (rotary_emb): Qwen2RotaryEmbedding() + ) + (mlp): Qwen2MLP( + (gate_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=4864, bias=False) + (up_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=4864, bias=False) + (down_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=4864, out_features=896, bias=False) + (act_fn): SiLU() + ) + (input_layernorm): Qwen2RMSNorm() + (post_attention_layernorm): Qwen2RMSNorm() + ) + ) + (norm): Qwen2RMSNorm() + ) + (lm_head): Linear(in_features=896, out_features=151936, bias=False) + ) + (vision_tower): SIGLIPVisionTower( + (_vision_tower): SiglipVisionModel( + (vision_model): SiglipVisionTransformer( + (embeddings): SiglipVisionEmbeddings( + (patch_embedding): Conv2d(3, 1152, kernel_size=(14, 14), stride=(14, 14), padding=valid) + (position_embedding): Embedding(729, 1152) + ) + (encoder): SiglipEncoder( + (layers): ModuleList( + (0-26): 27 x SiglipEncoderLayer( + (self_attn): SiglipAttention( + (k_proj): Linear(in_features=1152, out_features=1152, bias=True) + (v_proj): Linear(in_features=1152, out_features=1152, bias=True) + (q_proj): Linear(in_features=1152, out_features=1152, bias=True) + (out_proj): Linear(in_features=1152, out_features=1152, bias=True) + ) + (layer_norm1): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + (layer_norm2): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + ) + ) + ) + (post_layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (head): SiglipMultiheadAttentionPoolingHead( + (attention): MultiheadAttention( + (out_proj): NonDynamicallyQuantizableLinear(in_features=1152, out_features=1152, bias=True) + ) + (layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + ) + ) + ) + ) + (connector): MLPConnector( + (_connector): Sequential( + (0): SupermaskLinearSparsity_SoftForward_Normal(in_features=1152, out_features=896, bias=True) + (1): GELU(approximate='none') + (2): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=True) + ) + ) +) +Pre-training init language_model.model.layers.0.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.0.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.0.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.0.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.0.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.0.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.0.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.mlp.down_proj.scores: Mean=3.000000 +Pre-training init connector._connector.0.scores: Mean=3.000005 +Pre-training init connector._connector.2.scores: Mean=2.999970 +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +2025-10-12 05:29:20,085 | INFO: Total Parameters: 1283756736, Total Trainable Parameters: 359661568 +2025-10-12 05:29:20,092 | INFO: Trainable Parameters: +language_model.model.layers.0.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.0.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.0.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.0.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.0.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.0.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.0.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.1.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.1.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.1.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.1.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.1.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.1.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.1.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.2.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.2.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.2.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.2.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.2.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.2.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.2.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.3.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.3.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.3.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.3.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.3.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.3.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.3.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.4.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.4.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.4.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.4.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.4.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.4.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.4.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.5.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.5.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.5.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.5.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.5.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.5.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.5.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.6.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.6.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.6.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.6.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.6.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.6.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.6.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.7.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.7.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.7.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.7.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.7.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.7.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.7.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.8.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.8.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.8.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.8.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.8.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.8.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.8.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.9.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.9.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.9.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.9.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.9.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.9.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.9.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.10.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.10.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.10.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.10.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.10.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.10.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.10.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.11.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.11.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.11.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.11.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.11.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.11.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.11.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.12.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.12.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.12.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.12.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.12.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.12.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.12.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.13.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.13.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.13.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.13.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.13.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.13.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.13.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.14.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.14.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.14.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.14.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.14.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.14.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.14.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.15.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.15.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.15.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.15.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.15.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.15.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.15.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.16.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.16.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.16.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.16.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.16.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.16.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.16.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.17.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.17.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.17.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.17.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.17.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.17.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.17.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.18.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.18.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.18.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.18.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.18.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.18.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.18.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.19.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.19.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.19.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.19.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.19.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.19.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.19.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.20.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.20.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.20.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.20.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.20.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.20.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.20.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.21.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.21.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.21.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.21.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.21.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.21.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.21.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.22.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.22.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.22.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.22.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.22.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.22.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.22.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.23.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.23.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.23.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.23.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.23.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.23.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.23.mlp.down_proj.scores: 4358144 parameters +connector._connector.0.scores: 1032192 parameters +connector._connector.2.scores: 802816 parameters +Parameter Offload: Total persistent parameters: 486464 in 403 params + 0%| | 0/520 [00:003->2 [1] 4/-1/-1->3->2 [2] 4/-1/-1->3->2 [3] 4/-1/-1->3->2 [4] 4/-1/-1->3->2 [5] 4/-1/-1->3->2 [6] 4/-1/-1->3->2 [7] 4/-1/-1->3->2 [8] 4/-1/-1->3->2 [9] 4/-1/-1->3->2 [10] 4/-1/-1->3->2 [11] 4/-1/-1->3->2 [12] 4/-1/-1->3->2 [13] 4/-1/-1->3->2 [14] 4/-1/-1->3->2 [15] 4/-1/-1->3->2 [16] 4/-1/-1->3->2 [17] 4/-1/-1->3->2 [18] 4/-1/-1->3->2 [19] 4/-1/-1->3->2 [20] 4/-1/-1->3->2 [21] 4/-1/-1->3->2 [22] 4/-1/-1->3->2 [23] 4/-1/-1->3->2 +ywang29-vrdb-test2-worker-0:496293:502927 [6] NCCL INFO Trees [0] 7/-1/-1->6->5 [1] 7/-1/-1->6->5 [2] 7/-1/-1->6->5 [3] 7/-1/-1->6->5 [4] 7/-1/-1->6->5 [5] 7/-1/-1->6->5 [6] 7/-1/-1->6->5 [7] 7/-1/-1->6->5 [8] 7/-1/-1->6->5 [9] 7/-1/-1->6->5 [10] 7/-1/-1->6->5 [11] 7/-1/-1->6->5 [12] 7/-1/-1->6->5 [13] 7/-1/-1->6->5 [14] 7/-1/-1->6->5 [15] 7/-1/-1->6->5 [16] 7/-1/-1->6->5 [17] 7/-1/-1->6->5 [18] 7/-1/-1->6->5 [19] 7/-1/-1->6->5 [20] 7/-1/-1->6->5 [21] 7/-1/-1->6->5 [22] 7/-1/-1->6->5 [23] 7/-1/-1->6->5 +ywang29-vrdb-test2-worker-0:496294:502928 [7] NCCL INFO Trees [0] -1/-1/-1->7->6 [1] -1/-1/-1->7->6 [2] -1/-1/-1->7->6 [3] -1/-1/-1->7->6 [4] -1/-1/-1->7->6 [5] -1/-1/-1->7->6 [6] -1/-1/-1->7->6 [7] -1/-1/-1->7->6 [8] -1/-1/-1->7->6 [9] -1/-1/-1->7->6 [10] -1/-1/-1->7->6 [11] -1/-1/-1->7->6 [12] -1/-1/-1->7->6 [13] -1/-1/-1->7->6 [14] -1/-1/-1->7->6 [15] -1/-1/-1->7->6 [16] -1/-1/-1->7->6 [17] -1/-1/-1->7->6 [18] -1/-1/-1->7->6 [19] -1/-1/-1->7->6 [20] -1/-1/-1->7->6 [21] -1/-1/-1->7->6 [22] -1/-1/-1->7->6 [23] -1/-1/-1->7->6 +ywang29-vrdb-test2-worker-0:496287:502922 [0] NCCL INFO Channel 08/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:496293:502927 [6] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:496290:502924 [3] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:496294:502928 [7] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:496287:502922 [0] NCCL INFO Channel 09/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:496287:502922 [0] NCCL INFO Channel 10/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:496287:502922 [0] NCCL INFO Channel 11/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:496287:502922 [0] NCCL INFO Channel 12/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:496287:502922 [0] NCCL INFO Channel 13/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:496287:502922 [0] NCCL INFO Channel 14/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:496287:502922 [0] NCCL INFO Channel 15/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:496287:502922 [0] NCCL INFO Channel 16/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:496289:502925 [2] NCCL INFO Trees [0] 3/-1/-1->2->1 [1] 3/-1/-1->2->1 [2] 3/-1/-1->2->1 [3] 3/-1/-1->2->1 [4] 3/-1/-1->2->1 [5] 3/-1/-1->2->1 [6] 3/-1/-1->2->1 [7] 3/-1/-1->2->1 [8] 3/-1/-1->2->1 [9] 3/-1/-1->2->1 [10] 3/-1/-1->2->1 [11] 3/-1/-1->2->1 [12] 3/-1/-1->2->1 [13] 3/-1/-1->2->1 [14] 3/-1/-1->2->1 [15] 3/-1/-1->2->1 [16] 3/-1/-1->2->1 [17] 3/-1/-1->2->1 [18] 3/-1/-1->2->1 [19] 3/-1/-1->2->1 [20] 3/-1/-1->2->1 [21] 3/-1/-1->2->1 [22] 3/-1/-1->2->1 [23] 3/-1/-1->2->1 +ywang29-vrdb-test2-worker-0:496288:502926 [1] NCCL INFO Trees [0] 2/-1/-1->1->0 [1] 2/-1/-1->1->0 [2] 2/-1/-1->1->0 [3] 2/-1/-1->1->0 [4] 2/-1/-1->1->0 [5] 2/-1/-1->1->0 [6] 2/-1/-1->1->0 [7] 2/-1/-1->1->0 [8] 2/-1/-1->1->0 [9] 2/-1/-1->1->0 [10] 2/-1/-1->1->0 [11] 2/-1/-1->1->0 [12] 2/-1/-1->1->0 [13] 2/-1/-1->1->0 [14] 2/-1/-1->1->0 [15] 2/-1/-1->1->0 [16] 2/-1/-1->1->0 [17] 2/-1/-1->1->0 [18] 2/-1/-1->1->0 [19] 2/-1/-1->1->0 [20] 2/-1/-1->1->0 [21] 2/-1/-1->1->0 [22] 2/-1/-1->1->0 [23] 2/-1/-1->1->0 +ywang29-vrdb-test2-worker-0:496287:502922 [0] NCCL INFO Channel 17/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:496289:502925 [2] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:496288:502926 [1] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:496287:502922 [0] NCCL INFO Channel 18/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:496287:502922 [0] NCCL INFO Channel 19/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:496291:502923 [4] NCCL INFO Trees [0] 5/-1/-1->4->3 [1] 5/-1/-1->4->3 [2] 5/-1/-1->4->3 [3] 5/-1/-1->4->3 [4] 5/-1/-1->4->3 [5] 5/-1/-1->4->3 [6] 5/-1/-1->4->3 [7] 5/-1/-1->4->3 [8] 5/-1/-1->4->3 [9] 5/-1/-1->4->3 [10] 5/-1/-1->4->3 [11] 5/-1/-1->4->3 [12] 5/-1/-1->4->3 [13] 5/-1/-1->4->3 [14] 5/-1/-1->4->3 [15] 5/-1/-1->4->3 [16] 5/-1/-1->4->3 [17] 5/-1/-1->4->3 [18] 5/-1/-1->4->3 [19] 5/-1/-1->4->3 [20] 5/-1/-1->4->3 [21] 5/-1/-1->4->3 [22] 5/-1/-1->4->3 [23] 5/-1/-1->4->3 +ywang29-vrdb-test2-worker-0:496292:502929 [5] NCCL INFO Trees [0] 6/-1/-1->5->4 [1] 6/-1/-1->5->4 [2] 6/-1/-1->5->4 [3] 6/-1/-1->5->4 [4] 6/-1/-1->5->4 [5] 6/-1/-1->5->4 [6] 6/-1/-1->5->4 [7] 6/-1/-1->5->4 [8] 6/-1/-1->5->4 [9] 6/-1/-1->5->4 [10] 6/-1/-1->5->4 [11] 6/-1/-1->5->4 [12] 6/-1/-1->5->4 [13] 6/-1/-1->5->4 [14] 6/-1/-1->5->4 [15] 6/-1/-1->5->4 [16] 6/-1/-1->5->4 [17] 6/-1/-1->5->4 [18] 6/-1/-1->5->4 [19] 6/-1/-1->5->4 [20] 6/-1/-1->5->4 [21] 6/-1/-1->5->4 [22] 6/-1/-1->5->4 [23] 6/-1/-1->5->4 +ywang29-vrdb-test2-worker-0:496287:502922 [0] NCCL INFO Channel 20/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:496291:502923 [4] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:496287:502922 [0] NCCL INFO Channel 21/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:496292:502929 [5] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:496287:502922 [0] NCCL INFO Channel 22/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:496287:502922 [0] NCCL INFO Channel 23/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:496287:502922 [0] NCCL INFO Trees [0] 1/-1/-1->0->-1 [1] 1/-1/-1->0->-1 [2] 1/-1/-1->0->-1 [3] 1/-1/-1->0->-1 [4] 1/-1/-1->0->-1 [5] 1/-1/-1->0->-1 [6] 1/-1/-1->0->-1 [7] 1/-1/-1->0->-1 [8] 1/-1/-1->0->-1 [9] 1/-1/-1->0->-1 [10] 1/-1/-1->0->-1 [11] 1/-1/-1->0->-1 [12] 1/-1/-1->0->-1 [13] 1/-1/-1->0->-1 [14] 1/-1/-1->0->-1 [15] 1/-1/-1->0->-1 [16] 1/-1/-1->0->-1 [17] 1/-1/-1->0->-1 [18] 1/-1/-1->0->-1 [19] 1/-1/-1->0->-1 [20] 1/-1/-1->0->-1 [21] 1/-1/-1->0->-1 [22] 1/-1/-1->0->-1 [23] 1/-1/-1->0->-1 +ywang29-vrdb-test2-worker-0:496287:502922 [0] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:496287:502922 [0] NCCL INFO Channel 00/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496287:502922 [0] NCCL INFO Channel 01/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:502923 [4] NCCL INFO Channel 00/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496287:502922 [0] NCCL INFO Channel 02/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:502923 [4] NCCL INFO Channel 01/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496287:502922 [0] NCCL INFO Channel 03/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:502927 [6] NCCL INFO Channel 00/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:502923 [4] NCCL INFO Channel 02/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496287:502922 [0] NCCL INFO Channel 04/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:502927 [6] NCCL INFO Channel 01/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:502923 [4] NCCL INFO Channel 03/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496287:502922 [0] NCCL INFO Channel 05/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:502927 [6] NCCL INFO Channel 02/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:502923 [4] NCCL INFO Channel 04/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496287:502922 [0] NCCL INFO Channel 06/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:502927 [6] NCCL INFO Channel 03/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:502923 [4] NCCL INFO Channel 05/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496287:502922 [0] NCCL INFO Channel 07/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:502927 [6] NCCL INFO Channel 04/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:502923 [4] NCCL INFO Channel 06/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496287:502922 [0] NCCL INFO Channel 08/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:502927 [6] NCCL INFO Channel 05/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:502923 [4] NCCL INFO Channel 07/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496287:502922 [0] NCCL INFO Channel 09/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:502924 [3] NCCL INFO Channel 00/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:502927 [6] NCCL INFO Channel 06/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:502923 [4] NCCL INFO Channel 08/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496287:502922 [0] NCCL INFO Channel 10/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:502924 [3] NCCL INFO Channel 01/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:502927 [6] NCCL INFO Channel 07/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:502923 [4] NCCL INFO Channel 09/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496287:502922 [0] NCCL INFO Channel 11/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:502924 [3] NCCL INFO Channel 02/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:502927 [6] NCCL INFO Channel 08/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:502923 [4] NCCL INFO Channel 10/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496287:502922 [0] NCCL INFO Channel 12/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:502924 [3] NCCL INFO Channel 03/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:502927 [6] NCCL INFO Channel 09/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:502923 [4] NCCL INFO Channel 11/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496287:502922 [0] NCCL INFO Channel 13/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:502924 [3] NCCL INFO Channel 04/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:502927 [6] NCCL INFO Channel 10/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496287:502922 [0] NCCL INFO Channel 14/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:502923 [4] NCCL INFO Channel 12/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:502924 [3] NCCL INFO Channel 05/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:502927 [6] NCCL INFO Channel 11/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496287:502922 [0] NCCL INFO Channel 15/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:502923 [4] NCCL INFO Channel 13/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:502924 [3] NCCL INFO Channel 06/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:502927 [6] NCCL INFO Channel 12/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496287:502922 [0] NCCL INFO Channel 16/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:502923 [4] NCCL INFO Channel 14/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:502924 [3] NCCL INFO Channel 07/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:502927 [6] NCCL INFO Channel 13/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496287:502922 [0] NCCL INFO Channel 17/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:502923 [4] NCCL INFO Channel 15/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:502927 [6] NCCL INFO Channel 14/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:502924 [3] NCCL INFO Channel 08/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496287:502922 [0] NCCL INFO Channel 18/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:502923 [4] NCCL INFO Channel 16/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:502927 [6] NCCL INFO Channel 15/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:502924 [3] NCCL INFO Channel 09/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496287:502922 [0] NCCL INFO Channel 19/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:502923 [4] NCCL INFO Channel 17/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:502927 [6] NCCL INFO Channel 16/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:502924 [3] NCCL INFO Channel 10/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:502928 [7] NCCL INFO Channel 00/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496287:502922 [0] NCCL INFO Channel 20/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:502923 [4] NCCL INFO Channel 18/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:502927 [6] NCCL INFO Channel 17/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:502924 [3] NCCL INFO Channel 11/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:502928 [7] NCCL INFO Channel 01/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496287:502922 [0] NCCL INFO Channel 21/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:502923 [4] NCCL INFO Channel 19/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:502927 [6] NCCL INFO Channel 18/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:502924 [3] NCCL INFO Channel 12/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:502928 [7] NCCL INFO Channel 02/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496287:502922 [0] NCCL INFO Channel 22/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:502923 [4] NCCL INFO Channel 20/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:502927 [6] NCCL INFO Channel 19/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:502924 [3] NCCL INFO Channel 13/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:502928 [7] NCCL INFO Channel 03/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:502925 [2] NCCL INFO Channel 00/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496287:502922 [0] NCCL INFO Channel 23/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:502926 [1] NCCL INFO Channel 00/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:502927 [6] NCCL INFO Channel 20/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:502924 [3] NCCL INFO Channel 14/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:502928 [7] NCCL INFO Channel 04/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:502923 [4] NCCL INFO Channel 21/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:502925 [2] NCCL INFO Channel 01/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:502926 [1] NCCL INFO Channel 01/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:502927 [6] NCCL INFO Channel 21/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:502924 [3] NCCL INFO Channel 15/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:502928 [7] NCCL INFO Channel 05/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:502929 [5] NCCL INFO Channel 00/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:502923 [4] NCCL INFO Channel 22/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:502925 [2] NCCL INFO Channel 02/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:502926 [1] NCCL INFO Channel 02/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:502927 [6] NCCL INFO Channel 22/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:502924 [3] NCCL INFO Channel 16/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:502928 [7] NCCL INFO Channel 06/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:502929 [5] NCCL INFO Channel 01/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:502923 [4] NCCL INFO Channel 23/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:502926 [1] NCCL INFO Channel 03/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:502925 [2] NCCL INFO Channel 03/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:502927 [6] NCCL INFO Channel 23/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:502924 [3] NCCL INFO Channel 17/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:502928 [7] NCCL INFO Channel 07/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:502929 [5] NCCL INFO Channel 02/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:502926 [1] NCCL INFO Channel 04/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:502925 [2] NCCL INFO Channel 04/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:502924 [3] NCCL INFO Channel 18/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:502928 [7] NCCL INFO Channel 08/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:502929 [5] NCCL INFO Channel 03/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:502926 [1] NCCL INFO Channel 05/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:502924 [3] NCCL INFO Channel 19/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:502925 [2] NCCL INFO Channel 05/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:502928 [7] NCCL INFO Channel 09/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:502929 [5] NCCL INFO Channel 04/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:502926 [1] NCCL INFO Channel 06/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:502924 [3] NCCL INFO Channel 20/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:502928 [7] NCCL INFO Channel 10/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:502925 [2] NCCL INFO Channel 06/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:502929 [5] NCCL INFO Channel 05/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:502926 [1] NCCL INFO Channel 07/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:502924 [3] NCCL INFO Channel 21/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:502928 [7] NCCL INFO Channel 11/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:502925 [2] NCCL INFO Channel 07/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:502929 [5] NCCL INFO Channel 06/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:502924 [3] NCCL INFO Channel 22/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:502926 [1] NCCL INFO Channel 08/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:502928 [7] NCCL INFO Channel 12/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:502925 [2] NCCL INFO Channel 08/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:502929 [5] NCCL INFO Channel 07/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:502924 [3] NCCL INFO Channel 23/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:502926 [1] NCCL INFO Channel 09/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:502928 [7] NCCL INFO Channel 13/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:502925 [2] NCCL INFO Channel 09/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:502929 [5] NCCL INFO Channel 08/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:502928 [7] NCCL INFO Channel 14/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:502926 [1] NCCL INFO Channel 10/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:502929 [5] NCCL INFO Channel 09/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:502928 [7] NCCL INFO Channel 15/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:502925 [2] NCCL INFO Channel 10/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:502926 [1] NCCL INFO Channel 11/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:502929 [5] NCCL INFO Channel 10/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:502928 [7] NCCL INFO Channel 16/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:502925 [2] NCCL INFO Channel 11/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:502926 [1] NCCL INFO Channel 12/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:502929 [5] NCCL INFO Channel 11/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:502928 [7] NCCL INFO Channel 17/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:502925 [2] NCCL INFO Channel 12/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:502926 [1] NCCL INFO Channel 13/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:502928 [7] NCCL INFO Channel 18/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:502929 [5] NCCL INFO Channel 12/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:502925 [2] NCCL INFO Channel 13/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:502928 [7] NCCL INFO Channel 19/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:502926 [1] NCCL INFO Channel 14/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:502929 [5] NCCL INFO Channel 13/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:502925 [2] NCCL INFO Channel 14/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:502929 [5] NCCL INFO Channel 14/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:502928 [7] NCCL INFO Channel 20/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:502926 [1] NCCL INFO Channel 15/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:502929 [5] NCCL INFO Channel 15/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:502928 [7] NCCL INFO Channel 21/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:502925 [2] NCCL INFO Channel 15/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:502926 [1] NCCL INFO Channel 16/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:502929 [5] NCCL INFO Channel 16/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:502928 [7] NCCL INFO Channel 22/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:502925 [2] NCCL INFO Channel 16/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:502926 [1] NCCL INFO Channel 17/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:502929 [5] NCCL INFO Channel 17/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:502928 [7] NCCL INFO Channel 23/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:502925 [2] NCCL INFO Channel 17/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:502926 [1] NCCL INFO Channel 18/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:502929 [5] NCCL INFO Channel 18/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:502925 [2] NCCL INFO Channel 18/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:502926 [1] NCCL INFO Channel 19/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:502929 [5] NCCL INFO Channel 19/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:502925 [2] NCCL INFO Channel 19/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:502929 [5] NCCL INFO Channel 20/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:502926 [1] NCCL INFO Channel 20/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:502929 [5] NCCL INFO Channel 21/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:502925 [2] NCCL INFO Channel 20/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:502926 [1] NCCL INFO Channel 21/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:502929 [5] NCCL INFO Channel 22/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:502925 [2] NCCL INFO Channel 21/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:502926 [1] NCCL INFO Channel 22/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:502929 [5] NCCL INFO Channel 23/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:502925 [2] NCCL INFO Channel 22/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:502926 [1] NCCL INFO Channel 23/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:502925 [2] NCCL INFO Channel 23/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:502925 [2] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:496290:502924 [3] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:496288:502926 [1] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:496287:502922 [0] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:496291:502923 [4] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:496294:502928 [7] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:496294:502928 [7] NCCL INFO Channel 00/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:502929 [5] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:496293:502927 [6] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:496294:502928 [7] NCCL INFO Channel 01/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:502928 [7] NCCL INFO Channel 02/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:502928 [7] NCCL INFO Channel 03/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:502928 [7] NCCL INFO Channel 04/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:502928 [7] NCCL INFO Channel 05/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:502928 [7] NCCL INFO Channel 06/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:502928 [7] NCCL INFO Channel 07/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:502928 [7] NCCL INFO Channel 08/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:502924 [3] NCCL INFO Channel 00/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:502928 [7] NCCL INFO Channel 09/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:502924 [3] NCCL INFO Channel 01/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:502928 [7] NCCL INFO Channel 10/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:502925 [2] NCCL INFO Channel 00/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:502926 [1] NCCL INFO Channel 00/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:502924 [3] NCCL INFO Channel 02/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:502928 [7] NCCL INFO Channel 11/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:502925 [2] NCCL INFO Channel 01/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:502926 [1] NCCL INFO Channel 01/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:502924 [3] NCCL INFO Channel 03/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:502925 [2] NCCL INFO Channel 02/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:502926 [1] NCCL INFO Channel 02/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:502924 [3] NCCL INFO Channel 04/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:502925 [2] NCCL INFO Channel 03/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:502926 [1] NCCL INFO Channel 03/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:502924 [3] NCCL INFO Channel 05/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:502925 [2] NCCL INFO Channel 04/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:502928 [7] NCCL INFO Channel 12/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:502926 [1] NCCL INFO Channel 04/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:502924 [3] NCCL INFO Channel 06/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:502925 [2] NCCL INFO Channel 05/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:502928 [7] NCCL INFO Channel 13/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:502926 [1] NCCL INFO Channel 05/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:502924 [3] NCCL INFO Channel 07/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:502925 [2] NCCL INFO Channel 06/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:502928 [7] NCCL INFO Channel 14/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:502926 [1] NCCL INFO Channel 06/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:502924 [3] NCCL INFO Channel 08/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:502925 [2] NCCL INFO Channel 07/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:502928 [7] NCCL INFO Channel 15/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:502926 [1] NCCL INFO Channel 07/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:502924 [3] NCCL INFO Channel 09/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:502925 [2] NCCL INFO Channel 08/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:502928 [7] NCCL INFO Channel 16/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:502926 [1] NCCL INFO Channel 08/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:502924 [3] NCCL INFO Channel 10/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:502925 [2] NCCL INFO Channel 09/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:502928 [7] NCCL INFO Channel 17/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:502926 [1] NCCL INFO Channel 09/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:502924 [3] NCCL INFO Channel 11/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:502925 [2] NCCL INFO Channel 10/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:502928 [7] NCCL INFO Channel 18/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:502926 [1] NCCL INFO Channel 10/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:502924 [3] NCCL INFO Channel 12/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:502925 [2] NCCL INFO Channel 11/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:502928 [7] NCCL INFO Channel 19/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:502926 [1] NCCL INFO Channel 11/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:502924 [3] NCCL INFO Channel 13/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:502925 [2] NCCL INFO Channel 12/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:502923 [4] NCCL INFO Channel 00/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:502928 [7] NCCL INFO Channel 20/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:502926 [1] NCCL INFO Channel 12/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:502924 [3] NCCL INFO Channel 14/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:502925 [2] NCCL INFO Channel 13/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:502923 [4] NCCL INFO Channel 01/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:502928 [7] NCCL INFO Channel 21/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:502926 [1] NCCL INFO Channel 13/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:502924 [3] NCCL INFO Channel 15/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:502925 [2] NCCL INFO Channel 14/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:502927 [6] NCCL INFO Channel 00/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:502923 [4] NCCL INFO Channel 02/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:502928 [7] NCCL INFO Channel 22/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:502926 [1] NCCL INFO Channel 14/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:502924 [3] NCCL INFO Channel 16/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:502925 [2] NCCL INFO Channel 15/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:502927 [6] NCCL INFO Channel 01/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:502923 [4] NCCL INFO Channel 03/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:502926 [1] NCCL INFO Channel 15/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496294:502928 [7] NCCL INFO Channel 23/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:502925 [2] NCCL INFO Channel 16/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:502924 [3] NCCL INFO Channel 17/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:502927 [6] NCCL INFO Channel 02/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:502923 [4] NCCL INFO Channel 04/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:502926 [1] NCCL INFO Channel 16/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:502925 [2] NCCL INFO Channel 17/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:502924 [3] NCCL INFO Channel 18/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:502927 [6] NCCL INFO Channel 03/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:502926 [1] NCCL INFO Channel 17/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:502923 [4] NCCL INFO Channel 05/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:502925 [2] NCCL INFO Channel 18/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:502924 [3] NCCL INFO Channel 19/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:502927 [6] NCCL INFO Channel 04/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:502926 [1] NCCL INFO Channel 18/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:502923 [4] NCCL INFO Channel 06/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:502925 [2] NCCL INFO Channel 19/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:502924 [3] NCCL INFO Channel 20/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:502927 [6] NCCL INFO Channel 05/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:502926 [1] NCCL INFO Channel 19/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:502923 [4] NCCL INFO Channel 07/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:502925 [2] NCCL INFO Channel 20/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:502924 [3] NCCL INFO Channel 21/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:502927 [6] NCCL INFO Channel 06/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:502926 [1] NCCL INFO Channel 20/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:502923 [4] NCCL INFO Channel 08/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:502925 [2] NCCL INFO Channel 21/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:502927 [6] NCCL INFO Channel 07/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:502924 [3] NCCL INFO Channel 22/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:502926 [1] NCCL INFO Channel 21/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:502923 [4] NCCL INFO Channel 09/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:502925 [2] NCCL INFO Channel 22/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:502927 [6] NCCL INFO Channel 08/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:502926 [1] NCCL INFO Channel 22/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496290:502924 [3] NCCL INFO Channel 23/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:502923 [4] NCCL INFO Channel 10/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496289:502925 [2] NCCL INFO Channel 23/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:502927 [6] NCCL INFO Channel 09/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496288:502926 [1] NCCL INFO Channel 23/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:502923 [4] NCCL INFO Channel 11/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:502927 [6] NCCL INFO Channel 10/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:502923 [4] NCCL INFO Channel 12/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:502927 [6] NCCL INFO Channel 11/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:502923 [4] NCCL INFO Channel 13/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:502927 [6] NCCL INFO Channel 12/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:502923 [4] NCCL INFO Channel 14/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:502927 [6] NCCL INFO Channel 13/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:502927 [6] NCCL INFO Channel 14/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:502923 [4] NCCL INFO Channel 15/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:502929 [5] NCCL INFO Channel 00/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:502923 [4] NCCL INFO Channel 16/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:502927 [6] NCCL INFO Channel 15/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:502929 [5] NCCL INFO Channel 01/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:502927 [6] NCCL INFO Channel 16/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:502923 [4] NCCL INFO Channel 17/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:502929 [5] NCCL INFO Channel 02/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:502927 [6] NCCL INFO Channel 17/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:502923 [4] NCCL INFO Channel 18/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:502927 [6] NCCL INFO Channel 18/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:502929 [5] NCCL INFO Channel 03/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:502923 [4] NCCL INFO Channel 19/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:502927 [6] NCCL INFO Channel 19/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:502923 [4] NCCL INFO Channel 20/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:502927 [6] NCCL INFO Channel 20/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:502929 [5] NCCL INFO Channel 04/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:502923 [4] NCCL INFO Channel 21/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:502929 [5] NCCL INFO Channel 05/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:502927 [6] NCCL INFO Channel 21/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:502923 [4] NCCL INFO Channel 22/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496291:502923 [4] NCCL INFO Channel 23/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:502929 [5] NCCL INFO Channel 06/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:502927 [6] NCCL INFO Channel 22/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:502929 [5] NCCL INFO Channel 07/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:502929 [5] NCCL INFO Channel 08/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496293:502927 [6] NCCL INFO Channel 23/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:502929 [5] NCCL INFO Channel 09/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:502929 [5] NCCL INFO Channel 10/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:502929 [5] NCCL INFO Channel 11/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:502929 [5] NCCL INFO Channel 12/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:502929 [5] NCCL INFO Channel 13/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:502929 [5] NCCL INFO Channel 14/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:502929 [5] NCCL INFO Channel 15/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:502929 [5] NCCL INFO Channel 16/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:502929 [5] NCCL INFO Channel 17/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:502929 [5] NCCL INFO Channel 18/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:502929 [5] NCCL INFO Channel 19/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:502929 [5] NCCL INFO Channel 20/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:502929 [5] NCCL INFO Channel 21/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:502929 [5] NCCL INFO Channel 22/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496292:502929 [5] NCCL INFO Channel 23/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:496287:502922 [0] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:496287:502922 [0] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:496287:502922 [0] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:496288:502926 [1] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:496288:502926 [1] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:496288:502926 [1] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:496289:502925 [2] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:496289:502925 [2] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:496289:502925 [2] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:496290:502924 [3] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:496290:502924 [3] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:496290:502924 [3] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:496291:502923 [4] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:496291:502923 [4] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:496291:502923 [4] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:496292:502929 [5] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:496292:502929 [5] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:496292:502929 [5] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:496294:502928 [7] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:496293:502927 [6] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:496294:502928 [7] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:496294:502928 [7] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:496293:502927 [6] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:496293:502927 [6] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:496294:502928 [7] NCCL INFO ncclCommInitRank comm 0x7fae3006a840 rank 7 nranks 8 cudaDev 7 nvmlDev 7 busId a01d0 commId 0x27e97f43fd8218c9 - Init COMPLETE +ywang29-vrdb-test2-worker-0:496292:502929 [5] NCCL INFO ncclCommInitRank comm 0x7f198806a8a0 rank 5 nranks 8 cudaDev 5 nvmlDev 5 busId 901d0 commId 0x27e97f43fd8218c9 - Init COMPLETE +ywang29-vrdb-test2-worker-0:496288:502926 [1] NCCL INFO ncclCommInitRank comm 0x7fbdb006ac20 rank 1 nranks 8 cudaDev 1 nvmlDev 1 busId 101d0 commId 0x27e97f43fd8218c9 - Init COMPLETE +ywang29-vrdb-test2-worker-0:496290:502924 [3] NCCL INFO ncclCommInitRank comm 0x7f3cc806b0f0 rank 3 nranks 8 cudaDev 3 nvmlDev 3 busId 201d0 commId 0x27e97f43fd8218c9 - Init COMPLETE +ywang29-vrdb-test2-worker-0:496287:502922 [0] NCCL INFO ncclCommInitRank comm 0x7f0c8806b590 rank 0 nranks 8 cudaDev 0 nvmlDev 0 busId 101c0 commId 0x27e97f43fd8218c9 - Init COMPLETE +ywang29-vrdb-test2-worker-0:496289:502925 [2] NCCL INFO ncclCommInitRank comm 0x7f8fb406ad00 rank 2 nranks 8 cudaDev 2 nvmlDev 2 busId 201c0 commId 0x27e97f43fd8218c9 - Init COMPLETE +ywang29-vrdb-test2-worker-0:496293:502927 [6] NCCL INFO ncclCommInitRank comm 0x7f800006b1c0 rank 6 nranks 8 cudaDev 6 nvmlDev 6 busId a01c0 commId 0x27e97f43fd8218c9 - Init COMPLETE +ywang29-vrdb-test2-worker-0:496291:502923 [4] NCCL INFO ncclCommInitRank comm 0x7f07b406a990 rank 4 nranks 8 cudaDev 4 nvmlDev 4 busId 901c0 commId 0x27e97f43fd8218c9 - Init COMPLETE + 0%| | 1/520 [00:14<2:05:43, 14.53s/it] {'loss': 2.0453, 'grad_norm': 0.0048350191321738945, 'learning_rate': 0.03125, 'epoch': 0.0} + 0%| | 1/520 [00:14<2:05:43, 14.53s/it] 0%| | 2/520 [00:18<1:10:52, 8.21s/it] {'loss': 2.0549, 'grad_norm': 0.005247995502092988, 'learning_rate': 0.0625, 'epoch': 0.0} + 0%| | 2/520 [00:18<1:10:52, 8.21s/it] 1%| | 3/520 [00:21<52:41, 6.11s/it] {'loss': 2.1899, 'grad_norm': 0.006004399320151867, 'learning_rate': 0.09375, 'epoch': 0.01} + 1%| | 3/520 [00:21<52:41, 6.11s/it] 1%| | 4/520 [00:25<44:05, 5.13s/it] {'loss': 1.7201, 'grad_norm': 0.0017709612255715637, 'learning_rate': 0.125, 'epoch': 0.01} + 1%| | 4/520 [00:25<44:05, 5.13s/it] 1%| | 5/520 [00:29<39:21, 4.59s/it] {'loss': 1.7423, 'grad_norm': 0.0015196033571961993, 'learning_rate': 0.15625, 'epoch': 0.01} + 1%| | 5/520 [00:29<39:21, 4.59s/it] 1%| | 6/520 [00:32<36:33, 4.27s/it] {'loss': 1.4094, 'grad_norm': 0.0005342345188768363, 'learning_rate': 0.1875, 'epoch': 0.01} + 1%| | 6/520 [00:32<36:33, 4.27s/it] 1%|▏ | 7/520 [00:36<34:38, 4.05s/it] {'loss': 1.472, 'grad_norm': 0.0005274286038408722, 'learning_rate': 0.21875, 'epoch': 0.01} + 1%|▏ | 7/520 [00:36<34:38, 4.05s/it] 2%|▏ | 8/520 [00:40<35:08, 4.12s/it] {'loss': 1.4855, 'grad_norm': 0.0006231238565361639, 'learning_rate': 0.25, 'epoch': 0.02} + 2%|▏ | 8/520 [00:40<35:08, 4.12s/it] 2%|▏ | 9/520 [00:44<35:12, 4.13s/it] {'loss': 1.5483, 'grad_norm': 0.0006247074017264759, 'learning_rate': 0.28125, 'epoch': 0.02} + 2%|▏ | 9/520 [00:44<35:12, 4.13s/it] 2%|▏ | 10/520 [00:48<33:46, 3.97s/it] {'loss': 1.3751, 'grad_norm': 0.0007646334798139372, 'learning_rate': 0.3125, 'epoch': 0.02} + 2%|▏ | 10/520 [00:48<33:46, 3.97s/it] 2%|▏ | 11/520 [00:52<33:12, 3.91s/it] {'loss': 1.4311, 'grad_norm': 0.0008221866728204538, 'learning_rate': 0.34375, 'epoch': 0.02} + 2%|▏ | 11/520 [00:52<33:12, 3.91s/it] 2%|▏ | 12/520 [00:55<32:25, 3.83s/it] {'loss': 1.3267, 'grad_norm': 0.0012637017485482868, 'learning_rate': 0.375, 'epoch': 0.02} + 2%|▏ | 12/520 [00:55<32:25, 3.83s/it][2025-10-12 05:30:24,730] [WARNING] [stage3.py:2069:step] 1 pytorch allocator cache flushes since last step. this happens when there is high memory pressure and is detrimental to performance. if this is happening frequently consider adjusting settings to reduce memory consumption. If you are unable to make the cache flushes go away consider adding get_accelerator().empty_cache() calls in your training loop to ensure that all ranks flush their caches at the same time + 2%|▎ | 13/520 [01:00<33:35, 3.98s/it] {'loss': 1.3756, 'grad_norm': 0.0014962797057518696, 'learning_rate': 0.40625, 'epoch': 0.03} + 2%|▎ | 13/520 [01:00<33:35, 3.98s/it] 3%|▎ | 14/520 [01:03<32:44, 3.88s/it] {'loss': 1.4303, 'grad_norm': 0.0016182207218081363, 'learning_rate': 0.4375, 'epoch': 0.03} + 3%|▎ | 14/520 [01:03<32:44, 3.88s/it] 3%|▎ | 15/520 [01:07<32:00, 3.80s/it] {'loss': 1.3932, 'grad_norm': 0.0015760151155077107, 'learning_rate': 0.46875, 'epoch': 0.03} + 3%|▎ | 15/520 [01:07<32:00, 3.80s/it] 3%|▎ | 16/520 [01:11<31:35, 3.76s/it] {'loss': 1.3808, 'grad_norm': 0.0020300188382358078, 'learning_rate': 0.5, 'epoch': 0.03} + 3%|▎ | 16/520 [01:11<31:35, 3.76s/it] 3%|▎ | 17/520 [01:14<31:12, 3.72s/it] {'loss': 1.5106, 'grad_norm': 0.0024891700393680754, 'learning_rate': 0.4999951432328845, 'epoch': 0.03} + 3%|▎ | 17/520 [01:14<31:12, 3.72s/it] 3%|▎ | 18/520 [01:18<31:14, 3.73s/it] {'loss': 1.3974, 'grad_norm': 0.0031585088724308957, 'learning_rate': 0.4999805731202437, 'epoch': 0.03} + 3%|▎ | 18/520 [01:18<31:14, 3.73s/it] 4%|▎ | 19/520 [01:22<31:33, 3.78s/it] {'loss': 1.4456, 'grad_norm': 0.003228185720630258, 'learning_rate': 0.4999562902281866, 'epoch': 0.04} + 4%|▎ | 19/520 [01:22<31:33, 3.78s/it] 4%|▍ | 20/520 [01:26<31:49, 3.82s/it] {'loss': 1.4604, 'grad_norm': 0.0046953056759989585, 'learning_rate': 0.49992229550020406, 'epoch': 0.04} + 4%|▍ | 20/520 [01:26<31:49, 3.82s/it] 4%|▍ | 21/520 [01:30<31:58, 3.84s/it] {'loss': 1.6916, 'grad_norm': 0.012960070494433632, 'learning_rate': 0.4998785902571319, 'epoch': 0.04} + 4%|▍ | 21/520 [01:30<31:58, 3.84s/it] 4%|▍ | 22/520 [01:34<32:00, 3.86s/it] {'loss': 1.8679, 'grad_norm': 0.012542157590523396, 'learning_rate': 0.49982517619709965, 'epoch': 0.04} + 4%|▍ | 22/520 [01:34<32:00, 3.86s/it] 4%|▍ | 23/520 [01:38<31:59, 3.86s/it] {'loss': 1.6967, 'grad_norm': 0.006497733188858558, 'learning_rate': 0.4997620553954645, 'epoch': 0.04} + 4%|▍ | 23/520 [01:38<31:59, 3.86s/it] 5%|▍ | 24/520 [01:41<31:40, 3.83s/it] {'loss': 1.6369, 'grad_norm': 0.013495381849010337, 'learning_rate': 0.49968923030473056, 'epoch': 0.05} + 5%|▍ | 24/520 [01:41<31:40, 3.83s/it] 5%|▍ | 25/520 [01:45<31:05, 3.77s/it] {'loss': 1.7842, 'grad_norm': 0.013445805924670316, 'learning_rate': 0.49960670375445415, 'epoch': 0.05} + 5%|▍ | 25/520 [01:45<31:05, 3.77s/it] 5%|▌ | 26/520 [01:49<30:44, 3.73s/it] {'loss': 1.6885, 'grad_norm': 0.00628920532879043, 'learning_rate': 0.499514478951133, 'epoch': 0.05} + 5%|▌ | 26/520 [01:49<30:44, 3.73s/it] 5%|▌ | 27/520 [01:52<30:26, 3.70s/it] {'loss': 1.5762, 'grad_norm': 0.004407963293923971, 'learning_rate': 0.4994125594780822, 'epoch': 0.05} + 5%|▌ | 27/520 [01:52<30:26, 3.70s/it] 5%|▌ | 28/520 [01:56<30:09, 3.68s/it] {'loss': 1.5769, 'grad_norm': 0.00516889039391272, 'learning_rate': 0.499300949295295, 'epoch': 0.05} + 5%|▌ | 28/520 [01:56<30:09, 3.68s/it] 6%|▌ | 29/520 [01:59<30:02, 3.67s/it] {'loss': 1.6871, 'grad_norm': 0.014219057136325375, 'learning_rate': 0.49917965273928877, 'epoch': 0.06} + 6%|▌ | 29/520 [01:59<30:02, 3.67s/it] 6%|▌ | 30/520 [02:03<29:54, 3.66s/it] {'loss': 1.8912, 'grad_norm': 0.006166985037545043, 'learning_rate': 0.4990486745229364, 'epoch': 0.06} + 6%|▌ | 30/520 [02:03<29:54, 3.66s/it] 6%|▌ | 31/520 [02:07<29:44, 3.65s/it] {'loss': 1.6653, 'grad_norm': 0.009362478224730025, 'learning_rate': 0.4989080197352834, 'epoch': 0.06} + 6%|▌ | 31/520 [02:07<29:44, 3.65s/it] 6%|▌ | 32/520 [02:10<29:40, 3.65s/it] {'loss': 1.8081, 'grad_norm': 0.007211012766938582, 'learning_rate': 0.4987576938413504, 'epoch': 0.06} + 6%|▌ | 32/520 [02:10<29:40, 3.65s/it] 6%|▋ | 33/520 [02:14<29:40, 3.66s/it] {'loss': 1.6965, 'grad_norm': 0.031128390263136234, 'learning_rate': 0.49859770268191994, 'epoch': 0.06} + 6%|▋ | 33/520 [02:14<29:40, 3.66s/it] 7%|▋ | 34/520 [02:18<29:56, 3.70s/it] {'loss': 1.8156, 'grad_norm': 0.024041869985611492, 'learning_rate': 0.4984280524733107, 'epoch': 0.07} + 7%|▋ | 34/520 [02:18<29:56, 3.70s/it] 7%|▋ | 35/520 [02:22<29:51, 3.69s/it] {'loss': 1.8895, 'grad_norm': 0.014990081286728001, 'learning_rate': 0.49824874980713485, 'epoch': 0.07} + 7%|▋ | 35/520 [02:22<29:51, 3.69s/it] 7%|▋ | 36/520 [02:25<29:52, 3.70s/it] {'loss': 1.8186, 'grad_norm': 0.0044952445122991985, 'learning_rate': 0.49805980165004304, 'epoch': 0.07} + 7%|▋ | 36/520 [02:25<29:52, 3.70s/it] 7%|▋ | 37/520 [02:29<29:42, 3.69s/it] {'loss': 1.9585, 'grad_norm': 0.013027967338171305, 'learning_rate': 0.4978612153434526, 'epoch': 0.07} + 7%|▋ | 37/520 [02:29<29:42, 3.69s/it] 7%|▋ | 38/520 [02:33<29:29, 3.67s/it] {'loss': 1.8476, 'grad_norm': 0.0066559288413444335, 'learning_rate': 0.4976529986032632, 'epoch': 0.07} + 7%|▋ | 38/520 [02:33<29:29, 3.67s/it] 8%|▊ | 39/520 [02:36<29:16, 3.65s/it] {'loss': 1.7039, 'grad_norm': 0.027478614213411447, 'learning_rate': 0.4974351595195565, 'epoch': 0.07} + 8%|▊ | 39/520 [02:36<29:16, 3.65s/it] 8%|▊ | 40/520 [02:40<29:14, 3.66s/it] {'loss': 1.6794, 'grad_norm': 0.0029579225485126277, 'learning_rate': 0.4972077065562821, 'epoch': 0.08} + 8%|▊ | 40/520 [02:40<29:14, 3.66s/it] 8%|▊ | 41/520 [02:43<29:09, 3.65s/it] {'loss': 1.6381, 'grad_norm': 0.0034126213107306706, 'learning_rate': 0.49697064855092865, 'epoch': 0.08} + 8%|▊ | 41/520 [02:43<29:09, 3.65s/it] 8%|▊ | 42/520 [02:47<28:59, 3.64s/it] {'loss': 1.6739, 'grad_norm': 0.003404549373303303, 'learning_rate': 0.4967239947141803, 'epoch': 0.08} + 8%|▊ | 42/520 [02:47<28:59, 3.64s/it] 8%|▊ | 43/520 [02:51<29:12, 3.67s/it] {'loss': 1.819, 'grad_norm': 0.0057991051650208856, 'learning_rate': 0.49646775462955894, 'epoch': 0.08} + 8%|▊ | 43/520 [02:51<29:12, 3.67s/it] 8%|▊ | 44/520 [02:55<29:21, 3.70s/it] {'loss': 1.8945, 'grad_norm': 0.004901391892227444, 'learning_rate': 0.49620193825305203, 'epoch': 0.08} + 8%|▊ | 44/520 [02:55<29:21, 3.70s/it] 9%|▊ | 45/520 [02:58<29:19, 3.70s/it] {'loss': 1.6392, 'grad_norm': 0.0028453667276623334, 'learning_rate': 0.49592655591272533, 'epoch': 0.09} + 9%|▊ | 45/520 [02:58<29:19, 3.70s/it] 9%|▉ | 46/520 [03:02<29:25, 3.73s/it] {'loss': 1.9381, 'grad_norm': 0.0034848916289823345, 'learning_rate': 0.4956416183083221, 'epoch': 0.09} + 9%|▉ | 46/520 [03:02<29:25, 3.73s/it] 9%|▉ | 47/520 [03:06<29:16, 3.71s/it] {'loss': 1.6018, 'grad_norm': 0.002296444394723713, 'learning_rate': 0.49534713651084694, 'epoch': 0.09} + 9%|▉ | 47/520 [03:06<29:16, 3.71s/it] 9%|▉ | 48/520 [03:09<28:58, 3.68s/it] {'loss': 1.5695, 'grad_norm': 0.0021679296879417993, 'learning_rate': 0.49504312196213596, 'epoch': 0.09} + 9%|▉ | 48/520 [03:09<28:58, 3.68s/it] 9%|▉ | 49/520 [03:13<28:45, 3.66s/it] {'loss': 1.5858, 'grad_norm': 0.0023318659254503485, 'learning_rate': 0.4947295864744121, 'epoch': 0.09} + 9%|▉ | 49/520 [03:13<28:45, 3.66s/it] 10%|▉ | 50/520 [03:17<28:35, 3.65s/it] {'loss': 1.5596, 'grad_norm': 0.0018184707612209237, 'learning_rate': 0.49440654222982616, 'epoch': 0.1} + 10%|▉ | 50/520 [03:17<28:35, 3.65s/it] 10%|▉ | 51/520 [03:20<28:26, 3.64s/it] {'loss': 1.4886, 'grad_norm': 0.0021982178594285727, 'learning_rate': 0.49407400177998334, 'epoch': 0.1} + 10%|▉ | 51/520 [03:20<28:26, 3.64s/it] 10%|█ | 52/520 [03:24<28:18, 3.63s/it] {'loss': 1.6309, 'grad_norm': 0.0020663723024169625, 'learning_rate': 0.4937319780454559, 'epoch': 0.1} + 10%|█ | 52/520 [03:24<28:18, 3.63s/it] 10%|█ | 53/520 [03:27<28:09, 3.62s/it] {'loss': 1.6132, 'grad_norm': 0.0018086684568946357, 'learning_rate': 0.4933804843152808, 'epoch': 0.1} + 10%|█ | 53/520 [03:27<28:09, 3.62s/it] 10%|█ | 54/520 [03:31<28:10, 3.63s/it] {'loss': 1.5001, 'grad_norm': 0.0016815848523741804, 'learning_rate': 0.4930195342464436, 'epoch': 0.1} + 10%|█ | 54/520 [03:31<28:10, 3.63s/it] 11%|█ | 55/520 [03:35<28:09, 3.63s/it] {'loss': 1.4842, 'grad_norm': 0.0018441675450389275, 'learning_rate': 0.49264914186334774, 'epoch': 0.11} + 11%|█ | 55/520 [03:35<28:09, 3.63s/it] 11%|█ | 56/520 [03:38<28:06, 3.63s/it] {'loss': 1.6145, 'grad_norm': 0.0016945242354041107, 'learning_rate': 0.4922693215572695, 'epoch': 0.11} + 11%|█ | 56/520 [03:38<28:06, 3.63s/it] 11%|█ | 57/520 [03:42<28:10, 3.65s/it] {'loss': 1.4471, 'grad_norm': 0.0017372080431781356, 'learning_rate': 0.4918800880857991, 'epoch': 0.11} + 11%|█ | 57/520 [03:42<28:10, 3.65s/it] 11%|█ | 58/520 [03:46<28:36, 3.72s/it] {'loss': 1.6233, 'grad_norm': 0.0014664212806088132, 'learning_rate': 0.49148145657226705, 'epoch': 0.11} + 11%|█ | 58/520 [03:46<28:36, 3.72s/it] 11%|█▏ | 59/520 [03:50<28:39, 3.73s/it] {'loss': 1.6114, 'grad_norm': 0.0036110123301029804, 'learning_rate': 0.49107344250515694, 'epoch': 0.11} + 11%|█▏ | 59/520 [03:50<28:39, 3.73s/it] 12%|█▏ | 60/520 [03:53<28:28, 3.71s/it] {'loss': 1.5312, 'grad_norm': 0.0013255480680444249, 'learning_rate': 0.490656061737503, 'epoch': 0.12} + 12%|█▏ | 60/520 [03:53<28:28, 3.71s/it] 12%|█▏ | 61/520 [03:57<28:09, 3.68s/it] {'loss': 1.78, 'grad_norm': 0.002183843445592421, 'learning_rate': 0.49022933048627493, 'epoch': 0.12} + 12%|█▏ | 61/520 [03:57<28:09, 3.68s/it] 12%|█▏ | 62/520 [04:01<28:04, 3.68s/it] {'loss': 1.5, 'grad_norm': 0.0013921157990131245, 'learning_rate': 0.48979326533174694, 'epoch': 0.12} + 12%|█▏ | 62/520 [04:01<28:04, 3.68s/it] 12%|█▏ | 63/520 [04:04<28:00, 3.68s/it] {'loss': 1.4945, 'grad_norm': 0.0014551601060394214, 'learning_rate': 0.4893478832168545, 'epoch': 0.12} + 12%|█▏ | 63/520 [04:04<28:00, 3.68s/it] 12%|█▏ | 64/520 [04:08<27:47, 3.66s/it] {'loss': 1.5108, 'grad_norm': 0.0013660412585215366, 'learning_rate': 0.4888932014465352, 'epoch': 0.12} + 12%|█▏ | 64/520 [04:08<27:47, 3.66s/it] 12%|█▎ | 65/520 [04:12<27:47, 3.66s/it] {'loss': 1.5364, 'grad_norm': 0.001674238388410394, 'learning_rate': 0.4884292376870567, 'epoch': 0.12} + 12%|█▎ | 65/520 [04:12<27:47, 3.66s/it] 13%|█▎ | 66/520 [04:15<27:41, 3.66s/it] {'loss': 1.4897, 'grad_norm': 0.0013475674236556746, 'learning_rate': 0.4879560099653306, 'epoch': 0.13} + 13%|█▎ | 66/520 [04:15<27:41, 3.66s/it] 13%|█▎ | 67/520 [04:19<27:35, 3.65s/it] {'loss': 1.3585, 'grad_norm': 0.0012250385996308813, 'learning_rate': 0.48747353666821147, 'epoch': 0.13} + 13%|█▎ | 67/520 [04:19<27:35, 3.65s/it] 13%|█▎ | 68/520 [04:23<27:37, 3.67s/it] {'loss': 1.4063, 'grad_norm': 0.0012593918707512768, 'learning_rate': 0.48698183654178295, 'epoch': 0.13} + 13%|█▎ | 68/520 [04:23<27:37, 3.67s/it] 13%|█▎ | 69/520 [04:26<27:38, 3.68s/it] {'loss': 1.383, 'grad_norm': 0.0012360118881901414, 'learning_rate': 0.486480928690629, 'epoch': 0.13} + 13%|█▎ | 69/520 [04:26<27:38, 3.68s/it] 13%|█▎ | 70/520 [04:30<27:37, 3.68s/it] {'loss': 1.4399, 'grad_norm': 0.0013603814439225549, 'learning_rate': 0.4859708325770919, 'epoch': 0.13} + 13%|█▎ | 70/520 [04:30<27:37, 3.68s/it] 14%|█▎ | 71/520 [04:34<27:28, 3.67s/it] {'loss': 1.3483, 'grad_norm': 0.0011243308515908536, 'learning_rate': 0.48545156802051587, 'epoch': 0.14} + 14%|█▎ | 71/520 [04:34<27:28, 3.67s/it] 14%|█▍ | 72/520 [04:37<27:20, 3.66s/it] {'loss': 1.5079, 'grad_norm': 0.0011953517988414703, 'learning_rate': 0.4849231551964771, 'epoch': 0.14} + 14%|█▍ | 72/520 [04:37<27:20, 3.66s/it] 14%|█▍ | 73/520 [04:41<27:14, 3.66s/it] {'loss': 1.3421, 'grad_norm': 0.001175582148696314, 'learning_rate': 0.48438561463599983, 'epoch': 0.14} + 14%|█▍ | 73/520 [04:41<27:14, 3.66s/it] 14%|█▍ | 74/520 [04:45<27:11, 3.66s/it] {'loss': 1.452, 'grad_norm': 0.001248865107846754, 'learning_rate': 0.4838389672247585, 'epoch': 0.14} + 14%|█▍ | 74/520 [04:45<27:11, 3.66s/it] 14%|█▍ | 75/520 [04:48<27:01, 3.64s/it] {'loss': 1.35, 'grad_norm': 0.001149571311482633, 'learning_rate': 0.4832832342022666, 'epoch': 0.14} + 14%|█▍ | 75/520 [04:48<27:01, 3.64s/it] 15%|█▍ | 76/520 [04:52<26:56, 3.64s/it] {'loss': 1.765, 'grad_norm': 0.0032179650771626787, 'learning_rate': 0.48271843716105106, 'epoch': 0.15} + 15%|█▍ | 76/520 [04:52<26:56, 3.64s/it] 15%|█▍ | 77/520 [04:55<26:50, 3.64s/it] {'loss': 1.2663, 'grad_norm': 0.0012655886425706033, 'learning_rate': 0.4821445980458134, 'epoch': 0.15} + 15%|█▍ | 77/520 [04:55<26:50, 3.64s/it] 15%|█▌ | 78/520 [04:59<26:44, 3.63s/it] {'loss': 1.4145, 'grad_norm': 0.0011043138593043356, 'learning_rate': 0.48156173915257716, 'epoch': 0.15} + 15%|█▌ | 78/520 [04:59<26:44, 3.63s/it] 15%|█▌ | 79/520 [05:03<26:40, 3.63s/it] {'loss': 1.3851, 'grad_norm': 0.0009494049621623314, 'learning_rate': 0.4809698831278217, 'epoch': 0.15} + 15%|█▌ | 79/520 [05:03<26:40, 3.63s/it] 15%|█▌ | 80/520 [05:06<26:37, 3.63s/it] {'loss': 1.7319, 'grad_norm': 0.0021749338520303998, 'learning_rate': 0.4803690529676019, 'epoch': 0.15} + 15%|█▌ | 80/520 [05:06<26:37, 3.63s/it] 16%|█▌ | 81/520 [05:10<26:37, 3.64s/it] {'loss': 1.5364, 'grad_norm': 0.001269327645227412, 'learning_rate': 0.47975927201665514, 'epoch': 0.16} + 16%|█▌ | 81/520 [05:10<26:37, 3.64s/it] 16%|█▌ | 82/520 [05:14<26:34, 3.64s/it] {'loss': 1.4619, 'grad_norm': 0.0010728471449508827, 'learning_rate': 0.47914056396749405, 'epoch': 0.16} + 16%|█▌ | 82/520 [05:14<26:34, 3.64s/it] 16%|█▌ | 83/520 [05:17<26:21, 3.62s/it] {'loss': 1.5017, 'grad_norm': 0.0010779603768931259, 'learning_rate': 0.4785129528594858, 'epoch': 0.16} + 16%|█▌ | 83/520 [05:17<26:21, 3.62s/it] 16%|█▌ | 84/520 [05:21<26:16, 3.62s/it] {'loss': 1.4704, 'grad_norm': 0.0010782810410721161, 'learning_rate': 0.4778764630779183, 'epoch': 0.16} + 16%|█▌ | 84/520 [05:21<26:16, 3.62s/it] 16%|█▋ | 85/520 [05:24<26:15, 3.62s/it] {'loss': 1.4751, 'grad_norm': 0.0009285187330158229, 'learning_rate': 0.47723111935305274, 'epoch': 0.16} + 16%|█▋ | 85/520 [05:24<26:15, 3.62s/it] 17%|█▋ | 86/520 [05:28<26:08, 3.61s/it] {'loss': 1.5309, 'grad_norm': 0.0009714354553424872, 'learning_rate': 0.4765769467591625, 'epoch': 0.17} + 17%|█▋ | 86/520 [05:28<26:08, 3.61s/it] 17%|█▋ | 87/520 [05:32<26:07, 3.62s/it] {'loss': 1.6476, 'grad_norm': 0.001427170073541938, 'learning_rate': 0.4759139707135592, 'epoch': 0.17} + 17%|█▋ | 87/520 [05:32<26:07, 3.62s/it] 17%|█▋ | 88/520 [05:35<26:17, 3.65s/it] {'loss': 1.7012, 'grad_norm': 0.0012810422133682532, 'learning_rate': 0.47524221697560476, 'epoch': 0.17} + 17%|█▋ | 88/520 [05:35<26:17, 3.65s/it] 17%|█▋ | 89/520 [05:39<26:35, 3.70s/it] {'loss': 1.4398, 'grad_norm': 0.0009077796618701228, 'learning_rate': 0.47456171164571104, 'epoch': 0.17} + 17%|█▋ | 89/520 [05:39<26:35, 3.70s/it] 17%|█▋ | 90/520 [05:43<26:46, 3.74s/it] {'loss': 1.3739, 'grad_norm': 0.0010923322878835184, 'learning_rate': 0.47387248116432523, 'epoch': 0.17} + 17%|█▋ | 90/520 [05:43<26:46, 3.74s/it] 18%|█▊ | 91/520 [05:47<26:50, 3.75s/it] {'loss': 1.465, 'grad_norm': 0.0009822964818478317, 'learning_rate': 0.4731745523109029, 'epoch': 0.17} + 18%|█▊ | 91/520 [05:47<26:50, 3.75s/it] 18%|█▊ | 92/520 [05:51<26:57, 3.78s/it] {'loss': 1.4059, 'grad_norm': 0.0008584405380803537, 'learning_rate': 0.4724679522028672, 'epoch': 0.18} + 18%|█▊ | 92/520 [05:51<26:57, 3.78s/it] 18%|█▊ | 93/520 [05:54<26:59, 3.79s/it] {'loss': 1.3956, 'grad_norm': 0.0009686621103157136, 'learning_rate': 0.4717527082945554, 'epoch': 0.18} + 18%|█▊ | 93/520 [05:54<26:59, 3.79s/it] 18%|█▊ | 94/520 [05:58<26:55, 3.79s/it] {'loss': 1.5017, 'grad_norm': 0.0008577865933897866, 'learning_rate': 0.47102884837615244, 'epoch': 0.18} + 18%|█▊ | 94/520 [05:58<26:55, 3.79s/it] 18%|█▊ | 95/520 [06:02<26:55, 3.80s/it] {'loss': 1.3767, 'grad_norm': 0.0009761482363157921, 'learning_rate': 0.4702964005726106, 'epoch': 0.18} + 18%|█▊ | 95/520 [06:02<26:55, 3.80s/it] 18%|█▊ | 96/520 [06:06<26:57, 3.82s/it] {'loss': 1.3938, 'grad_norm': 0.0010089735608678643, 'learning_rate': 0.46955539334255714, 'epoch': 0.18} + 18%|█▊ | 96/520 [06:06<26:57, 3.82s/it] 19%|█▊ | 97/520 [06:10<26:53, 3.81s/it] {'loss': 1.3455, 'grad_norm': 0.000861060211854971, 'learning_rate': 0.46880585547718845, 'epoch': 0.19} + 19%|█▊ | 97/520 [06:10<26:53, 3.81s/it] 19%|█▉ | 98/520 [06:14<26:49, 3.81s/it] {'loss': 1.3551, 'grad_norm': 0.0006903864605535305, 'learning_rate': 0.46804781609915136, 'epoch': 0.19} + 19%|█▉ | 98/520 [06:14<26:49, 3.81s/it] 19%|█▉ | 99/520 [06:17<26:46, 3.82s/it] {'loss': 1.3789, 'grad_norm': 0.0009258948874760369, 'learning_rate': 0.4672813046614116, 'epoch': 0.19} + 19%|█▉ | 99/520 [06:17<26:46, 3.82s/it] 19%|█▉ | 100/520 [06:21<26:44, 3.82s/it] {'loss': 1.4986, 'grad_norm': 0.0008119021039269241, 'learning_rate': 0.4665063509461097, 'epoch': 0.19} + 19%|█▉ | 100/520 [06:21<26:44, 3.82s/it] 19%|█▉ | 101/520 [06:25<26:42, 3.82s/it] {'loss': 1.3837, 'grad_norm': 0.0008420486738650505, 'learning_rate': 0.46572298506340326, 'epoch': 0.19} + 19%|█▉ | 101/520 [06:25<26:42, 3.82s/it] 20%|█▉ | 102/520 [06:29<26:34, 3.82s/it] {'loss': 1.371, 'grad_norm': 0.0008208927326748219, 'learning_rate': 0.4649312374502976, 'epoch': 0.2} + 20%|█▉ | 102/520 [06:29<26:34, 3.82s/it] 20%|█▉ | 103/520 [06:33<26:32, 3.82s/it] {'loss': 1.3218, 'grad_norm': 0.0008631937395504688, 'learning_rate': 0.4641311388694629, 'epoch': 0.2} + 20%|█▉ | 103/520 [06:33<26:32, 3.82s/it] 20%|██ | 104/520 [06:36<26:24, 3.81s/it] {'loss': 1.3953, 'grad_norm': 0.0008065481313748699, 'learning_rate': 0.4633227204080389, 'epoch': 0.2} + 20%|██ | 104/520 [06:36<26:24, 3.81s/it] 20%|██ | 105/520 [06:40<26:16, 3.80s/it] {'loss': 1.3933, 'grad_norm': 0.0007100786356673988, 'learning_rate': 0.4625060134764273, 'epoch': 0.2} + 20%|██ | 105/520 [06:40<26:16, 3.80s/it] 20%|██ | 106/520 [06:44<26:13, 3.80s/it] {'loss': 1.5111, 'grad_norm': 0.0009176930849225255, 'learning_rate': 0.46168104980707103, 'epoch': 0.2} + 20%|██ | 106/520 [06:44<26:13, 3.80s/it] 21%|██ | 107/520 [06:48<26:11, 3.81s/it] {'loss': 1.4761, 'grad_norm': 0.0008084372218646724, 'learning_rate': 0.46084786145322143, 'epoch': 0.21} + 21%|██ | 107/520 [06:48<26:11, 3.81s/it] 21%|██ | 108/520 [06:52<26:12, 3.82s/it] {'loss': 1.3233, 'grad_norm': 0.0007803607390399065, 'learning_rate': 0.4600064807876929, 'epoch': 0.21} + 21%|██ | 108/520 [06:52<26:12, 3.82s/it] 21%|██ | 109/520 [06:56<26:11, 3.82s/it] {'loss': 1.4823, 'grad_norm': 0.0007627894167218986, 'learning_rate': 0.45915694050160494, 'epoch': 0.21} + 21%|██ | 109/520 [06:56<26:11, 3.82s/it] 21%|██ | 110/520 [06:59<26:05, 3.82s/it] {'loss': 1.5201, 'grad_norm': 0.0009084523817193894, 'learning_rate': 0.45829927360311223, 'epoch': 0.21} + 21%|██ | 110/520 [06:59<26:05, 3.82s/it] 21%|██▏ | 111/520 [07:03<26:02, 3.82s/it] {'loss': 1.5162, 'grad_norm': 0.0008210333923563192, 'learning_rate': 0.4574335134161219, 'epoch': 0.21} + 21%|██▏ | 111/520 [07:03<26:02, 3.82s/it] 22%|██▏ | 112/520 [07:07<25:55, 3.81s/it] {'loss': 1.4069, 'grad_norm': 0.0007541997568416729, 'learning_rate': 0.4565596935789987, 'epoch': 0.22} + 22%|██▏ | 112/520 [07:07<25:55, 3.81s/it] 22%|██▏ | 113/520 [07:11<25:53, 3.82s/it] {'loss': 1.2818, 'grad_norm': 0.0009203785154359074, 'learning_rate': 0.45567784804325834, 'epoch': 0.22} + 22%|██▏ | 113/520 [07:11<25:53, 3.82s/it] 22%|██▏ | 114/520 [07:15<25:46, 3.81s/it] {'loss': 1.3929, 'grad_norm': 0.0007148933285790398, 'learning_rate': 0.454788011072248, 'epoch': 0.22} + 22%|██▏ | 114/520 [07:15<25:46, 3.81s/it] 22%|██▏ | 115/520 [07:18<25:43, 3.81s/it] {'loss': 1.511, 'grad_norm': 0.0008770033687124297, 'learning_rate': 0.45389021723981504, 'epoch': 0.22} + 22%|██▏ | 115/520 [07:18<25:43, 3.81s/it] 22%|██▏ | 116/520 [07:22<25:31, 3.79s/it] {'loss': 1.5017, 'grad_norm': 0.000751309063249999, 'learning_rate': 0.4529845014289642, 'epoch': 0.22} + 22%|██▏ | 116/520 [07:22<25:31, 3.79s/it] 22%|██▎ | 117/520 [07:26<25:13, 3.75s/it] {'loss': 1.4713, 'grad_norm': 0.0007625743622111709, 'learning_rate': 0.45207089883050133, 'epoch': 0.23} + 22%|██▎ | 117/520 [07:26<25:13, 3.75s/it] 23%|██▎ | 118/520 [07:29<25:01, 3.74s/it] {'loss': 1.3695, 'grad_norm': 0.0007662741356513081, 'learning_rate': 0.4511494449416671, 'epoch': 0.23} + 23%|██▎ | 118/520 [07:29<25:01, 3.74s/it] 23%|██▎ | 119/520 [07:33<24:50, 3.72s/it] {'loss': 1.3168, 'grad_norm': 0.0007546353496376597, 'learning_rate': 0.450220175564757, 'epoch': 0.23} + 23%|██▎ | 119/520 [07:33<24:50, 3.72s/it] 23%|██▎ | 120/520 [07:37<24:51, 3.73s/it] {'loss': 1.3534, 'grad_norm': 0.0008550992685723979, 'learning_rate': 0.4492831268057306, 'epoch': 0.23} + 23%|██▎ | 120/520 [07:37<24:51, 3.73s/it] 23%|██▎ | 121/520 [07:41<24:39, 3.71s/it] {'loss': 1.3981, 'grad_norm': 0.0010048216675046266, 'learning_rate': 0.4483383350728088, 'epoch': 0.23} + 23%|██▎ | 121/520 [07:41<24:39, 3.71s/it] 23%|██▎ | 122/520 [07:44<24:28, 3.69s/it] {'loss': 1.2783, 'grad_norm': 0.0007091477264636573, 'learning_rate': 0.44738583707505886, 'epoch': 0.23} + 23%|██▎ | 122/520 [07:44<24:28, 3.69s/it] 24%|██▎ | 123/520 [07:48<24:23, 3.69s/it] {'loss': 1.5437, 'grad_norm': 0.0008911698287248476, 'learning_rate': 0.44642566982096843, 'epoch': 0.24} + 24%|██▎ | 123/520 [07:48<24:23, 3.69s/it] 24%|██▍ | 124/520 [07:52<24:17, 3.68s/it] {'loss': 1.368, 'grad_norm': 0.000973657924293641, 'learning_rate': 0.4454578706170075, 'epoch': 0.24} + 24%|██▍ | 124/520 [07:52<24:17, 3.68s/it] 24%|██▍ | 125/520 [07:55<24:13, 3.68s/it] {'loss': 1.3466, 'grad_norm': 0.0006931126788995097, 'learning_rate': 0.4444824770661787, 'epoch': 0.24} + 24%|██▍ | 125/520 [07:55<24:13, 3.68s/it] 24%|██▍ | 126/520 [08:00<25:24, 3.87s/it] {'loss': 1.4636, 'grad_norm': 0.0006074337106804465, 'learning_rate': 0.4434995270665569, 'epoch': 0.24} + 24%|██▍ | 126/520 [08:00<25:24, 3.87s/it] 24%|██▍ | 127/520 [08:03<24:56, 3.81s/it] {'loss': 1.3269, 'grad_norm': 0.000678353396459727, 'learning_rate': 0.4425090588098157, 'epoch': 0.24} + 24%|██▍ | 127/520 [08:03<24:56, 3.81s/it] 25%|██▍ | 128/520 [08:07<24:49, 3.80s/it] {'loss': 1.3865, 'grad_norm': 0.0007904723559580812, 'learning_rate': 0.4415111107797445, 'epoch': 0.25} + 25%|██▍ | 128/520 [08:07<24:49, 3.80s/it] 25%|██▍ | 129/520 [08:11<24:49, 3.81s/it] {'loss': 1.3186, 'grad_norm': 0.0005518740531683538, 'learning_rate': 0.4405057217507527, 'epoch': 0.25} + 25%|██▍ | 129/520 [08:11<24:49, 3.81s/it] 25%|██▌ | 130/520 [08:15<24:46, 3.81s/it] {'loss': 1.3833, 'grad_norm': 0.0006220852814472575, 'learning_rate': 0.43949293078636326, 'epoch': 0.25} + 25%|██▌ | 130/520 [08:15<24:46, 3.81s/it] 25%|██▌ | 131/520 [08:18<24:42, 3.81s/it] {'loss': 1.4275, 'grad_norm': 0.0006222629143781254, 'learning_rate': 0.43847277723769496, 'epoch': 0.25} + 25%|██▌ | 131/520 [08:18<24:42, 3.81s/it] 25%|██▌ | 132/520 [08:22<24:36, 3.81s/it] {'loss': 1.4233, 'grad_norm': 0.0008216697392849457, 'learning_rate': 0.43744530074193355, 'epoch': 0.25} + 25%|██▌ | 132/520 [08:22<24:36, 3.81s/it] 26%|██▌ | 133/520 [08:26<24:44, 3.84s/it] {'loss': 1.3355, 'grad_norm': 0.0006495993688545539, 'learning_rate': 0.43641054122079137, 'epoch': 0.26} + 26%|██▌ | 133/520 [08:26<24:44, 3.84s/it] 26%|██▌ | 134/520 [08:30<24:19, 3.78s/it] {'loss': 1.4067, 'grad_norm': 0.00060446240708835, 'learning_rate': 0.43536853887895666, 'epoch': 0.26} + 26%|██▌ | 134/520 [08:30<24:19, 3.78s/it] 26%|██▌ | 135/520 [08:33<23:58, 3.74s/it] {'loss': 1.4829, 'grad_norm': 0.0006415544538125546, 'learning_rate': 0.434319334202531, 'epoch': 0.26} + 26%|██▌ | 135/520 [08:33<23:58, 3.74s/it] 26%|██▌ | 136/520 [08:37<23:43, 3.71s/it] {'loss': 1.4137, 'grad_norm': 0.00072573383784569, 'learning_rate': 0.43326296795745656, 'epoch': 0.26} + 26%|██▌ | 136/520 [08:37<23:43, 3.71s/it] 26%|██▋ | 137/520 [08:41<23:33, 3.69s/it] {'loss': 1.3286, 'grad_norm': 0.0008655827604180089, 'learning_rate': 0.4321994811879321, 'epoch': 0.26} + 26%|██▋ | 137/520 [08:41<23:33, 3.69s/it] 27%|██▋ | 138/520 [08:44<23:25, 3.68s/it] {'loss': 1.3426, 'grad_norm': 0.0005829027214342389, 'learning_rate': 0.43112891521481816, 'epoch': 0.27} + 27%|██▋ | 138/520 [08:44<23:25, 3.68s/it] 27%|██▋ | 139/520 [08:48<23:15, 3.66s/it] {'loss': 1.3329, 'grad_norm': 0.0006098680250596359, 'learning_rate': 0.43005131163403165, 'epoch': 0.27} + 27%|██▋ | 139/520 [08:48<23:15, 3.66s/it] 27%|██▋ | 140/520 [08:52<23:09, 3.66s/it] {'loss': 1.4717, 'grad_norm': 0.0007655328882823075, 'learning_rate': 0.4289667123149296, 'epoch': 0.27} + 27%|██▋ | 140/520 [08:52<23:09, 3.66s/it] 27%|██▋ | 141/520 [08:55<23:04, 3.65s/it] {'loss': 1.4494, 'grad_norm': 0.0005273018752277501, 'learning_rate': 0.42787515939868254, 'epoch': 0.27} + 27%|██▋ | 141/520 [08:55<23:04, 3.65s/it] 27%|██▋ | 142/520 [08:59<22:55, 3.64s/it] {'loss': 1.5274, 'grad_norm': 0.0009148338157727112, 'learning_rate': 0.42677669529663687, 'epoch': 0.27} + 27%|██▋ | 142/520 [08:59<22:55, 3.64s/it] 28%|██▊ | 143/520 [09:03<22:50, 3.63s/it] {'loss': 1.3732, 'grad_norm': 0.0009367152350295844, 'learning_rate': 0.4256713626886673, 'epoch': 0.28} + 28%|██▊ | 143/520 [09:03<22:50, 3.63s/it] 28%|██▊ | 144/520 [09:06<22:42, 3.62s/it] {'loss': 1.3245, 'grad_norm': 0.0006865939150974411, 'learning_rate': 0.4245592045215182, 'epoch': 0.28} + 28%|██▊ | 144/520 [09:06<22:42, 3.62s/it] 28%|██▊ | 145/520 [09:10<22:39, 3.63s/it] {'loss': 1.2561, 'grad_norm': 0.0005094509459239976, 'learning_rate': 0.4234402640071354, 'epoch': 0.28} + 28%|██▊ | 145/520 [09:10<22:39, 3.63s/it] 28%|██▊ | 146/520 [09:13<22:34, 3.62s/it] {'loss': 1.552, 'grad_norm': 0.0006714815141171796, 'learning_rate': 0.42231458462098675, 'epoch': 0.28} + 28%|██▊ | 146/520 [09:13<22:34, 3.62s/it] 28%|██▊ | 147/520 [09:17<22:30, 3.62s/it] {'loss': 1.2972, 'grad_norm': 0.0006034719002867865, 'learning_rate': 0.4211822101003734, 'epoch': 0.28} + 28%|██▊ | 147/520 [09:17<22:30, 3.62s/it] 28%|██▊ | 148/520 [09:21<22:26, 3.62s/it] {'loss': 1.3355, 'grad_norm': 0.0006461387605371646, 'learning_rate': 0.4200431844427298, 'epoch': 0.28} + 28%|██▊ | 148/520 [09:21<22:26, 3.62s/it] 29%|██▊ | 149/520 [09:24<22:22, 3.62s/it] {'loss': 1.2855, 'grad_norm': 0.0005696710963240196, 'learning_rate': 0.4188975519039151, 'epoch': 0.29} + 29%|██▊ | 149/520 [09:24<22:22, 3.62s/it] 29%|██▉ | 150/520 [09:28<22:17, 3.61s/it] {'loss': 1.4949, 'grad_norm': 0.0005763035841768691, 'learning_rate': 0.4177453569964925, 'epoch': 0.29} + 29%|██▉ | 150/520 [09:28<22:17, 3.61s/it] 29%|██▉ | 151/520 [09:31<22:14, 3.62s/it] {'loss': 1.324, 'grad_norm': 0.0006582432424764241, 'learning_rate': 0.416586644488001, 'epoch': 0.29} + 29%|██▉ | 151/520 [09:31<22:14, 3.62s/it] 29%|██▉ | 152/520 [09:35<22:11, 3.62s/it] {'loss': 1.2849, 'grad_norm': 0.0007414422476963738, 'learning_rate': 0.4154214593992149, 'epoch': 0.29} + 29%|██▉ | 152/520 [09:35<22:11, 3.62s/it] 29%|██▉ | 153/520 [09:39<22:04, 3.61s/it] {'loss': 1.3307, 'grad_norm': 0.0006346696858794366, 'learning_rate': 0.41424984700239514, 'epoch': 0.29} + 29%|██▉ | 153/520 [09:39<22:04, 3.61s/it] 30%|██▉ | 154/520 [09:42<22:01, 3.61s/it] {'loss': 1.4147, 'grad_norm': 0.0005413991265607093, 'learning_rate': 0.4130718528195303, 'epoch': 0.3} + 30%|██▉ | 154/520 [09:42<22:01, 3.61s/it] 30%|██▉ | 155/520 [09:46<21:57, 3.61s/it] {'loss': 1.3091, 'grad_norm': 0.0005848317430683962, 'learning_rate': 0.4118875226205676, 'epoch': 0.3} + 30%|██▉ | 155/520 [09:46<21:57, 3.61s/it] 30%|███ | 156/520 [09:49<21:51, 3.60s/it] {'loss': 1.356, 'grad_norm': 0.0006738140816241061, 'learning_rate': 0.4106969024216348, 'epoch': 0.3} + 30%|███ | 156/520 [09:49<21:51, 3.60s/it] 30%|███ | 157/520 [09:53<21:50, 3.61s/it] {'loss': 1.5489, 'grad_norm': 0.0007505646826407663, 'learning_rate': 0.40950003848325217, 'epoch': 0.3} + 30%|███ | 157/520 [09:53<21:50, 3.61s/it] 30%|███ | 158/520 [09:57<21:50, 3.62s/it] {'loss': 1.3266, 'grad_norm': 0.0007301164188513719, 'learning_rate': 0.408296977308535, 'epoch': 0.3} + 30%|███ | 158/520 [09:57<21:50, 3.62s/it] 31%|███ | 159/520 [10:00<21:51, 3.63s/it] {'loss': 1.3472, 'grad_norm': 0.0008333822937532313, 'learning_rate': 0.4070877656413868, 'epoch': 0.31} + 31%|███ | 159/520 [10:00<21:51, 3.63s/it] 31%|███ | 160/520 [10:04<21:49, 3.64s/it] {'loss': 1.3909, 'grad_norm': 0.0006598470884431461, 'learning_rate': 0.4058724504646834, 'epoch': 0.31} + 31%|███ | 160/520 [10:04<21:49, 3.64s/it] 31%|███ | 161/520 [10:08<21:47, 3.64s/it] {'loss': 1.351, 'grad_norm': 0.0005509901946910249, 'learning_rate': 0.40465107899844704, 'epoch': 0.31} + 31%|███ | 161/520 [10:08<21:47, 3.64s/it] 31%|███ | 162/520 [10:11<21:42, 3.64s/it] {'loss': 1.4725, 'grad_norm': 0.0007111828470342798, 'learning_rate': 0.40342369869801187, 'epoch': 0.31} + 31%|███ | 162/520 [10:11<21:42, 3.64s/it] 31%|███▏ | 163/520 [10:15<21:34, 3.63s/it] {'loss': 1.2474, 'grad_norm': 0.0006708905689076092, 'learning_rate': 0.40219035725218016, 'epoch': 0.31} + 31%|███▏ | 163/520 [10:15<21:34, 3.63s/it] 32%|███▏ | 164/520 [10:19<21:29, 3.62s/it] {'loss': 1.2235, 'grad_norm': 0.0005481752189364698, 'learning_rate': 0.40095110258136935, 'epoch': 0.32} + 32%|███▏ | 164/520 [10:19<21:29, 3.62s/it] 32%|███▏ | 165/520 [10:22<21:22, 3.61s/it] {'loss': 1.3518, 'grad_norm': 0.0005014629333054306, 'learning_rate': 0.3997059828357501, 'epoch': 0.32} + 32%|███▏ | 165/520 [10:22<21:22, 3.61s/it] 32%|███▏ | 166/520 [10:26<21:18, 3.61s/it] {'loss': 1.3339, 'grad_norm': 0.0005918725368924636, 'learning_rate': 0.39845504639337537, 'epoch': 0.32} + 32%|███▏ | 166/520 [10:26<21:18, 3.61s/it] 32%|███▏ | 167/520 [10:29<21:17, 3.62s/it] {'loss': 1.3183, 'grad_norm': 0.0005195764264637016, 'learning_rate': 0.39719834185830116, 'epoch': 0.32} + 32%|███▏ | 167/520 [10:29<21:17, 3.62s/it] 32%|███▏ | 168/520 [10:33<21:16, 3.63s/it] {'loss': 1.2796, 'grad_norm': 0.0005273366493208508, 'learning_rate': 0.3959359180586975, 'epoch': 0.32} + 32%|███▏ | 168/520 [10:33<21:16, 3.63s/it] 32%|███▎ | 169/520 [10:37<21:21, 3.65s/it] {'loss': 1.3499, 'grad_norm': 0.0005571557716799046, 'learning_rate': 0.3946678240449515, 'epoch': 0.33} + 32%|███▎ | 169/520 [10:37<21:21, 3.65s/it] 33%|███▎ | 170/520 [10:41<21:36, 3.70s/it] {'loss': 1.3945, 'grad_norm': 0.0006621382842112868, 'learning_rate': 0.3933941090877615, 'epoch': 0.33} + 33%|███▎ | 170/520 [10:41<21:36, 3.70s/it] 33%|███▎ | 171/520 [10:44<21:38, 3.72s/it] {'loss': 1.29, 'grad_norm': 0.0005609265888253222, 'learning_rate': 0.3921148226762231, 'epoch': 0.33} + 33%|███▎ | 171/520 [10:44<21:38, 3.72s/it] 33%|███▎ | 172/520 [10:48<21:39, 3.74s/it] {'loss': 1.3662, 'grad_norm': 0.0005202340609825367, 'learning_rate': 0.3908300145159055, 'epoch': 0.33} + 33%|███▎ | 172/520 [10:48<21:39, 3.74s/it] 33%|███▎ | 173/520 [10:52<21:40, 3.75s/it] {'loss': 1.2892, 'grad_norm': 0.0008665052183570846, 'learning_rate': 0.38953973452692103, 'epoch': 0.33} + 33%|███▎ | 173/520 [10:52<21:40, 3.75s/it] 33%|███▎ | 174/520 [10:56<21:39, 3.76s/it] {'loss': 1.3576, 'grad_norm': 0.0008098688259184392, 'learning_rate': 0.38824403284198483, 'epoch': 0.33} + 33%|███▎ | 174/520 [10:56<21:39, 3.76s/it] 34%|███▎ | 175/520 [10:59<21:38, 3.76s/it] {'loss': 1.2639, 'grad_norm': 0.0010285300422498832, 'learning_rate': 0.38694295980446786, 'epoch': 0.34} + 34%|███▎ | 175/520 [10:59<21:38, 3.76s/it] 34%|███▍ | 176/520 [11:03<21:36, 3.77s/it] {'loss': 1.4914, 'grad_norm': 0.0006424195676219644, 'learning_rate': 0.38563656596643986, 'epoch': 0.34} + 34%|███▍ | 176/520 [11:03<21:36, 3.77s/it] 34%|███▍ | 177/520 [11:07<21:38, 3.79s/it] {'loss': 1.3637, 'grad_norm': 0.0005606816449643465, 'learning_rate': 0.384324902086706, 'epoch': 0.34} + 34%|███▍ | 177/520 [11:07<21:38, 3.79s/it] 34%|███▍ | 178/520 [11:11<21:32, 3.78s/it] {'loss': 1.3455, 'grad_norm': 0.0005582470153386239, 'learning_rate': 0.38300801912883414, 'epoch': 0.34} + 34%|███▍ | 178/520 [11:11<21:32, 3.78s/it] 34%|███▍ | 179/520 [11:15<21:29, 3.78s/it] {'loss': 1.4251, 'grad_norm': 0.000659412910078437, 'learning_rate': 0.38168596825917517, 'epoch': 0.34} + 34%|███▍ | 179/520 [11:15<21:29, 3.78s/it] 35%|███▍ | 180/520 [11:18<21:25, 3.78s/it] {'loss': 1.3217, 'grad_norm': 0.000723004562970165, 'learning_rate': 0.3803588008448745, 'epoch': 0.35} + 35%|███▍ | 180/520 [11:18<21:25, 3.78s/it] 35%|███▍ | 181/520 [11:22<21:21, 3.78s/it] {'loss': 1.3244, 'grad_norm': 0.0006452286701898244, 'learning_rate': 0.37902656845187666, 'epoch': 0.35} + 35%|███▍ | 181/520 [11:22<21:21, 3.78s/it] 35%|███▌ | 182/520 [11:26<21:19, 3.79s/it] {'loss': 1.3192, 'grad_norm': 0.0005702168058122251, 'learning_rate': 0.37768932284292145, 'epoch': 0.35} + 35%|███▌ | 182/520 [11:26<21:19, 3.79s/it] 35%|███▌ | 183/520 [11:30<21:15, 3.79s/it] {'loss': 1.3511, 'grad_norm': 0.0006812126127999836, 'learning_rate': 0.3763471159755327, 'epoch': 0.35} + 35%|███▌ | 183/520 [11:30<21:15, 3.79s/it] 35%|███▌ | 184/520 [11:33<21:06, 3.77s/it] {'loss': 1.272, 'grad_norm': 0.0005971088484406683, 'learning_rate': 0.375, 'epoch': 0.35} + 35%|███▌ | 184/520 [11:33<21:06, 3.77s/it] 36%|███▌ | 185/520 [11:37<20:55, 3.75s/it] {'loss': 1.4308, 'grad_norm': 0.0006402625090712889, 'learning_rate': 0.37364802725735186, 'epoch': 0.36} + 36%|███▌ | 185/520 [11:37<20:55, 3.75s/it] 36%|███▌ | 186/520 [11:41<20:43, 3.72s/it] {'loss': 1.2905, 'grad_norm': 0.0005854384474101797, 'learning_rate': 0.37229125027732235, 'epoch': 0.36} + 36%|███▌ | 186/520 [11:41<20:43, 3.72s/it] 36%|███▌ | 187/520 [11:44<20:30, 3.69s/it] {'loss': 1.2863, 'grad_norm': 0.0006509498102824706, 'learning_rate': 0.37092972177630995, 'epoch': 0.36} + 36%|███▌ | 187/520 [11:44<20:30, 3.69s/it] 36%|███▌ | 188/520 [11:48<20:20, 3.68s/it] {'loss': 1.3762, 'grad_norm': 0.0006786187023820654, 'learning_rate': 0.36956349465532956, 'epoch': 0.36} + 36%|███▌ | 188/520 [11:48<20:20, 3.68s/it] 36%|███▋ | 189/520 [11:52<20:14, 3.67s/it] {'loss': 1.3908, 'grad_norm': 0.0008330698508293802, 'learning_rate': 0.36819262199795677, 'epoch': 0.36} + 36%|███▋ | 189/520 [11:52<20:14, 3.67s/it] 37%|███▋ | 190/520 [11:55<20:05, 3.65s/it] {'loss': 1.2949, 'grad_norm': 0.0006988041561720816, 'learning_rate': 0.3668171570682655, 'epoch': 0.37} + 37%|███▋ | 190/520 [11:55<20:05, 3.65s/it] 37%|███▋ | 191/520 [11:59<20:01, 3.65s/it] {'loss': 1.2632, 'grad_norm': 0.0005720878995110236, 'learning_rate': 0.36543715330875853, 'epoch': 0.37} + 37%|███▋ | 191/520 [11:59<20:01, 3.65s/it] 37%|███▋ | 192/520 [12:03<20:04, 3.67s/it] {'loss': 1.351, 'grad_norm': 0.0005364724734532762, 'learning_rate': 0.36405266433829075, 'epoch': 0.37} + 37%|███▋ | 192/520 [12:03<20:04, 3.67s/it] 37%|███▋ | 193/520 [12:06<19:56, 3.66s/it] {'loss': 1.4277, 'grad_norm': 0.0008227362744572845, 'learning_rate': 0.36266374394998635, 'epoch': 0.37} + 37%|███▋ | 193/520 [12:06<19:56, 3.66s/it] 37%|███▋ | 194/520 [12:10<19:50, 3.65s/it] {'loss': 1.317, 'grad_norm': 0.0005832998494940012, 'learning_rate': 0.36127044610914805, 'epoch': 0.37} + 37%|███▋ | 194/520 [12:10<19:50, 3.65s/it] 38%|███▊ | 195/520 [12:14<19:46, 3.65s/it] {'loss': 1.3592, 'grad_norm': 0.0005881929294300234, 'learning_rate': 0.35987282495116124, 'epoch': 0.38} + 38%|███▊ | 195/520 [12:14<19:46, 3.65s/it] 38%|███▊ | 196/520 [12:17<19:38, 3.64s/it] {'loss': 1.3417, 'grad_norm': 0.0007705694945627778, 'learning_rate': 0.3584709347793895, 'epoch': 0.38} + 38%|███▊ | 196/520 [12:17<19:38, 3.64s/it] 38%|███▊ | 197/520 [12:21<19:35, 3.64s/it] {'loss': 1.296, 'grad_norm': 0.0005305839289376428, 'learning_rate': 0.35706483006306566, 'epoch': 0.38} + 38%|███▊ | 197/520 [12:21<19:35, 3.64s/it] 38%|███▊ | 198/520 [12:24<19:29, 3.63s/it] {'loss': 1.3726, 'grad_norm': 0.0007538722224010206, 'learning_rate': 0.3556545654351749, 'epoch': 0.38} + 38%|███▊ | 198/520 [12:24<19:29, 3.63s/it] 38%|███▊ | 199/520 [12:28<19:27, 3.64s/it] {'loss': 1.2835, 'grad_norm': 0.0005113741609197054, 'learning_rate': 0.35424019569033205, 'epoch': 0.38} + 38%|███▊ | 199/520 [12:28<19:27, 3.64s/it] 38%|███▊ | 200/520 [12:32<19:27, 3.65s/it] {'loss': 1.3709, 'grad_norm': 0.0005779578954734121, 'learning_rate': 0.35282177578265295, 'epoch': 0.38} + 38%|███▊ | 200/520 [12:32<19:27, 3.65s/it] 39%|███▊ | 201/520 [12:35<19:24, 3.65s/it] {'loss': 1.3867, 'grad_norm': 0.0004760936738391124, 'learning_rate': 0.3513993608236188, 'epoch': 0.39} + 39%|███▊ | 201/520 [12:35<19:24, 3.65s/it] 39%|███▉ | 202/520 [12:39<19:19, 3.65s/it] {'loss': 1.2774, 'grad_norm': 0.0005157253800302079, 'learning_rate': 0.3499730060799352, 'epoch': 0.39} + 39%|███▉ | 202/520 [12:39<19:19, 3.65s/it] 39%|███▉ | 203/520 [12:43<19:13, 3.64s/it] {'loss': 1.3323, 'grad_norm': 0.0006669835132811949, 'learning_rate': 0.34854276697138487, 'epoch': 0.39} + 39%|███▉ | 203/520 [12:43<19:13, 3.64s/it] 39%|███▉ | 204/520 [12:46<19:10, 3.64s/it] {'loss': 1.3556, 'grad_norm': 0.0004917114234029344, 'learning_rate': 0.3471086990686737, 'epoch': 0.39} + 39%|███▉ | 204/520 [12:46<19:10, 3.64s/it] 39%|███▉ | 205/520 [12:50<19:05, 3.64s/it] {'loss': 1.3999, 'grad_norm': 0.000542879571462809, 'learning_rate': 0.34567085809127246, 'epoch': 0.39} + 39%|███▉ | 205/520 [12:50<19:05, 3.64s/it] 40%|███▉ | 206/520 [12:54<19:02, 3.64s/it] {'loss': 1.4002, 'grad_norm': 0.0006223187580271043, 'learning_rate': 0.3442292999052513, 'epoch': 0.4} + 40%|███▉ | 206/520 [12:54<19:02, 3.64s/it] 40%|███▉ | 207/520 [12:57<18:58, 3.64s/it] {'loss': 1.3646, 'grad_norm': 0.0005900487438473084, 'learning_rate': 0.3427840805211095, 'epoch': 0.4} + 40%|███▉ | 207/520 [12:57<18:58, 3.64s/it] 40%|████ | 208/520 [13:01<18:52, 3.63s/it] {'loss': 1.3544, 'grad_norm': 0.0006307904972519447, 'learning_rate': 0.3413352560915988, 'epoch': 0.4} + 40%|████ | 208/520 [13:01<18:52, 3.63s/it] 40%|████ | 209/520 [13:05<18:52, 3.64s/it] {'loss': 1.2925, 'grad_norm': 0.0004817094733079617, 'learning_rate': 0.3398828829095419, 'epoch': 0.4} + 40%|████ | 209/520 [13:05<18:52, 3.64s/it] 40%|████ | 210/520 [13:08<18:48, 3.64s/it] {'loss': 1.3483, 'grad_norm': 0.0005920622240403865, 'learning_rate': 0.33842701740564535, 'epoch': 0.4} + 40%|████ | 210/520 [13:08<18:48, 3.64s/it] 41%|████ | 211/520 [13:12<18:46, 3.65s/it] {'loss': 1.372, 'grad_norm': 0.0006070953258438848, 'learning_rate': 0.3369677161463068, 'epoch': 0.41} + 41%|████ | 211/520 [13:12<18:46, 3.65s/it] 41%|████ | 212/520 [13:15<18:40, 3.64s/it] {'loss': 1.3575, 'grad_norm': 0.0005691091035656534, 'learning_rate': 0.3355050358314172, 'epoch': 0.41} + 41%|████ | 212/520 [13:15<18:40, 3.64s/it] 41%|████ | 213/520 [13:19<18:35, 3.63s/it] {'loss': 1.3035, 'grad_norm': 0.0007389980993151245, 'learning_rate': 0.3340390332921577, 'epoch': 0.41} + 41%|████ | 213/520 [13:19<18:35, 3.63s/it] 41%|████ | 214/520 [13:23<18:31, 3.63s/it] {'loss': 1.3127, 'grad_norm': 0.0009451077475575871, 'learning_rate': 0.3325697654887918, 'epoch': 0.41} + 41%|████ | 214/520 [13:23<18:31, 3.63s/it] 41%|████▏ | 215/520 [13:26<18:29, 3.64s/it] {'loss': 1.308, 'grad_norm': 0.00048605064821581335, 'learning_rate': 0.3310972895084518, 'epoch': 0.41} + 41%|████▏ | 215/520 [13:26<18:29, 3.64s/it] 42%|████▏ | 216/520 [13:30<18:23, 3.63s/it] {'loss': 1.2115, 'grad_norm': 0.0005102327159204213, 'learning_rate': 0.32962166256292114, 'epoch': 0.42} + 42%|████▏ | 216/520 [13:30<18:23, 3.63s/it] 42%|████▏ | 217/520 [13:34<18:23, 3.64s/it] {'loss': 1.3516, 'grad_norm': 0.0005323638950226611, 'learning_rate': 0.32814294198641114, 'epoch': 0.42} + 42%|████▏ | 217/520 [13:34<18:23, 3.64s/it] 42%|████▏ | 218/520 [13:37<18:16, 3.63s/it] {'loss': 1.3469, 'grad_norm': 0.0006546758923118947, 'learning_rate': 0.3266611852333336, 'epoch': 0.42} + 42%|████▏ | 218/520 [13:37<18:16, 3.63s/it] 42%|████▏ | 219/520 [13:41<18:12, 3.63s/it] {'loss': 1.3297, 'grad_norm': 0.0004856252742318247, 'learning_rate': 0.32517644987606825, 'epoch': 0.42} + 42%|████▏ | 219/520 [13:41<18:12, 3.63s/it] 42%|████▏ | 220/520 [13:44<18:06, 3.62s/it] {'loss': 1.3704, 'grad_norm': 0.0005022349262374555, 'learning_rate': 0.32368879360272607, 'epoch': 0.42} + 42%|████▏ | 220/520 [13:44<18:06, 3.62s/it] 42%|████▎ | 221/520 [13:48<18:03, 3.62s/it] {'loss': 1.3487, 'grad_norm': 0.0005622278119067065, 'learning_rate': 0.3221982742149075, 'epoch': 0.42} + 42%|████▎ | 221/520 [13:48<18:03, 3.62s/it] 43%|████▎ | 222/520 [13:52<17:58, 3.62s/it] {'loss': 1.2623, 'grad_norm': 0.0004769264502741241, 'learning_rate': 0.32070494962545687, 'epoch': 0.43} + 43%|████▎ | 222/520 [13:52<17:58, 3.62s/it] 43%|████▎ | 223/520 [13:55<17:54, 3.62s/it] {'loss': 1.2646, 'grad_norm': 0.0005108979710238484, 'learning_rate': 0.3192088778562123, 'epoch': 0.43} + 43%|████▎ | 223/520 [13:55<17:54, 3.62s/it] 43%|████▎ | 224/520 [13:59<17:49, 3.61s/it] {'loss': 1.5708, 'grad_norm': 0.0007352793234845265, 'learning_rate': 0.3177101170357513, 'epoch': 0.43} + 43%|████▎ | 224/520 [13:59<17:49, 3.61s/it] 43%|████▎ | 225/520 [14:03<17:50, 3.63s/it] {'loss': 1.2638, 'grad_norm': 0.0005546232890176265, 'learning_rate': 0.3162087253971318, 'epoch': 0.43} + 43%|████▎ | 225/520 [14:03<17:50, 3.63s/it] 43%|████▎ | 226/520 [14:06<17:47, 3.63s/it] {'loss': 1.3724, 'grad_norm': 0.00047782111945009144, 'learning_rate': 0.3147047612756302, 'epoch': 0.43} + 43%|████▎ | 226/520 [14:06<17:47, 3.63s/it] 44%|████▎ | 227/520 [14:10<17:44, 3.63s/it] {'loss': 1.3614, 'grad_norm': 0.00047605710421158565, 'learning_rate': 0.31319828310647435, 'epoch': 0.44} + 44%|████▎ | 227/520 [14:10<17:44, 3.63s/it] 44%|████▍ | 228/520 [14:14<17:43, 3.64s/it] {'loss': 1.4934, 'grad_norm': 0.0005227251905285143, 'learning_rate': 0.31168934942257337, 'epoch': 0.44} + 44%|████▍ | 228/520 [14:14<17:43, 3.64s/it] 44%|████▍ | 229/520 [14:17<17:39, 3.64s/it] {'loss': 1.3434, 'grad_norm': 0.00043896402423767195, 'learning_rate': 0.3101780188522433, 'epoch': 0.44} + 44%|████▍ | 229/520 [14:17<17:39, 3.64s/it] 44%|████▍ | 230/520 [14:21<17:44, 3.67s/it] {'loss': 1.2172, 'grad_norm': 0.0006357674937785741, 'learning_rate': 0.30866435011692883, 'epoch': 0.44} + 44%|████▍ | 230/520 [14:21<17:44, 3.67s/it] 44%|████▍ | 231/520 [14:25<17:59, 3.74s/it] {'loss': 1.2863, 'grad_norm': 0.0006660145533357579, 'learning_rate': 0.3071484020289224, 'epoch': 0.44} + 44%|████▍ | 231/520 [14:25<17:59, 3.74s/it] 45%|████▍ | 232/520 [14:29<18:07, 3.77s/it] {'loss': 1.5321, 'grad_norm': 0.000553218802626062, 'learning_rate': 0.3056302334890786, 'epoch': 0.45} + 45%|████▍ | 232/520 [14:29<18:07, 3.77s/it] 45%|████▍ | 233/520 [14:32<18:06, 3.79s/it] {'loss': 1.4031, 'grad_norm': 0.0004925996112646419, 'learning_rate': 0.3041099034845257, 'epoch': 0.45} + 45%|████▍ | 233/520 [14:32<18:06, 3.79s/it] 45%|████▌ | 234/520 [14:36<18:05, 3.79s/it] {'loss': 1.2433, 'grad_norm': 0.0004999139337756515, 'learning_rate': 0.30258747108637396, 'epoch': 0.45} + 45%|████▌ | 234/520 [14:36<18:05, 3.79s/it] 45%|████▌ | 235/520 [14:40<18:01, 3.79s/it] {'loss': 1.2924, 'grad_norm': 0.00046505435933613833, 'learning_rate': 0.3010629954474201, 'epoch': 0.45} + 45%|████▌ | 235/520 [14:40<18:01, 3.79s/it] 45%|████▌ | 236/520 [14:44<18:03, 3.81s/it] {'loss': 1.3843, 'grad_norm': 0.000511929194685905, 'learning_rate': 0.2995365357998494, 'epoch': 0.45} + 45%|████▌ | 236/520 [14:44<18:03, 3.81s/it] 46%|████▌ | 237/520 [14:48<18:00, 3.82s/it] {'loss': 1.3524, 'grad_norm': 0.00047330108029133145, 'learning_rate': 0.2980081514529341, 'epoch': 0.46} + 46%|████▌ | 237/520 [14:48<18:00, 3.82s/it] 46%|████▌ | 238/520 [14:51<17:46, 3.78s/it] {'loss': 1.3129, 'grad_norm': 0.0005471850427555895, 'learning_rate': 0.2964779017907287, 'epoch': 0.46} + 46%|████▌ | 238/520 [14:51<17:46, 3.78s/it] 46%|████▌ | 239/520 [14:55<17:33, 3.75s/it] {'loss': 1.3949, 'grad_norm': 0.0006246891962407655, 'learning_rate': 0.29494584626976317, 'epoch': 0.46} + 46%|████▌ | 239/520 [14:55<17:33, 3.75s/it] 46%|████▌ | 240/520 [14:59<17:23, 3.73s/it] {'loss': 1.1804, 'grad_norm': 0.0004217964431068179, 'learning_rate': 0.29341204441673263, 'epoch': 0.46} + 46%|████▌ | 240/520 [14:59<17:23, 3.73s/it] 46%|████▋ | 241/520 [15:02<17:15, 3.71s/it] {'loss': 1.2549, 'grad_norm': 0.0004001821875930727, 'learning_rate': 0.2918765558261841, 'epoch': 0.46} + 46%|████▋ | 241/520 [15:02<17:15, 3.71s/it] 47%|████▋ | 242/520 [15:06<17:08, 3.70s/it] {'loss': 1.2808, 'grad_norm': 0.0005505812525291796, 'learning_rate': 0.2903394401582017, 'epoch': 0.47} + 47%|████▋ | 242/520 [15:06<17:08, 3.70s/it] 47%|████▋ | 243/520 [15:10<17:01, 3.69s/it] {'loss': 1.2756, 'grad_norm': 0.00043562841338075435, 'learning_rate': 0.2888007571360879, 'epoch': 0.47} + 47%|████▋ | 243/520 [15:10<17:01, 3.69s/it] 47%|████▋ | 244/520 [15:14<16:58, 3.69s/it] {'loss': 1.3958, 'grad_norm': 0.0005627559322295305, 'learning_rate': 0.2872605665440436, 'epoch': 0.47} + 47%|████▋ | 244/520 [15:14<16:58, 3.69s/it] 47%|████▋ | 245/520 [15:17<17:04, 3.72s/it] {'loss': 1.2697, 'grad_norm': 0.0005440162529745525, 'learning_rate': 0.285718928224845, 'epoch': 0.47} + 47%|████▋ | 245/520 [15:17<17:04, 3.72s/it] 47%|████▋ | 246/520 [15:21<17:09, 3.76s/it] {'loss': 1.5141, 'grad_norm': 0.0004983999331358455, 'learning_rate': 0.28417590207751836, 'epoch': 0.47} + 47%|████▋ | 246/520 [15:21<17:09, 3.76s/it] 48%|████▊ | 247/520 [15:25<17:08, 3.77s/it] {'loss': 1.4265, 'grad_norm': 0.000649045124212033, 'learning_rate': 0.28263154805501295, 'epoch': 0.47} + 48%|████▊ | 247/520 [15:25<17:08, 3.77s/it] 48%|████▊ | 248/520 [15:29<17:07, 3.78s/it] {'loss': 1.2505, 'grad_norm': 0.000489838416195826, 'learning_rate': 0.28108592616187134, 'epoch': 0.48} + 48%|████▊ | 248/520 [15:29<17:07, 3.78s/it] 48%|████▊ | 249/520 [15:33<17:08, 3.79s/it] {'loss': 1.3699, 'grad_norm': 0.0006193705763924629, 'learning_rate': 0.27953909645189823, 'epoch': 0.48} + 48%|████▊ | 249/520 [15:33<17:08, 3.79s/it] 48%|████▊ | 250/520 [15:36<17:07, 3.81s/it] {'loss': 1.3179, 'grad_norm': 0.0005042220190829777, 'learning_rate': 0.27799111902582696, 'epoch': 0.48} + 48%|████▊ | 250/520 [15:36<17:07, 3.81s/it] 48%|████▊ | 251/520 [15:40<17:07, 3.82s/it] {'loss': 1.3584, 'grad_norm': 0.00041823752247377113, 'learning_rate': 0.2764420540289845, 'epoch': 0.48} + 48%|████▊ | 251/520 [15:40<17:07, 3.82s/it] 48%|████▊ | 252/520 [15:44<17:07, 3.83s/it] {'loss': 1.3847, 'grad_norm': 0.0005192186463953096, 'learning_rate': 0.2748919616489542, 'epoch': 0.48} + 48%|████▊ | 252/520 [15:44<17:07, 3.83s/it] 49%|████▊ | 253/520 [15:48<17:07, 3.85s/it] {'loss': 1.3597, 'grad_norm': 0.0004785236592462828, 'learning_rate': 0.27334090211323764, 'epoch': 0.49} + 49%|████▊ | 253/520 [15:48<17:07, 3.85s/it] 49%|████▉ | 254/520 [15:52<17:01, 3.84s/it] {'loss': 1.2764, 'grad_norm': 0.0004045622252282362, 'learning_rate': 0.2717889356869146, 'epoch': 0.49} + 49%|████▉ | 254/520 [15:52<17:01, 3.84s/it] 49%|████▉ | 255/520 [15:56<16:56, 3.84s/it] {'loss': 1.2956, 'grad_norm': 0.0006083180927147534, 'learning_rate': 0.27023612267030084, 'epoch': 0.49} + 49%|████▉ | 255/520 [15:56<16:56, 3.84s/it] 49%|████▉ | 256/520 [15:59<16:51, 3.83s/it] {'loss': 1.3628, 'grad_norm': 0.000461834318310395, 'learning_rate': 0.2686825233966061, 'epoch': 0.49} + 49%|████▉ | 256/520 [15:59<16:51, 3.83s/it] 49%|████▉ | 257/520 [16:03<16:47, 3.83s/it] {'loss': 1.3401, 'grad_norm': 0.00047862668639274457, 'learning_rate': 0.26712819822958916, 'epoch': 0.49} + 49%|████▉ | 257/520 [16:03<16:47, 3.83s/it] 50%|████▉ | 258/520 [16:07<16:37, 3.81s/it] {'loss': 1.3521, 'grad_norm': 0.000547127255158418, 'learning_rate': 0.26557320756121305, 'epoch': 0.5} + 50%|████▉ | 258/520 [16:07<16:37, 3.81s/it] 50%|████▉ | 259/520 [16:11<16:18, 3.75s/it] {'loss': 1.4091, 'grad_norm': 0.0004959327442319288, 'learning_rate': 0.26401761180929795, 'epoch': 0.5} + 50%|████▉ | 259/520 [16:11<16:18, 3.75s/it] 50%|█████ | 260/520 [16:14<16:17, 3.76s/it] {'loss': 1.4808, 'grad_norm': 0.0005938834962529418, 'learning_rate': 0.2624614714151743, 'epoch': 0.5} + 50%|█████ | 260/520 [16:14<16:17, 3.76s/it] 50%|█████ | 261/520 [16:18<16:12, 3.75s/it] {'loss': 1.4163, 'grad_norm': 0.0007675166833470595, 'learning_rate': 0.260904846841334, 'epoch': 0.5} + 50%|█████ | 261/520 [16:18<16:12, 3.75s/it] 50%|█████ | 262/520 [16:22<15:59, 3.72s/it] {'loss': 1.2784, 'grad_norm': 0.0004572347083151796, 'learning_rate': 0.25934779856908147, 'epoch': 0.5} + 50%|█████ | 262/520 [16:22<15:59, 3.72s/it] 51%|█████ | 263/520 [16:25<15:45, 3.68s/it] {'loss': 1.4184, 'grad_norm': 0.0004929385045222917, 'learning_rate': 0.2577903870961833, 'epoch': 0.51} + 51%|█████ | 263/520 [16:25<15:45, 3.68s/it] 51%|█████ | 264/520 [16:29<15:39, 3.67s/it] {'loss': 1.3804, 'grad_norm': 0.00046952313405786394, 'learning_rate': 0.25623267293451824, 'epoch': 0.51} + 51%|█████ | 264/520 [16:29<15:39, 3.67s/it] 51%|█████ | 265/520 [16:33<15:38, 3.68s/it] {'loss': 1.2899, 'grad_norm': 0.0005609049023032974, 'learning_rate': 0.25467471660772556, 'epoch': 0.51} + 51%|█████ | 265/520 [16:33<15:38, 3.68s/it] 51%|█████ | 266/520 [16:36<15:30, 3.66s/it] {'loss': 1.1473, 'grad_norm': 0.0003967703516030346, 'learning_rate': 0.2531165786488538, 'epoch': 0.51} + 51%|█████ | 266/520 [16:36<15:30, 3.66s/it] 51%|█████▏ | 267/520 [16:40<15:26, 3.66s/it] {'loss': 1.2872, 'grad_norm': 0.0004943691661670576, 'learning_rate': 0.2515583195980084, 'epoch': 0.51} + 51%|█████▏ | 267/520 [16:40<15:26, 3.66s/it] 52%|█████▏ | 268/520 [16:44<15:21, 3.66s/it] {'loss': 1.5437, 'grad_norm': 0.0008519126710662879, 'learning_rate': 0.25, 'epoch': 0.52} + 52%|█████▏ | 268/520 [16:44<15:21, 3.66s/it] 52%|█████▏ | 269/520 [16:47<15:15, 3.65s/it] {'loss': 1.3772, 'grad_norm': 0.0005651897333721065, 'learning_rate': 0.2484416804019916, 'epoch': 0.52} + 52%|█████▏ | 269/520 [16:47<15:15, 3.65s/it] 52%|█████▏ | 270/520 [16:51<15:10, 3.64s/it] {'loss': 1.3484, 'grad_norm': 0.0004705130064974907, 'learning_rate': 0.24688342135114624, 'epoch': 0.52} + 52%|█████▏ | 270/520 [16:51<15:10, 3.64s/it] 52%|█████▏ | 271/520 [16:55<15:04, 3.63s/it] {'loss': 1.3659, 'grad_norm': 0.00047571617618867666, 'learning_rate': 0.2453252833922745, 'epoch': 0.52} + 52%|█████▏ | 271/520 [16:55<15:04, 3.63s/it] 52%|█████▏ | 272/520 [16:58<15:02, 3.64s/it] {'loss': 1.3463, 'grad_norm': 0.000515067295346631, 'learning_rate': 0.24376732706548182, 'epoch': 0.52} + 52%|█████▏ | 272/520 [16:58<15:02, 3.64s/it] 52%|█████▎ | 273/520 [17:02<14:59, 3.64s/it] {'loss': 1.4621, 'grad_norm': 0.0005537508158144429, 'learning_rate': 0.24220961290381676, 'epoch': 0.53} + 52%|█████▎ | 273/520 [17:02<14:59, 3.64s/it] 53%|█████▎ | 274/520 [17:05<14:54, 3.64s/it] {'loss': 1.3423, 'grad_norm': 0.000668058171284698, 'learning_rate': 0.24065220143091862, 'epoch': 0.53} + 53%|█████▎ | 274/520 [17:06<14:54, 3.64s/it] 53%|█████▎ | 275/520 [17:09<14:51, 3.64s/it] {'loss': 1.2802, 'grad_norm': 0.0006168531737791672, 'learning_rate': 0.23909515315866603, 'epoch': 0.53} + 53%|█████▎ | 275/520 [17:09<14:51, 3.64s/it] 53%|█████▎ | 276/520 [17:13<14:56, 3.67s/it] {'loss': 1.3532, 'grad_norm': 0.0005049241829275313, 'learning_rate': 0.2375385285848257, 'epoch': 0.53} + 53%|█████▎ | 276/520 [17:13<14:56, 3.67s/it] 53%|█████▎ | 277/520 [17:17<14:51, 3.67s/it] {'loss': 1.4794, 'grad_norm': 0.0005094608266202402, 'learning_rate': 0.23598238819070202, 'epoch': 0.53} + 53%|█████▎ | 277/520 [17:17<14:51, 3.67s/it] 53%|█████▎ | 278/520 [17:20<14:45, 3.66s/it] {'loss': 1.2424, 'grad_norm': 0.00047621057240103745, 'learning_rate': 0.23442679243878697, 'epoch': 0.53} + 53%|█████▎ | 278/520 [17:20<14:45, 3.66s/it] 54%|█████▎ | 279/520 [17:24<14:39, 3.65s/it] {'loss': 1.3783, 'grad_norm': 0.0005977417962868176, 'learning_rate': 0.23287180177041084, 'epoch': 0.54} + 54%|█████▎ | 279/520 [17:24<14:39, 3.65s/it] 54%|█████▍ | 280/520 [17:27<14:36, 3.65s/it] {'loss': 1.2839, 'grad_norm': 0.0005103127789929424, 'learning_rate': 0.23131747660339394, 'epoch': 0.54} + 54%|█████▍ | 280/520 [17:27<14:36, 3.65s/it] 54%|█████▍ | 281/520 [17:31<14:31, 3.65s/it] {'loss': 1.3863, 'grad_norm': 0.000537857771567801, 'learning_rate': 0.22976387732969916, 'epoch': 0.54} + 54%|█████▍ | 281/520 [17:31<14:31, 3.65s/it] 54%|█████▍ | 282/520 [17:35<14:27, 3.65s/it] {'loss': 1.2537, 'grad_norm': 0.0004633695427644894, 'learning_rate': 0.22821106431308544, 'epoch': 0.54} + 54%|█████▍ | 282/520 [17:35<14:27, 3.65s/it] 54%|█████▍ | 283/520 [17:38<14:30, 3.67s/it] {'loss': 1.4176, 'grad_norm': 0.0004961054710184318, 'learning_rate': 0.22665909788676236, 'epoch': 0.54} + 54%|█████▍ | 283/520 [17:38<14:30, 3.67s/it] 55%|█████▍ | 284/520 [17:42<14:39, 3.73s/it] {'loss': 1.3533, 'grad_norm': 0.00057216597149076, 'learning_rate': 0.22510803835104587, 'epoch': 0.55} + 55%|█████▍ | 284/520 [17:42<14:39, 3.73s/it] 55%|█████▍ | 285/520 [17:46<14:40, 3.75s/it] {'loss': 1.2777, 'grad_norm': 0.0005080132762230548, 'learning_rate': 0.22355794597101558, 'epoch': 0.55} + 55%|█████▍ | 285/520 [17:46<14:40, 3.75s/it] 55%|█████▌ | 286/520 [17:50<14:40, 3.76s/it] {'loss': 1.149, 'grad_norm': 0.0004721457013474809, 'learning_rate': 0.22200888097417304, 'epoch': 0.55} + 55%|█████▌ | 286/520 [17:50<14:40, 3.76s/it] 55%|█████▌ | 287/520 [17:54<14:40, 3.78s/it] {'loss': 1.3926, 'grad_norm': 0.0004236847015743194, 'learning_rate': 0.2204609035481018, 'epoch': 0.55} + 55%|█████▌ | 287/520 [17:54<14:40, 3.78s/it] 55%|█████▌ | 288/520 [17:58<14:38, 3.79s/it] {'loss': 1.4359, 'grad_norm': 0.0006729423662029276, 'learning_rate': 0.21891407383812878, 'epoch': 0.55} + 55%|█████▌ | 288/520 [17:58<14:38, 3.79s/it] 56%|█████▌ | 289/520 [18:01<14:39, 3.81s/it] {'loss': 1.2953, 'grad_norm': 0.0005114811191988865, 'learning_rate': 0.21736845194498716, 'epoch': 0.56} + 56%|█████▌ | 289/520 [18:01<14:39, 3.81s/it] 56%|█████▌ | 290/520 [18:05<14:34, 3.80s/it] {'loss': 1.2209, 'grad_norm': 0.0004834539871880715, 'learning_rate': 0.2158240979224817, 'epoch': 0.56} + 56%|█████▌ | 290/520 [18:05<14:34, 3.80s/it] 56%|█████▌ | 291/520 [18:09<14:22, 3.77s/it] {'loss': 1.2606, 'grad_norm': 0.0005441819665670266, 'learning_rate': 0.21428107177515504, 'epoch': 0.56} + 56%|█████▌ | 291/520 [18:09<14:22, 3.77s/it] 56%|█████▌ | 292/520 [18:13<14:11, 3.73s/it] {'loss': 1.3255, 'grad_norm': 0.0005509060146577794, 'learning_rate': 0.21273943345595636, 'epoch': 0.56} + 56%|█████▌ | 292/520 [18:13<14:11, 3.73s/it] 56%|█████▋ | 293/520 [18:16<14:02, 3.71s/it] {'loss': 1.2572, 'grad_norm': 0.0005227498878725473, 'learning_rate': 0.2111992428639121, 'epoch': 0.56} + 56%|█████▋ | 293/520 [18:16<14:02, 3.71s/it] 57%|█████▋ | 294/520 [18:20<13:55, 3.70s/it] {'loss': 1.2878, 'grad_norm': 0.0005575465755251511, 'learning_rate': 0.20966055984179832, 'epoch': 0.57} + 57%|█████▋ | 294/520 [18:20<13:55, 3.70s/it] 57%|█████▋ | 295/520 [18:24<13:49, 3.69s/it] {'loss': 1.4477, 'grad_norm': 0.0005374641868344732, 'learning_rate': 0.20812344417381592, 'epoch': 0.57} + 57%|█████▋ | 295/520 [18:24<13:49, 3.69s/it] 57%|█████▋ | 296/520 [18:27<13:42, 3.67s/it] {'loss': 1.2438, 'grad_norm': 0.00043588929062918116, 'learning_rate': 0.20658795558326742, 'epoch': 0.57} + 57%|█████▋ | 296/520 [18:27<13:42, 3.67s/it] 57%|█████▋ | 297/520 [18:31<13:48, 3.72s/it] {'loss': 1.3722, 'grad_norm': 0.0004898387967716852, 'learning_rate': 0.20505415373023683, 'epoch': 0.57} + 57%|█████▋ | 297/520 [18:31<13:48, 3.72s/it] 57%|█████▋ | 298/520 [18:35<13:52, 3.75s/it] {'loss': 1.3422, 'grad_norm': 0.00043673487355320995, 'learning_rate': 0.20352209820927136, 'epoch': 0.57} + 57%|█████▋ | 298/520 [18:35<13:52, 3.75s/it] 57%|█████▊ | 299/520 [18:39<13:51, 3.76s/it] {'loss': 1.4443, 'grad_norm': 0.0005493390104323044, 'learning_rate': 0.20199184854706598, 'epoch': 0.57} + 57%|█████▊ | 299/520 [18:39<13:51, 3.76s/it] 58%|█████▊ | 300/520 [18:42<13:51, 3.78s/it] {'loss': 1.388, 'grad_norm': 0.0004594202809124367, 'learning_rate': 0.20046346420015065, 'epoch': 0.58} + 58%|█████▊ | 300/520 [18:42<13:51, 3.78s/it] 58%|█████▊ | 301/520 [18:46<13:49, 3.79s/it] {'loss': 1.3847, 'grad_norm': 0.00042509439886404503, 'learning_rate': 0.19893700455257995, 'epoch': 0.58} + 58%|█████▊ | 301/520 [18:46<13:49, 3.79s/it] 58%|█████▊ | 302/520 [18:50<13:39, 3.76s/it] {'loss': 1.4589, 'grad_norm': 0.0004721522897400503, 'learning_rate': 0.1974125289136261, 'epoch': 0.58} + 58%|█████▊ | 302/520 [18:50<13:39, 3.76s/it] 58%|█████▊ | 303/520 [18:54<13:29, 3.73s/it] {'loss': 1.2957, 'grad_norm': 0.0005094932428148304, 'learning_rate': 0.1958900965154743, 'epoch': 0.58} + 58%|█████▊ | 303/520 [18:54<13:29, 3.73s/it] 58%|█████▊ | 304/520 [18:57<13:21, 3.71s/it] {'loss': 1.3602, 'grad_norm': 0.0008059079063740703, 'learning_rate': 0.19436976651092142, 'epoch': 0.58} + 58%|█████▊ | 304/520 [18:57<13:21, 3.71s/it] 59%|█████▊ | 305/520 [19:01<13:15, 3.70s/it] {'loss': 1.4248, 'grad_norm': 0.0006965268021991243, 'learning_rate': 0.19285159797107765, 'epoch': 0.59} + 59%|█████▊ | 305/520 [19:01<13:15, 3.70s/it] 59%|█████▉ | 306/520 [19:05<13:09, 3.69s/it] {'loss': 1.3404, 'grad_norm': 0.00047655023257603114, 'learning_rate': 0.19133564988307125, 'epoch': 0.59} + 59%|█████▉ | 306/520 [19:05<13:09, 3.69s/it] 59%|█████▉ | 307/520 [19:09<13:24, 3.78s/it] {'loss': 1.2904, 'grad_norm': 0.0004679860384018704, 'learning_rate': 0.1898219811477568, 'epoch': 0.59} + 59%|█████▉ | 307/520 [19:09<13:24, 3.78s/it] 59%|█████▉ | 308/520 [19:12<13:12, 3.74s/it] {'loss': 1.4063, 'grad_norm': 0.0005363418008560086, 'learning_rate': 0.18831065057742657, 'epoch': 0.59} + 59%|█████▉ | 308/520 [19:12<13:12, 3.74s/it] 59%|█████▉ | 309/520 [19:16<13:02, 3.71s/it] {'loss': 1.2885, 'grad_norm': 0.0005410076114510583, 'learning_rate': 0.1868017168935256, 'epoch': 0.59} + 59%|█████▉ | 309/520 [19:16<13:02, 3.71s/it] 60%|█████▉ | 310/520 [19:20<12:54, 3.69s/it] {'loss': 1.2589, 'grad_norm': 0.0009705347637170786, 'learning_rate': 0.1852952387243698, 'epoch': 0.6} + 60%|█████▉ | 310/520 [19:20<12:54, 3.69s/it] 60%|█████▉ | 311/520 [19:23<12:51, 3.69s/it] {'loss': 1.2416, 'grad_norm': 0.000598189864320441, 'learning_rate': 0.18379127460286818, 'epoch': 0.6} + 60%|█████▉ | 311/520 [19:23<12:51, 3.69s/it] 60%|██████ | 312/520 [19:27<12:46, 3.69s/it] {'loss': 1.2268, 'grad_norm': 0.0006289159831217087, 'learning_rate': 0.18228988296424875, 'epoch': 0.6} + 60%|██████ | 312/520 [19:27<12:46, 3.69s/it] 60%|██████ | 313/520 [19:31<12:47, 3.71s/it] {'loss': 1.2123, 'grad_norm': 0.0003921007824807057, 'learning_rate': 0.18079112214378767, 'epoch': 0.6} + 60%|██████ | 313/520 [19:31<12:47, 3.71s/it] 60%|██████ | 314/520 [19:35<13:22, 3.90s/it] {'loss': 1.2451, 'grad_norm': 0.0004065620643055171, 'learning_rate': 0.17929505037454313, 'epoch': 0.6} + 60%|██████ | 314/520 [19:35<13:22, 3.90s/it] 61%|██████ | 315/520 [19:39<13:04, 3.83s/it] {'loss': 1.4475, 'grad_norm': 0.0005984136941805349, 'learning_rate': 0.17780172578509257, 'epoch': 0.61} + 61%|██████ | 315/520 [19:39<13:04, 3.83s/it] 61%|██████ | 316/520 [19:43<13:18, 3.92s/it] {'loss': 1.2297, 'grad_norm': 0.0004285844434950497, 'learning_rate': 0.17631120639727393, 'epoch': 0.61} + 61%|██████ | 316/520 [19:43<13:18, 3.92s/it] 61%|██████ | 317/520 [19:46<13:02, 3.85s/it] {'loss': 1.2402, 'grad_norm': 0.00042190130334670325, 'learning_rate': 0.17482355012393175, 'epoch': 0.61} + 61%|██████ | 317/520 [19:46<13:02, 3.85s/it] 61%|██████ | 318/520 [19:50<12:48, 3.80s/it] {'loss': 1.3694, 'grad_norm': 0.0009705992246418527, 'learning_rate': 0.17333881476666646, 'epoch': 0.61} + 61%|██████ | 318/520 [19:50<12:48, 3.80s/it] 61%|██████▏ | 319/520 [19:54<12:58, 3.87s/it] {'loss': 1.2351, 'grad_norm': 0.0006228354678083541, 'learning_rate': 0.17185705801358891, 'epoch': 0.61} + 61%|██████▏ | 319/520 [19:54<12:58, 3.87s/it] 62%|██████▏ | 320/520 [19:58<12:45, 3.83s/it] {'loss': 1.1743, 'grad_norm': 0.0004296130527905815, 'learning_rate': 0.17037833743707892, 'epoch': 0.62} + 62%|██████▏ | 320/520 [19:58<12:45, 3.83s/it] 62%|██████▏ | 321/520 [20:02<12:31, 3.78s/it] {'loss': 1.3886, 'grad_norm': 0.0006015014241410016, 'learning_rate': 0.16890271049154826, 'epoch': 0.62} + 62%|██████▏ | 321/520 [20:02<12:31, 3.78s/it] 62%|██████▏ | 322/520 [20:05<12:20, 3.74s/it] {'loss': 1.291, 'grad_norm': 0.0004314701410648533, 'learning_rate': 0.1674302345112083, 'epoch': 0.62} + 62%|██████▏ | 322/520 [20:05<12:20, 3.74s/it] 62%|██████▏ | 323/520 [20:09<12:11, 3.71s/it] {'loss': 1.3777, 'grad_norm': 0.0005760835891488661, 'learning_rate': 0.16596096670784236, 'epoch': 0.62} + 62%|██████▏ | 323/520 [20:09<12:11, 3.71s/it] 62%|██████▏ | 324/520 [20:13<12:05, 3.70s/it] {'loss': 1.3198, 'grad_norm': 0.0004396525693166416, 'learning_rate': 0.16449496416858284, 'epoch': 0.62} + 62%|██████▏ | 324/520 [20:13<12:05, 3.70s/it] 62%|██████▎ | 325/520 [20:16<11:59, 3.69s/it] {'loss': 1.3255, 'grad_norm': 0.0004813666880203096, 'learning_rate': 0.16303228385369317, 'epoch': 0.62} + 62%|██████▎ | 325/520 [20:16<11:59, 3.69s/it] 63%|██████▎ | 326/520 [20:20<11:53, 3.68s/it] {'loss': 1.3069, 'grad_norm': 0.0004655284947091194, 'learning_rate': 0.16157298259435465, 'epoch': 0.63} + 63%|██████▎ | 326/520 [20:20<11:53, 3.68s/it] 63%|██████▎ | 327/520 [20:24<11:48, 3.67s/it] {'loss': 1.43, 'grad_norm': 0.0004815607542201367, 'learning_rate': 0.16011711709045812, 'epoch': 0.63} + 63%|██████▎ | 327/520 [20:24<11:48, 3.67s/it] 63%|██████▎ | 328/520 [20:27<11:42, 3.66s/it] {'loss': 1.3716, 'grad_norm': 0.0004970170763641753, 'learning_rate': 0.15866474390840124, 'epoch': 0.63} + 63%|██████▎ | 328/520 [20:27<11:42, 3.66s/it] 63%|██████▎ | 329/520 [20:31<11:38, 3.66s/it] {'loss': 1.2322, 'grad_norm': 0.0003795956684141727, 'learning_rate': 0.15721591947889052, 'epoch': 0.63} + 63%|██████▎ | 329/520 [20:31<11:38, 3.66s/it] 63%|██████▎ | 330/520 [20:34<11:33, 3.65s/it] {'loss': 1.3143, 'grad_norm': 0.00044268800978916026, 'learning_rate': 0.1557707000947487, 'epoch': 0.63} + 63%|██████▎ | 330/520 [20:34<11:33, 3.65s/it] 64%|██████▎ | 331/520 [20:38<11:28, 3.64s/it] {'loss': 1.2658, 'grad_norm': 0.00042582855919154545, 'learning_rate': 0.15432914190872757, 'epoch': 0.64} + 64%|██████▎ | 331/520 [20:38<11:28, 3.64s/it] 64%|██████▍ | 332/520 [20:42<11:24, 3.64s/it] {'loss': 1.4448, 'grad_norm': 0.0004720969325627767, 'learning_rate': 0.15289130093132633, 'epoch': 0.64} + 64%|██████▍ | 332/520 [20:42<11:24, 3.64s/it] 64%|██████▍ | 333/520 [20:45<11:20, 3.64s/it] {'loss': 1.4317, 'grad_norm': 0.0005549062040696464, 'learning_rate': 0.1514572330286152, 'epoch': 0.64} + 64%|██████▍ | 333/520 [20:45<11:20, 3.64s/it] 64%|██████▍ | 334/520 [20:49<11:17, 3.64s/it] {'loss': 1.32, 'grad_norm': 0.0004511057465401887, 'learning_rate': 0.1500269939200648, 'epoch': 0.64} + 64%|██████▍ | 334/520 [20:49<11:17, 3.64s/it] 64%|██████▍ | 335/520 [20:53<11:14, 3.65s/it] {'loss': 1.3279, 'grad_norm': 0.0004813748364892179, 'learning_rate': 0.14860063917638128, 'epoch': 0.64} + 64%|██████▍ | 335/520 [20:53<11:14, 3.65s/it] 65%|██████▍ | 336/520 [20:56<11:11, 3.65s/it] {'loss': 1.2192, 'grad_norm': 0.0005863671801451039, 'learning_rate': 0.14717822421734716, 'epoch': 0.65} + 65%|██████▍ | 336/520 [20:56<11:11, 3.65s/it] 65%|██████▍ | 337/520 [21:00<11:07, 3.65s/it] {'loss': 1.2087, 'grad_norm': 0.0004198787640665937, 'learning_rate': 0.14575980430966806, 'epoch': 0.65} + 65%|██████▍ | 337/520 [21:00<11:07, 3.65s/it] 65%|██████▌ | 338/520 [21:04<11:05, 3.66s/it] {'loss': 1.337, 'grad_norm': 0.0006148044175433274, 'learning_rate': 0.14434543456482518, 'epoch': 0.65} + 65%|██████▌ | 338/520 [21:04<11:05, 3.66s/it] 65%|██████▌ | 339/520 [21:07<11:03, 3.66s/it] {'loss': 1.2811, 'grad_norm': 0.0004923444105027091, 'learning_rate': 0.14293516993693428, 'epoch': 0.65} + 65%|██████▌ | 339/520 [21:07<11:03, 3.66s/it] 65%|██████▌ | 340/520 [21:11<11:00, 3.67s/it] {'loss': 1.2583, 'grad_norm': 0.0007953488557380372, 'learning_rate': 0.14152906522061048, 'epoch': 0.65} + 65%|██████▌ | 340/520 [21:11<11:00, 3.67s/it] 66%|██████▌ | 341/520 [21:15<10:56, 3.67s/it] {'loss': 1.2776, 'grad_norm': 0.0005414837382165337, 'learning_rate': 0.14012717504883873, 'epoch': 0.66} + 66%|██████▌ | 341/520 [21:15<10:56, 3.67s/it] 66%|██████▌ | 342/520 [21:18<10:52, 3.67s/it] {'loss': 1.4216, 'grad_norm': 0.0005141912207381263, 'learning_rate': 0.1387295538908519, 'epoch': 0.66} + 66%|██████▌ | 342/520 [21:18<10:52, 3.67s/it] 66%|██████▌ | 343/520 [21:22<10:48, 3.66s/it] {'loss': 1.3832, 'grad_norm': 0.0003852870369873143, 'learning_rate': 0.13733625605001365, 'epoch': 0.66} + 66%|██████▌ | 343/520 [21:22<10:48, 3.66s/it] 66%|██████▌ | 344/520 [21:26<10:45, 3.67s/it] {'loss': 1.2352, 'grad_norm': 0.0004077631098210358, 'learning_rate': 0.13594733566170925, 'epoch': 0.66} + 66%|██████▌ | 344/520 [21:26<10:45, 3.67s/it] 66%|██████▋ | 345/520 [21:29<10:49, 3.71s/it] {'loss': 1.3506, 'grad_norm': 0.00048437486687532335, 'learning_rate': 0.13456284669124158, 'epoch': 0.66} + 66%|██████▋ | 345/520 [21:29<10:49, 3.71s/it] 67%|██████▋ | 346/520 [21:33<10:52, 3.75s/it] {'loss': 1.3758, 'grad_norm': 0.0004271851511070874, 'learning_rate': 0.1331828429317345, 'epoch': 0.67} + 67%|██████▋ | 346/520 [21:33<10:52, 3.75s/it] 67%|██████▋ | 347/520 [21:37<10:56, 3.79s/it] {'loss': 1.2621, 'grad_norm': 0.00038319109189114363, 'learning_rate': 0.13180737800204329, 'epoch': 0.67} + 67%|██████▋ | 347/520 [21:37<10:56, 3.79s/it]Token indices sequence length is longer than the specified maximum sequence length for this model (2778 > 2048). Running this sequence through the model will result in indexing errors + 67%|██████▋ | 348/520 [21:41<10:58, 3.83s/it] {'loss': 1.233, 'grad_norm': 0.0006157758026779868, 'learning_rate': 0.13043650534467052, 'epoch': 0.67} + 67%|██████▋ | 348/520 [21:41<10:58, 3.83s/it] 67%|██████▋ | 349/520 [21:45<10:46, 3.78s/it] {'loss': 1.2634, 'grad_norm': 0.0004965878162436011, 'learning_rate': 0.12907027822369005, 'epoch': 0.67} + 67%|██████▋ | 349/520 [21:45<10:46, 3.78s/it] 67%|██████▋ | 350/520 [21:48<10:36, 3.75s/it] {'loss': 1.3024, 'grad_norm': 0.0005611636070048727, 'learning_rate': 0.12770874972267776, 'epoch': 0.67} + 67%|██████▋ | 350/520 [21:48<10:36, 3.75s/it] 68%|██████▊ | 351/520 [21:52<10:34, 3.76s/it] {'loss': 1.2109, 'grad_norm': 0.0004253282889488441, 'learning_rate': 0.12635197274264814, 'epoch': 0.68} + 68%|██████▊ | 351/520 [21:52<10:34, 3.76s/it] 68%|██████▊ | 352/520 [21:56<10:28, 3.74s/it] {'loss': 1.3362, 'grad_norm': 0.0004042206691744572, 'learning_rate': 0.12500000000000006, 'epoch': 0.68} + 68%|██████▊ | 352/520 [21:56<10:28, 3.74s/it] 68%|██████▊ | 353/520 [22:00<10:22, 3.73s/it] {'loss': 1.3436, 'grad_norm': 0.0005328561760186368, 'learning_rate': 0.12365288402446739, 'epoch': 0.68} + 68%|██████▊ | 353/520 [22:00<10:22, 3.73s/it] 68%|██████▊ | 354/520 [22:03<10:12, 3.69s/it] {'loss': 1.4764, 'grad_norm': 0.00043943477267839663, 'learning_rate': 0.12231067715707866, 'epoch': 0.68} + 68%|██████▊ | 354/520 [22:03<10:12, 3.69s/it] 68%|██████▊ | 355/520 [22:07<10:06, 3.67s/it] {'loss': 1.2711, 'grad_norm': 0.00044812109681345206, 'learning_rate': 0.12097343154812332, 'epoch': 0.68} + 68%|██████▊ | 355/520 [22:07<10:06, 3.67s/it] 68%|██████▊ | 356/520 [22:11<10:01, 3.67s/it] {'loss': 1.2757, 'grad_norm': 0.00041657486378038984, 'learning_rate': 0.1196411991551255, 'epoch': 0.68} + 68%|██████▊ | 356/520 [22:11<10:01, 3.67s/it] 69%|██████▊ | 357/520 [22:14<10:02, 3.70s/it] {'loss': 1.2909, 'grad_norm': 0.0003952810463161782, 'learning_rate': 0.1183140317408248, 'epoch': 0.69} + 69%|██████▊ | 357/520 [22:14<10:02, 3.70s/it] 69%|██████▉ | 358/520 [22:18<10:01, 3.71s/it] {'loss': 1.2355, 'grad_norm': 0.0005425743929262464, 'learning_rate': 0.11699198087116588, 'epoch': 0.69} + 69%|██████▉ | 358/520 [22:18<10:01, 3.71s/it] 69%|██████▉ | 359/520 [22:22<09:54, 3.69s/it] {'loss': 1.4042, 'grad_norm': 0.00048665108052980255, 'learning_rate': 0.115675097913294, 'epoch': 0.69} + 69%|██████▉ | 359/520 [22:22<09:54, 3.69s/it] 69%|██████▉ | 360/520 [22:25<09:48, 3.68s/it] {'loss': 1.4131, 'grad_norm': 0.0005475712528750736, 'learning_rate': 0.11436343403356017, 'epoch': 0.69} + 69%|██████▉ | 360/520 [22:25<09:48, 3.68s/it] 69%|██████▉ | 361/520 [22:29<09:41, 3.66s/it] {'loss': 1.4141, 'grad_norm': 0.0005194422296544079, 'learning_rate': 0.1130570401955322, 'epoch': 0.69} + 69%|██████▉ | 361/520 [22:29<09:41, 3.66s/it] 70%|██████▉ | 362/520 [22:33<09:36, 3.65s/it] {'loss': 1.2791, 'grad_norm': 0.0005959289218612312, 'learning_rate': 0.11175596715801514, 'epoch': 0.7} + 70%|██████▉ | 362/520 [22:33<09:36, 3.65s/it] 70%|██████▉ | 363/520 [22:36<09:31, 3.64s/it] {'loss': 1.3417, 'grad_norm': 0.000519319681001639, 'learning_rate': 0.11046026547307905, 'epoch': 0.7} + 70%|██████▉ | 363/520 [22:36<09:31, 3.64s/it] 70%|███████ | 364/520 [22:40<09:27, 3.64s/it] {'loss': 1.4281, 'grad_norm': 0.000497956207344362, 'learning_rate': 0.10916998548409448, 'epoch': 0.7} + 70%|███████ | 364/520 [22:40<09:27, 3.64s/it] 70%|███████ | 365/520 [22:43<09:24, 3.64s/it] {'loss': 1.3897, 'grad_norm': 0.0005236971414312448, 'learning_rate': 0.10788517732377695, 'epoch': 0.7} + 70%|███████ | 365/520 [22:43<09:24, 3.64s/it] 70%|███████ | 366/520 [22:47<09:28, 3.69s/it] {'loss': 1.3431, 'grad_norm': 0.00039296651153242706, 'learning_rate': 0.10660589091223854, 'epoch': 0.7} + 70%|███████ | 366/520 [22:47<09:28, 3.69s/it] 71%|███████ | 367/520 [22:51<09:31, 3.74s/it] {'loss': 1.3435, 'grad_norm': 0.0005743108779549897, 'learning_rate': 0.10533217595504857, 'epoch': 0.71} + 71%|███████ | 367/520 [22:51<09:31, 3.74s/it] 71%|███████ | 368/520 [22:55<09:33, 3.77s/it] {'loss': 1.1955, 'grad_norm': 0.0005093677432427576, 'learning_rate': 0.1040640819413026, 'epoch': 0.71} + 71%|███████ | 368/520 [22:55<09:33, 3.77s/it] 71%|███████ | 369/520 [22:59<09:31, 3.78s/it] {'loss': 1.3907, 'grad_norm': 0.000798867880135867, 'learning_rate': 0.10280165814169884, 'epoch': 0.71} + 71%|███████ | 369/520 [22:59<09:31, 3.78s/it] 71%|███████ | 370/520 [23:03<09:30, 3.80s/it] {'loss': 1.2461, 'grad_norm': 0.0004423121297608829, 'learning_rate': 0.10154495360662463, 'epoch': 0.71} + 71%|███████ | 370/520 [23:03<09:30, 3.80s/it] 71%|███████▏ | 371/520 [23:06<09:27, 3.81s/it] {'loss': 1.2363, 'grad_norm': 0.000455426258026778, 'learning_rate': 0.10029401716424993, 'epoch': 0.71} + 71%|███████▏ | 371/520 [23:06<09:27, 3.81s/it] 72%|███████▏ | 372/520 [23:10<09:25, 3.82s/it] {'loss': 1.4795, 'grad_norm': 0.0006807617766423851, 'learning_rate': 0.0990488974186306, 'epoch': 0.72} + 72%|███████▏ | 372/520 [23:10<09:25, 3.82s/it] 72%|███████▏ | 373/520 [23:14<09:22, 3.83s/it] {'loss': 1.3652, 'grad_norm': 0.0005703958397858211, 'learning_rate': 0.09780964274781984, 'epoch': 0.72} + 72%|███████▏ | 373/520 [23:14<09:22, 3.83s/it] 72%|███████▏ | 374/520 [23:18<09:14, 3.80s/it] {'loss': 1.3339, 'grad_norm': 0.0004736695270555238, 'learning_rate': 0.09657630130198819, 'epoch': 0.72} + 72%|███████▏ | 374/520 [23:18<09:14, 3.80s/it] 72%|███████▏ | 375/520 [23:22<09:04, 3.76s/it] {'loss': 1.2369, 'grad_norm': 0.00042046668335293, 'learning_rate': 0.09534892100155296, 'epoch': 0.72} + 72%|███████▏ | 375/520 [23:22<09:04, 3.76s/it] 72%|███████▏ | 376/520 [23:25<08:56, 3.72s/it] {'loss': 1.352, 'grad_norm': 0.00039674251285156455, 'learning_rate': 0.09412754953531663, 'epoch': 0.72} + 72%|███████▏ | 376/520 [23:25<08:56, 3.72s/it] 72%|███████▎ | 377/520 [23:29<08:49, 3.70s/it] {'loss': 1.3, 'grad_norm': 0.0007722729980186801, 'learning_rate': 0.09291223435861318, 'epoch': 0.72} + 72%|███████▎ | 377/520 [23:29<08:49, 3.70s/it] 73%|███████▎ | 378/520 [23:32<08:42, 3.68s/it] {'loss': 1.3598, 'grad_norm': 0.00042575021309939254, 'learning_rate': 0.09170302269146507, 'epoch': 0.73} + 73%|███████▎ | 378/520 [23:32<08:42, 3.68s/it] 73%|███████▎ | 379/520 [23:36<08:40, 3.69s/it] {'loss': 1.3251, 'grad_norm': 0.0003919253793707493, 'learning_rate': 0.09049996151674788, 'epoch': 0.73} + 73%|███████▎ | 379/520 [23:36<08:40, 3.69s/it] 73%|███████▎ | 380/520 [23:40<08:34, 3.67s/it] {'loss': 1.4748, 'grad_norm': 0.00048117818654789813, 'learning_rate': 0.08930309757836516, 'epoch': 0.73} + 73%|███████▎ | 380/520 [23:40<08:34, 3.67s/it] 73%|███████▎ | 381/520 [23:43<08:27, 3.65s/it] {'loss': 1.3384, 'grad_norm': 0.0005074004000020823, 'learning_rate': 0.08811247737943242, 'epoch': 0.73} + 73%|███████▎ | 381/520 [23:43<08:27, 3.65s/it] 73%|███████▎ | 382/520 [23:47<08:23, 3.65s/it] {'loss': 1.4026, 'grad_norm': 0.00047464409441250275, 'learning_rate': 0.08692814718046979, 'epoch': 0.73} + 73%|███████▎ | 382/520 [23:47<08:23, 3.65s/it] 74%|███████▎ | 383/520 [23:51<08:25, 3.69s/it] {'loss': 1.1739, 'grad_norm': 0.0005450117740583133, 'learning_rate': 0.08575015299760491, 'epoch': 0.74} + 74%|███████▎ | 383/520 [23:51<08:25, 3.69s/it] 74%|███████▍ | 384/520 [23:55<08:27, 3.73s/it] {'loss': 1.5448, 'grad_norm': 0.0006627729594847213, 'learning_rate': 0.08457854060078521, 'epoch': 0.74} + 74%|███████▍ | 384/520 [23:55<08:27, 3.73s/it] 74%|███████▍ | 385/520 [23:58<08:25, 3.75s/it] {'loss': 1.3135, 'grad_norm': 0.0004225997756453681, 'learning_rate': 0.08341335551199902, 'epoch': 0.74} + 74%|███████▍ | 385/520 [23:58<08:25, 3.75s/it] 74%|███████▍ | 386/520 [24:02<08:23, 3.76s/it] {'loss': 1.2583, 'grad_norm': 0.00038116754356807853, 'learning_rate': 0.08225464300350752, 'epoch': 0.74} + 74%|███████▍ | 386/520 [24:02<08:23, 3.76s/it] 74%|███████▍ | 387/520 [24:06<08:20, 3.76s/it] {'loss': 1.4891, 'grad_norm': 0.00043884637810214877, 'learning_rate': 0.08110244809608494, 'epoch': 0.74} + 74%|███████▍ | 387/520 [24:06<08:20, 3.76s/it] 75%|███████▍ | 388/520 [24:10<08:17, 3.77s/it] {'loss': 1.2182, 'grad_norm': 0.0004635249496621264, 'learning_rate': 0.0799568155572701, 'epoch': 0.75} + 75%|███████▍ | 388/520 [24:10<08:17, 3.77s/it] 75%|███████▍ | 389/520 [24:14<08:17, 3.80s/it] {'loss': 1.2801, 'grad_norm': 0.0006542008325933946, 'learning_rate': 0.07881778989962662, 'epoch': 0.75} + 75%|███████▍ | 389/520 [24:14<08:17, 3.80s/it] 75%|███████▌ | 390/520 [24:17<08:15, 3.81s/it] {'loss': 1.336, 'grad_norm': 0.0005007995644153289, 'learning_rate': 0.07768541537901325, 'epoch': 0.75} + 75%|███████▌ | 390/520 [24:17<08:15, 3.81s/it] 75%|███████▌ | 391/520 [24:21<08:13, 3.82s/it] {'loss': 1.4243, 'grad_norm': 0.0005323896117084931, 'learning_rate': 0.07655973599286459, 'epoch': 0.75} + 75%|███████▌ | 391/520 [24:21<08:13, 3.82s/it] 75%|███████▌ | 392/520 [24:25<08:10, 3.83s/it] {'loss': 1.2261, 'grad_norm': 0.0004331175596778587, 'learning_rate': 0.07544079547848181, 'epoch': 0.75} + 75%|███████▌ | 392/520 [24:25<08:10, 3.83s/it] 76%|███████▌ | 393/520 [24:29<08:06, 3.83s/it] {'loss': 1.2986, 'grad_norm': 0.0003912582780771376, 'learning_rate': 0.07432863731133271, 'epoch': 0.76} + 76%|███████▌ | 393/520 [24:29<08:06, 3.83s/it] 76%|███████▌ | 394/520 [24:33<08:02, 3.83s/it] {'loss': 1.2956, 'grad_norm': 0.00055313759374018, 'learning_rate': 0.07322330470336313, 'epoch': 0.76} + 76%|███████▌ | 394/520 [24:33<08:02, 3.83s/it] 76%|███████▌ | 395/520 [24:37<07:58, 3.83s/it] {'loss': 1.258, 'grad_norm': 0.0004543039139614778, 'learning_rate': 0.07212484060131752, 'epoch': 0.76} + 76%|███████▌ | 395/520 [24:37<07:58, 3.83s/it] 76%|███████▌ | 396/520 [24:40<07:54, 3.83s/it] {'loss': 1.3336, 'grad_norm': 0.0005305170772758131, 'learning_rate': 0.07103328768507039, 'epoch': 0.76} + 76%|███████▌ | 396/520 [24:41<07:54, 3.83s/it] 76%|███████▋ | 397/520 [24:44<07:51, 3.83s/it] {'loss': 1.32, 'grad_norm': 0.0006116436942218197, 'learning_rate': 0.0699486883659684, 'epoch': 0.76} + 76%|███████▋ | 397/520 [24:44<07:51, 3.83s/it] 77%|███████▋ | 398/520 [24:48<07:40, 3.78s/it] {'loss': 1.2959, 'grad_norm': 0.0004903882253093521, 'learning_rate': 0.06887108478518184, 'epoch': 0.77} + 77%|███████▋ | 398/520 [24:48<07:40, 3.78s/it] 77%|███████▋ | 399/520 [24:52<07:32, 3.74s/it] {'loss': 1.3437, 'grad_norm': 0.0007042604036705451, 'learning_rate': 0.06780051881206792, 'epoch': 0.77} + 77%|███████▋ | 399/520 [24:52<07:32, 3.74s/it] 77%|███████▋ | 400/520 [24:55<07:24, 3.71s/it] {'loss': 1.3882, 'grad_norm': 0.0005007255732100227, 'learning_rate': 0.06673703204254347, 'epoch': 0.77} + 77%|███████▋ | 400/520 [24:55<07:24, 3.71s/it] 77%|███████▋ | 401/520 [24:59<07:17, 3.68s/it] {'loss': 1.1387, 'grad_norm': 0.0005060444176450576, 'learning_rate': 0.065680665797469, 'epoch': 0.77} + 77%|███████▋ | 401/520 [24:59<07:17, 3.68s/it] 77%|███████▋ | 402/520 [25:02<07:11, 3.66s/it] {'loss': 1.2614, 'grad_norm': 0.00044936766679420497, 'learning_rate': 0.06463146112104332, 'epoch': 0.77} + 77%|███████▋ | 402/520 [25:02<07:11, 3.66s/it] 78%|███████▊ | 403/520 [25:06<07:07, 3.65s/it] {'loss': 1.3074, 'grad_norm': 0.0005765715593572758, 'learning_rate': 0.0635894587792086, 'epoch': 0.78} + 78%|███████▊ | 403/520 [25:06<07:07, 3.65s/it] 78%|███████▊ | 404/520 [25:10<07:03, 3.65s/it] {'loss': 1.2067, 'grad_norm': 0.0006684899106826506, 'learning_rate': 0.06255469925806642, 'epoch': 0.78} + 78%|███████▊ | 404/520 [25:10<07:03, 3.65s/it] 78%|███████▊ | 405/520 [25:13<07:00, 3.65s/it] {'loss': 1.3541, 'grad_norm': 0.00045922161533108773, 'learning_rate': 0.061527222762305045, 'epoch': 0.78} + 78%|███████▊ | 405/520 [25:13<07:00, 3.65s/it] 78%|███████▊ | 406/520 [25:17<06:56, 3.65s/it] {'loss': 1.2938, 'grad_norm': 0.0006907446920073234, 'learning_rate': 0.060507069213636716, 'epoch': 0.78} + 78%|███████▊ | 406/520 [25:17<06:56, 3.65s/it] 78%|███████▊ | 407/520 [25:21<06:53, 3.66s/it] {'loss': 1.3922, 'grad_norm': 0.0004910723355013919, 'learning_rate': 0.0594942782492473, 'epoch': 0.78} + 78%|███████▊ | 407/520 [25:21<06:53, 3.66s/it] 78%|███████▊ | 408/520 [25:24<06:50, 3.66s/it] {'loss': 1.2852, 'grad_norm': 0.0005789198694863336, 'learning_rate': 0.058488889220255524, 'epoch': 0.78} + 78%|███████▊ | 408/520 [25:24<06:50, 3.66s/it] 79%|███████▊ | 409/520 [25:28<06:47, 3.67s/it] {'loss': 1.4269, 'grad_norm': 0.0006936418831668173, 'learning_rate': 0.0574909411901843, 'epoch': 0.79} + 79%|███████▊ | 409/520 [25:28<06:47, 3.67s/it] 79%|███████▉ | 410/520 [25:32<06:44, 3.68s/it] {'loss': 1.1306, 'grad_norm': 0.000546042505611479, 'learning_rate': 0.05650047293344315, 'epoch': 0.79} + 79%|███████▉ | 410/520 [25:32<06:44, 3.68s/it] 79%|███████▉ | 411/520 [25:36<06:44, 3.71s/it] {'loss': 1.3806, 'grad_norm': 0.00045819432483448633, 'learning_rate': 0.05551752293382131, 'epoch': 0.79} + 79%|███████▉ | 411/520 [25:36<06:44, 3.71s/it] 79%|███████▉ | 412/520 [25:39<06:45, 3.75s/it] {'loss': 1.2994, 'grad_norm': 0.0005095841558704691, 'learning_rate': 0.05454212938299255, 'epoch': 0.79} + 79%|███████▉ | 412/520 [25:39<06:45, 3.75s/it] 79%|███████▉ | 413/520 [25:43<06:45, 3.79s/it] {'loss': 1.4017, 'grad_norm': 0.0005036754892103234, 'learning_rate': 0.053574330179031626, 'epoch': 0.79} + 79%|███████▉ | 413/520 [25:43<06:45, 3.79s/it] 80%|███████▉ | 414/520 [25:47<06:43, 3.81s/it] {'loss': 1.1725, 'grad_norm': 0.00045298707913232116, 'learning_rate': 0.05261416292494117, 'epoch': 0.8} + 80%|███████▉ | 414/520 [25:47<06:43, 3.81s/it] 80%|███████▉ | 415/520 [25:51<06:36, 3.77s/it] {'loss': 1.2716, 'grad_norm': 0.0004283915036172017, 'learning_rate': 0.051661664927191236, 'epoch': 0.8} + 80%|███████▉ | 415/520 [25:51<06:36, 3.77s/it] 80%|████████ | 416/520 [25:55<06:28, 3.73s/it] {'loss': 1.1869, 'grad_norm': 0.000572807664070659, 'learning_rate': 0.05071687319426946, 'epoch': 0.8} + 80%|████████ | 416/520 [25:55<06:28, 3.73s/it] 80%|████████ | 417/520 [25:58<06:21, 3.71s/it] {'loss': 1.3449, 'grad_norm': 0.0005075056566789036, 'learning_rate': 0.049779824435243036, 'epoch': 0.8} + 80%|████████ | 417/520 [25:58<06:21, 3.71s/it] 80%|████████ | 418/520 [26:02<06:16, 3.69s/it] {'loss': 1.3483, 'grad_norm': 0.000459661903437559, 'learning_rate': 0.04885055505833291, 'epoch': 0.8} + 80%|████████ | 418/520 [26:02<06:16, 3.69s/it] 81%|████████ | 419/520 [26:05<06:11, 3.67s/it] {'loss': 1.3354, 'grad_norm': 0.0005614046842666578, 'learning_rate': 0.047929101169498695, 'epoch': 0.81} + 81%|████████ | 419/520 [26:05<06:11, 3.67s/it] 81%|████████ | 420/520 [26:09<06:06, 3.67s/it] {'loss': 1.2106, 'grad_norm': 0.0004986543781047508, 'learning_rate': 0.047015498571035874, 'epoch': 0.81} + 81%|████████ | 420/520 [26:09<06:06, 3.67s/it] 81%|████████ | 421/520 [26:13<06:02, 3.66s/it] {'loss': 1.1411, 'grad_norm': 0.0005941624100310515, 'learning_rate': 0.04610978276018496, 'epoch': 0.81} + 81%|████████ | 421/520 [26:13<06:02, 3.66s/it] 81%|████████ | 422/520 [26:16<05:58, 3.66s/it] {'loss': 1.2806, 'grad_norm': 0.0006601881243363389, 'learning_rate': 0.04521198892775202, 'epoch': 0.81} + 81%|████████ | 422/520 [26:16<05:58, 3.66s/it] 81%|████████▏ | 423/520 [26:20<05:54, 3.66s/it] {'loss': 1.2585, 'grad_norm': 0.0005045635479965424, 'learning_rate': 0.04432215195674166, 'epoch': 0.81} + 81%|████████▏ | 423/520 [26:20<05:54, 3.66s/it] 82%|████████▏ | 424/520 [26:24<05:51, 3.66s/it] {'loss': 1.5046, 'grad_norm': 0.0004935712164718136, 'learning_rate': 0.04344030642100133, 'epoch': 0.82} + 82%|████████▏ | 424/520 [26:24<05:51, 3.66s/it] 82%|████████▏ | 425/520 [26:27<05:50, 3.69s/it] {'loss': 1.2594, 'grad_norm': 0.0004858293940053346, 'learning_rate': 0.04256648658387813, 'epoch': 0.82} + 82%|████████▏ | 425/520 [26:27<05:50, 3.69s/it] 82%|████████▏ | 426/520 [26:31<05:50, 3.73s/it] {'loss': 1.3158, 'grad_norm': 0.0008329459236423462, 'learning_rate': 0.041700726396887794, 'epoch': 0.82} + 82%|████████▏ | 426/520 [26:31<05:50, 3.73s/it] 82%|████████▏ | 427/520 [26:35<05:50, 3.77s/it] {'loss': 1.2006, 'grad_norm': 0.00047343798963172694, 'learning_rate': 0.04084305949839506, 'epoch': 0.82} + 82%|████████▏ | 427/520 [26:35<05:50, 3.77s/it] 82%|████████▏ | 428/520 [26:39<05:43, 3.73s/it] {'loss': 1.1822, 'grad_norm': 0.00046192543052533795, 'learning_rate': 0.03999351921230715, 'epoch': 0.82} + 82%|████████▏ | 428/520 [26:39<05:43, 3.73s/it] 82%|████████▎ | 429/520 [26:42<05:38, 3.72s/it] {'loss': 1.2957, 'grad_norm': 0.0004565869135707613, 'learning_rate': 0.039152138546778625, 'epoch': 0.82} + 82%|████████▎ | 429/520 [26:42<05:38, 3.72s/it]Token indices sequence length is longer than the specified maximum sequence length for this model (2076 > 2048). Running this sequence through the model will result in indexing errors + 83%|████████▎ | 430/520 [26:46<05:33, 3.70s/it] {'loss': 1.2837, 'grad_norm': 0.00045261881349153234, 'learning_rate': 0.03831895019292897, 'epoch': 0.83} + 83%|████████▎ | 430/520 [26:46<05:33, 3.70s/it] 83%|████████▎ | 431/520 [26:50<05:27, 3.68s/it] {'loss': 1.3734, 'grad_norm': 0.0005950096017181923, 'learning_rate': 0.03749398652357272, 'epoch': 0.83} + 83%|████████▎ | 431/520 [26:50<05:27, 3.68s/it] 83%|████████▎ | 432/520 [26:53<05:23, 3.68s/it] {'loss': 1.1892, 'grad_norm': 0.0004477303059712554, 'learning_rate': 0.0366772795919611, 'epoch': 0.83} + 83%|████████▎ | 432/520 [26:53<05:23, 3.68s/it] 83%|████████▎ | 433/520 [26:57<05:19, 3.67s/it] {'loss': 1.3331, 'grad_norm': 0.0004931801058441286, 'learning_rate': 0.035868861130537166, 'epoch': 0.83} + 83%|████████▎ | 433/520 [26:57<05:19, 3.67s/it] 83%|████████▎ | 434/520 [27:01<05:15, 3.67s/it] {'loss': 1.0748, 'grad_norm': 0.0005948893335496149, 'learning_rate': 0.035068762549702426, 'epoch': 0.83} + 83%|████████▎ | 434/520 [27:01<05:15, 3.67s/it] 84%|████████▎ | 435/520 [27:04<05:11, 3.66s/it] {'loss': 1.3849, 'grad_norm': 0.0005473924621299065, 'learning_rate': 0.03427701493659674, 'epoch': 0.84} + 84%|████████▎ | 435/520 [27:04<05:11, 3.66s/it] 84%|████████▍ | 436/520 [27:08<05:06, 3.65s/it] {'loss': 1.168, 'grad_norm': 0.0005741521554165338, 'learning_rate': 0.03349364905389032, 'epoch': 0.84} + 84%|████████▍ | 436/520 [27:08<05:06, 3.65s/it] 84%|████████▍ | 437/520 [27:12<05:03, 3.65s/it] {'loss': 1.396, 'grad_norm': 0.0004602805782989352, 'learning_rate': 0.0327186953385884, 'epoch': 0.84} + 84%|████████▍ | 437/520 [27:12<05:03, 3.65s/it] 84%|████████▍ | 438/520 [27:15<04:59, 3.66s/it] {'loss': 1.1894, 'grad_norm': 0.0004386012830075433, 'learning_rate': 0.03195218390084867, 'epoch': 0.84} + 84%|████████▍ | 438/520 [27:15<04:59, 3.66s/it] 84%|████████▍ | 439/520 [27:19<04:55, 3.65s/it] {'loss': 1.325, 'grad_norm': 0.0003800311058165404, 'learning_rate': 0.03119414452281158, 'epoch': 0.84} + 84%|████████▍ | 439/520 [27:19<04:55, 3.65s/it] 85%|████████▍ | 440/520 [27:23<04:52, 3.65s/it] {'loss': 1.2524, 'grad_norm': 0.0006669292703739447, 'learning_rate': 0.030444606657442835, 'epoch': 0.85} + 85%|████████▍ | 440/520 [27:23<04:52, 3.65s/it] 85%|████████▍ | 441/520 [27:26<04:48, 3.65s/it] {'loss': 1.3646, 'grad_norm': 0.0004658571344646486, 'learning_rate': 0.0297035994273894, 'epoch': 0.85} + 85%|████████▍ | 441/520 [27:26<04:48, 3.65s/it] 85%|████████▌ | 442/520 [27:30<04:44, 3.65s/it] {'loss': 1.3065, 'grad_norm': 0.0007221070819692379, 'learning_rate': 0.028971151623847585, 'epoch': 0.85} + 85%|████████▌ | 442/520 [27:30<04:44, 3.65s/it] 85%|████████▌ | 443/520 [27:34<04:41, 3.65s/it] {'loss': 1.3157, 'grad_norm': 0.0004429262272062958, 'learning_rate': 0.02824729170544457, 'epoch': 0.85} + 85%|████████▌ | 443/520 [27:34<04:41, 3.65s/it] 85%|████████▌ | 444/520 [27:37<04:37, 3.65s/it] {'loss': 1.2877, 'grad_norm': 0.0005082342752790778, 'learning_rate': 0.027532047797132864, 'epoch': 0.85} + 85%|████████▌ | 444/520 [27:37<04:37, 3.65s/it] 86%|████████▌ | 445/520 [27:41<04:34, 3.66s/it] {'loss': 1.2041, 'grad_norm': 0.0005045840637963218, 'learning_rate': 0.02682544768909717, 'epoch': 0.86} + 86%|████████▌ | 445/520 [27:41<04:34, 3.66s/it] 86%|████████▌ | 446/520 [27:45<04:30, 3.66s/it] {'loss': 1.4263, 'grad_norm': 0.0004463153705102194, 'learning_rate': 0.026127518835674768, 'epoch': 0.86} + 86%|████████▌ | 446/520 [27:45<04:30, 3.66s/it] 86%|████████▌ | 447/520 [27:48<04:27, 3.66s/it] {'loss': 1.289, 'grad_norm': 0.000472002045048552, 'learning_rate': 0.02543828835428899, 'epoch': 0.86} + 86%|████████▌ | 447/520 [27:48<04:27, 3.66s/it] 86%|████████▌ | 448/520 [27:52<04:23, 3.67s/it] {'loss': 1.2698, 'grad_norm': 0.0008144832824369589, 'learning_rate': 0.02475778302439524, 'epoch': 0.86} + 86%|████████▌ | 448/520 [27:52<04:23, 3.67s/it] 86%|████████▋ | 449/520 [27:56<04:19, 3.66s/it] {'loss': 1.4139, 'grad_norm': 0.0006029707180275999, 'learning_rate': 0.02408602928644088, 'epoch': 0.86} + 86%|████████▋ | 449/520 [27:56<04:19, 3.66s/it] 87%|████████▋ | 450/520 [27:59<04:15, 3.65s/it] {'loss': 1.3191, 'grad_norm': 0.0004964276378590791, 'learning_rate': 0.023423053240837516, 'epoch': 0.87} + 87%|████████▋ | 450/520 [27:59<04:15, 3.65s/it] 87%|████████▋ | 451/520 [28:03<04:12, 3.65s/it] {'loss': 1.3257, 'grad_norm': 0.0005826425356174742, 'learning_rate': 0.022768880646947265, 'epoch': 0.87} + 87%|████████▋ | 451/520 [28:03<04:12, 3.65s/it] 87%|████████▋ | 452/520 [28:07<04:08, 3.65s/it] {'loss': 1.4266, 'grad_norm': 0.00042184686138845256, 'learning_rate': 0.022123536922081716, 'epoch': 0.87} + 87%|████████▋ | 452/520 [28:07<04:08, 3.65s/it] 87%|████████▋ | 453/520 [28:10<04:04, 3.65s/it] {'loss': 1.406, 'grad_norm': 0.000655450996411973, 'learning_rate': 0.021487047140514248, 'epoch': 0.87} + 87%|████████▋ | 453/520 [28:10<04:04, 3.65s/it] 87%|████████▋ | 454/520 [28:14<04:01, 3.65s/it] {'loss': 1.2146, 'grad_norm': 0.00042340829995080216, 'learning_rate': 0.02085943603250595, 'epoch': 0.87} + 87%|████████▋ | 454/520 [28:14<04:01, 3.65s/it] 88%|████████▊ | 455/520 [28:17<03:57, 3.65s/it] {'loss': 1.3599, 'grad_norm': 0.0005779347490950407, 'learning_rate': 0.020240727983344836, 'epoch': 0.88} + 88%|████████▊ | 455/520 [28:18<03:57, 3.65s/it] 88%|████████▊ | 456/520 [28:21<03:53, 3.65s/it] {'loss': 1.2799, 'grad_norm': 0.0004432022120271323, 'learning_rate': 0.019630947032398066, 'epoch': 0.88} + 88%|████████▊ | 456/520 [28:21<03:53, 3.65s/it] 88%|████████▊ | 457/520 [28:25<03:50, 3.65s/it] {'loss': 1.3882, 'grad_norm': 0.000555833855780623, 'learning_rate': 0.019030116872178315, 'epoch': 0.88} + 88%|████████▊ | 457/520 [28:25<03:50, 3.65s/it] 88%|████████▊ | 458/520 [28:28<03:46, 3.65s/it] {'loss': 1.431, 'grad_norm': 0.0005528820323463165, 'learning_rate': 0.018438260847422838, 'epoch': 0.88} + 88%|████████▊ | 458/520 [28:28<03:46, 3.65s/it] 88%|████████▊ | 459/520 [28:32<03:43, 3.66s/it] {'loss': 1.3466, 'grad_norm': 0.0005793876399063554, 'learning_rate': 0.01785540195418661, 'epoch': 0.88} + 88%|████████▊ | 459/520 [28:32<03:43, 3.66s/it] 88%|████████▊ | 460/520 [28:36<03:39, 3.66s/it] {'loss': 1.2184, 'grad_norm': 0.0004506388411777561, 'learning_rate': 0.017281562838948966, 'epoch': 0.88} + 88%|████████▊ | 460/520 [28:36<03:39, 3.66s/it] 89%|████████▊ | 461/520 [28:39<03:35, 3.65s/it] {'loss': 1.4769, 'grad_norm': 0.00046964872848218443, 'learning_rate': 0.016716765797733374, 'epoch': 0.89} + 89%|████████▊ | 461/520 [28:39<03:35, 3.65s/it] 89%|████████▉ | 462/520 [28:43<03:31, 3.65s/it] {'loss': 1.4759, 'grad_norm': 0.000502800578911677, 'learning_rate': 0.0161610327752415, 'epoch': 0.89} + 89%|████████▉ | 462/520 [28:43<03:31, 3.65s/it] 89%|████████▉ | 463/520 [28:47<03:28, 3.65s/it] {'loss': 1.185, 'grad_norm': 0.00047575459359830343, 'learning_rate': 0.015614385364000227, 'epoch': 0.89} + 89%|████████▉ | 463/520 [28:47<03:28, 3.65s/it] 89%|████████▉ | 464/520 [28:50<03:24, 3.65s/it] {'loss': 1.3378, 'grad_norm': 0.0004916747303960815, 'learning_rate': 0.01507684480352292, 'epoch': 0.89} + 89%|████████▉ | 464/520 [28:50<03:24, 3.65s/it] 89%|████████▉ | 465/520 [28:54<03:20, 3.65s/it] {'loss': 1.4368, 'grad_norm': 0.0005033044018233399, 'learning_rate': 0.014548431979484133, 'epoch': 0.89} + 89%|████████▉ | 465/520 [28:54<03:20, 3.65s/it] 90%|████████▉ | 466/520 [28:58<03:16, 3.65s/it] {'loss': 1.3349, 'grad_norm': 0.0004204280274173982, 'learning_rate': 0.014029167422908106, 'epoch': 0.9} + 90%|████████▉ | 466/520 [28:58<03:16, 3.65s/it] 90%|████████▉ | 467/520 [29:01<03:13, 3.65s/it] {'loss': 1.3578, 'grad_norm': 0.00043854159889458386, 'learning_rate': 0.013519071309370995, 'epoch': 0.9} + 90%|████████▉ | 467/520 [29:01<03:13, 3.65s/it] 90%|█████████ | 468/520 [29:05<03:09, 3.65s/it] {'loss': 1.2982, 'grad_norm': 0.0005306848240928154, 'learning_rate': 0.013018163458217075, 'epoch': 0.9} + 90%|█████████ | 468/520 [29:05<03:09, 3.65s/it] 90%|█████████ | 469/520 [29:09<03:05, 3.65s/it] {'loss': 1.3648, 'grad_norm': 0.0005187263578188781, 'learning_rate': 0.012526463331788501, 'epoch': 0.9} + 90%|█████████ | 469/520 [29:09<03:05, 3.65s/it] 90%|█████████ | 470/520 [29:12<03:02, 3.65s/it] {'loss': 1.2205, 'grad_norm': 0.0004370716671877161, 'learning_rate': 0.01204399003466941, 'epoch': 0.9} + 90%|█████████ | 470/520 [29:12<03:02, 3.65s/it] 91%|█████████ | 471/520 [29:16<02:58, 3.65s/it] {'loss': 1.2756, 'grad_norm': 0.0006714822754883725, 'learning_rate': 0.011570762312943295, 'epoch': 0.91} + 91%|█████████ | 471/520 [29:16<02:58, 3.65s/it] 91%|█████████ | 472/520 [29:20<02:55, 3.66s/it] {'loss': 1.2263, 'grad_norm': 0.0004737356725347295, 'learning_rate': 0.011106798553464803, 'epoch': 0.91} + 91%|█████████ | 472/520 [29:20<02:55, 3.66s/it] 91%|█████████ | 473/520 [29:23<02:52, 3.66s/it] {'loss': 1.3041, 'grad_norm': 0.0004601468506341803, 'learning_rate': 0.010652116783145482, 'epoch': 0.91} + 91%|█████████ | 473/520 [29:23<02:52, 3.66s/it] 91%|█████████ | 474/520 [29:27<02:50, 3.71s/it] {'loss': 1.406, 'grad_norm': 0.00047994319301738184, 'learning_rate': 0.010206734668253059, 'epoch': 0.91} + 91%|█████████ | 474/520 [29:27<02:50, 3.71s/it] 91%|█████████▏| 475/520 [29:31<02:46, 3.70s/it] {'loss': 1.3059, 'grad_norm': 0.00048146498076018864, 'learning_rate': 0.009770669513725128, 'epoch': 0.91} + 91%|█████████▏| 475/520 [29:31<02:46, 3.70s/it] 92%|█████████▏| 476/520 [29:34<02:42, 3.69s/it] {'loss': 1.2718, 'grad_norm': 0.0005182216688188496, 'learning_rate': 0.009343938262496992, 'epoch': 0.92} + 92%|█████████▏| 476/520 [29:34<02:42, 3.69s/it] 92%|█████████▏| 477/520 [29:38<02:38, 3.68s/it] {'loss': 1.2718, 'grad_norm': 0.000676970716160501, 'learning_rate': 0.008926557494843085, 'epoch': 0.92} + 92%|█████████▏| 477/520 [29:38<02:38, 3.68s/it] 92%|█████████▏| 478/520 [29:42<02:34, 3.68s/it] {'loss': 1.2119, 'grad_norm': 0.0005057219264849662, 'learning_rate': 0.00851854342773295, 'epoch': 0.92} + 92%|█████████▏| 478/520 [29:42<02:34, 3.68s/it] 92%|█████████▏| 479/520 [29:45<02:30, 3.67s/it] {'loss': 1.3917, 'grad_norm': 0.0005628338453576699, 'learning_rate': 0.008119911914200972, 'epoch': 0.92} + 92%|█████████▏| 479/520 [29:45<02:30, 3.67s/it] 92%|█████████▏| 480/520 [29:49<02:26, 3.67s/it] {'loss': 1.3936, 'grad_norm': 0.0004885204733846416, 'learning_rate': 0.0077306784427305375, 'epoch': 0.92} + 92%|█████████▏| 480/520 [29:49<02:26, 3.67s/it] 92%|█████████▎| 481/520 [29:53<02:23, 3.67s/it] {'loss': 1.4172, 'grad_norm': 0.0004973816412630443, 'learning_rate': 0.007350858136652261, 'epoch': 0.93} + 92%|█████████▎| 481/520 [29:53<02:23, 3.67s/it] 93%|█████████▎| 482/520 [29:56<02:19, 3.66s/it] {'loss': 1.4177, 'grad_norm': 0.0004553592039175797, 'learning_rate': 0.006980465753556375, 'epoch': 0.93} + 93%|█████████▎| 482/520 [29:56<02:19, 3.66s/it] 93%|█████████▎| 483/520 [30:00<02:15, 3.66s/it] {'loss': 1.304, 'grad_norm': 0.0004748723900427597, 'learning_rate': 0.006619515684719163, 'epoch': 0.93} + 93%|█████████▎| 483/520 [30:00<02:15, 3.66s/it] 93%|█████████▎| 484/520 [30:04<02:11, 3.67s/it] {'loss': 1.2965, 'grad_norm': 0.000463684407717344, 'learning_rate': 0.006268021954544095, 'epoch': 0.93} + 93%|█████████▎| 484/520 [30:04<02:11, 3.67s/it] 93%|█████████▎| 485/520 [30:07<02:08, 3.67s/it] {'loss': 1.246, 'grad_norm': 0.0005770659022328152, 'learning_rate': 0.0059259982200166594, 'epoch': 0.93} + 93%|█████████▎| 485/520 [30:07<02:08, 3.67s/it] 93%|█████████▎| 486/520 [30:11<02:05, 3.68s/it] {'loss': 1.3737, 'grad_norm': 0.0004948246462584944, 'learning_rate': 0.005593457770173865, 'epoch': 0.93} + 93%|█████████▎| 486/520 [30:11<02:05, 3.68s/it] 94%|█████████▎| 487/520 [30:15<02:01, 3.69s/it] {'loss': 1.2305, 'grad_norm': 0.0004567288399527241, 'learning_rate': 0.005270413525587908, 'epoch': 0.94} + 94%|█████████▎| 487/520 [30:15<02:01, 3.69s/it] 94%|█████████▍| 488/520 [30:18<01:57, 3.68s/it] {'loss': 1.169, 'grad_norm': 0.0005216251711962053, 'learning_rate': 0.004956878037864043, 'epoch': 0.94} + 94%|█████████▍| 488/520 [30:18<01:57, 3.68s/it] 94%|█████████▍| 489/520 [30:22<01:53, 3.67s/it] {'loss': 1.4042, 'grad_norm': 0.00044891732035993195, 'learning_rate': 0.004652863489153086, 'epoch': 0.94} + 94%|█████████▍| 489/520 [30:22<01:53, 3.67s/it] 94%|█████████▍| 490/520 [30:26<01:50, 3.67s/it] {'loss': 1.2897, 'grad_norm': 0.00048187622178432913, 'learning_rate': 0.004358381691677932, 'epoch': 0.94} + 94%|█████████▍| 490/520 [30:26<01:50, 3.67s/it] 94%|█████████▍| 491/520 [30:29<01:46, 3.66s/it] {'loss': 1.2409, 'grad_norm': 0.0004905003812704163, 'learning_rate': 0.004073444087274669, 'epoch': 0.94} + 94%|█████████▍| 491/520 [30:29<01:46, 3.66s/it] 95%|█████████▍| 492/520 [30:33<01:42, 3.67s/it] {'loss': 1.3668, 'grad_norm': 0.0005847166161663524, 'learning_rate': 0.003798061746947995, 'epoch': 0.95} + 95%|█████████▍| 492/520 [30:33<01:42, 3.67s/it] 95%|█████████▍| 493/520 [30:37<01:39, 3.70s/it] {'loss': 1.458, 'grad_norm': 0.0005560819611859807, 'learning_rate': 0.0035322453704410284, 'epoch': 0.95} + 95%|█████████▍| 493/520 [30:37<01:39, 3.70s/it] 95%|█████████▌| 494/520 [30:41<01:37, 3.74s/it] {'loss': 1.3084, 'grad_norm': 0.0004930876910774732, 'learning_rate': 0.0032760052858197275, 'epoch': 0.95} + 95%|█████████▌| 494/520 [30:41<01:37, 3.74s/it] 95%|█████████▌| 495/520 [30:45<01:34, 3.77s/it] {'loss': 1.2692, 'grad_norm': 0.0004930200802329888, 'learning_rate': 0.0030293514490713214, 'epoch': 0.95} + 95%|█████████▌| 495/520 [30:45<01:34, 3.77s/it] 95%|█████████▌| 496/520 [30:48<01:30, 3.79s/it] {'loss': 1.204, 'grad_norm': 0.0004639013635929623, 'learning_rate': 0.0027922934437178693, 'epoch': 0.95} + 95%|█████████▌| 496/520 [30:48<01:30, 3.79s/it] 96%|█████████▌| 497/520 [30:52<01:27, 3.81s/it] {'loss': 1.3157, 'grad_norm': 0.00044734326926917754, 'learning_rate': 0.002564840480443503, 'epoch': 0.96} + 96%|█████████▌| 497/520 [30:52<01:27, 3.81s/it] 96%|█████████▌| 498/520 [30:56<01:23, 3.81s/it] {'loss': 1.2716, 'grad_norm': 0.0006522588298072334, 'learning_rate': 0.0023470013967367975, 'epoch': 0.96} + 96%|█████████▌| 498/520 [30:56<01:23, 3.81s/it] 96%|█████████▌| 499/520 [31:00<01:20, 3.82s/it] {'loss': 1.4825, 'grad_norm': 0.0005575467470603889, 'learning_rate': 0.0021387846565474045, 'epoch': 0.96} + 96%|█████████▌| 499/520 [31:00<01:20, 3.82s/it] 96%|█████████▌| 500/520 [31:04<01:16, 3.83s/it] {'loss': 1.3963, 'grad_norm': 0.0006869748531611493, 'learning_rate': 0.0019401983499569841, 'epoch': 0.96} + 96%|█████████▌| 500/520 [31:04<01:16, 3.83s/it] 96%|█████████▋| 501/520 [31:08<01:12, 3.83s/it] {'loss': 1.3815, 'grad_norm': 0.0005641773244334516, 'learning_rate': 0.0017512501928650948, 'epoch': 0.96} + 96%|█████████▋| 501/520 [31:08<01:12, 3.83s/it] 97%|█████████▋| 502/520 [31:11<01:09, 3.84s/it] {'loss': 1.2939, 'grad_norm': 0.0005140705761433629, 'learning_rate': 0.001571947526689349, 'epoch': 0.97} + 97%|█████████▋| 502/520 [31:12<01:09, 3.84s/it] 97%|█████████▋| 503/520 [31:15<01:05, 3.84s/it] {'loss': 1.3429, 'grad_norm': 0.0005024961668185098, 'learning_rate': 0.001402297318080059, 'epoch': 0.97} + 97%|█████████▋| 503/520 [31:15<01:05, 3.84s/it] 97%|█████████▋| 504/520 [31:19<01:01, 3.85s/it] {'loss': 1.2983, 'grad_norm': 0.0005217383324692644, 'learning_rate': 0.0012423061586496476, 'epoch': 0.97} + 97%|█████████▋| 504/520 [31:19<01:01, 3.85s/it] 97%|█████████▋| 505/520 [31:23<00:57, 3.84s/it] {'loss': 1.3442, 'grad_norm': 0.0005430538331783174, 'learning_rate': 0.0010919802647165466, 'epoch': 0.97} + 97%|█████████▋| 505/520 [31:23<00:57, 3.84s/it] 97%|█████████▋| 506/520 [31:27<00:53, 3.81s/it] {'loss': 1.2548, 'grad_norm': 0.0004769150963408378, 'learning_rate': 0.0009513254770636137, 'epoch': 0.97} + 97%|█████████▋| 506/520 [31:27<00:53, 3.81s/it] 98%|█████████▊| 507/520 [31:30<00:48, 3.77s/it] {'loss': 1.5177, 'grad_norm': 0.0005009889564023181, 'learning_rate': 0.0008203472607112294, 'epoch': 0.97} + 98%|█████████▊| 507/520 [31:30<00:48, 3.77s/it] 98%|█████████▊| 508/520 [31:34<00:44, 3.75s/it] {'loss': 1.3789, 'grad_norm': 0.0004966686136579567, 'learning_rate': 0.0006990507047049677, 'epoch': 0.98} + 98%|█████████▊| 508/520 [31:34<00:44, 3.75s/it] 98%|█████████▊| 509/520 [31:38<00:41, 3.75s/it] {'loss': 1.3478, 'grad_norm': 0.0004947180822586158, 'learning_rate': 0.0005874405219177814, 'epoch': 0.98} + 98%|█████████▊| 509/520 [31:38<00:41, 3.75s/it] 98%|█████████▊| 510/520 [31:42<00:37, 3.74s/it] {'loss': 1.297, 'grad_norm': 0.00045845154296242173, 'learning_rate': 0.00048552104886703806, 'epoch': 0.98} + 98%|█████████▊| 510/520 [31:42<00:37, 3.74s/it] 98%|█████████▊| 511/520 [31:45<00:33, 3.73s/it] {'loss': 1.2645, 'grad_norm': 0.0005608968666672025, 'learning_rate': 0.00039329624554584885, 'epoch': 0.98} + 98%|█████████▊| 511/520 [31:45<00:33, 3.73s/it] 98%|█████████▊| 512/520 [31:49<00:29, 3.71s/it] {'loss': 1.1451, 'grad_norm': 0.0004933782133072741, 'learning_rate': 0.0003107696952694139, 'epoch': 0.98} + 98%|█████████▊| 512/520 [31:49<00:29, 3.71s/it] 99%|█████████▊| 513/520 [31:53<00:25, 3.69s/it] {'loss': 1.3654, 'grad_norm': 0.0005861716610351249, 'learning_rate': 0.00023794460453555044, 'epoch': 0.99} + 99%|█████████▊| 513/520 [31:53<00:25, 3.69s/it] 99%|█████████▉| 514/520 [31:56<00:22, 3.70s/it] {'loss': 1.3366, 'grad_norm': 0.0004442123765426053, 'learning_rate': 0.00017482380290034794, 'epoch': 0.99} + 99%|█████████▉| 514/520 [31:56<00:22, 3.70s/it] 99%|█████████▉| 515/520 [32:00<00:18, 3.70s/it] {'loss': 1.3922, 'grad_norm': 0.0006888785032915229, 'learning_rate': 0.00012140974286808937, 'epoch': 0.99} + 99%|█████████▉| 515/520 [32:00<00:18, 3.70s/it] 99%|█████████▉| 516/520 [32:04<00:14, 3.68s/it] {'loss': 1.2717, 'grad_norm': 0.0004529481050208175, 'learning_rate': 7.770449979593863e-05, 'epoch': 0.99} + 99%|█████████▉| 516/520 [32:04<00:14, 3.68s/it] 99%|█████████▉| 517/520 [32:07<00:10, 3.66s/it] {'loss': 1.4166, 'grad_norm': 0.0005118490244105063, 'learning_rate': 4.370977181339386e-05, 'epoch': 0.99} + 99%|█████████▉| 517/520 [32:07<00:10, 3.66s/it] 100%|█████████▉| 518/520 [32:11<00:07, 3.64s/it] {'loss': 1.3055, 'grad_norm': 0.0005097219246675865, 'learning_rate': 1.9426879756284654e-05, 'epoch': 1.0} + 100%|█████████▉| 518/520 [32:11<00:07, 3.64s/it] 100%|█████████▉| 519/520 [32:14<00:03, 3.63s/it] {'loss': 1.3824, 'grad_norm': 0.0007005151033038591, 'learning_rate': 4.856767115452021e-06, 'epoch': 1.0} + 100%|█████████▉| 519/520 [32:14<00:03, 3.63s/it] 100%|██████████| 520/520 [32:19<00:00, 3.91s/it] {'loss': 1.4224, 'grad_norm': 0.0006298737076528061, 'learning_rate': 0.0, 'epoch': 1.0} + 100%|██████████| 520/520 [32:19<00:00, 3.91s/it] {'train_runtime': 1939.5614, 'train_samples_per_second': 34.301, 'train_steps_per_second': 0.268, 'train_loss': 1.381853013084485, 'epoch': 1.0} + 100%|██████████| 520/520 [32:19<00:00, 3.91s/it] 100%|██████████| 520/520 [32:19<00:00, 3.73s/it] +[2025-10-12 06:01:50,158] [INFO] [launch.py:348:main] Process 496294 exits successfully. +[2025-10-12 06:01:50,159] [INFO] [launch.py:348:main] Process 496288 exits successfully. +[2025-10-12 06:01:50,159] [INFO] [launch.py:348:main] Process 496289 exits successfully. +[2025-10-12 06:01:51,160] [INFO] [launch.py:348:main] Process 496293 exits successfully. +[2025-10-12 06:01:51,161] [INFO] [launch.py:348:main] Process 496290 exits successfully. +[2025-10-12 06:01:51,161] [INFO] [launch.py:348:main] Process 496291 exits successfully. +[2025-10-12 06:01:51,161] [INFO] [launch.py:348:main] Process 496292 exits successfully. +[2025-10-12 06:01:54,165] [INFO] [launch.py:348:main] Process 496287 exits successfully. +==== EXPERIMENT COMPLETED: qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_5e-1_connector-3.0_0.5_5e-1_ablation ==== +Log File: qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_5e-1_connector-3.0_0.5_5e-1_ablation_20251012_052751.log +Timestamp: 2025-10-12 06:01:56 +===================================== diff --git a/logs_oct11/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_5e-2_connector-3.0_0.5_5e-2_ablation_20251012_113222.log b/logs_oct11/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_5e-2_connector-3.0_0.5_5e-2_ablation_20251012_113222.log new file mode 100644 index 0000000000000000000000000000000000000000..69a4f6199063fd416dd66acee9b451726572d7f7 --- /dev/null +++ b/logs_oct11/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_5e-2_connector-3.0_0.5_5e-2_ablation_20251012_113222.log @@ -0,0 +1,2325 @@ +==== STARTING EXPERIMENT: qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_5e-2_connector-3.0_0.5_5e-2_ablation ==== +Log File: qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_5e-2_connector-3.0_0.5_5e-2_ablation_20251012_113222.log +Timestamp: 2025-10-12 11:32:22 +===================================== +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 11:32:25,124] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 11:32:28,751] [WARNING] [runner.py:202:fetch_hostfile] Unable to find hostfile, will proceed with training with local resources only. +[2025-10-12 11:32:28,753] [INFO] [runner.py:568:main] cmd = /opt/conda/envs/tinyllava/bin/python3.10 -u -m deepspeed.launcher.launch --world_info=eyJsb2NhbGhvc3QiOiBbMCwgMSwgMiwgMywgNCwgNSwgNiwgN119 --master_addr=127.0.0.1 --master_port=29501 --enable_each_rank_log=None tinyllava/train/train.py --deepspeed ./scripts/zero3.json --data_path /s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json --image_folder /s3-code/ywang29/datasets/tinyllava --is_multimodal True --conv_version qwen2_base --model_name_or_path Qwen/Qwen2.5-0.5B --vision_tower google/siglip-so400m-patch14-384 --vision_tower2 --connector_type mlp2x_gelu --mm_vision_select_layer -2 --image_aspect_ratio square --attn_implementation flash_attention_2 --bf16 True --training_recipe common --tune_type_llm full --tune_type_vision_tower frozen --tune_vision_tower_from_layer 0 --tune_type_connector full --group_by_modality_length True --pretrained_model_path /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain --output_dir /nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_5e-2_connector-3.0_0.5_5e-2_ablation --num_train_epochs 1 --per_device_train_batch_size 4 --per_device_eval_batch_size 4 --gradient_accumulation_steps 4 --evaluation_strategy no --learning_rate 5e-2 --weight_decay 0. --warmup_ratio 0.03 --lr_scheduler_type cosine --logging_steps 1 --tf32 False --model_max_length 2048 --gradient_checkpointing True --dataloader_num_workers 8 --lazy_preprocess True --report_to tensorboard --tokenizer_use_fast False --run_name tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune --subnet_mode_text both --subnet_type_text None --mask_type_text soft --init_mean_text 3.0 --temperature_attn_text 0.5 --temperature_mlp_text 0.5 --backward_type_text normal --masked_layers_text all --subnet_mode_vision both --subnet_type_vision None --mask_type_vision soft --init_mean_vision 3.0 --temperature_attn_vision 0.5 --temperature_mlp_vision 0.5 --backward_type_vision normal --masked_layers_vision all --subnet_type_connector global --mask_type_connector soft --init_mean_connector 3.0 --temperature_connector 0.5 --backward_type_connector normal --mm_projector_lr 5e-2 --seed 42 --mask_model llm-connector --save_strategy steps --save_steps 50000 --save_total_limit 1 --train_data_ratio 0.1 +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 11:32:31,329] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 11:32:32,371] [INFO] [launch.py:138:main] 0 NCCL_VERSION=2.21.5 +[2025-10-12 11:32:32,371] [INFO] [launch.py:138:main] 0 NCCL_SOCKET_IFNAME=eth +[2025-10-12 11:32:32,372] [INFO] [launch.py:145:main] WORLD INFO DICT: {'localhost': [0, 1, 2, 3, 4, 5, 6, 7]} +[2025-10-12 11:32:32,372] [INFO] [launch.py:151:main] nnodes=1, num_local_procs=8, node_rank=0 +[2025-10-12 11:32:32,372] [INFO] [launch.py:162:main] global_rank_mapping=defaultdict(, {'localhost': [0, 1, 2, 3, 4, 5, 6, 7]}) +[2025-10-12 11:32:32,372] [INFO] [launch.py:163:main] dist_world_size=8 +[2025-10-12 11:32:32,372] [INFO] [launch.py:165:main] Setting CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 +[2025-10-12 11:32:32,374] [INFO] [launch.py:253:main] process 720428 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=0', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_5e-2_connector-3.0_0.5_5e-2_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '5e-2', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '5e-2', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 11:32:32,376] [INFO] [launch.py:253:main] process 720429 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=1', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_5e-2_connector-3.0_0.5_5e-2_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '5e-2', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '5e-2', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 11:32:32,378] [INFO] [launch.py:253:main] process 720430 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=2', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_5e-2_connector-3.0_0.5_5e-2_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '5e-2', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '5e-2', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 11:32:32,380] [INFO] [launch.py:253:main] process 720431 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=3', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_5e-2_connector-3.0_0.5_5e-2_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '5e-2', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '5e-2', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 11:32:32,381] [INFO] [launch.py:253:main] process 720432 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=4', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_5e-2_connector-3.0_0.5_5e-2_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '5e-2', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '5e-2', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 11:32:32,383] [INFO] [launch.py:253:main] process 720433 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=5', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_5e-2_connector-3.0_0.5_5e-2_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '5e-2', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '5e-2', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 11:32:32,385] [INFO] [launch.py:253:main] process 720434 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=6', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_5e-2_connector-3.0_0.5_5e-2_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '5e-2', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '5e-2', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 11:32:32,387] [INFO] [launch.py:253:main] process 720435 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=7', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_5e-2_connector-3.0_0.5_5e-2_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '5e-2', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '5e-2', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 11:32:39,265] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 11:32:39,355] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 11:32:39,356] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 11:32:39,380] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 11:32:39,380] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 11:32:39,419] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 11:32:39,419] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 11:32:39,420] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 11:32:39,874] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 11:32:39,874] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 11:32:39,874] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 11:32:39,874] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 11:32:39,874] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 11:32:39,874] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 11:32:39,874] [INFO] [comm.py:668:init_distributed] Initializing TorchBackend in DeepSpeed with backend nccl +[2025-10-12 11:32:39,881] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 11:32:39,881] [INFO] [comm.py:637:init_distributed] cdb=None +Apply masks for the following modules: ['llm', 'connector'] +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +{'llm': {'model_name_or_path': 'Qwen/Qwen2.5-0.5B', 'cache_dir': None, 'attn_implementation': 'flash_attention_2', 'subnet_mode': 'both', 'subnet_type': 'None', 'sparsity_attn': None, 'sparsity_mlp': None, 'threshold_attn': None, 'threshold_mlp': None, 'temperature_attn': 0.5, 'temperature_mlp': 0.5, 'masked_layers': 'all', 'mask_type': 'soft', 'backward_type': 'normal'}, 'vision_tower': {'model_name_or_path': 'google/siglip-so400m-patch14-384'}, 'connector': {'connector_type': 'mlp2x_gelu', 'subnet_type': 'global', 'threshold': None, 'sparsity': None, 'temperature': 0.5, 'mask_type': 'soft', 'backward_type': 'normal'}} +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +TinyLlavaConfig { + "backward_type_connector": "normal", + "cache_dir": null, + "connector_type": "mlp2x_gelu", + "hidden_size": 896, + "ignore_index": -100, + "image_aspect_ratio": "square", + "image_token_index": -200, + "llm_model_name_or_path": "Qwen/Qwen2.5-0.5B", + "mask_model": [ + "llm", + "connector" + ], + "mask_type_connector": "soft", + "model_type": "tinyllava", + "num_queries": 128, + "num_resampler_layers": 3, + "pad_token": null, + "resampler_hidden_size": 768, + "sparsity_connector": null, + "subnet_type_connector": "global", + "temperature_connector": 0.5, + "text_config": { + "_name_or_path": "Qwen/Qwen2.5-0.5B", + "architectures": [ + "Qwen2ForCausalLM" + ], + "backward_type": "normal", + "bos_token_id": 151643, + "eos_token_id": 151643, + "hidden_size": 896, + "intermediate_size": 4864, + "mask_type": "soft", + "masked_layers": "all", + "max_position_embeddings": 32768, + "max_window_layers": 24, + "model_type": "qwen2", + "num_attention_heads": 14, + "num_hidden_layers": 24, + "num_key_value_heads": 2, + "rope_theta": 1000000.0, + "sliding_window": 32768, + "subnet_mode": "both", + "subnet_type": "None", + "temperature_attn": 0.5, + "temperature_mlp": 0.5, + "tie_word_embeddings": true, + "torch_dtype": "bfloat16", + "use_mrope": false, + "use_sliding_window": false, + "vocab_size": 151936 + }, + "threshold_connector": null, + "tokenizer_model_max_length": 2048, + "tokenizer_name_or_path": "Qwen/Qwen2.5-0.5B", + "tokenizer_padding_side": "right", + "tokenizer_use_fast": false, + "transformers_version": "4.40.1", + "tune_type_connector": "frozen", + "tune_type_llm": "frozen", + "tune_type_vision_tower": "frozen", + "tune_vision_tower_from_layer": -1, + "use_cache": false, + "vision_config": { + "hidden_act": "gelu_pytorch_tanh", + "hidden_size": 1152, + "image_size": 384, + "intermediate_size": 4304, + "layer_norm_eps": 1e-06, + "model_name_or_path": "google/siglip-so400m-patch14-384", + "model_name_or_path2": "", + "model_type": "siglip_vision_model", + "num_attention_heads": 16, + "num_hidden_layers": 27, + "patch_size": 14 + }, + "vision_feature_layer": -2, + "vision_feature_select_strategy": "patch", + "vision_hidden_size": 1152, + "vision_model_name_or_path": "google/siglip-so400m-patch14-384", + "vision_model_name_or_path2": "", + "vocab_size": 151936 +} + +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +ywang29-vrdb-test2-worker-0:720428:720428 [0] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:720428:720428 [0] NCCL INFO Bootstrap : Using eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:720428:720428 [0] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test2-worker-0:720428:720428 [0] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test2-worker-0:720428:720428 [0] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test2-worker-0:720428:720428 [0] NCCL INFO cudaDriverVersion 12040 +NCCL version 2.21.5+cuda12.1 +ywang29-vrdb-test2-worker-0:720430:720430 [2] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test2-worker-0:720430:720430 [2] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:720432:720432 [4] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test2-worker-0:720430:720430 [2] NCCL INFO Bootstrap : Using eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:720432:720432 [4] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:720432:720432 [4] NCCL INFO Bootstrap : Using eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:720430:720430 [2] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test2-worker-0:720430:720430 [2] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test2-worker-0:720430:720430 [2] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test2-worker-0:720432:720432 [4] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test2-worker-0:720432:720432 [4] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test2-worker-0:720432:720432 [4] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test2-worker-0:720428:722034 [0] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:720428:722034 [0] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test2-worker-0:720428:722034 [0] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:720428:722034 [0] NCCL INFO NET/Socket : Using [0]eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:720428:722034 [0] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test2-worker-0:720428:722034 [0] NCCL INFO Using network Socket +ywang29-vrdb-test2-worker-0:720430:722035 [2] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:720430:722035 [2] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test2-worker-0:720430:722035 [2] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:720430:722035 [2] NCCL INFO NET/Socket : Using [0]eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:720430:722035 [2] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test2-worker-0:720430:722035 [2] NCCL INFO Using network Socket +ywang29-vrdb-test2-worker-0:720432:722036 [4] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:720432:722036 [4] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test2-worker-0:720432:722036 [4] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:720432:722036 [4] NCCL INFO NET/Socket : Using [0]eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:720432:722036 [4] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test2-worker-0:720432:722036 [4] NCCL INFO Using network Socket +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test2-worker-0:720434:720434 [6] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test2-worker-0:720434:720434 [6] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:720434:720434 [6] NCCL INFO Bootstrap : Using eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:720433:720433 [5] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test2-worker-0:720433:720433 [5] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:720433:720433 [5] NCCL INFO Bootstrap : Using eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:720434:720434 [6] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test2-worker-0:720434:720434 [6] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test2-worker-0:720434:720434 [6] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test2-worker-0:720433:720433 [5] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test2-worker-0:720433:720433 [5] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test2-worker-0:720433:720433 [5] NCCL INFO NET/Plugin: Using internal network plugin. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test2-worker-0:720429:720429 [1] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test2-worker-0:720429:720429 [1] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:720429:720429 [1] NCCL INFO Bootstrap : Using eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:720429:720429 [1] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test2-worker-0:720429:720429 [1] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test2-worker-0:720429:720429 [1] NCCL INFO NET/Plugin: Using internal network plugin. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test2-worker-0:720433:722038 [5] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:720433:722038 [5] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test2-worker-0:720433:722038 [5] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:720433:722038 [5] NCCL INFO NET/Socket : Using [0]eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:720433:722038 [5] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test2-worker-0:720433:722038 [5] NCCL INFO Using network Socket +ywang29-vrdb-test2-worker-0:720434:722037 [6] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:720434:722037 [6] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test2-worker-0:720434:722037 [6] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:720434:722037 [6] NCCL INFO NET/Socket : Using [0]eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:720434:722037 [6] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test2-worker-0:720434:722037 [6] NCCL INFO Using network Socket +ywang29-vrdb-test2-worker-0:720431:720431 [3] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test2-worker-0:720431:720431 [3] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:720431:720431 [3] NCCL INFO Bootstrap : Using eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:720431:720431 [3] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test2-worker-0:720431:720431 [3] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test2-worker-0:720431:720431 [3] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test2-worker-0:720429:722039 [1] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:720429:722039 [1] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test2-worker-0:720429:722039 [1] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:720429:722039 [1] NCCL INFO NET/Socket : Using [0]eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:720429:722039 [1] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test2-worker-0:720429:722039 [1] NCCL INFO Using network Socket +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test2-worker-0:720435:720435 [7] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test2-worker-0:720435:720435 [7] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:720435:720435 [7] NCCL INFO Bootstrap : Using eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:720435:720435 [7] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test2-worker-0:720435:720435 [7] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test2-worker-0:720435:720435 [7] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test2-worker-0:720431:722040 [3] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:720431:722040 [3] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test2-worker-0:720431:722040 [3] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:720431:722040 [3] NCCL INFO NET/Socket : Using [0]eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:720431:722040 [3] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test2-worker-0:720431:722040 [3] NCCL INFO Using network Socket +ywang29-vrdb-test2-worker-0:720435:722041 [7] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:720435:722041 [7] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test2-worker-0:720435:722041 [7] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:720435:722041 [7] NCCL INFO NET/Socket : Using [0]eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:720435:722041 [7] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test2-worker-0:720435:722041 [7] NCCL INFO Using network Socket +ywang29-vrdb-test2-worker-0:720435:722041 [7] NCCL INFO ncclCommInitRank comm 0x55ba9f4bb640 rank 7 nranks 8 cudaDev 7 nvmlDev 7 busId a01d0 commId 0x149f63fecb3e4a5d - Init START +ywang29-vrdb-test2-worker-0:720434:722037 [6] NCCL INFO ncclCommInitRank comm 0x5569a6daac30 rank 6 nranks 8 cudaDev 6 nvmlDev 6 busId a01c0 commId 0x149f63fecb3e4a5d - Init START +ywang29-vrdb-test2-worker-0:720429:722039 [1] NCCL INFO ncclCommInitRank comm 0x560824237a50 rank 1 nranks 8 cudaDev 1 nvmlDev 1 busId 101d0 commId 0x149f63fecb3e4a5d - Init START +ywang29-vrdb-test2-worker-0:720432:722036 [4] NCCL INFO ncclCommInitRank comm 0x556b8d5713b0 rank 4 nranks 8 cudaDev 4 nvmlDev 4 busId 901c0 commId 0x149f63fecb3e4a5d - Init START +ywang29-vrdb-test2-worker-0:720428:722034 [0] NCCL INFO ncclCommInitRank comm 0x55d7da9e9650 rank 0 nranks 8 cudaDev 0 nvmlDev 0 busId 101c0 commId 0x149f63fecb3e4a5d - Init START +ywang29-vrdb-test2-worker-0:720433:722038 [5] NCCL INFO ncclCommInitRank comm 0x55c945a6d9b0 rank 5 nranks 8 cudaDev 5 nvmlDev 5 busId 901d0 commId 0x149f63fecb3e4a5d - Init START +ywang29-vrdb-test2-worker-0:720431:722040 [3] NCCL INFO ncclCommInitRank comm 0x56492cd58b90 rank 3 nranks 8 cudaDev 3 nvmlDev 3 busId 201d0 commId 0x149f63fecb3e4a5d - Init START +ywang29-vrdb-test2-worker-0:720430:722035 [2] NCCL INFO ncclCommInitRank comm 0x56110a1eff40 rank 2 nranks 8 cudaDev 2 nvmlDev 2 busId 201c0 commId 0x149f63fecb3e4a5d - Init START +ywang29-vrdb-test2-worker-0:720429:722039 [1] NCCL INFO Setting affinity for GPU 1 to ff,ffff0000,00ffffff +ywang29-vrdb-test2-worker-0:720434:722037 [6] NCCL INFO Setting affinity for GPU 6 to ffffff00,0000ffff,ff000000 +ywang29-vrdb-test2-worker-0:720429:722039 [1] NCCL INFO NVLS multicast support is not available on dev 1 +ywang29-vrdb-test2-worker-0:720434:722037 [6] NCCL INFO NVLS multicast support is not available on dev 6 +ywang29-vrdb-test2-worker-0:720431:722040 [3] NCCL INFO Setting affinity for GPU 3 to ff,ffff0000,00ffffff +ywang29-vrdb-test2-worker-0:720431:722040 [3] NCCL INFO NVLS multicast support is not available on dev 3 +ywang29-vrdb-test2-worker-0:720430:722035 [2] NCCL INFO Setting affinity for GPU 2 to ff,ffff0000,00ffffff +ywang29-vrdb-test2-worker-0:720435:722041 [7] NCCL INFO Setting affinity for GPU 7 to ffffff00,0000ffff,ff000000 +ywang29-vrdb-test2-worker-0:720435:722041 [7] NCCL INFO NVLS multicast support is not available on dev 7 +ywang29-vrdb-test2-worker-0:720430:722035 [2] NCCL INFO NVLS multicast support is not available on dev 2 +ywang29-vrdb-test2-worker-0:720432:722036 [4] NCCL INFO Setting affinity for GPU 4 to ffffff00,0000ffff,ff000000 +ywang29-vrdb-test2-worker-0:720432:722036 [4] NCCL INFO NVLS multicast support is not available on dev 4 +ywang29-vrdb-test2-worker-0:720428:722034 [0] NCCL INFO Setting affinity for GPU 0 to ff,ffff0000,00ffffff +ywang29-vrdb-test2-worker-0:720428:722034 [0] NCCL INFO NVLS multicast support is not available on dev 0 +ywang29-vrdb-test2-worker-0:720433:722038 [5] NCCL INFO Setting affinity for GPU 5 to ffffff00,0000ffff,ff000000 +ywang29-vrdb-test2-worker-0:720433:722038 [5] NCCL INFO NVLS multicast support is not available on dev 5 +ywang29-vrdb-test2-worker-0:720433:722038 [5] NCCL INFO comm 0x55c945a6d9b0 rank 5 nRanks 8 nNodes 1 localRanks 8 localRank 5 MNNVL 0 +ywang29-vrdb-test2-worker-0:720432:722036 [4] NCCL INFO comm 0x556b8d5713b0 rank 4 nRanks 8 nNodes 1 localRanks 8 localRank 4 MNNVL 0 +ywang29-vrdb-test2-worker-0:720431:722040 [3] NCCL INFO comm 0x56492cd58b90 rank 3 nRanks 8 nNodes 1 localRanks 8 localRank 3 MNNVL 0 +ywang29-vrdb-test2-worker-0:720434:722037 [6] NCCL INFO comm 0x5569a6daac30 rank 6 nRanks 8 nNodes 1 localRanks 8 localRank 6 MNNVL 0 +ywang29-vrdb-test2-worker-0:720430:722035 [2] NCCL INFO comm 0x56110a1eff40 rank 2 nRanks 8 nNodes 1 localRanks 8 localRank 2 MNNVL 0 +ywang29-vrdb-test2-worker-0:720429:722039 [1] NCCL INFO comm 0x560824237a50 rank 1 nRanks 8 nNodes 1 localRanks 8 localRank 1 MNNVL 0 +ywang29-vrdb-test2-worker-0:720428:722034 [0] NCCL INFO comm 0x55d7da9e9650 rank 0 nRanks 8 nNodes 1 localRanks 8 localRank 0 MNNVL 0 +ywang29-vrdb-test2-worker-0:720435:722041 [7] NCCL INFO comm 0x55ba9f4bb640 rank 7 nRanks 8 nNodes 1 localRanks 8 localRank 7 MNNVL 0 +ywang29-vrdb-test2-worker-0:720428:722034 [0] NCCL INFO Channel 00/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:720433:722038 [5] NCCL INFO Trees [0] 6/-1/-1->5->4 [1] 6/-1/-1->5->4 [2] 6/-1/-1->5->4 [3] 6/-1/-1->5->4 [4] 6/-1/-1->5->4 [5] 6/-1/-1->5->4 [6] 6/-1/-1->5->4 [7] 6/-1/-1->5->4 [8] 6/-1/-1->5->4 [9] 6/-1/-1->5->4 [10] 6/-1/-1->5->4 [11] 6/-1/-1->5->4 [12] 6/-1/-1->5->4 [13] 6/-1/-1->5->4 [14] 6/-1/-1->5->4 [15] 6/-1/-1->5->4 [16] 6/-1/-1->5->4 [17] 6/-1/-1->5->4 [18] 6/-1/-1->5->4 [19] 6/-1/-1->5->4 [20] 6/-1/-1->5->4 [21] 6/-1/-1->5->4 [22] 6/-1/-1->5->4 [23] 6/-1/-1->5->4 +ywang29-vrdb-test2-worker-0:720432:722036 [4] NCCL INFO Trees [0] 5/-1/-1->4->3 [1] 5/-1/-1->4->3 [2] 5/-1/-1->4->3 [3] 5/-1/-1->4->3 [4] 5/-1/-1->4->3 [5] 5/-1/-1->4->3 [6] 5/-1/-1->4->3 [7] 5/-1/-1->4->3 [8] 5/-1/-1->4->3 [9] 5/-1/-1->4->3 [10] 5/-1/-1->4->3 [11] 5/-1/-1->4->3 [12] 5/-1/-1->4->3 [13] 5/-1/-1->4->3 [14] 5/-1/-1->4->3 [15] 5/-1/-1->4->3 [16] 5/-1/-1->4->3 [17] 5/-1/-1->4->3 [18] 5/-1/-1->4->3 [19] 5/-1/-1->4->3 [20] 5/-1/-1->4->3 [21] 5/-1/-1->4->3 [22] 5/-1/-1->4->3 [23] 5/-1/-1->4->3 +ywang29-vrdb-test2-worker-0:720428:722034 [0] NCCL INFO Channel 01/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:720433:722038 [5] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:720431:722040 [3] NCCL INFO Trees [0] 4/-1/-1->3->2 [1] 4/-1/-1->3->2 [2] 4/-1/-1->3->2 [3] 4/-1/-1->3->2 [4] 4/-1/-1->3->2 [5] 4/-1/-1->3->2 [6] 4/-1/-1->3->2 [7] 4/-1/-1->3->2 [8] 4/-1/-1->3->2 [9] 4/-1/-1->3->2 [10] 4/-1/-1->3->2 [11] 4/-1/-1->3->2 [12] 4/-1/-1->3->2 [13] 4/-1/-1->3->2 [14] 4/-1/-1->3->2 [15] 4/-1/-1->3->2 [16] 4/-1/-1->3->2 [17] 4/-1/-1->3->2 [18] 4/-1/-1->3->2 [19] 4/-1/-1->3->2 [20] 4/-1/-1->3->2 [21] 4/-1/-1->3->2 [22] 4/-1/-1->3->2 [23] 4/-1/-1->3->2 +ywang29-vrdb-test2-worker-0:720428:722034 [0] NCCL INFO Channel 02/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:720432:722036 [4] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:720430:722035 [2] NCCL INFO Trees [0] 3/-1/-1->2->1 [1] 3/-1/-1->2->1 [2] 3/-1/-1->2->1 [3] 3/-1/-1->2->1 [4] 3/-1/-1->2->1 [5] 3/-1/-1->2->1 [6] 3/-1/-1->2->1 [7] 3/-1/-1->2->1 [8] 3/-1/-1->2->1 [9] 3/-1/-1->2->1 [10] 3/-1/-1->2->1 [11] 3/-1/-1->2->1 [12] 3/-1/-1->2->1 [13] 3/-1/-1->2->1 [14] 3/-1/-1->2->1 [15] 3/-1/-1->2->1 [16] 3/-1/-1->2->1 [17] 3/-1/-1->2->1 [18] 3/-1/-1->2->1 [19] 3/-1/-1->2->1 [20] 3/-1/-1->2->1 [21] 3/-1/-1->2->1 [22] 3/-1/-1->2->1 [23] 3/-1/-1->2->1 +ywang29-vrdb-test2-worker-0:720429:722039 [1] NCCL INFO Trees [0] 2/-1/-1->1->0 [1] 2/-1/-1->1->0 [2] 2/-1/-1->1->0 [3] 2/-1/-1->1->0 [4] 2/-1/-1->1->0 [5] 2/-1/-1->1->0 [6] 2/-1/-1->1->0 [7] 2/-1/-1->1->0 [8] 2/-1/-1->1->0 [9] 2/-1/-1->1->0 [10] 2/-1/-1->1->0 [11] 2/-1/-1->1->0 [12] 2/-1/-1->1->0 [13] 2/-1/-1->1->0 [14] 2/-1/-1->1->0 [15] 2/-1/-1->1->0 [16] 2/-1/-1->1->0 [17] 2/-1/-1->1->0 [18] 2/-1/-1->1->0 [19] 2/-1/-1->1->0 [20] 2/-1/-1->1->0 [21] 2/-1/-1->1->0 [22] 2/-1/-1->1->0 [23] 2/-1/-1->1->0 +ywang29-vrdb-test2-worker-0:720428:722034 [0] NCCL INFO Channel 03/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:720431:722040 [3] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:720428:722034 [0] NCCL INFO Channel 04/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:720430:722035 [2] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:720429:722039 [1] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:720428:722034 [0] NCCL INFO Channel 05/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:720428:722034 [0] NCCL INFO Channel 06/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:720428:722034 [0] NCCL INFO Channel 07/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:720428:722034 [0] NCCL INFO Channel 08/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:720428:722034 [0] NCCL INFO Channel 09/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:720428:722034 [0] NCCL INFO Channel 10/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:720428:722034 [0] NCCL INFO Channel 11/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:720428:722034 [0] NCCL INFO Channel 12/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:720428:722034 [0] NCCL INFO Channel 13/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:720428:722034 [0] NCCL INFO Channel 14/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:720428:722034 [0] NCCL INFO Channel 15/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:720428:722034 [0] NCCL INFO Channel 16/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:720428:722034 [0] NCCL INFO Channel 17/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:720434:722037 [6] NCCL INFO Trees [0] 7/-1/-1->6->5 [1] 7/-1/-1->6->5 [2] 7/-1/-1->6->5 [3] 7/-1/-1->6->5 [4] 7/-1/-1->6->5 [5] 7/-1/-1->6->5 [6] 7/-1/-1->6->5 [7] 7/-1/-1->6->5 [8] 7/-1/-1->6->5 [9] 7/-1/-1->6->5 [10] 7/-1/-1->6->5 [11] 7/-1/-1->6->5 [12] 7/-1/-1->6->5 [13] 7/-1/-1->6->5 [14] 7/-1/-1->6->5 [15] 7/-1/-1->6->5 [16] 7/-1/-1->6->5 [17] 7/-1/-1->6->5 [18] 7/-1/-1->6->5 [19] 7/-1/-1->6->5 [20] 7/-1/-1->6->5 [21] 7/-1/-1->6->5 [22] 7/-1/-1->6->5 [23] 7/-1/-1->6->5 +ywang29-vrdb-test2-worker-0:720428:722034 [0] NCCL INFO Channel 18/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:720428:722034 [0] NCCL INFO Channel 19/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:720435:722041 [7] NCCL INFO Trees [0] -1/-1/-1->7->6 [1] -1/-1/-1->7->6 [2] -1/-1/-1->7->6 [3] -1/-1/-1->7->6 [4] -1/-1/-1->7->6 [5] -1/-1/-1->7->6 [6] -1/-1/-1->7->6 [7] -1/-1/-1->7->6 [8] -1/-1/-1->7->6 [9] -1/-1/-1->7->6 [10] -1/-1/-1->7->6 [11] -1/-1/-1->7->6 [12] -1/-1/-1->7->6 [13] -1/-1/-1->7->6 [14] -1/-1/-1->7->6 [15] -1/-1/-1->7->6 [16] -1/-1/-1->7->6 [17] -1/-1/-1->7->6 [18] -1/-1/-1->7->6 [19] -1/-1/-1->7->6 [20] -1/-1/-1->7->6 [21] -1/-1/-1->7->6 [22] -1/-1/-1->7->6 [23] -1/-1/-1->7->6 +ywang29-vrdb-test2-worker-0:720428:722034 [0] NCCL INFO Channel 20/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:720428:722034 [0] NCCL INFO Channel 21/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:720434:722037 [6] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:720428:722034 [0] NCCL INFO Channel 22/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:720428:722034 [0] NCCL INFO Channel 23/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:720435:722041 [7] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:720428:722034 [0] NCCL INFO Trees [0] 1/-1/-1->0->-1 [1] 1/-1/-1->0->-1 [2] 1/-1/-1->0->-1 [3] 1/-1/-1->0->-1 [4] 1/-1/-1->0->-1 [5] 1/-1/-1->0->-1 [6] 1/-1/-1->0->-1 [7] 1/-1/-1->0->-1 [8] 1/-1/-1->0->-1 [9] 1/-1/-1->0->-1 [10] 1/-1/-1->0->-1 [11] 1/-1/-1->0->-1 [12] 1/-1/-1->0->-1 [13] 1/-1/-1->0->-1 [14] 1/-1/-1->0->-1 [15] 1/-1/-1->0->-1 [16] 1/-1/-1->0->-1 [17] 1/-1/-1->0->-1 [18] 1/-1/-1->0->-1 [19] 1/-1/-1->0->-1 [20] 1/-1/-1->0->-1 [21] 1/-1/-1->0->-1 [22] 1/-1/-1->0->-1 [23] 1/-1/-1->0->-1 +ywang29-vrdb-test2-worker-0:720428:722034 [0] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:720432:722036 [4] NCCL INFO Channel 00/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:722035 [2] NCCL INFO Channel 00/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:722040 [3] NCCL INFO Channel 00/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:722039 [1] NCCL INFO Channel 00/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:722038 [5] NCCL INFO Channel 00/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:722036 [4] NCCL INFO Channel 01/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720428:722034 [0] NCCL INFO Channel 00/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:722035 [2] NCCL INFO Channel 01/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:722038 [5] NCCL INFO Channel 01/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:722040 [3] NCCL INFO Channel 01/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:722039 [1] NCCL INFO Channel 01/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:722036 [4] NCCL INFO Channel 02/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:722041 [7] NCCL INFO Channel 00/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720428:722034 [0] NCCL INFO Channel 01/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:722035 [2] NCCL INFO Channel 02/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:722038 [5] NCCL INFO Channel 02/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:722040 [3] NCCL INFO Channel 02/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:722039 [1] NCCL INFO Channel 02/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:722036 [4] NCCL INFO Channel 03/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:722041 [7] NCCL INFO Channel 01/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720428:722034 [0] NCCL INFO Channel 02/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:722035 [2] NCCL INFO Channel 03/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:722038 [5] NCCL INFO Channel 03/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:722040 [3] NCCL INFO Channel 03/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:722039 [1] NCCL INFO Channel 03/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:722036 [4] NCCL INFO Channel 04/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:722037 [6] NCCL INFO Channel 00/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720428:722034 [0] NCCL INFO Channel 03/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:722041 [7] NCCL INFO Channel 02/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:722035 [2] NCCL INFO Channel 04/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:722038 [5] NCCL INFO Channel 04/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:722040 [3] NCCL INFO Channel 04/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:722039 [1] NCCL INFO Channel 04/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:722036 [4] NCCL INFO Channel 05/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:722037 [6] NCCL INFO Channel 01/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720428:722034 [0] NCCL INFO Channel 04/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:722041 [7] NCCL INFO Channel 03/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:722038 [5] NCCL INFO Channel 05/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:722035 [2] NCCL INFO Channel 05/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:722040 [3] NCCL INFO Channel 05/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:722039 [1] NCCL INFO Channel 05/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:722036 [4] NCCL INFO Channel 06/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:722037 [6] NCCL INFO Channel 02/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720428:722034 [0] NCCL INFO Channel 05/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:722041 [7] NCCL INFO Channel 04/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:722038 [5] NCCL INFO Channel 06/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:722035 [2] NCCL INFO Channel 06/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:722040 [3] NCCL INFO Channel 06/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:722039 [1] NCCL INFO Channel 06/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:722036 [4] NCCL INFO Channel 07/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:722037 [6] NCCL INFO Channel 03/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720428:722034 [0] NCCL INFO Channel 06/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:722041 [7] NCCL INFO Channel 05/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:722038 [5] NCCL INFO Channel 07/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:722035 [2] NCCL INFO Channel 07/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:722040 [3] NCCL INFO Channel 07/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:722039 [1] NCCL INFO Channel 07/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:722036 [4] NCCL INFO Channel 08/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:722037 [6] NCCL INFO Channel 04/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720428:722034 [0] NCCL INFO Channel 07/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:722040 [3] NCCL INFO Channel 08/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:722038 [5] NCCL INFO Channel 08/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:722041 [7] NCCL INFO Channel 06/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:722035 [2] NCCL INFO Channel 08/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:722039 [1] NCCL INFO Channel 08/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:722036 [4] NCCL INFO Channel 09/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720428:722034 [0] NCCL INFO Channel 08/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:722037 [6] NCCL INFO Channel 05/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:722040 [3] NCCL INFO Channel 09/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:722038 [5] NCCL INFO Channel 09/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:722035 [2] NCCL INFO Channel 09/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:722039 [1] NCCL INFO Channel 09/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:722041 [7] NCCL INFO Channel 07/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:722036 [4] NCCL INFO Channel 10/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720428:722034 [0] NCCL INFO Channel 09/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:722037 [6] NCCL INFO Channel 06/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:722040 [3] NCCL INFO Channel 10/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:722038 [5] NCCL INFO Channel 10/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:722035 [2] NCCL INFO Channel 10/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:722039 [1] NCCL INFO Channel 10/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:722041 [7] NCCL INFO Channel 08/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:722036 [4] NCCL INFO Channel 11/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720428:722034 [0] NCCL INFO Channel 10/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:722037 [6] NCCL INFO Channel 07/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:722040 [3] NCCL INFO Channel 11/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:722038 [5] NCCL INFO Channel 11/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:722035 [2] NCCL INFO Channel 11/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:722039 [1] NCCL INFO Channel 11/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:722036 [4] NCCL INFO Channel 12/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:722041 [7] NCCL INFO Channel 09/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720428:722034 [0] NCCL INFO Channel 11/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:722037 [6] NCCL INFO Channel 08/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:722040 [3] NCCL INFO Channel 12/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:722038 [5] NCCL INFO Channel 12/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:722035 [2] NCCL INFO Channel 12/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:722039 [1] NCCL INFO Channel 12/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:722036 [4] NCCL INFO Channel 13/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:722041 [7] NCCL INFO Channel 10/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720428:722034 [0] NCCL INFO Channel 12/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:722037 [6] NCCL INFO Channel 09/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:722040 [3] NCCL INFO Channel 13/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:722038 [5] NCCL INFO Channel 13/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:722035 [2] NCCL INFO Channel 13/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:722039 [1] NCCL INFO Channel 13/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:722036 [4] NCCL INFO Channel 14/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:722041 [7] NCCL INFO Channel 11/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720428:722034 [0] NCCL INFO Channel 13/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:722040 [3] NCCL INFO Channel 14/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:722037 [6] NCCL INFO Channel 10/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:722038 [5] NCCL INFO Channel 14/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:722036 [4] NCCL INFO Channel 15/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:722035 [2] NCCL INFO Channel 14/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:722039 [1] NCCL INFO Channel 14/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:722041 [7] NCCL INFO Channel 12/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720428:722034 [0] NCCL INFO Channel 14/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:722040 [3] NCCL INFO Channel 15/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:722036 [4] NCCL INFO Channel 16/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:722037 [6] NCCL INFO Channel 11/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:722038 [5] NCCL INFO Channel 15/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:722035 [2] NCCL INFO Channel 15/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:722039 [1] NCCL INFO Channel 15/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:722041 [7] NCCL INFO Channel 13/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720428:722034 [0] NCCL INFO Channel 15/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:722040 [3] NCCL INFO Channel 16/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:722036 [4] NCCL INFO Channel 17/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:722037 [6] NCCL INFO Channel 12/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:722038 [5] NCCL INFO Channel 16/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:722035 [2] NCCL INFO Channel 16/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:722039 [1] NCCL INFO Channel 16/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:722041 [7] NCCL INFO Channel 14/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720428:722034 [0] NCCL INFO Channel 16/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:722040 [3] NCCL INFO Channel 17/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:722036 [4] NCCL INFO Channel 18/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:722037 [6] NCCL INFO Channel 13/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:722038 [5] NCCL INFO Channel 17/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:722035 [2] NCCL INFO Channel 17/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:722039 [1] NCCL INFO Channel 17/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:722041 [7] NCCL INFO Channel 15/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720428:722034 [0] NCCL INFO Channel 17/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:722040 [3] NCCL INFO Channel 18/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:722036 [4] NCCL INFO Channel 19/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:722037 [6] NCCL INFO Channel 14/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:722038 [5] NCCL INFO Channel 18/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:722035 [2] NCCL INFO Channel 18/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:722039 [1] NCCL INFO Channel 18/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:722041 [7] NCCL INFO Channel 16/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720428:722034 [0] NCCL INFO Channel 18/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:722040 [3] NCCL INFO Channel 19/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:722036 [4] NCCL INFO Channel 20/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:722037 [6] NCCL INFO Channel 15/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:722038 [5] NCCL INFO Channel 19/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:722035 [2] NCCL INFO Channel 19/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:722039 [1] NCCL INFO Channel 19/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:722041 [7] NCCL INFO Channel 17/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720428:722034 [0] NCCL INFO Channel 19/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:722040 [3] NCCL INFO Channel 20/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:722036 [4] NCCL INFO Channel 21/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:722037 [6] NCCL INFO Channel 16/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:722038 [5] NCCL INFO Channel 20/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:722035 [2] NCCL INFO Channel 20/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:722039 [1] NCCL INFO Channel 20/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:722041 [7] NCCL INFO Channel 18/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720428:722034 [0] NCCL INFO Channel 20/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:722040 [3] NCCL INFO Channel 21/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:722036 [4] NCCL INFO Channel 22/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:722037 [6] NCCL INFO Channel 17/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:722038 [5] NCCL INFO Channel 21/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:722035 [2] NCCL INFO Channel 21/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:722039 [1] NCCL INFO Channel 21/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:722041 [7] NCCL INFO Channel 19/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720428:722034 [0] NCCL INFO Channel 21/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:722040 [3] NCCL INFO Channel 22/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:722036 [4] NCCL INFO Channel 23/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:722038 [5] NCCL INFO Channel 22/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:722037 [6] NCCL INFO Channel 18/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:722035 [2] NCCL INFO Channel 22/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:722039 [1] NCCL INFO Channel 22/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720428:722034 [0] NCCL INFO Channel 22/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:722040 [3] NCCL INFO Channel 23/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:722041 [7] NCCL INFO Channel 20/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:722038 [5] NCCL INFO Channel 23/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:722037 [6] NCCL INFO Channel 19/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:722035 [2] NCCL INFO Channel 23/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:722039 [1] NCCL INFO Channel 23/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720428:722034 [0] NCCL INFO Channel 23/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:722041 [7] NCCL INFO Channel 21/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:722037 [6] NCCL INFO Channel 20/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:722041 [7] NCCL INFO Channel 22/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:722037 [6] NCCL INFO Channel 21/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:722041 [7] NCCL INFO Channel 23/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:722037 [6] NCCL INFO Channel 22/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:722037 [6] NCCL INFO Channel 23/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:722035 [2] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:720431:722040 [3] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:720429:722039 [1] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:720428:722034 [0] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:720435:722041 [7] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:720435:722041 [7] NCCL INFO Channel 00/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:722041 [7] NCCL INFO Channel 01/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:722041 [7] NCCL INFO Channel 02/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:722041 [7] NCCL INFO Channel 03/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:722041 [7] NCCL INFO Channel 04/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:722041 [7] NCCL INFO Channel 05/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:722037 [6] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:720432:722036 [4] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:720433:722038 [5] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:720435:722041 [7] NCCL INFO Channel 06/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:722041 [7] NCCL INFO Channel 07/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:722041 [7] NCCL INFO Channel 08/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:722035 [2] NCCL INFO Channel 00/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:722041 [7] NCCL INFO Channel 09/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:722041 [7] NCCL INFO Channel 10/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:722035 [2] NCCL INFO Channel 01/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:722041 [7] NCCL INFO Channel 11/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:722035 [2] NCCL INFO Channel 02/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:722041 [7] NCCL INFO Channel 12/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:722035 [2] NCCL INFO Channel 03/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:722041 [7] NCCL INFO Channel 13/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:722035 [2] NCCL INFO Channel 04/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:722041 [7] NCCL INFO Channel 14/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:722040 [3] NCCL INFO Channel 00/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:722035 [2] NCCL INFO Channel 05/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:722041 [7] NCCL INFO Channel 15/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:722040 [3] NCCL INFO Channel 01/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:722035 [2] NCCL INFO Channel 06/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:722041 [7] NCCL INFO Channel 16/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:722040 [3] NCCL INFO Channel 02/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:722035 [2] NCCL INFO Channel 07/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:722041 [7] NCCL INFO Channel 17/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:722040 [3] NCCL INFO Channel 03/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:722035 [2] NCCL INFO Channel 08/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:722041 [7] NCCL INFO Channel 18/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:722040 [3] NCCL INFO Channel 04/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:722035 [2] NCCL INFO Channel 09/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:722040 [3] NCCL INFO Channel 05/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:722035 [2] NCCL INFO Channel 10/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:722040 [3] NCCL INFO Channel 06/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:722035 [2] NCCL INFO Channel 11/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:722040 [3] NCCL INFO Channel 07/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:722035 [2] NCCL INFO Channel 12/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:722040 [3] NCCL INFO Channel 08/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:722035 [2] NCCL INFO Channel 13/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:722040 [3] NCCL INFO Channel 09/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:722039 [1] NCCL INFO Channel 00/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:722040 [3] NCCL INFO Channel 10/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:722035 [2] NCCL INFO Channel 14/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:722039 [1] NCCL INFO Channel 01/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:722040 [3] NCCL INFO Channel 11/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:722035 [2] NCCL INFO Channel 15/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:722039 [1] NCCL INFO Channel 02/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:722040 [3] NCCL INFO Channel 12/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:722035 [2] NCCL INFO Channel 16/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:722039 [1] NCCL INFO Channel 03/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:722040 [3] NCCL INFO Channel 13/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:722035 [2] NCCL INFO Channel 17/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:722039 [1] NCCL INFO Channel 04/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:722037 [6] NCCL INFO Channel 00/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:722040 [3] NCCL INFO Channel 14/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:722035 [2] NCCL INFO Channel 18/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:722039 [1] NCCL INFO Channel 05/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:722037 [6] NCCL INFO Channel 01/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:722040 [3] NCCL INFO Channel 15/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:722039 [1] NCCL INFO Channel 06/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:722035 [2] NCCL INFO Channel 19/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:722037 [6] NCCL INFO Channel 02/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:722040 [3] NCCL INFO Channel 16/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:722041 [7] NCCL INFO Channel 19/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:722038 [5] NCCL INFO Channel 00/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:722039 [1] NCCL INFO Channel 07/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:722035 [2] NCCL INFO Channel 20/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:722037 [6] NCCL INFO Channel 03/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:722040 [3] NCCL INFO Channel 17/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:722041 [7] NCCL INFO Channel 20/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:722038 [5] NCCL INFO Channel 01/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:722039 [1] NCCL INFO Channel 08/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:722035 [2] NCCL INFO Channel 21/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:722040 [3] NCCL INFO Channel 18/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:722041 [7] NCCL INFO Channel 21/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:722038 [5] NCCL INFO Channel 02/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:722039 [1] NCCL INFO Channel 09/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:722035 [2] NCCL INFO Channel 22/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:722040 [3] NCCL INFO Channel 19/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:722041 [7] NCCL INFO Channel 22/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:722038 [5] NCCL INFO Channel 03/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:722039 [1] NCCL INFO Channel 10/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:722035 [2] NCCL INFO Channel 23/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:722040 [3] NCCL INFO Channel 20/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:722041 [7] NCCL INFO Channel 23/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:722038 [5] NCCL INFO Channel 04/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:722039 [1] NCCL INFO Channel 11/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:722040 [3] NCCL INFO Channel 21/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:722038 [5] NCCL INFO Channel 05/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:722039 [1] NCCL INFO Channel 12/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:722040 [3] NCCL INFO Channel 22/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:722037 [6] NCCL INFO Channel 04/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:722038 [5] NCCL INFO Channel 06/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:722039 [1] NCCL INFO Channel 13/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:722040 [3] NCCL INFO Channel 23/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:722037 [6] NCCL INFO Channel 05/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:722038 [5] NCCL INFO Channel 07/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:722039 [1] NCCL INFO Channel 14/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:722037 [6] NCCL INFO Channel 06/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:722038 [5] NCCL INFO Channel 08/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:722039 [1] NCCL INFO Channel 15/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:722037 [6] NCCL INFO Channel 07/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:722038 [5] NCCL INFO Channel 09/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:722039 [1] NCCL INFO Channel 16/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:722037 [6] NCCL INFO Channel 08/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:722038 [5] NCCL INFO Channel 10/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:722039 [1] NCCL INFO Channel 17/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:722037 [6] NCCL INFO Channel 09/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:722038 [5] NCCL INFO Channel 11/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:722039 [1] NCCL INFO Channel 18/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:722036 [4] NCCL INFO Channel 00/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:722037 [6] NCCL INFO Channel 10/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:722038 [5] NCCL INFO Channel 12/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:722039 [1] NCCL INFO Channel 19/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:722036 [4] NCCL INFO Channel 01/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:722037 [6] NCCL INFO Channel 11/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:722038 [5] NCCL INFO Channel 13/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:722039 [1] NCCL INFO Channel 20/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:722036 [4] NCCL INFO Channel 02/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:722037 [6] NCCL INFO Channel 12/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:722038 [5] NCCL INFO Channel 14/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:722039 [1] NCCL INFO Channel 21/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:722036 [4] NCCL INFO Channel 03/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:722037 [6] NCCL INFO Channel 13/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:722038 [5] NCCL INFO Channel 15/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:722039 [1] NCCL INFO Channel 22/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:722036 [4] NCCL INFO Channel 04/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:722037 [6] NCCL INFO Channel 14/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:722038 [5] NCCL INFO Channel 16/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:722039 [1] NCCL INFO Channel 23/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:722036 [4] NCCL INFO Channel 05/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:722038 [5] NCCL INFO Channel 17/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:722037 [6] NCCL INFO Channel 15/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:722036 [4] NCCL INFO Channel 06/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:722038 [5] NCCL INFO Channel 18/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:722037 [6] NCCL INFO Channel 16/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:722036 [4] NCCL INFO Channel 07/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:722038 [5] NCCL INFO Channel 19/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:722037 [6] NCCL INFO Channel 17/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:722036 [4] NCCL INFO Channel 08/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:722038 [5] NCCL INFO Channel 20/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:722037 [6] NCCL INFO Channel 18/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:722038 [5] NCCL INFO Channel 21/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:722037 [6] NCCL INFO Channel 19/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:722036 [4] NCCL INFO Channel 09/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:722038 [5] NCCL INFO Channel 22/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:722037 [6] NCCL INFO Channel 20/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:722036 [4] NCCL INFO Channel 10/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:722038 [5] NCCL INFO Channel 23/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:722037 [6] NCCL INFO Channel 21/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:722036 [4] NCCL INFO Channel 11/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:722036 [4] NCCL INFO Channel 12/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:722036 [4] NCCL INFO Channel 13/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:722037 [6] NCCL INFO Channel 22/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:722036 [4] NCCL INFO Channel 14/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:722037 [6] NCCL INFO Channel 23/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:722036 [4] NCCL INFO Channel 15/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:722036 [4] NCCL INFO Channel 16/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:722036 [4] NCCL INFO Channel 17/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:722036 [4] NCCL INFO Channel 18/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:722036 [4] NCCL INFO Channel 19/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:722036 [4] NCCL INFO Channel 20/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:722036 [4] NCCL INFO Channel 21/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:722036 [4] NCCL INFO Channel 22/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:722036 [4] NCCL INFO Channel 23/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720428:722034 [0] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:720428:722034 [0] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:720428:722034 [0] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:720429:722039 [1] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:720429:722039 [1] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:720429:722039 [1] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:720430:722035 [2] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:720430:722035 [2] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:720430:722035 [2] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:720435:722041 [7] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:720434:722037 [6] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:720435:722041 [7] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:720435:722041 [7] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:720434:722037 [6] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:720434:722037 [6] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:720431:722040 [3] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:720431:722040 [3] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:720431:722040 [3] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:720433:722038 [5] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:720433:722038 [5] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:720433:722038 [5] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:720432:722036 [4] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:720432:722036 [4] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:720432:722036 [4] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:720432:722036 [4] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test2-worker-0:720433:722038 [5] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test2-worker-0:720434:722037 [6] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test2-worker-0:720432:722036 [4] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test2-worker-0:720433:722038 [5] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test2-worker-0:720434:722037 [6] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test2-worker-0:720433:722038 [5] NCCL INFO ncclCommInitRank comm 0x55c945a6d9b0 rank 5 nranks 8 cudaDev 5 nvmlDev 5 busId 901d0 commId 0x149f63fecb3e4a5d - Init COMPLETE +ywang29-vrdb-test2-worker-0:720432:722036 [4] NCCL INFO ncclCommInitRank comm 0x556b8d5713b0 rank 4 nranks 8 cudaDev 4 nvmlDev 4 busId 901c0 commId 0x149f63fecb3e4a5d - Init COMPLETE +ywang29-vrdb-test2-worker-0:720434:722037 [6] NCCL INFO ncclCommInitRank comm 0x5569a6daac30 rank 6 nranks 8 cudaDev 6 nvmlDev 6 busId a01c0 commId 0x149f63fecb3e4a5d - Init COMPLETE +ywang29-vrdb-test2-worker-0:720435:722041 [7] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test2-worker-0:720429:722039 [1] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test2-worker-0:720435:722041 [7] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test2-worker-0:720429:722039 [1] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test2-worker-0:720428:722034 [0] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test2-worker-0:720431:722040 [3] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test2-worker-0:720435:722041 [7] NCCL INFO ncclCommInitRank comm 0x55ba9f4bb640 rank 7 nranks 8 cudaDev 7 nvmlDev 7 busId a01d0 commId 0x149f63fecb3e4a5d - Init COMPLETE +ywang29-vrdb-test2-worker-0:720429:722039 [1] NCCL INFO ncclCommInitRank comm 0x560824237a50 rank 1 nranks 8 cudaDev 1 nvmlDev 1 busId 101d0 commId 0x149f63fecb3e4a5d - Init COMPLETE +ywang29-vrdb-test2-worker-0:720428:722034 [0] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test2-worker-0:720431:722040 [3] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test2-worker-0:720431:722040 [3] NCCL INFO ncclCommInitRank comm 0x56492cd58b90 rank 3 nranks 8 cudaDev 3 nvmlDev 3 busId 201d0 commId 0x149f63fecb3e4a5d - Init COMPLETE +ywang29-vrdb-test2-worker-0:720428:722034 [0] NCCL INFO ncclCommInitRank comm 0x55d7da9e9650 rank 0 nranks 8 cudaDev 0 nvmlDev 0 busId 101c0 commId 0x149f63fecb3e4a5d - Init COMPLETE +ywang29-vrdb-test2-worker-0:720430:722035 [2] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test2-worker-0:720430:722035 [2] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test2-worker-0:720430:722035 [2] NCCL INFO ncclCommInitRank comm 0x56110a1eff40 rank 2 nranks 8 cudaDev 2 nvmlDev 2 busId 201c0 commId 0x149f63fecb3e4a5d - Init COMPLETE +[2025-10-12 11:33:22,814] [INFO] [partition_parameters.py:343:__exit__] finished initializing model - num_params = 459, num_elems = 0.99B +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.laSome weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +loading language model from loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model + /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +yers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. + /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +[2025-10-12 11:34:30,198] [INFO] [partition_parameters.py:343:__exit__] finished initializing model - num_params = 907, num_elems = 1.42B +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +TinyLlavaForConditionalGeneration( + (language_model): Qwen2ForCausalLM( + (model): Qwen2Model( + (embed_tokens): Embedding(151936, 896) + (layers): ModuleList( + (0-23): 24 x Qwen2DecoderLayer( + (self_attn): Qwen2FlashAttention2( + (q_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=True) + (k_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=128, bias=True) + (v_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=128, bias=True) + (o_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=False) + (rotary_emb): Qwen2RotaryEmbedding() + ) + (mlp): Qwen2MLP( + (gate_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=4864, bias=False) + (up_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=4864, bias=False) + (down_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=4864, out_features=896, bias=False) + (act_fn): SiLU() + ) + (input_layernorm): Qwen2RMSNorm() + (post_attention_layernorm): Qwen2RMSNorm() + ) + ) + (norm): Qwen2RMSNorm() + ) + (lm_head): Linear(in_features=896, out_features=151936, bias=False) + ) + (vision_tower): SIGLIPVisionTower( + (_vision_tower): SiglipVisionModel( + (vision_model): SiglipVisionTransformer( + (embeddings): SiglipVisionEmbeddings( + (patch_embedding): Conv2d(3, 1152, kernel_size=(14, 14), stride=(14, 14), padding=valid) + (position_embedding): Embedding(729, 1152) + ) + (encoder): SiglipEncoder( + (layers): ModuleList( + (0-26): 27 x SiglipEncoderLayer( + (self_attn): SiglipAttention( + (k_proj): Linear(in_features=1152, out_features=1152, bias=True) + (v_proj): Linear(in_features=1152, out_features=1152, bias=True) + (q_proj): Linear(in_features=1152, out_features=1152, bias=True) + (out_proj): Linear(in_features=1152, out_features=1152, bias=True) + ) + (layer_norm1): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + (layer_norm2): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + ) + ) + ) + (post_layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (head): SiglipMultiheadAttentionPoolingHead( + (attention): MultiheadAttention( + (out_proj): NonDynamicallyQuantizableLinear(in_features=1152, out_features=1152, bias=True) + ) + (layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + ) + ) + ) + ) + (connector): MLPConnector( + (_connector): Sequential( + (0): SupermaskLinearSparsity_SoftForward_Normal(in_features=1152, out_features=896, bias=True) + (1): GELU(approximate='none') + (2): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=True) + ) + ) +) +Pre-training init language_model.model.layers.0.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.0.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.0.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.0.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.0.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.0.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.0.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.mlp.down_proj.scores: Mean=3.000000 +Pre-training init connector._connector.0.scores: Mean=3.000005 +Pre-training init connector._connector.2.scores: Mean=2.999970 +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +2025-10-12 11:34:47,641 | INFO: Total Parameters: 1283756736, Total Trainable Parameters: 359661568 +2025-10-12 11:34:47,646 | INFO: Trainable Parameters: +language_model.model.layers.0.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.0.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.0.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.0.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.0.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.0.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.0.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.1.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.1.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.1.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.1.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.1.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.1.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.1.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.2.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.2.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.2.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.2.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.2.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.2.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.2.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.3.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.3.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.3.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.3.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.3.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.3.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.3.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.4.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.4.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.4.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.4.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.4.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.4.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.4.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.5.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.5.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.5.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.5.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.5.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.5.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.5.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.6.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.6.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.6.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.6.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.6.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.6.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.6.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.7.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.7.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.7.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.7.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.7.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.7.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.7.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.8.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.8.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.8.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.8.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.8.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.8.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.8.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.9.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.9.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.9.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.9.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.9.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.9.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.9.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.10.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.10.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.10.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.10.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.10.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.10.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.10.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.11.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.11.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.11.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.11.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.11.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.11.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.11.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.12.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.12.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.12.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.12.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.12.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.12.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.12.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.13.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.13.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.13.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.13.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.13.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.13.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.13.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.14.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.14.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.14.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.14.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.14.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.14.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.14.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.15.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.15.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.15.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.15.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.15.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.15.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.15.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.16.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.16.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.16.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.16.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.16.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.16.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.16.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.17.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.17.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.17.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.17.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.17.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.17.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.17.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.18.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.18.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.18.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.18.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.18.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.18.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.18.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.19.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.19.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.19.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.19.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.19.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.19.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.19.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.20.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.20.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.20.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.20.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.20.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.20.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.20.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.21.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.21.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.21.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.21.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.21.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.21.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.21.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.22.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.22.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.22.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.22.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.22.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.22.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.22.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.23.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.23.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.23.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.23.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.23.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.23.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.23.mlp.down_proj.scores: 4358144 parameters +connector._connector.0.scores: 1032192 parameters +connector._connector.2.scores: 802816 parameters +Parameter Offload: Total persistent parameters: 486464 in 403 params + 0%| | 0/520 [00:001->0 [1] 2/-1/-1->1->0 [2] 2/-1/-1->1->0 [3] 2/-1/-1->1->0 [4] 2/-1/-1->1->0 [5] 2/-1/-1->1->0 [6] 2/-1/-1->1->0 [7] 2/-1/-1->1->0 [8] 2/-1/-1->1->0 [9] 2/-1/-1->1->0 [10] 2/-1/-1->1->0 [11] 2/-1/-1->1->0 [12] 2/-1/-1->1->0 [13] 2/-1/-1->1->0 [14] 2/-1/-1->1->0 [15] 2/-1/-1->1->0 [16] 2/-1/-1->1->0 [17] 2/-1/-1->1->0 [18] 2/-1/-1->1->0 [19] 2/-1/-1->1->0 [20] 2/-1/-1->1->0 [21] 2/-1/-1->1->0 [22] 2/-1/-1->1->0 [23] 2/-1/-1->1->0 +ywang29-vrdb-test2-worker-0:720428:727092 [0] NCCL INFO Channel 07/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:720429:727095 [1] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:720428:727092 [0] NCCL INFO Channel 08/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:720433:727096 [5] NCCL INFO Trees [0] 6/-1/-1->5->4 [1] 6/-1/-1->5->4 [2] 6/-1/-1->5->4 [3] 6/-1/-1->5->4 [4] 6/-1/-1->5->4 [5] 6/-1/-1->5->4 [6] 6/-1/-1->5->4 [7] 6/-1/-1->5->4 [8] 6/-1/-1->5->4 [9] 6/-1/-1->5->4 [10] 6/-1/-1->5->4 [11] 6/-1/-1->5->4 [12] 6/-1/-1->5->4 [13] 6/-1/-1->5->4 [14] 6/-1/-1->5->4 [15] 6/-1/-1->5->4 [16] 6/-1/-1->5->4 [17] 6/-1/-1->5->4 [18] 6/-1/-1->5->4 [19] 6/-1/-1->5->4 [20] 6/-1/-1->5->4 [21] 6/-1/-1->5->4 [22] 6/-1/-1->5->4 [23] 6/-1/-1->5->4 +ywang29-vrdb-test2-worker-0:720428:727092 [0] NCCL INFO Channel 09/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:720433:727096 [5] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:720428:727092 [0] NCCL INFO Channel 10/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:720430:727094 [2] NCCL INFO Trees [0] 3/-1/-1->2->1 [1] 3/-1/-1->2->1 [2] 3/-1/-1->2->1 [3] 3/-1/-1->2->1 [4] 3/-1/-1->2->1 [5] 3/-1/-1->2->1 [6] 3/-1/-1->2->1 [7] 3/-1/-1->2->1 [8] 3/-1/-1->2->1 [9] 3/-1/-1->2->1 [10] 3/-1/-1->2->1 [11] 3/-1/-1->2->1 [12] 3/-1/-1->2->1 [13] 3/-1/-1->2->1 [14] 3/-1/-1->2->1 [15] 3/-1/-1->2->1 [16] 3/-1/-1->2->1 [17] 3/-1/-1->2->1 [18] 3/-1/-1->2->1 [19] 3/-1/-1->2->1 [20] 3/-1/-1->2->1 [21] 3/-1/-1->2->1 [22] 3/-1/-1->2->1 [23] 3/-1/-1->2->1 +ywang29-vrdb-test2-worker-0:720435:727098 [7] NCCL INFO Trees [0] -1/-1/-1->7->6 [1] -1/-1/-1->7->6 [2] -1/-1/-1->7->6 [3] -1/-1/-1->7->6 [4] -1/-1/-1->7->6 [5] -1/-1/-1->7->6 [6] -1/-1/-1->7->6 [7] -1/-1/-1->7->6 [8] -1/-1/-1->7->6 [9] -1/-1/-1->7->6 [10] -1/-1/-1->7->6 [11] -1/-1/-1->7->6 [12] -1/-1/-1->7->6 [13] -1/-1/-1->7->6 [14] -1/-1/-1->7->6 [15] -1/-1/-1->7->6 [16] -1/-1/-1->7->6 [17] -1/-1/-1->7->6 [18] -1/-1/-1->7->6 [19] -1/-1/-1->7->6 [20] -1/-1/-1->7->6 [21] -1/-1/-1->7->6 [22] -1/-1/-1->7->6 [23] -1/-1/-1->7->6 +ywang29-vrdb-test2-worker-0:720434:727097 [6] NCCL INFO Trees [0] 7/-1/-1->6->5 [1] 7/-1/-1->6->5 [2] 7/-1/-1->6->5 [3] 7/-1/-1->6->5 [4] 7/-1/-1->6->5 [5] 7/-1/-1->6->5 [6] 7/-1/-1->6->5 [7] 7/-1/-1->6->5 [8] 7/-1/-1->6->5 [9] 7/-1/-1->6->5 [10] 7/-1/-1->6->5 [11] 7/-1/-1->6->5 [12] 7/-1/-1->6->5 [13] 7/-1/-1->6->5 [14] 7/-1/-1->6->5 [15] 7/-1/-1->6->5 [16] 7/-1/-1->6->5 [17] 7/-1/-1->6->5 [18] 7/-1/-1->6->5 [19] 7/-1/-1->6->5 [20] 7/-1/-1->6->5 [21] 7/-1/-1->6->5 [22] 7/-1/-1->6->5 [23] 7/-1/-1->6->5 +ywang29-vrdb-test2-worker-0:720428:727092 [0] NCCL INFO Channel 11/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:720430:727094 [2] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:720431:727099 [3] NCCL INFO Trees [0] 4/-1/-1->3->2 [1] 4/-1/-1->3->2 [2] 4/-1/-1->3->2 [3] 4/-1/-1->3->2 [4] 4/-1/-1->3->2 [5] 4/-1/-1->3->2 [6] 4/-1/-1->3->2 [7] 4/-1/-1->3->2 [8] 4/-1/-1->3->2 [9] 4/-1/-1->3->2 [10] 4/-1/-1->3->2 [11] 4/-1/-1->3->2 [12] 4/-1/-1->3->2 [13] 4/-1/-1->3->2 [14] 4/-1/-1->3->2 [15] 4/-1/-1->3->2 [16] 4/-1/-1->3->2 [17] 4/-1/-1->3->2 [18] 4/-1/-1->3->2 [19] 4/-1/-1->3->2 [20] 4/-1/-1->3->2 [21] 4/-1/-1->3->2 [22] 4/-1/-1->3->2 [23] 4/-1/-1->3->2 +ywang29-vrdb-test2-worker-0:720432:727093 [4] NCCL INFO Trees [0] 5/-1/-1->4->3 [1] 5/-1/-1->4->3 [2] 5/-1/-1->4->3 [3] 5/-1/-1->4->3 [4] 5/-1/-1->4->3 [5] 5/-1/-1->4->3 [6] 5/-1/-1->4->3 [7] 5/-1/-1->4->3 [8] 5/-1/-1->4->3 [9] 5/-1/-1->4->3 [10] 5/-1/-1->4->3 [11] 5/-1/-1->4->3 [12] 5/-1/-1->4->3 [13] 5/-1/-1->4->3 [14] 5/-1/-1->4->3 [15] 5/-1/-1->4->3 [16] 5/-1/-1->4->3 [17] 5/-1/-1->4->3 [18] 5/-1/-1->4->3 [19] 5/-1/-1->4->3 [20] 5/-1/-1->4->3 [21] 5/-1/-1->4->3 [22] 5/-1/-1->4->3 [23] 5/-1/-1->4->3 +ywang29-vrdb-test2-worker-0:720435:727098 [7] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:720434:727097 [6] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:720428:727092 [0] NCCL INFO Channel 12/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:720431:727099 [3] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:720432:727093 [4] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:720428:727092 [0] NCCL INFO Channel 13/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:720428:727092 [0] NCCL INFO Channel 14/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:720428:727092 [0] NCCL INFO Channel 15/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:720428:727092 [0] NCCL INFO Channel 16/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:720428:727092 [0] NCCL INFO Channel 17/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:720428:727092 [0] NCCL INFO Channel 18/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:720428:727092 [0] NCCL INFO Channel 19/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:720428:727092 [0] NCCL INFO Channel 20/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:720428:727092 [0] NCCL INFO Channel 21/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:720428:727092 [0] NCCL INFO Channel 22/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:720428:727092 [0] NCCL INFO Channel 23/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:720428:727092 [0] NCCL INFO Trees [0] 1/-1/-1->0->-1 [1] 1/-1/-1->0->-1 [2] 1/-1/-1->0->-1 [3] 1/-1/-1->0->-1 [4] 1/-1/-1->0->-1 [5] 1/-1/-1->0->-1 [6] 1/-1/-1->0->-1 [7] 1/-1/-1->0->-1 [8] 1/-1/-1->0->-1 [9] 1/-1/-1->0->-1 [10] 1/-1/-1->0->-1 [11] 1/-1/-1->0->-1 [12] 1/-1/-1->0->-1 [13] 1/-1/-1->0->-1 [14] 1/-1/-1->0->-1 [15] 1/-1/-1->0->-1 [16] 1/-1/-1->0->-1 [17] 1/-1/-1->0->-1 [18] 1/-1/-1->0->-1 [19] 1/-1/-1->0->-1 [20] 1/-1/-1->0->-1 [21] 1/-1/-1->0->-1 [22] 1/-1/-1->0->-1 [23] 1/-1/-1->0->-1 +ywang29-vrdb-test2-worker-0:720428:727092 [0] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:720431:727099 [3] NCCL INFO Channel 00/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:727096 [5] NCCL INFO Channel 00/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:727093 [4] NCCL INFO Channel 00/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:727095 [1] NCCL INFO Channel 00/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720428:727092 [0] NCCL INFO Channel 00/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:727099 [3] NCCL INFO Channel 01/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:727098 [7] NCCL INFO Channel 00/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:727096 [5] NCCL INFO Channel 01/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:727093 [4] NCCL INFO Channel 01/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:727095 [1] NCCL INFO Channel 01/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720428:727092 [0] NCCL INFO Channel 01/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:727099 [3] NCCL INFO Channel 02/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:727098 [7] NCCL INFO Channel 01/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:727096 [5] NCCL INFO Channel 02/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:727093 [4] NCCL INFO Channel 02/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720428:727092 [0] NCCL INFO Channel 02/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:727095 [1] NCCL INFO Channel 02/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:727099 [3] NCCL INFO Channel 03/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:727098 [7] NCCL INFO Channel 02/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:727096 [5] NCCL INFO Channel 03/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:727093 [4] NCCL INFO Channel 03/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720428:727092 [0] NCCL INFO Channel 03/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:727095 [1] NCCL INFO Channel 03/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:727099 [3] NCCL INFO Channel 04/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:727098 [7] NCCL INFO Channel 03/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:727096 [5] NCCL INFO Channel 04/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:727093 [4] NCCL INFO Channel 04/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720428:727092 [0] NCCL INFO Channel 04/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:727095 [1] NCCL INFO Channel 04/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:727099 [3] NCCL INFO Channel 05/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:727098 [7] NCCL INFO Channel 04/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:727096 [5] NCCL INFO Channel 05/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:727093 [4] NCCL INFO Channel 05/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720428:727092 [0] NCCL INFO Channel 05/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:727095 [1] NCCL INFO Channel 05/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:727099 [3] NCCL INFO Channel 06/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:727098 [7] NCCL INFO Channel 05/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:727096 [5] NCCL INFO Channel 06/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:727093 [4] NCCL INFO Channel 06/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720428:727092 [0] NCCL INFO Channel 06/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:727095 [1] NCCL INFO Channel 06/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:727099 [3] NCCL INFO Channel 07/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:727098 [7] NCCL INFO Channel 06/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:727096 [5] NCCL INFO Channel 07/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:727093 [4] NCCL INFO Channel 07/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720428:727092 [0] NCCL INFO Channel 07/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:727095 [1] NCCL INFO Channel 07/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:727099 [3] NCCL INFO Channel 08/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:727098 [7] NCCL INFO Channel 07/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:727096 [5] NCCL INFO Channel 08/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:727093 [4] NCCL INFO Channel 08/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720428:727092 [0] NCCL INFO Channel 08/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:727095 [1] NCCL INFO Channel 08/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:727099 [3] NCCL INFO Channel 09/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:727098 [7] NCCL INFO Channel 08/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:727096 [5] NCCL INFO Channel 09/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:727093 [4] NCCL INFO Channel 09/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720428:727092 [0] NCCL INFO Channel 09/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:727095 [1] NCCL INFO Channel 09/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:727099 [3] NCCL INFO Channel 10/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:727098 [7] NCCL INFO Channel 09/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:727096 [5] NCCL INFO Channel 10/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:727093 [4] NCCL INFO Channel 10/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720428:727092 [0] NCCL INFO Channel 10/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:727095 [1] NCCL INFO Channel 10/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:727099 [3] NCCL INFO Channel 11/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:727098 [7] NCCL INFO Channel 10/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:727096 [5] NCCL INFO Channel 11/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:727093 [4] NCCL INFO Channel 11/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720428:727092 [0] NCCL INFO Channel 11/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:727095 [1] NCCL INFO Channel 11/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:727099 [3] NCCL INFO Channel 12/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:727098 [7] NCCL INFO Channel 11/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:727096 [5] NCCL INFO Channel 12/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:727093 [4] NCCL INFO Channel 12/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720428:727092 [0] NCCL INFO Channel 12/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:727095 [1] NCCL INFO Channel 12/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:727099 [3] NCCL INFO Channel 13/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:727098 [7] NCCL INFO Channel 12/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:727096 [5] NCCL INFO Channel 13/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:727093 [4] NCCL INFO Channel 13/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:727094 [2] NCCL INFO Channel 00/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720428:727092 [0] NCCL INFO Channel 13/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:727095 [1] NCCL INFO Channel 13/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:727097 [6] NCCL INFO Channel 00/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:727099 [3] NCCL INFO Channel 14/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:727098 [7] NCCL INFO Channel 13/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:727096 [5] NCCL INFO Channel 14/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:727093 [4] NCCL INFO Channel 14/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:727094 [2] NCCL INFO Channel 01/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720428:727092 [0] NCCL INFO Channel 14/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:727095 [1] NCCL INFO Channel 14/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:727099 [3] NCCL INFO Channel 15/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:727097 [6] NCCL INFO Channel 01/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:727098 [7] NCCL INFO Channel 14/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:727096 [5] NCCL INFO Channel 15/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:727093 [4] NCCL INFO Channel 15/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:727094 [2] NCCL INFO Channel 02/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720428:727092 [0] NCCL INFO Channel 15/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:727095 [1] NCCL INFO Channel 15/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:727099 [3] NCCL INFO Channel 16/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:727097 [6] NCCL INFO Channel 02/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:727098 [7] NCCL INFO Channel 15/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:727096 [5] NCCL INFO Channel 16/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:727093 [4] NCCL INFO Channel 16/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:727094 [2] NCCL INFO Channel 03/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720428:727092 [0] NCCL INFO Channel 16/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:727095 [1] NCCL INFO Channel 16/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:727099 [3] NCCL INFO Channel 17/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:727097 [6] NCCL INFO Channel 03/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:727098 [7] NCCL INFO Channel 16/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:727096 [5] NCCL INFO Channel 17/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:727093 [4] NCCL INFO Channel 17/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:727094 [2] NCCL INFO Channel 04/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720428:727092 [0] NCCL INFO Channel 17/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:727095 [1] NCCL INFO Channel 17/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:727099 [3] NCCL INFO Channel 18/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:727097 [6] NCCL INFO Channel 04/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:727098 [7] NCCL INFO Channel 17/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:727096 [5] NCCL INFO Channel 18/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:727093 [4] NCCL INFO Channel 18/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:727094 [2] NCCL INFO Channel 05/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720428:727092 [0] NCCL INFO Channel 18/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:727095 [1] NCCL INFO Channel 18/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:727099 [3] NCCL INFO Channel 19/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:727097 [6] NCCL INFO Channel 05/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:727098 [7] NCCL INFO Channel 18/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:727096 [5] NCCL INFO Channel 19/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:727093 [4] NCCL INFO Channel 19/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720428:727092 [0] NCCL INFO Channel 19/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:727094 [2] NCCL INFO Channel 06/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:727095 [1] NCCL INFO Channel 19/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:727098 [7] NCCL INFO Channel 19/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:727099 [3] NCCL INFO Channel 20/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:727097 [6] NCCL INFO Channel 06/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:727096 [5] NCCL INFO Channel 20/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:727093 [4] NCCL INFO Channel 20/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720428:727092 [0] NCCL INFO Channel 20/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:727094 [2] NCCL INFO Channel 07/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:727095 [1] NCCL INFO Channel 20/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:727098 [7] NCCL INFO Channel 20/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:727099 [3] NCCL INFO Channel 21/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:727097 [6] NCCL INFO Channel 07/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:727096 [5] NCCL INFO Channel 21/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:727093 [4] NCCL INFO Channel 21/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720428:727092 [0] NCCL INFO Channel 21/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:727094 [2] NCCL INFO Channel 08/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:727095 [1] NCCL INFO Channel 21/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:727098 [7] NCCL INFO Channel 21/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:727099 [3] NCCL INFO Channel 22/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:727097 [6] NCCL INFO Channel 08/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:727096 [5] NCCL INFO Channel 22/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:727093 [4] NCCL INFO Channel 22/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720428:727092 [0] NCCL INFO Channel 22/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:727094 [2] NCCL INFO Channel 09/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:727095 [1] NCCL INFO Channel 22/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:727098 [7] NCCL INFO Channel 22/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:727099 [3] NCCL INFO Channel 23/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:727097 [6] NCCL INFO Channel 09/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:727096 [5] NCCL INFO Channel 23/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:727093 [4] NCCL INFO Channel 23/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720428:727092 [0] NCCL INFO Channel 23/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:727094 [2] NCCL INFO Channel 10/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:727095 [1] NCCL INFO Channel 23/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:727098 [7] NCCL INFO Channel 23/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:727097 [6] NCCL INFO Channel 10/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:727094 [2] NCCL INFO Channel 11/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:727097 [6] NCCL INFO Channel 11/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:727094 [2] NCCL INFO Channel 12/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:727097 [6] NCCL INFO Channel 12/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:727094 [2] NCCL INFO Channel 13/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:727097 [6] NCCL INFO Channel 13/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:727094 [2] NCCL INFO Channel 14/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:727097 [6] NCCL INFO Channel 14/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:727094 [2] NCCL INFO Channel 15/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:727097 [6] NCCL INFO Channel 15/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:727094 [2] NCCL INFO Channel 16/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:727097 [6] NCCL INFO Channel 16/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:727094 [2] NCCL INFO Channel 17/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:727097 [6] NCCL INFO Channel 17/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:727094 [2] NCCL INFO Channel 18/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:727097 [6] NCCL INFO Channel 18/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:727094 [2] NCCL INFO Channel 19/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:727097 [6] NCCL INFO Channel 19/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:727094 [2] NCCL INFO Channel 20/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:727097 [6] NCCL INFO Channel 20/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:727094 [2] NCCL INFO Channel 21/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:727097 [6] NCCL INFO Channel 21/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:727094 [2] NCCL INFO Channel 22/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:727097 [6] NCCL INFO Channel 22/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:727094 [2] NCCL INFO Channel 23/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:727097 [6] NCCL INFO Channel 23/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:727095 [1] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:720428:727092 [0] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:720431:727099 [3] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:720430:727094 [2] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:720429:727095 [1] NCCL INFO Channel 00/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:727095 [1] NCCL INFO Channel 01/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:727095 [1] NCCL INFO Channel 02/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:727095 [1] NCCL INFO Channel 03/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:727095 [1] NCCL INFO Channel 04/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:727095 [1] NCCL INFO Channel 05/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:727095 [1] NCCL INFO Channel 06/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:727093 [4] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:720429:727095 [1] NCCL INFO Channel 07/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:727095 [1] NCCL INFO Channel 08/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:727095 [1] NCCL INFO Channel 09/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:727095 [1] NCCL INFO Channel 10/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:727099 [3] NCCL INFO Channel 00/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:727095 [1] NCCL INFO Channel 11/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:727095 [1] NCCL INFO Channel 12/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:727099 [3] NCCL INFO Channel 01/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:727099 [3] NCCL INFO Channel 02/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:727095 [1] NCCL INFO Channel 13/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:727098 [7] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:720435:727098 [7] NCCL INFO Channel 00/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:727096 [5] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:720434:727097 [6] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:720431:727099 [3] NCCL INFO Channel 03/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:727095 [1] NCCL INFO Channel 14/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:727098 [7] NCCL INFO Channel 01/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:727094 [2] NCCL INFO Channel 00/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:727099 [3] NCCL INFO Channel 04/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:727095 [1] NCCL INFO Channel 15/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:727094 [2] NCCL INFO Channel 01/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:727098 [7] NCCL INFO Channel 02/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:727099 [3] NCCL INFO Channel 05/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:727095 [1] NCCL INFO Channel 16/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:727094 [2] NCCL INFO Channel 02/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:727098 [7] NCCL INFO Channel 03/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:727099 [3] NCCL INFO Channel 06/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:727095 [1] NCCL INFO Channel 17/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:727094 [2] NCCL INFO Channel 03/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:727098 [7] NCCL INFO Channel 04/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:727099 [3] NCCL INFO Channel 07/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:727095 [1] NCCL INFO Channel 18/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:727094 [2] NCCL INFO Channel 04/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:727098 [7] NCCL INFO Channel 05/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:727099 [3] NCCL INFO Channel 08/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:727095 [1] NCCL INFO Channel 19/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:727094 [2] NCCL INFO Channel 05/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:727099 [3] NCCL INFO Channel 09/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:727095 [1] NCCL INFO Channel 20/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:727094 [2] NCCL INFO Channel 06/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:727099 [3] NCCL INFO Channel 10/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:727095 [1] NCCL INFO Channel 21/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:727094 [2] NCCL INFO Channel 07/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:727099 [3] NCCL INFO Channel 11/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:727095 [1] NCCL INFO Channel 22/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:727094 [2] NCCL INFO Channel 08/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:727099 [3] NCCL INFO Channel 12/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720429:727095 [1] NCCL INFO Channel 23/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:727094 [2] NCCL INFO Channel 09/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:727099 [3] NCCL INFO Channel 13/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:727094 [2] NCCL INFO Channel 10/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:727099 [3] NCCL INFO Channel 14/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:727094 [2] NCCL INFO Channel 11/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:727098 [7] NCCL INFO Channel 06/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:727099 [3] NCCL INFO Channel 15/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:727094 [2] NCCL INFO Channel 12/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:727098 [7] NCCL INFO Channel 07/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:727099 [3] NCCL INFO Channel 16/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:727098 [7] NCCL INFO Channel 08/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:727099 [3] NCCL INFO Channel 17/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:727098 [7] NCCL INFO Channel 09/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:727094 [2] NCCL INFO Channel 13/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:727099 [3] NCCL INFO Channel 18/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:727098 [7] NCCL INFO Channel 10/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:727094 [2] NCCL INFO Channel 14/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:727099 [3] NCCL INFO Channel 19/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:727093 [4] NCCL INFO Channel 00/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:727093 [4] NCCL INFO Channel 01/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:727093 [4] NCCL INFO Channel 02/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:727093 [4] NCCL INFO Channel 03/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:727093 [4] NCCL INFO Channel 04/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:727098 [7] NCCL INFO Channel 11/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:727093 [4] NCCL INFO Channel 05/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:727093 [4] NCCL INFO Channel 06/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:727094 [2] NCCL INFO Channel 15/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:727098 [7] NCCL INFO Channel 12/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:727099 [3] NCCL INFO Channel 20/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:727098 [7] NCCL INFO Channel 13/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:727099 [3] NCCL INFO Channel 21/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:727098 [7] NCCL INFO Channel 14/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:727099 [3] NCCL INFO Channel 22/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:727098 [7] NCCL INFO Channel 15/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720431:727099 [3] NCCL INFO Channel 23/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:727093 [4] NCCL INFO Channel 07/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:727094 [2] NCCL INFO Channel 16/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:727093 [4] NCCL INFO Channel 08/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:727094 [2] NCCL INFO Channel 17/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:727093 [4] NCCL INFO Channel 09/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:727094 [2] NCCL INFO Channel 18/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:727093 [4] NCCL INFO Channel 10/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:727098 [7] NCCL INFO Channel 16/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:727094 [2] NCCL INFO Channel 19/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:727093 [4] NCCL INFO Channel 11/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:727098 [7] NCCL INFO Channel 17/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:727094 [2] NCCL INFO Channel 20/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:727093 [4] NCCL INFO Channel 12/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:727094 [2] NCCL INFO Channel 21/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:727098 [7] NCCL INFO Channel 18/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:727097 [6] NCCL INFO Channel 00/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:727093 [4] NCCL INFO Channel 13/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:727094 [2] NCCL INFO Channel 22/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:727098 [7] NCCL INFO Channel 19/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:727097 [6] NCCL INFO Channel 01/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:727093 [4] NCCL INFO Channel 14/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720430:727094 [2] NCCL INFO Channel 23/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:727097 [6] NCCL INFO Channel 02/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:727098 [7] NCCL INFO Channel 20/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:727096 [5] NCCL INFO Channel 00/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:727097 [6] NCCL INFO Channel 03/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:727098 [7] NCCL INFO Channel 21/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:727096 [5] NCCL INFO Channel 01/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:727098 [7] NCCL INFO Channel 22/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:727097 [6] NCCL INFO Channel 04/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:727096 [5] NCCL INFO Channel 02/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720435:727098 [7] NCCL INFO Channel 23/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:727097 [6] NCCL INFO Channel 05/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:727097 [6] NCCL INFO Channel 06/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:727096 [5] NCCL INFO Channel 03/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:727097 [6] NCCL INFO Channel 07/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:727096 [5] NCCL INFO Channel 04/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:727097 [6] NCCL INFO Channel 08/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:727096 [5] NCCL INFO Channel 05/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:727093 [4] NCCL INFO Channel 15/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:727096 [5] NCCL INFO Channel 06/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:727097 [6] NCCL INFO Channel 09/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:727093 [4] NCCL INFO Channel 16/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:727096 [5] NCCL INFO Channel 07/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:727097 [6] NCCL INFO Channel 10/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:727093 [4] NCCL INFO Channel 17/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:727096 [5] NCCL INFO Channel 08/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:727097 [6] NCCL INFO Channel 11/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:727093 [4] NCCL INFO Channel 18/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:727096 [5] NCCL INFO Channel 09/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:727097 [6] NCCL INFO Channel 12/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:727093 [4] NCCL INFO Channel 19/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:727096 [5] NCCL INFO Channel 10/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:727097 [6] NCCL INFO Channel 13/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:727093 [4] NCCL INFO Channel 20/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:727096 [5] NCCL INFO Channel 11/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:727093 [4] NCCL INFO Channel 21/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:727096 [5] NCCL INFO Channel 12/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:727097 [6] NCCL INFO Channel 14/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:727093 [4] NCCL INFO Channel 22/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:727097 [6] NCCL INFO Channel 15/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:727096 [5] NCCL INFO Channel 13/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720432:727093 [4] NCCL INFO Channel 23/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:727096 [5] NCCL INFO Channel 14/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:727097 [6] NCCL INFO Channel 16/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:727096 [5] NCCL INFO Channel 15/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:727097 [6] NCCL INFO Channel 17/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:727096 [5] NCCL INFO Channel 16/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:727097 [6] NCCL INFO Channel 18/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:727097 [6] NCCL INFO Channel 19/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:727096 [5] NCCL INFO Channel 17/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:727097 [6] NCCL INFO Channel 20/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:727096 [5] NCCL INFO Channel 18/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:727096 [5] NCCL INFO Channel 19/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:727097 [6] NCCL INFO Channel 21/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:727097 [6] NCCL INFO Channel 22/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:727096 [5] NCCL INFO Channel 20/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720434:727097 [6] NCCL INFO Channel 23/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:727096 [5] NCCL INFO Channel 21/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:727096 [5] NCCL INFO Channel 22/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720433:727096 [5] NCCL INFO Channel 23/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:720428:727092 [0] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:720428:727092 [0] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:720428:727092 [0] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:720430:727094 [2] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:720430:727094 [2] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:720430:727094 [2] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:720429:727095 [1] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:720429:727095 [1] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:720429:727095 [1] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:720431:727099 [3] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:720431:727099 [3] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:720431:727099 [3] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:720432:727093 [4] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:720432:727093 [4] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:720432:727093 [4] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:720433:727096 [5] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:720433:727096 [5] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:720433:727096 [5] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:720435:727098 [7] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:720435:727098 [7] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:720434:727097 [6] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:720435:727098 [7] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:720434:727097 [6] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:720434:727097 [6] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:720428:727092 [0] NCCL INFO ncclCommInitRank comm 0x7f8da006aff0 rank 0 nranks 8 cudaDev 0 nvmlDev 0 busId 101c0 commId 0x50c7af48d68024d5 - Init COMPLETE +ywang29-vrdb-test2-worker-0:720430:727094 [2] NCCL INFO ncclCommInitRank comm 0x7f927406abe0 rank 2 nranks 8 cudaDev 2 nvmlDev 2 busId 201c0 commId 0x50c7af48d68024d5 - Init COMPLETE +ywang29-vrdb-test2-worker-0:720432:727093 [4] NCCL INFO ncclCommInitRank comm 0x7fe78c06ae30 rank 4 nranks 8 cudaDev 4 nvmlDev 4 busId 901c0 commId 0x50c7af48d68024d5 - Init COMPLETE +ywang29-vrdb-test2-worker-0:720434:727097 [6] NCCL INFO ncclCommInitRank comm 0x7f234806b080 rank 6 nranks 8 cudaDev 6 nvmlDev 6 busId a01c0 commId 0x50c7af48d68024d5 - Init COMPLETE +ywang29-vrdb-test2-worker-0:720431:727099 [3] NCCL INFO ncclCommInitRank comm 0x7efe9c06a550 rank 3 nranks 8 cudaDev 3 nvmlDev 3 busId 201d0 commId 0x50c7af48d68024d5 - Init COMPLETE +ywang29-vrdb-test2-worker-0:720435:727098 [7] NCCL INFO ncclCommInitRank comm 0x7fa86c069af0 rank 7 nranks 8 cudaDev 7 nvmlDev 7 busId a01d0 commId 0x50c7af48d68024d5 - Init COMPLETE +ywang29-vrdb-test2-worker-0:720433:727096 [5] NCCL INFO ncclCommInitRank comm 0x7f1a5c06a640 rank 5 nranks 8 cudaDev 5 nvmlDev 5 busId 901d0 commId 0x50c7af48d68024d5 - Init COMPLETE +ywang29-vrdb-test2-worker-0:720429:727095 [1] NCCL INFO ncclCommInitRank comm 0x7f73fc06aa60 rank 1 nranks 8 cudaDev 1 nvmlDev 1 busId 101d0 commId 0x50c7af48d68024d5 - Init COMPLETE + 0%| | 1/520 [00:14<2:06:03, 14.57s/it] {'loss': 2.0453, 'grad_norm': 0.004834199985190469, 'learning_rate': 0.003125, 'epoch': 0.0} + 0%| | 1/520 [00:14<2:06:03, 14.57s/it] 0%| | 2/520 [00:18<1:11:15, 8.25s/it] {'loss': 2.0549, 'grad_norm': 0.005248831280669753, 'learning_rate': 0.00625, 'epoch': 0.0} + 0%| | 2/520 [00:18<1:11:15, 8.25s/it] 1%| | 3/520 [00:22<53:45, 6.24s/it] {'loss': 2.1899, 'grad_norm': 0.00600777480657454, 'learning_rate': 0.009375000000000001, 'epoch': 0.01} + 1%| | 3/520 [00:22<53:45, 6.24s/it] 1%| | 4/520 [00:26<45:37, 5.31s/it] {'loss': 2.0656, 'grad_norm': 0.004960502301914123, 'learning_rate': 0.0125, 'epoch': 0.01} + 1%| | 4/520 [00:26<45:37, 5.31s/it] 1%| | 5/520 [00:30<41:15, 4.81s/it] {'loss': 2.2333, 'grad_norm': 0.00548168690979133, 'learning_rate': 0.015625, 'epoch': 0.01} + 1%| | 5/520 [00:30<41:15, 4.81s/it] 1%| | 6/520 [00:33<38:12, 4.46s/it] {'loss': 1.6754, 'grad_norm': 0.0028029244302125505, 'learning_rate': 0.018750000000000003, 'epoch': 0.01} + 1%| | 6/520 [00:33<38:12, 4.46s/it] 1%|▏ | 7/520 [00:37<35:44, 4.18s/it] {'loss': 2.0776, 'grad_norm': 0.005415172484845658, 'learning_rate': 0.021875000000000002, 'epoch': 0.01} + 1%|▏ | 7/520 [00:37<35:44, 4.18s/it] 2%|▏ | 8/520 [00:41<35:56, 4.21s/it] {'loss': 2.0541, 'grad_norm': 0.004572123168919833, 'learning_rate': 0.025, 'epoch': 0.02} + 2%|▏ | 8/520 [00:41<35:56, 4.21s/it] 2%|▏ | 9/520 [00:45<35:49, 4.21s/it] {'loss': 2.1841, 'grad_norm': 0.004984627364807639, 'learning_rate': 0.028125, 'epoch': 0.02} + 2%|▏ | 9/520 [00:45<35:49, 4.21s/it] 2%|▏ | 10/520 [00:49<34:10, 4.02s/it] {'loss': 1.6626, 'grad_norm': 0.0027148621480709598, 'learning_rate': 0.03125, 'epoch': 0.02} + 2%|▏ | 10/520 [00:49<34:10, 4.02s/it] 2%|▏ | 11/520 [00:53<33:29, 3.95s/it] {'loss': 1.5813, 'grad_norm': 0.0010587442573068643, 'learning_rate': 0.034375, 'epoch': 0.02} + 2%|▏ | 11/520 [00:53<33:29, 3.95s/it] 2%|▏ | 12/520 [00:56<32:34, 3.85s/it] {'loss': 1.4307, 'grad_norm': 0.0006603654455268385, 'learning_rate': 0.037500000000000006, 'epoch': 0.02} + 2%|▏ | 12/520 [00:56<32:34, 3.85s/it][2025-10-12 11:35:54,062] [WARNING] [stage3.py:2069:step] 1 pytorch allocator cache flushes since last step. this happens when there is high memory pressure and is detrimental to performance. if this is happening frequently consider adjusting settings to reduce memory consumption. If you are unable to make the cache flushes go away consider adding get_accelerator().empty_cache() calls in your training loop to ensure that all ranks flush their caches at the same time + 2%|▎ | 13/520 [01:01<33:39, 3.98s/it] {'loss': 1.5419, 'grad_norm': 0.0007064934384140015, 'learning_rate': 0.040625, 'epoch': 0.03} + 2%|▎ | 13/520 [01:01<33:39, 3.98s/it] 3%|▎ | 14/520 [01:04<32:40, 3.88s/it] {'loss': 1.5559, 'grad_norm': 0.0005810748511104428, 'learning_rate': 0.043750000000000004, 'epoch': 0.03} + 3%|▎ | 14/520 [01:04<32:40, 3.88s/it] 3%|▎ | 15/520 [01:08<31:56, 3.80s/it] {'loss': 1.4628, 'grad_norm': 0.000457180306810859, 'learning_rate': 0.046875, 'epoch': 0.03} + 3%|▎ | 15/520 [01:08<31:56, 3.80s/it] 3%|▎ | 16/520 [01:12<31:46, 3.78s/it] {'loss': 1.4266, 'grad_norm': 0.0004426308520878069, 'learning_rate': 0.05, 'epoch': 0.03} + 3%|▎ | 16/520 [01:12<31:46, 3.78s/it] 3%|▎ | 17/520 [01:15<31:30, 3.76s/it] {'loss': 1.5834, 'grad_norm': 0.0004716571789022116, 'learning_rate': 0.04999951432328845, 'epoch': 0.03} + 3%|▎ | 17/520 [01:15<31:30, 3.76s/it] 3%|▎ | 18/520 [01:19<31:03, 3.71s/it] {'loss': 1.4478, 'grad_norm': 0.0004439312059682303, 'learning_rate': 0.04999805731202437, 'epoch': 0.03} + 3%|▎ | 18/520 [01:19<31:03, 3.71s/it] 4%|▎ | 19/520 [01:23<30:45, 3.68s/it] {'loss': 1.4241, 'grad_norm': 0.00039525863911811537, 'learning_rate': 0.04999562902281866, 'epoch': 0.04} + 4%|▎ | 19/520 [01:23<30:45, 3.68s/it] 4%|▍ | 20/520 [01:26<30:29, 3.66s/it] {'loss': 1.4121, 'grad_norm': 0.00041497841623067795, 'learning_rate': 0.04999222955002041, 'epoch': 0.04} + 4%|▍ | 20/520 [01:26<30:29, 3.66s/it] 4%|▍ | 21/520 [01:30<30:23, 3.65s/it] {'loss': 1.4115, 'grad_norm': 0.0003959782235709905, 'learning_rate': 0.04998785902571319, 'epoch': 0.04} + 4%|▍ | 21/520 [01:30<30:23, 3.65s/it] 4%|▍ | 22/520 [01:33<30:14, 3.64s/it] {'loss': 1.5339, 'grad_norm': 0.0003880575858693781, 'learning_rate': 0.04998251761970997, 'epoch': 0.04} + 4%|▍ | 22/520 [01:33<30:14, 3.64s/it] 4%|▍ | 23/520 [01:37<30:05, 3.63s/it] {'loss': 1.4677, 'grad_norm': 0.0003533962927283482, 'learning_rate': 0.04997620553954645, 'epoch': 0.04} + 4%|▍ | 23/520 [01:37<30:05, 3.63s/it] 5%|▍ | 24/520 [01:41<30:03, 3.64s/it] {'loss': 1.3799, 'grad_norm': 0.00036435199801384947, 'learning_rate': 0.04996892303047306, 'epoch': 0.05} + 5%|▍ | 24/520 [01:41<30:03, 3.64s/it] 5%|▍ | 25/520 [01:44<30:03, 3.64s/it] {'loss': 1.5007, 'grad_norm': 0.000474033974090886, 'learning_rate': 0.049960670375445417, 'epoch': 0.05} + 5%|▍ | 25/520 [01:44<30:03, 3.64s/it] 5%|▌ | 26/520 [01:48<29:54, 3.63s/it] {'loss': 1.4082, 'grad_norm': 0.0003652030659762915, 'learning_rate': 0.0499514478951133, 'epoch': 0.05} + 5%|▌ | 26/520 [01:48<29:54, 3.63s/it] 5%|▌ | 27/520 [01:52<29:44, 3.62s/it] {'loss': 1.3452, 'grad_norm': 0.0003973678333221638, 'learning_rate': 0.049941255947808225, 'epoch': 0.05} + 5%|▌ | 27/520 [01:52<29:44, 3.62s/it] 5%|▌ | 28/520 [01:55<29:43, 3.62s/it] {'loss': 1.3797, 'grad_norm': 0.0004390755021330575, 'learning_rate': 0.0499300949295295, 'epoch': 0.05} + 5%|▌ | 28/520 [01:55<29:43, 3.62s/it] 6%|▌ | 29/520 [01:59<29:40, 3.63s/it] {'loss': 1.3856, 'grad_norm': 0.0004402085678646073, 'learning_rate': 0.04991796527392888, 'epoch': 0.06} + 6%|▌ | 29/520 [01:59<29:40, 3.63s/it] 6%|▌ | 30/520 [02:03<29:42, 3.64s/it] {'loss': 1.443, 'grad_norm': 0.00037270312433353663, 'learning_rate': 0.04990486745229364, 'epoch': 0.06} + 6%|▌ | 30/520 [02:03<29:42, 3.64s/it] 6%|▌ | 31/520 [02:06<30:04, 3.69s/it] {'loss': 1.3489, 'grad_norm': 0.0003919725858062764, 'learning_rate': 0.04989080197352835, 'epoch': 0.06} + 6%|▌ | 31/520 [02:06<30:04, 3.69s/it] 6%|▌ | 32/520 [02:10<29:57, 3.68s/it] {'loss': 1.2425, 'grad_norm': 0.00035327761965861027, 'learning_rate': 0.049875769384135044, 'epoch': 0.06} + 6%|▌ | 32/520 [02:10<29:57, 3.68s/it] 6%|▋ | 33/520 [02:14<29:45, 3.67s/it] {'loss': 1.3575, 'grad_norm': 0.0005118386903194881, 'learning_rate': 0.049859770268191994, 'epoch': 0.06} + 6%|▋ | 33/520 [02:14<29:45, 3.67s/it] 7%|▋ | 34/520 [02:17<29:36, 3.66s/it] {'loss': 1.35, 'grad_norm': 0.0004893512384980576, 'learning_rate': 0.04984280524733107, 'epoch': 0.07} + 7%|▋ | 34/520 [02:17<29:36, 3.66s/it] 7%|▋ | 35/520 [02:21<29:25, 3.64s/it] {'loss': 1.3554, 'grad_norm': 0.0005035833235267077, 'learning_rate': 0.04982487498071349, 'epoch': 0.07} + 7%|▋ | 35/520 [02:21<29:25, 3.64s/it] 7%|▋ | 36/520 [02:25<29:29, 3.66s/it] {'loss': 1.4481, 'grad_norm': 0.00042522009898242986, 'learning_rate': 0.04980598016500431, 'epoch': 0.07} + 7%|▋ | 36/520 [02:25<29:29, 3.66s/it] 7%|▋ | 37/520 [02:28<29:20, 3.64s/it] {'loss': 1.3986, 'grad_norm': 0.00036508280192678325, 'learning_rate': 0.049786121534345265, 'epoch': 0.07} + 7%|▋ | 37/520 [02:28<29:20, 3.64s/it] 7%|▋ | 38/520 [02:32<29:17, 3.65s/it] {'loss': 1.5071, 'grad_norm': 0.00041297162699089913, 'learning_rate': 0.04976529986032632, 'epoch': 0.07} + 7%|▋ | 38/520 [02:32<29:17, 3.65s/it] 8%|▊ | 39/520 [02:35<29:06, 3.63s/it] {'loss': 1.391, 'grad_norm': 0.0004924474574181165, 'learning_rate': 0.049743515951955655, 'epoch': 0.07} + 8%|▊ | 39/520 [02:35<29:06, 3.63s/it] 8%|▊ | 40/520 [02:39<29:03, 3.63s/it] {'loss': 1.4039, 'grad_norm': 0.0004073477518398799, 'learning_rate': 0.049720770655628216, 'epoch': 0.08} + 8%|▊ | 40/520 [02:39<29:03, 3.63s/it] 8%|▊ | 41/520 [02:43<29:07, 3.65s/it] {'loss': 1.384, 'grad_norm': 0.0004447557629587047, 'learning_rate': 0.04969706485509287, 'epoch': 0.08} + 8%|▊ | 41/520 [02:43<29:07, 3.65s/it] 8%|▊ | 42/520 [02:46<29:04, 3.65s/it] {'loss': 1.3739, 'grad_norm': 0.0005589781778418665, 'learning_rate': 0.04967239947141803, 'epoch': 0.08} + 8%|▊ | 42/520 [02:46<29:04, 3.65s/it] 8%|▊ | 43/520 [02:50<29:00, 3.65s/it] {'loss': 1.2904, 'grad_norm': 0.0003823265935521392, 'learning_rate': 0.0496467754629559, 'epoch': 0.08} + 8%|▊ | 43/520 [02:50<29:00, 3.65s/it] 8%|▊ | 44/520 [02:54<29:00, 3.66s/it] {'loss': 1.3899, 'grad_norm': 0.0004338704422575009, 'learning_rate': 0.049620193825305206, 'epoch': 0.08} + 8%|▊ | 44/520 [02:54<29:00, 3.66s/it] 9%|▊ | 45/520 [02:57<28:55, 3.65s/it] {'loss': 1.3983, 'grad_norm': 0.0005354390704394461, 'learning_rate': 0.04959265559127254, 'epoch': 0.09} + 9%|▊ | 45/520 [02:57<28:55, 3.65s/it] 9%|▉ | 46/520 [03:01<28:51, 3.65s/it] {'loss': 1.4219, 'grad_norm': 0.0004208880665187292, 'learning_rate': 0.04956416183083221, 'epoch': 0.09} + 9%|▉ | 46/520 [03:01<28:51, 3.65s/it] 9%|▉ | 47/520 [03:05<28:43, 3.64s/it] {'loss': 1.3563, 'grad_norm': 0.0005604469818651931, 'learning_rate': 0.049534713651084694, 'epoch': 0.09} + 9%|▉ | 47/520 [03:05<28:43, 3.64s/it] 9%|▉ | 48/520 [03:08<28:32, 3.63s/it] {'loss': 1.3796, 'grad_norm': 0.0006152301569310692, 'learning_rate': 0.049504312196213596, 'epoch': 0.09} + 9%|▉ | 48/520 [03:08<28:32, 3.63s/it] 9%|▉ | 49/520 [03:12<28:32, 3.64s/it] {'loss': 1.3949, 'grad_norm': 0.0004986568971027748, 'learning_rate': 0.049472958647441216, 'epoch': 0.09} + 9%|▉ | 49/520 [03:12<28:32, 3.64s/it] 10%|▉ | 50/520 [03:15<28:26, 3.63s/it] {'loss': 1.3945, 'grad_norm': 0.0005086180607896283, 'learning_rate': 0.04944065422298262, 'epoch': 0.1} + 10%|▉ | 50/520 [03:16<28:26, 3.63s/it] 10%|▉ | 51/520 [03:19<28:21, 3.63s/it] {'loss': 1.3236, 'grad_norm': 0.0006066773267816338, 'learning_rate': 0.049407400177998335, 'epoch': 0.1} + 10%|▉ | 51/520 [03:19<28:21, 3.63s/it] 10%|█ | 52/520 [03:23<28:19, 3.63s/it] {'loss': 1.4522, 'grad_norm': 0.0006606528616382212, 'learning_rate': 0.04937319780454559, 'epoch': 0.1} + 10%|█ | 52/520 [03:23<28:19, 3.63s/it] 10%|█ | 53/520 [03:26<28:14, 3.63s/it] {'loss': 1.4215, 'grad_norm': 0.0005294168490805362, 'learning_rate': 0.049338048431528087, 'epoch': 0.1} + 10%|█ | 53/520 [03:26<28:14, 3.63s/it] 10%|█ | 54/520 [03:30<28:09, 3.63s/it] {'loss': 1.3636, 'grad_norm': 0.0005466397325103735, 'learning_rate': 0.049301953424644364, 'epoch': 0.1} + 10%|█ | 54/520 [03:30<28:09, 3.63s/it] 11%|█ | 55/520 [03:34<28:05, 3.63s/it] {'loss': 1.3084, 'grad_norm': 0.0006080512043349245, 'learning_rate': 0.049264914186334774, 'epoch': 0.11} + 11%|█ | 55/520 [03:34<28:05, 3.63s/it] 11%|█ | 56/520 [03:37<28:23, 3.67s/it] {'loss': 1.4364, 'grad_norm': 0.0005785073428137049, 'learning_rate': 0.04922693215572695, 'epoch': 0.11} + 11%|█ | 56/520 [03:37<28:23, 3.67s/it] 11%|█ | 57/520 [03:41<28:31, 3.70s/it] {'loss': 1.3209, 'grad_norm': 0.0006734220807047592, 'learning_rate': 0.04918800880857991, 'epoch': 0.11} + 11%|█ | 57/520 [03:41<28:31, 3.70s/it] 11%|█ | 58/520 [03:45<28:39, 3.72s/it] {'loss': 1.443, 'grad_norm': 0.000526701961032109, 'learning_rate': 0.04914814565722671, 'epoch': 0.11} + 11%|█ | 58/520 [03:45<28:39, 3.72s/it] 11%|█▏ | 59/520 [03:49<28:35, 3.72s/it] {'loss': 1.2559, 'grad_norm': 0.000554745046057998, 'learning_rate': 0.049107344250515696, 'epoch': 0.11} + 11%|█▏ | 59/520 [03:49<28:35, 3.72s/it] 12%|█▏ | 60/520 [03:52<28:20, 3.70s/it] {'loss': 1.3587, 'grad_norm': 0.0005663164738348194, 'learning_rate': 0.049065606173750304, 'epoch': 0.12} + 12%|█▏ | 60/520 [03:52<28:20, 3.70s/it] 12%|█▏ | 61/520 [03:56<28:29, 3.72s/it] {'loss': 1.3268, 'grad_norm': 0.0006265363951243053, 'learning_rate': 0.049022933048627496, 'epoch': 0.12} + 12%|█▏ | 61/520 [03:56<28:29, 3.72s/it] 12%|█▏ | 62/520 [04:00<28:37, 3.75s/it] {'loss': 1.3431, 'grad_norm': 0.0006340088134362661, 'learning_rate': 0.048979326533174695, 'epoch': 0.12} + 12%|█▏ | 62/520 [04:00<28:37, 3.75s/it] 12%|█▏ | 63/520 [04:04<28:44, 3.77s/it] {'loss': 1.3464, 'grad_norm': 0.0006786144866687667, 'learning_rate': 0.048934788321685456, 'epoch': 0.12} + 12%|█▏ | 63/520 [04:04<28:44, 3.77s/it] 12%|█▏ | 64/520 [04:08<28:45, 3.78s/it] {'loss': 1.3647, 'grad_norm': 0.0006543167959448877, 'learning_rate': 0.048889320144653525, 'epoch': 0.12} + 12%|█▏ | 64/520 [04:08<28:45, 3.78s/it] 12%|█▎ | 65/520 [04:11<28:44, 3.79s/it] {'loss': 1.3716, 'grad_norm': 0.0006940994251025098, 'learning_rate': 0.04884292376870567, 'epoch': 0.12} + 12%|█▎ | 65/520 [04:11<28:44, 3.79s/it] 13%|█▎ | 66/520 [04:15<28:45, 3.80s/it] {'loss': 1.3045, 'grad_norm': 0.000588220690861794, 'learning_rate': 0.048795600996533066, 'epoch': 0.13} + 13%|█▎ | 66/520 [04:15<28:45, 3.80s/it] 13%|█▎ | 67/520 [04:19<28:39, 3.80s/it] {'loss': 1.2369, 'grad_norm': 0.0006664133156316464, 'learning_rate': 0.04874735366682115, 'epoch': 0.13} + 13%|█▎ | 67/520 [04:19<28:39, 3.80s/it] 13%|█▎ | 68/520 [04:23<28:37, 3.80s/it] {'loss': 1.2991, 'grad_norm': 0.0006654246202845667, 'learning_rate': 0.0486981836541783, 'epoch': 0.13} + 13%|█▎ | 68/520 [04:23<28:37, 3.80s/it] 13%|█▎ | 69/520 [04:27<28:34, 3.80s/it] {'loss': 1.2764, 'grad_norm': 0.0008392652133856227, 'learning_rate': 0.0486480928690629, 'epoch': 0.13} + 13%|█▎ | 69/520 [04:27<28:34, 3.80s/it] 13%|█▎ | 70/520 [04:30<28:32, 3.81s/it] {'loss': 1.3057, 'grad_norm': 0.00079291986783912, 'learning_rate': 0.048597083257709194, 'epoch': 0.13} + 13%|█▎ | 70/520 [04:30<28:32, 3.81s/it] 14%|█▎ | 71/520 [04:34<28:15, 3.78s/it] {'loss': 1.2506, 'grad_norm': 0.0006951545465440202, 'learning_rate': 0.04854515680205159, 'epoch': 0.14} + 14%|█▎ | 71/520 [04:34<28:15, 3.78s/it] 14%|█▍ | 72/520 [04:38<27:49, 3.73s/it] {'loss': 1.3929, 'grad_norm': 0.0007074115927877894, 'learning_rate': 0.048492315519647715, 'epoch': 0.14} + 14%|█▍ | 72/520 [04:38<27:49, 3.73s/it] 14%|█▍ | 73/520 [04:41<27:34, 3.70s/it] {'loss': 1.2158, 'grad_norm': 0.0007685205542089958, 'learning_rate': 0.04843856146359998, 'epoch': 0.14} + 14%|█▍ | 73/520 [04:41<27:34, 3.70s/it] 14%|█▍ | 74/520 [04:45<27:20, 3.68s/it] {'loss': 1.3194, 'grad_norm': 0.0007598886976942846, 'learning_rate': 0.04838389672247585, 'epoch': 0.14} + 14%|█▍ | 74/520 [04:45<27:20, 3.68s/it] 14%|█▍ | 75/520 [04:49<27:12, 3.67s/it] {'loss': 1.2443, 'grad_norm': 0.0006864103714263882, 'learning_rate': 0.048328323420226665, 'epoch': 0.14} + 14%|█▍ | 75/520 [04:49<27:12, 3.67s/it] 15%|█▍ | 76/520 [04:52<26:59, 3.65s/it] {'loss': 1.3515, 'grad_norm': 0.0006032150548250471, 'learning_rate': 0.04827184371610511, 'epoch': 0.15} + 15%|█▍ | 76/520 [04:52<26:59, 3.65s/it] 15%|█▍ | 77/520 [04:56<26:54, 3.64s/it] {'loss': 1.1776, 'grad_norm': 0.0007928318308586095, 'learning_rate': 0.04821445980458134, 'epoch': 0.15} + 15%|█▍ | 77/520 [04:56<26:54, 3.64s/it] 15%|█▌ | 78/520 [04:59<26:49, 3.64s/it] {'loss': 1.2681, 'grad_norm': 0.0007235843155465202, 'learning_rate': 0.04815617391525772, 'epoch': 0.15} + 15%|█▌ | 78/520 [04:59<26:49, 3.64s/it] 15%|█▌ | 79/520 [05:03<26:44, 3.64s/it] {'loss': 1.2578, 'grad_norm': 0.0007638316705650676, 'learning_rate': 0.04809698831278217, 'epoch': 0.15} + 15%|█▌ | 79/520 [05:03<26:44, 3.64s/it] 15%|█▌ | 80/520 [05:07<26:39, 3.64s/it] {'loss': 1.328, 'grad_norm': 0.0008086470960475704, 'learning_rate': 0.04803690529676019, 'epoch': 0.15} + 15%|█▌ | 80/520 [05:07<26:39, 3.64s/it] 16%|█▌ | 81/520 [05:10<26:31, 3.63s/it] {'loss': 1.4028, 'grad_norm': 0.000996265163889237, 'learning_rate': 0.04797592720166552, 'epoch': 0.16} + 16%|█▌ | 81/520 [05:10<26:31, 3.63s/it] 16%|█▌ | 82/520 [05:14<26:27, 3.62s/it] {'loss': 1.3272, 'grad_norm': 0.0007369153329289499, 'learning_rate': 0.04791405639674941, 'epoch': 0.16} + 16%|█▌ | 82/520 [05:14<26:27, 3.62s/it] 16%|█▌ | 83/520 [05:18<26:23, 3.62s/it] {'loss': 1.3292, 'grad_norm': 0.000906177304548837, 'learning_rate': 0.047851295285948586, 'epoch': 0.16} + 16%|█▌ | 83/520 [05:18<26:23, 3.62s/it] 16%|█▌ | 84/520 [05:21<26:18, 3.62s/it] {'loss': 1.352, 'grad_norm': 0.0010133276491196685, 'learning_rate': 0.04778764630779184, 'epoch': 0.16} + 16%|█▌ | 84/520 [05:21<26:18, 3.62s/it] 16%|█▋ | 85/520 [05:25<26:17, 3.63s/it] {'loss': 1.3985, 'grad_norm': 0.0008303064100874082, 'learning_rate': 0.047723111935305276, 'epoch': 0.16} + 16%|█▋ | 85/520 [05:25<26:17, 3.63s/it] 17%|█▋ | 86/520 [05:28<26:15, 3.63s/it] {'loss': 1.3735, 'grad_norm': 0.0007997084828636547, 'learning_rate': 0.047657694675916254, 'epoch': 0.17} + 17%|█▋ | 86/520 [05:28<26:15, 3.63s/it] 17%|█▋ | 87/520 [05:32<26:18, 3.65s/it] {'loss': 1.286, 'grad_norm': 0.000759028136514078, 'learning_rate': 0.04759139707135592, 'epoch': 0.17} + 17%|█▋ | 87/520 [05:32<26:18, 3.65s/it] 17%|█▋ | 88/520 [05:36<26:18, 3.65s/it] {'loss': 1.2087, 'grad_norm': 0.000662761534113232, 'learning_rate': 0.04752422169756048, 'epoch': 0.17} + 17%|█▋ | 88/520 [05:36<26:18, 3.65s/it] 17%|█▋ | 89/520 [05:39<26:08, 3.64s/it] {'loss': 1.3354, 'grad_norm': 0.0008637855007503425, 'learning_rate': 0.047456171164571105, 'epoch': 0.17} + 17%|█▋ | 89/520 [05:39<26:08, 3.64s/it] 17%|█▋ | 90/520 [05:43<26:11, 3.66s/it] {'loss': 1.2732, 'grad_norm': 0.0008625677238199397, 'learning_rate': 0.04738724811643252, 'epoch': 0.17} + 17%|█▋ | 90/520 [05:43<26:11, 3.66s/it] 18%|█▊ | 91/520 [05:47<26:04, 3.65s/it] {'loss': 1.3328, 'grad_norm': 0.000804591392706582, 'learning_rate': 0.047317455231090294, 'epoch': 0.17} + 18%|█▊ | 91/520 [05:47<26:04, 3.65s/it] 18%|█▊ | 92/520 [05:50<25:55, 3.63s/it] {'loss': 1.2681, 'grad_norm': 0.0008572774078000106, 'learning_rate': 0.04724679522028672, 'epoch': 0.18} + 18%|█▊ | 92/520 [05:50<25:55, 3.63s/it] 18%|█▊ | 93/520 [05:54<25:52, 3.64s/it] {'loss': 1.2898, 'grad_norm': 0.0009254526265623268, 'learning_rate': 0.047175270829455544, 'epoch': 0.18} + 18%|█▊ | 93/520 [05:54<25:52, 3.64s/it] 18%|█▊ | 94/520 [05:58<25:47, 3.63s/it] {'loss': 1.3622, 'grad_norm': 0.0009986905923564547, 'learning_rate': 0.047102884837615246, 'epoch': 0.18} + 18%|█▊ | 94/520 [05:58<25:47, 3.63s/it] 18%|█▊ | 95/520 [06:01<25:39, 3.62s/it] {'loss': 1.2773, 'grad_norm': 0.0010350942517036313, 'learning_rate': 0.04702964005726106, 'epoch': 0.18} + 18%|█▊ | 95/520 [06:01<25:39, 3.62s/it] 18%|█▊ | 96/520 [06:05<25:35, 3.62s/it] {'loss': 1.2888, 'grad_norm': 0.0007829551705125049, 'learning_rate': 0.04695553933425572, 'epoch': 0.18} + 18%|█▊ | 96/520 [06:05<25:35, 3.62s/it] 19%|█▊ | 97/520 [06:08<25:32, 3.62s/it] {'loss': 1.2644, 'grad_norm': 0.0010070165695596055, 'learning_rate': 0.046880585547718845, 'epoch': 0.19} + 19%|█▊ | 97/520 [06:08<25:32, 3.62s/it] 19%|█▉ | 98/520 [06:12<25:29, 3.63s/it] {'loss': 1.26, 'grad_norm': 0.0007425042173259353, 'learning_rate': 0.046804781609915136, 'epoch': 0.19} + 19%|█▉ | 98/520 [06:12<25:29, 3.63s/it] 19%|█▉ | 99/520 [06:16<25:27, 3.63s/it] {'loss': 1.2526, 'grad_norm': 0.0009166754624807925, 'learning_rate': 0.04672813046614116, 'epoch': 0.19} + 19%|█▉ | 99/520 [06:16<25:27, 3.63s/it] 19%|█▉ | 100/520 [06:19<25:21, 3.62s/it] {'loss': 1.2402, 'grad_norm': 0.0008258645366460271, 'learning_rate': 0.046650635094610975, 'epoch': 0.19} + 19%|█▉ | 100/520 [06:19<25:21, 3.62s/it] 19%|█▉ | 101/520 [06:23<25:21, 3.63s/it] {'loss': 1.2861, 'grad_norm': 0.0009128814606134884, 'learning_rate': 0.04657229850634033, 'epoch': 0.19} + 19%|█▉ | 101/520 [06:23<25:21, 3.63s/it] 20%|█▉ | 102/520 [06:27<25:17, 3.63s/it] {'loss': 1.2798, 'grad_norm': 0.0009295929581858612, 'learning_rate': 0.04649312374502976, 'epoch': 0.2} + 20%|█▉ | 102/520 [06:27<25:17, 3.63s/it] 20%|█▉ | 103/520 [06:30<25:12, 3.63s/it] {'loss': 1.2275, 'grad_norm': 0.0008612455387123186, 'learning_rate': 0.04641311388694629, 'epoch': 0.2} + 20%|█▉ | 103/520 [06:30<25:12, 3.63s/it] 20%|██ | 104/520 [06:34<25:12, 3.64s/it] {'loss': 1.2952, 'grad_norm': 0.0009558986126139576, 'learning_rate': 0.046332272040803896, 'epoch': 0.2} + 20%|██ | 104/520 [06:34<25:12, 3.64s/it] 20%|██ | 105/520 [06:38<25:06, 3.63s/it] {'loss': 1.2841, 'grad_norm': 0.0008661746296258313, 'learning_rate': 0.04625060134764273, 'epoch': 0.2} + 20%|██ | 105/520 [06:38<25:06, 3.63s/it] 20%|██ | 106/520 [06:41<25:05, 3.64s/it] {'loss': 1.2582, 'grad_norm': 0.0008092815564509136, 'learning_rate': 0.04616810498070711, 'epoch': 0.2} + 20%|██ | 106/520 [06:41<25:05, 3.64s/it] 21%|██ | 107/520 [06:45<25:02, 3.64s/it] {'loss': 1.2351, 'grad_norm': 0.0008894501605605228, 'learning_rate': 0.046084786145322146, 'epoch': 0.21} + 21%|██ | 107/520 [06:45<25:02, 3.64s/it] 21%|██ | 108/520 [06:48<25:01, 3.64s/it] {'loss': 1.2382, 'grad_norm': 0.0009694801325282927, 'learning_rate': 0.046000648078769295, 'epoch': 0.21} + 21%|██ | 108/520 [06:48<25:01, 3.64s/it] 21%|██ | 109/520 [06:52<24:56, 3.64s/it] {'loss': 1.2062, 'grad_norm': 0.0008504999058318814, 'learning_rate': 0.045915694050160494, 'epoch': 0.21} + 21%|██ | 109/520 [06:52<24:56, 3.64s/it] 21%|██ | 110/520 [06:56<24:54, 3.65s/it] {'loss': 1.4064, 'grad_norm': 0.0009361173037164769, 'learning_rate': 0.045829927360311225, 'epoch': 0.21} + 21%|██ | 110/520 [06:56<24:54, 3.65s/it] 21%|██▏ | 111/520 [06:59<24:47, 3.64s/it] {'loss': 1.4051, 'grad_norm': 0.0009614918990709081, 'learning_rate': 0.04574335134161219, 'epoch': 0.21} + 21%|██▏ | 111/520 [06:59<24:47, 3.64s/it] 22%|██▏ | 112/520 [07:03<24:43, 3.64s/it] {'loss': 1.2948, 'grad_norm': 0.0009165122309948802, 'learning_rate': 0.045655969357899874, 'epoch': 0.22} + 22%|██▏ | 112/520 [07:03<24:43, 3.64s/it] 22%|██▏ | 113/520 [07:07<24:34, 3.62s/it] {'loss': 1.1994, 'grad_norm': 0.0008926783439976004, 'learning_rate': 0.045567784804325835, 'epoch': 0.22} + 22%|██▏ | 113/520 [07:07<24:34, 3.62s/it] 22%|██▏ | 114/520 [07:10<24:32, 3.63s/it] {'loss': 1.2965, 'grad_norm': 0.0009090392176069725, 'learning_rate': 0.0454788011072248, 'epoch': 0.22} + 22%|██▏ | 114/520 [07:10<24:32, 3.63s/it] 22%|██▏ | 115/520 [07:14<24:29, 3.63s/it] {'loss': 1.3679, 'grad_norm': 0.0009246762430395531, 'learning_rate': 0.04538902172398151, 'epoch': 0.22} + 22%|██▏ | 115/520 [07:14<24:29, 3.63s/it] 22%|██▏ | 116/520 [07:18<24:27, 3.63s/it] {'loss': 1.3855, 'grad_norm': 0.0008731750729465077, 'learning_rate': 0.04529845014289642, 'epoch': 0.22} + 22%|██▏ | 116/520 [07:18<24:27, 3.63s/it] 22%|██▎ | 117/520 [07:21<24:21, 3.63s/it] {'loss': 1.3502, 'grad_norm': 0.0009728438625610388, 'learning_rate': 0.04520708988305014, 'epoch': 0.23} + 22%|██▎ | 117/520 [07:21<24:21, 3.63s/it] 23%|██▎ | 118/520 [07:25<24:15, 3.62s/it] {'loss': 1.2862, 'grad_norm': 0.0008756558027285282, 'learning_rate': 0.04511494449416671, 'epoch': 0.23} + 23%|██▎ | 118/520 [07:25<24:15, 3.62s/it] 23%|██▎ | 119/520 [07:28<24:10, 3.62s/it] {'loss': 1.2377, 'grad_norm': 0.00097965378994526, 'learning_rate': 0.0450220175564757, 'epoch': 0.23} + 23%|██▎ | 119/520 [07:28<24:10, 3.62s/it] 23%|██▎ | 120/520 [07:32<24:10, 3.63s/it] {'loss': 1.2535, 'grad_norm': 0.0012085336459360344, 'learning_rate': 0.04492831268057307, 'epoch': 0.23} + 23%|██▎ | 120/520 [07:32<24:10, 3.63s/it] 23%|██▎ | 121/520 [07:36<24:05, 3.62s/it] {'loss': 1.2927, 'grad_norm': 0.0009967680029934018, 'learning_rate': 0.04483383350728088, 'epoch': 0.23} + 23%|██▎ | 121/520 [07:36<24:05, 3.62s/it] 23%|██▎ | 122/520 [07:39<24:03, 3.63s/it] {'loss': 1.2177, 'grad_norm': 0.0009359382050524199, 'learning_rate': 0.044738583707505886, 'epoch': 0.23} + 23%|██▎ | 122/520 [07:39<24:03, 3.63s/it] 24%|██▎ | 123/520 [07:43<24:02, 3.63s/it] {'loss': 1.2837, 'grad_norm': 0.0008960865924610417, 'learning_rate': 0.04464256698209684, 'epoch': 0.24} + 24%|██▎ | 123/520 [07:43<24:02, 3.63s/it] 24%|██▍ | 124/520 [07:47<24:00, 3.64s/it] {'loss': 1.2604, 'grad_norm': 0.0010628809737840087, 'learning_rate': 0.04454578706170075, 'epoch': 0.24} + 24%|██▍ | 124/520 [07:47<24:00, 3.64s/it] 24%|██▍ | 125/520 [07:50<23:57, 3.64s/it] {'loss': 1.2577, 'grad_norm': 0.0009868658375824988, 'learning_rate': 0.04444824770661787, 'epoch': 0.24} + 24%|██▍ | 125/520 [07:50<23:57, 3.64s/it] 24%|██▍ | 126/520 [07:54<25:12, 3.84s/it] {'loss': 1.2169, 'grad_norm': 0.000823801544614067, 'learning_rate': 0.04434995270665569, 'epoch': 0.24} + 24%|██▍ | 126/520 [07:54<25:12, 3.84s/it] 24%|██▍ | 127/520 [07:58<24:46, 3.78s/it] {'loss': 1.2552, 'grad_norm': 0.0011596272343995843, 'learning_rate': 0.04425090588098157, 'epoch': 0.24} + 24%|██▍ | 127/520 [07:58<24:46, 3.78s/it] 25%|██▍ | 128/520 [08:02<24:25, 3.74s/it] {'loss': 1.2804, 'grad_norm': 0.0009488661301956661, 'learning_rate': 0.04415111107797445, 'epoch': 0.25} + 25%|██▍ | 128/520 [08:02<24:25, 3.74s/it] 25%|██▍ | 129/520 [08:05<24:10, 3.71s/it] {'loss': 1.2563, 'grad_norm': 0.0008958085901964377, 'learning_rate': 0.04405057217507527, 'epoch': 0.25} + 25%|██▍ | 129/520 [08:05<24:10, 3.71s/it] 25%|██▌ | 130/520 [08:09<23:58, 3.69s/it] {'loss': 1.2761, 'grad_norm': 0.0009201089638534103, 'learning_rate': 0.04394929307863633, 'epoch': 0.25} + 25%|██▌ | 130/520 [08:09<23:58, 3.69s/it] 25%|██▌ | 131/520 [08:13<23:45, 3.67s/it] {'loss': 1.1852, 'grad_norm': 0.0008486646060080358, 'learning_rate': 0.0438472777237695, 'epoch': 0.25} + 25%|██▌ | 131/520 [08:13<23:45, 3.67s/it] 25%|██▌ | 132/520 [08:16<23:35, 3.65s/it] {'loss': 1.3291, 'grad_norm': 0.0011060428133955649, 'learning_rate': 0.04374453007419336, 'epoch': 0.25} + 25%|██▌ | 132/520 [08:16<23:35, 3.65s/it] 26%|██▌ | 133/520 [08:20<23:24, 3.63s/it] {'loss': 1.2401, 'grad_norm': 0.0011246455627726443, 'learning_rate': 0.04364105412207914, 'epoch': 0.26} + 26%|██▌ | 133/520 [08:20<23:24, 3.63s/it] 26%|██▌ | 134/520 [08:24<23:23, 3.64s/it] {'loss': 1.3103, 'grad_norm': 0.0010074116057302422, 'learning_rate': 0.04353685388789567, 'epoch': 0.26} + 26%|██▌ | 134/520 [08:24<23:23, 3.64s/it] 26%|██▌ | 135/520 [08:27<23:14, 3.62s/it] {'loss': 1.3602, 'grad_norm': 0.0010841982278226732, 'learning_rate': 0.0434319334202531, 'epoch': 0.26} + 26%|██▌ | 135/520 [08:27<23:14, 3.62s/it] 26%|██▌ | 136/520 [08:31<23:14, 3.63s/it] {'loss': 1.3309, 'grad_norm': 0.001073026825029318, 'learning_rate': 0.04332629679574566, 'epoch': 0.26} + 26%|██▌ | 136/520 [08:31<23:14, 3.63s/it] 26%|██▋ | 137/520 [08:34<23:21, 3.66s/it] {'loss': 1.2427, 'grad_norm': 0.0011328957190751595, 'learning_rate': 0.04321994811879321, 'epoch': 0.26} + 26%|██▋ | 137/520 [08:34<23:21, 3.66s/it] 27%|██▋ | 138/520 [08:38<23:20, 3.67s/it] {'loss': 1.2475, 'grad_norm': 0.0009252563919391036, 'learning_rate': 0.043112891521481816, 'epoch': 0.27} + 27%|██▋ | 138/520 [08:38<23:20, 3.67s/it] 27%|██▋ | 139/520 [08:42<23:20, 3.67s/it] {'loss': 1.0988, 'grad_norm': 0.0009341359455905236, 'learning_rate': 0.04300513116340317, 'epoch': 0.27} + 27%|██▋ | 139/520 [08:42<23:20, 3.67s/it] 27%|██▋ | 140/520 [08:45<23:11, 3.66s/it] {'loss': 1.2408, 'grad_norm': 0.0008941522518570024, 'learning_rate': 0.04289667123149296, 'epoch': 0.27} + 27%|██▋ | 140/520 [08:45<23:11, 3.66s/it] 27%|██▋ | 141/520 [08:49<23:07, 3.66s/it] {'loss': 1.3307, 'grad_norm': 0.0009133786636080142, 'learning_rate': 0.04278751593986826, 'epoch': 0.27} + 27%|██▋ | 141/520 [08:49<23:07, 3.66s/it] 27%|██▋ | 142/520 [08:53<23:02, 3.66s/it] {'loss': 1.2487, 'grad_norm': 0.0009037805713967151, 'learning_rate': 0.04267766952966369, 'epoch': 0.27} + 27%|██▋ | 142/520 [08:53<23:02, 3.66s/it] 28%|██▊ | 143/520 [08:56<22:56, 3.65s/it] {'loss': 1.2659, 'grad_norm': 0.0010699153824529335, 'learning_rate': 0.04256713626886673, 'epoch': 0.28} + 28%|██▊ | 143/520 [08:56<22:56, 3.65s/it] 28%|██▊ | 144/520 [09:00<22:54, 3.66s/it] {'loss': 1.2574, 'grad_norm': 0.0010878831782604869, 'learning_rate': 0.042455920452151824, 'epoch': 0.28} + 28%|██▊ | 144/520 [09:00<22:54, 3.66s/it] 28%|██▊ | 145/520 [09:04<23:00, 3.68s/it] {'loss': 1.1759, 'grad_norm': 0.0009777334782596994, 'learning_rate': 0.042344026400713546, 'epoch': 0.28} + 28%|██▊ | 145/520 [09:04<23:00, 3.68s/it] 28%|██▊ | 146/520 [09:08<22:54, 3.68s/it] {'loss': 1.2951, 'grad_norm': 0.0009990274133302112, 'learning_rate': 0.04223145846209868, 'epoch': 0.28} + 28%|██▊ | 146/520 [09:08<22:54, 3.68s/it] 28%|██▊ | 147/520 [09:11<23:02, 3.71s/it] {'loss': 1.2252, 'grad_norm': 0.001084663910576141, 'learning_rate': 0.04211822101003734, 'epoch': 0.28} + 28%|██▊ | 147/520 [09:11<23:02, 3.71s/it] 28%|██▊ | 148/520 [09:15<23:05, 3.72s/it] {'loss': 1.255, 'grad_norm': 0.001017824059192031, 'learning_rate': 0.042004318444272985, 'epoch': 0.28} + 28%|██▊ | 148/520 [09:15<23:05, 3.72s/it] 29%|██▊ | 149/520 [09:19<22:58, 3.72s/it] {'loss': 1.1836, 'grad_norm': 0.000995491962124548, 'learning_rate': 0.04188975519039151, 'epoch': 0.29} + 29%|██▊ | 149/520 [09:19<22:58, 3.72s/it] 29%|██▉ | 150/520 [09:22<22:48, 3.70s/it] {'loss': 1.3946, 'grad_norm': 0.0010335756994501147, 'learning_rate': 0.04177453569964926, 'epoch': 0.29} + 29%|██▉ | 150/520 [09:22<22:48, 3.70s/it] 29%|██▉ | 151/520 [09:26<22:32, 3.67s/it] {'loss': 1.2404, 'grad_norm': 0.001064391473775573, 'learning_rate': 0.0416586644488001, 'epoch': 0.29} + 29%|██▉ | 151/520 [09:26<22:32, 3.67s/it] 29%|██▉ | 152/520 [09:30<22:26, 3.66s/it] {'loss': 1.2063, 'grad_norm': 0.0011390269322737918, 'learning_rate': 0.04154214593992149, 'epoch': 0.29} + 29%|██▉ | 152/520 [09:30<22:26, 3.66s/it] 29%|██▉ | 153/520 [09:33<22:33, 3.69s/it] {'loss': 1.2352, 'grad_norm': 0.0010248870224467024, 'learning_rate': 0.04142498470023952, 'epoch': 0.29} + 29%|██▉ | 153/520 [09:33<22:33, 3.69s/it] 30%|██▉ | 154/520 [09:37<22:31, 3.69s/it] {'loss': 1.3165, 'grad_norm': 0.0009783464763418069, 'learning_rate': 0.041307185281953034, 'epoch': 0.3} + 30%|██▉ | 154/520 [09:37<22:31, 3.69s/it] 30%|██▉ | 155/520 [09:41<22:26, 3.69s/it] {'loss': 1.2344, 'grad_norm': 0.001123030764250693, 'learning_rate': 0.04118875226205676, 'epoch': 0.3} + 30%|██▉ | 155/520 [09:41<22:26, 3.69s/it] 30%|███ | 156/520 [09:44<22:25, 3.70s/it] {'loss': 1.2597, 'grad_norm': 0.0011892853579669948, 'learning_rate': 0.04106969024216348, 'epoch': 0.3} + 30%|███ | 156/520 [09:44<22:25, 3.70s/it] 30%|███ | 157/520 [09:48<22:28, 3.71s/it] {'loss': 1.293, 'grad_norm': 0.0009931916132295193, 'learning_rate': 0.04095000384832522, 'epoch': 0.3} + 30%|███ | 157/520 [09:48<22:28, 3.71s/it] 30%|███ | 158/520 [09:52<22:24, 3.71s/it] {'loss': 1.2423, 'grad_norm': 0.0010991993605669792, 'learning_rate': 0.0408296977308535, 'epoch': 0.3} + 30%|███ | 158/520 [09:52<22:24, 3.71s/it] 31%|███ | 159/520 [09:56<22:23, 3.72s/it] {'loss': 1.2884, 'grad_norm': 0.0010407981706060708, 'learning_rate': 0.04070877656413868, 'epoch': 0.31} + 31%|███ | 159/520 [09:56<22:23, 3.72s/it] 31%|███ | 160/520 [09:59<22:15, 3.71s/it] {'loss': 1.2938, 'grad_norm': 0.0010797242994068485, 'learning_rate': 0.040587245046468344, 'epoch': 0.31} + 31%|███ | 160/520 [09:59<22:15, 3.71s/it] 31%|███ | 161/520 [10:03<22:17, 3.72s/it] {'loss': 1.2688, 'grad_norm': 0.0010834750100326663, 'learning_rate': 0.040465107899844706, 'epoch': 0.31} + 31%|███ | 161/520 [10:03<22:17, 3.72s/it] 31%|███ | 162/520 [10:07<22:07, 3.71s/it] {'loss': 1.2318, 'grad_norm': 0.0009827115219441908, 'learning_rate': 0.04034236986980119, 'epoch': 0.31} + 31%|███ | 162/520 [10:07<22:07, 3.71s/it] 31%|███▏ | 163/520 [10:10<22:01, 3.70s/it] {'loss': 1.1762, 'grad_norm': 0.0011835885410327457, 'learning_rate': 0.04021903572521802, 'epoch': 0.31} + 31%|███▏ | 163/520 [10:10<22:01, 3.70s/it] 32%|███▏ | 164/520 [10:14<21:56, 3.70s/it] {'loss': 1.137, 'grad_norm': 0.001056283223805921, 'learning_rate': 0.04009511025813694, 'epoch': 0.32} + 32%|███▏ | 164/520 [10:14<21:56, 3.70s/it] 32%|███▏ | 165/520 [10:18<21:50, 3.69s/it] {'loss': 1.2976, 'grad_norm': 0.001051043129693678, 'learning_rate': 0.039970598283575014, 'epoch': 0.32} + 32%|███▏ | 165/520 [10:18<21:50, 3.69s/it] 32%|███▏ | 166/520 [10:22<21:45, 3.69s/it] {'loss': 1.2524, 'grad_norm': 0.0012194222001343843, 'learning_rate': 0.03984550463933754, 'epoch': 0.32} + 32%|███▏ | 166/520 [10:22<21:45, 3.69s/it] 32%|███▏ | 167/520 [10:25<21:45, 3.70s/it] {'loss': 1.2492, 'grad_norm': 0.0010857160345771235, 'learning_rate': 0.03971983418583012, 'epoch': 0.32} + 32%|███▏ | 167/520 [10:25<21:45, 3.70s/it] 32%|███▏ | 168/520 [10:29<21:37, 3.69s/it] {'loss': 1.1898, 'grad_norm': 0.0010593511274171934, 'learning_rate': 0.03959359180586975, 'epoch': 0.32} + 32%|███▏ | 168/520 [10:29<21:37, 3.69s/it] 32%|███▎ | 169/520 [10:33<21:34, 3.69s/it] {'loss': 1.2613, 'grad_norm': 0.0010363445355952771, 'learning_rate': 0.03946678240449515, 'epoch': 0.33} + 32%|███▎ | 169/520 [10:33<21:34, 3.69s/it] 33%|███▎ | 170/520 [10:36<21:27, 3.68s/it] {'loss': 1.1885, 'grad_norm': 0.0009028091339220126, 'learning_rate': 0.03933941090877616, 'epoch': 0.33} + 33%|███▎ | 170/520 [10:36<21:27, 3.68s/it] 33%|███▎ | 171/520 [10:40<21:22, 3.67s/it] {'loss': 1.2122, 'grad_norm': 0.0011850362484430801, 'learning_rate': 0.03921148226762231, 'epoch': 0.33} + 33%|███▎ | 171/520 [10:40<21:22, 3.67s/it] 33%|███▎ | 172/520 [10:44<21:19, 3.68s/it] {'loss': 1.2808, 'grad_norm': 0.0010540521683815032, 'learning_rate': 0.03908300145159055, 'epoch': 0.33} + 33%|███▎ | 172/520 [10:44<21:19, 3.68s/it] 33%|███▎ | 173/520 [10:47<21:13, 3.67s/it] {'loss': 1.2081, 'grad_norm': 0.0010854866867968833, 'learning_rate': 0.03895397345269211, 'epoch': 0.33} + 33%|███▎ | 173/520 [10:47<21:13, 3.67s/it] 33%|███▎ | 174/520 [10:51<21:09, 3.67s/it] {'loss': 1.2651, 'grad_norm': 0.0011161152664475858, 'learning_rate': 0.038824403284198485, 'epoch': 0.33} + 33%|███▎ | 174/520 [10:51<21:09, 3.67s/it] 34%|███▎ | 175/520 [10:55<21:04, 3.66s/it] {'loss': 1.1752, 'grad_norm': 0.00102833216530038, 'learning_rate': 0.03869429598044679, 'epoch': 0.34} + 34%|███▎ | 175/520 [10:55<21:04, 3.66s/it] 34%|███▍ | 176/520 [10:58<20:55, 3.65s/it] {'loss': 1.2488, 'grad_norm': 0.0010252494704463442, 'learning_rate': 0.038563656596643986, 'epoch': 0.34} + 34%|███▍ | 176/520 [10:58<20:55, 3.65s/it] 34%|███▍ | 177/520 [11:02<21:03, 3.68s/it] {'loss': 1.1432, 'grad_norm': 0.0011416196577763939, 'learning_rate': 0.0384324902086706, 'epoch': 0.34} + 34%|███▍ | 177/520 [11:02<21:03, 3.68s/it] 34%|███▍ | 178/520 [11:06<20:57, 3.68s/it] {'loss': 1.2443, 'grad_norm': 0.0011778033565367429, 'learning_rate': 0.03830080191288342, 'epoch': 0.34} + 34%|███▍ | 178/520 [11:06<20:57, 3.68s/it] 34%|███▍ | 179/520 [11:09<20:48, 3.66s/it] {'loss': 1.3213, 'grad_norm': 0.001052358855414811, 'learning_rate': 0.03816859682591752, 'epoch': 0.34} + 34%|███▍ | 179/520 [11:09<20:48, 3.66s/it] 35%|███▍ | 180/520 [11:13<20:46, 3.67s/it] {'loss': 1.2365, 'grad_norm': 0.0010854423599617124, 'learning_rate': 0.03803588008448745, 'epoch': 0.35} + 35%|███▍ | 180/520 [11:13<20:46, 3.67s/it] 35%|███▍ | 181/520 [11:17<20:56, 3.71s/it] {'loss': 1.2185, 'grad_norm': 0.0010090948486174414, 'learning_rate': 0.03790265684518767, 'epoch': 0.35} + 35%|███▍ | 181/520 [11:17<20:56, 3.71s/it] 35%|███▌ | 182/520 [11:21<21:03, 3.74s/it] {'loss': 1.2306, 'grad_norm': 0.001129566949598125, 'learning_rate': 0.03776893228429215, 'epoch': 0.35} + 35%|███▌ | 182/520 [11:21<21:03, 3.74s/it] 35%|███▌ | 183/520 [11:24<21:05, 3.76s/it] {'loss': 1.2445, 'grad_norm': 0.0010505094067954168, 'learning_rate': 0.03763471159755327, 'epoch': 0.35} + 35%|███▌ | 183/520 [11:24<21:05, 3.76s/it] 35%|███▌ | 184/520 [11:28<21:05, 3.77s/it] {'loss': 1.2028, 'grad_norm': 0.001221693215078871, 'learning_rate': 0.037500000000000006, 'epoch': 0.35} + 35%|███▌ | 184/520 [11:28<21:05, 3.77s/it] 36%|███▌ | 185/520 [11:32<20:56, 3.75s/it] {'loss': 1.3069, 'grad_norm': 0.0010878515834446742, 'learning_rate': 0.03736480272573519, 'epoch': 0.36} + 36%|███▌ | 185/520 [11:32<20:56, 3.75s/it] 36%|███▌ | 186/520 [11:35<20:41, 3.72s/it] {'loss': 1.2211, 'grad_norm': 0.0011906142725710916, 'learning_rate': 0.037229125027732235, 'epoch': 0.36} + 36%|███▌ | 186/520 [11:35<20:41, 3.72s/it] 36%|███▌ | 187/520 [11:39<20:25, 3.68s/it] {'loss': 1.2113, 'grad_norm': 0.001263648467977501, 'learning_rate': 0.037092972177630995, 'epoch': 0.36} + 36%|███▌ | 187/520 [11:39<20:25, 3.68s/it] 36%|███▌ | 188/520 [11:43<20:18, 3.67s/it] {'loss': 1.3073, 'grad_norm': 0.0011694247567790755, 'learning_rate': 0.03695634946553296, 'epoch': 0.36} + 36%|███▌ | 188/520 [11:43<20:18, 3.67s/it] 36%|███▋ | 189/520 [11:46<20:10, 3.66s/it] {'loss': 1.2997, 'grad_norm': 0.001029501755333484, 'learning_rate': 0.03681926219979568, 'epoch': 0.36} + 36%|███▋ | 189/520 [11:46<20:10, 3.66s/it] 37%|███▋ | 190/520 [11:50<20:11, 3.67s/it] {'loss': 1.2198, 'grad_norm': 0.0012351730644864995, 'learning_rate': 0.03668171570682655, 'epoch': 0.37} + 37%|███▋ | 190/520 [11:50<20:11, 3.67s/it] 37%|███▋ | 191/520 [11:54<20:14, 3.69s/it] {'loss': 1.1841, 'grad_norm': 0.0010938927516329147, 'learning_rate': 0.03654371533087585, 'epoch': 0.37} + 37%|███▋ | 191/520 [11:54<20:14, 3.69s/it] 37%|███▋ | 192/520 [11:57<20:07, 3.68s/it] {'loss': 1.2612, 'grad_norm': 0.0010863342762977377, 'learning_rate': 0.036405266433829075, 'epoch': 0.37} + 37%|███▋ | 192/520 [11:57<20:07, 3.68s/it] 37%|███▋ | 193/520 [12:01<20:00, 3.67s/it] {'loss': 1.2017, 'grad_norm': 0.0011804555893682664, 'learning_rate': 0.036266374394998635, 'epoch': 0.37} + 37%|███▋ | 193/520 [12:01<20:00, 3.67s/it] 37%|███▋ | 194/520 [12:05<19:53, 3.66s/it] {'loss': 1.0915, 'grad_norm': 0.0010074160650965964, 'learning_rate': 0.03612704461091481, 'epoch': 0.37} + 37%|███▋ | 194/520 [12:05<19:53, 3.66s/it] 38%|███▊ | 195/520 [12:08<19:52, 3.67s/it] {'loss': 1.2718, 'grad_norm': 0.0010785227900420116, 'learning_rate': 0.035987282495116124, 'epoch': 0.38} + 38%|███▊ | 195/520 [12:08<19:52, 3.67s/it] 38%|███▊ | 196/520 [12:12<20:05, 3.72s/it] {'loss': 1.2651, 'grad_norm': 0.0013192970628956879, 'learning_rate': 0.035847093477938956, 'epoch': 0.38} + 38%|███▊ | 196/520 [12:12<20:05, 3.72s/it] 38%|███▊ | 197/520 [12:16<20:14, 3.76s/it] {'loss': 1.2085, 'grad_norm': 0.0011818218286009548, 'learning_rate': 0.03570648300630657, 'epoch': 0.38} + 38%|███▊ | 197/520 [12:16<20:14, 3.76s/it] 38%|███▊ | 198/520 [12:20<20:08, 3.75s/it] {'loss': 1.2836, 'grad_norm': 0.001273236937549413, 'learning_rate': 0.03556545654351749, 'epoch': 0.38} + 38%|███▊ | 198/520 [12:20<20:08, 3.75s/it] 38%|███▊ | 199/520 [12:24<20:00, 3.74s/it] {'loss': 1.1899, 'grad_norm': 0.0011592180398208541, 'learning_rate': 0.035424019569033204, 'epoch': 0.38} + 38%|███▊ | 199/520 [12:24<20:00, 3.74s/it] 38%|███▊ | 200/520 [12:27<19:49, 3.72s/it] {'loss': 1.1573, 'grad_norm': 0.0011244885862263221, 'learning_rate': 0.035282177578265295, 'epoch': 0.38} + 38%|███▊ | 200/520 [12:27<19:49, 3.72s/it] 39%|███▊ | 201/520 [12:31<19:40, 3.70s/it] {'loss': 1.169, 'grad_norm': 0.0009755971463305182, 'learning_rate': 0.03513993608236188, 'epoch': 0.39} + 39%|███▊ | 201/520 [12:31<19:40, 3.70s/it] 39%|███▉ | 202/520 [12:35<19:34, 3.69s/it] {'loss': 1.2144, 'grad_norm': 0.0011975042972247874, 'learning_rate': 0.03499730060799352, 'epoch': 0.39} + 39%|███▉ | 202/520 [12:35<19:34, 3.69s/it] 39%|███▉ | 203/520 [12:38<19:24, 3.67s/it] {'loss': 1.2495, 'grad_norm': 0.0012094483746176016, 'learning_rate': 0.03485427669713849, 'epoch': 0.39} + 39%|███▉ | 203/520 [12:38<19:24, 3.67s/it] 39%|███▉ | 204/520 [12:42<19:20, 3.67s/it] {'loss': 1.2614, 'grad_norm': 0.0012169751044363865, 'learning_rate': 0.03471086990686737, 'epoch': 0.39} + 39%|███▉ | 204/520 [12:42<19:20, 3.67s/it] 39%|███▉ | 205/520 [12:46<19:15, 3.67s/it] {'loss': 1.1708, 'grad_norm': 0.0010555096167676975, 'learning_rate': 0.03456708580912725, 'epoch': 0.39} + 39%|███▉ | 205/520 [12:46<19:15, 3.67s/it] 40%|███▉ | 206/520 [12:49<19:09, 3.66s/it] {'loss': 1.2783, 'grad_norm': 0.0011148540015990852, 'learning_rate': 0.03442292999052513, 'epoch': 0.4} + 40%|███▉ | 206/520 [12:49<19:09, 3.66s/it] 40%|███▉ | 207/520 [12:53<19:05, 3.66s/it] {'loss': 1.1318, 'grad_norm': 0.0009602207938652538, 'learning_rate': 0.03427840805211095, 'epoch': 0.4} + 40%|███▉ | 207/520 [12:53<19:05, 3.66s/it] 40%|████ | 208/520 [12:56<19:00, 3.65s/it] {'loss': 1.2828, 'grad_norm': 0.0012530491559166286, 'learning_rate': 0.03413352560915988, 'epoch': 0.4} + 40%|████ | 208/520 [12:56<19:00, 3.65s/it] 40%|████ | 209/520 [13:00<18:57, 3.66s/it] {'loss': 1.1958, 'grad_norm': 0.0011130429450639204, 'learning_rate': 0.03398828829095419, 'epoch': 0.4} + 40%|████ | 209/520 [13:00<18:57, 3.66s/it] 40%|████ | 210/520 [13:04<18:52, 3.65s/it] {'loss': 1.2636, 'grad_norm': 0.0011726688304498278, 'learning_rate': 0.033842701740564536, 'epoch': 0.4} + 40%|████ | 210/520 [13:04<18:52, 3.65s/it] 41%|████ | 211/520 [13:07<18:48, 3.65s/it] {'loss': 1.2601, 'grad_norm': 0.001075059717989367, 'learning_rate': 0.03369677161463068, 'epoch': 0.41} + 41%|████ | 211/520 [13:07<18:48, 3.65s/it] 41%|████ | 212/520 [13:11<18:41, 3.64s/it] {'loss': 1.2665, 'grad_norm': 0.0011194940659534688, 'learning_rate': 0.033550503583141725, 'epoch': 0.41} + 41%|████ | 212/520 [13:11<18:41, 3.64s/it] 41%|████ | 213/520 [13:15<18:38, 3.64s/it] {'loss': 1.2217, 'grad_norm': 0.0013123337597974807, 'learning_rate': 0.03340390332921577, 'epoch': 0.41} + 41%|████ | 213/520 [13:15<18:38, 3.64s/it] 41%|████ | 214/520 [13:18<18:35, 3.65s/it] {'loss': 1.2191, 'grad_norm': 0.0011929019407636821, 'learning_rate': 0.03325697654887918, 'epoch': 0.41} + 41%|████ | 214/520 [13:18<18:35, 3.65s/it] 41%|████▏ | 215/520 [13:22<18:35, 3.66s/it] {'loss': 1.1083, 'grad_norm': 0.001118886800145629, 'learning_rate': 0.03310972895084518, 'epoch': 0.41} + 41%|████▏ | 215/520 [13:22<18:35, 3.66s/it] 42%|████▏ | 216/520 [13:26<18:30, 3.65s/it] {'loss': 1.1313, 'grad_norm': 0.0011381816863178555, 'learning_rate': 0.032962166256292114, 'epoch': 0.42} + 42%|████▏ | 216/520 [13:26<18:30, 3.65s/it] 42%|████▏ | 217/520 [13:29<18:27, 3.65s/it] {'loss': 1.2599, 'grad_norm': 0.0011925191517396434, 'learning_rate': 0.03281429419864112, 'epoch': 0.42} + 42%|████▏ | 217/520 [13:29<18:27, 3.65s/it] 42%|████▏ | 218/520 [13:33<18:23, 3.65s/it] {'loss': 1.2362, 'grad_norm': 0.0012786819402634518, 'learning_rate': 0.032666118523333364, 'epoch': 0.42} + 42%|████▏ | 218/520 [13:33<18:23, 3.65s/it] 42%|████▏ | 219/520 [13:37<18:21, 3.66s/it] {'loss': 1.2575, 'grad_norm': 0.0011065960558829628, 'learning_rate': 0.03251764498760683, 'epoch': 0.42} + 42%|████▏ | 219/520 [13:37<18:21, 3.66s/it] 42%|████▏ | 220/520 [13:40<18:16, 3.66s/it] {'loss': 1.1511, 'grad_norm': 0.0010824258107650089, 'learning_rate': 0.03236887936027261, 'epoch': 0.42} + 42%|████▏ | 220/520 [13:40<18:16, 3.66s/it] 42%|████▎ | 221/520 [13:44<18:13, 3.66s/it] {'loss': 1.249, 'grad_norm': 0.001198814549011231, 'learning_rate': 0.03221982742149075, 'epoch': 0.42} + 42%|████▎ | 221/520 [13:44<18:13, 3.66s/it] 43%|████▎ | 222/520 [13:48<18:08, 3.65s/it] {'loss': 1.1866, 'grad_norm': 0.001186425408688233, 'learning_rate': 0.03207049496254569, 'epoch': 0.43} + 43%|████▎ | 222/520 [13:48<18:08, 3.65s/it] 43%|████▎ | 223/520 [13:51<18:02, 3.64s/it] {'loss': 1.1956, 'grad_norm': 0.0011472117502028647, 'learning_rate': 0.03192088778562124, 'epoch': 0.43} + 43%|████▎ | 223/520 [13:51<18:02, 3.64s/it] 43%|████▎ | 224/520 [13:55<17:58, 3.64s/it] {'loss': 1.1869, 'grad_norm': 0.0010247019454351753, 'learning_rate': 0.03177101170357513, 'epoch': 0.43} + 43%|████▎ | 224/520 [13:55<17:58, 3.64s/it] 43%|████▎ | 225/520 [13:58<17:54, 3.64s/it] {'loss': 1.1903, 'grad_norm': 0.0012326416106951225, 'learning_rate': 0.03162087253971318, 'epoch': 0.43} + 43%|████▎ | 225/520 [13:58<17:54, 3.64s/it] 43%|████▎ | 226/520 [14:02<17:59, 3.67s/it] {'loss': 1.2961, 'grad_norm': 0.0011583439739273008, 'learning_rate': 0.03147047612756302, 'epoch': 0.43} + 43%|████▎ | 226/520 [14:02<17:59, 3.67s/it] 44%|████▎ | 227/520 [14:06<17:57, 3.68s/it] {'loss': 1.271, 'grad_norm': 0.0011781893929706003, 'learning_rate': 0.03131982831064744, 'epoch': 0.44} + 44%|████▎ | 227/520 [14:06<17:57, 3.68s/it] 44%|████▍ | 228/520 [14:10<17:50, 3.67s/it] {'loss': 1.2377, 'grad_norm': 0.0011181035562639, 'learning_rate': 0.031168934942257337, 'epoch': 0.44} + 44%|████▍ | 228/520 [14:10<17:50, 3.67s/it] 44%|████▍ | 229/520 [14:13<17:45, 3.66s/it] {'loss': 1.2488, 'grad_norm': 0.001081970490050859, 'learning_rate': 0.03101780188522433, 'epoch': 0.44} + 44%|████▍ | 229/520 [14:13<17:45, 3.66s/it] 44%|████▍ | 230/520 [14:17<17:40, 3.66s/it] {'loss': 1.1442, 'grad_norm': 0.0011123502506865514, 'learning_rate': 0.030866435011692884, 'epoch': 0.44} + 44%|████▍ | 230/520 [14:17<17:40, 3.66s/it] 44%|████▍ | 231/520 [14:21<17:36, 3.66s/it] {'loss': 1.2146, 'grad_norm': 0.0011080213756691122, 'learning_rate': 0.03071484020289224, 'epoch': 0.44} + 44%|████▍ | 231/520 [14:21<17:36, 3.66s/it] 45%|████▍ | 232/520 [14:24<17:32, 3.65s/it] {'loss': 1.2788, 'grad_norm': 0.0011286889474519419, 'learning_rate': 0.03056302334890786, 'epoch': 0.45} + 45%|████▍ | 232/520 [14:24<17:32, 3.65s/it] 45%|████▍ | 233/520 [14:28<17:27, 3.65s/it] {'loss': 1.1763, 'grad_norm': 0.0012346035120945824, 'learning_rate': 0.03041099034845257, 'epoch': 0.45} + 45%|████▍ | 233/520 [14:28<17:27, 3.65s/it] 45%|████▌ | 234/520 [14:31<17:23, 3.65s/it] {'loss': 1.166, 'grad_norm': 0.0012345204381198072, 'learning_rate': 0.030258747108637397, 'epoch': 0.45} + 45%|████▌ | 234/520 [14:31<17:23, 3.65s/it] 45%|████▌ | 235/520 [14:35<17:19, 3.65s/it] {'loss': 1.2217, 'grad_norm': 0.0012218064269946041, 'learning_rate': 0.03010629954474201, 'epoch': 0.45} + 45%|████▌ | 235/520 [14:35<17:19, 3.65s/it] 45%|████▌ | 236/520 [14:39<17:14, 3.64s/it] {'loss': 1.2811, 'grad_norm': 0.0011525154602254981, 'learning_rate': 0.029953653579984942, 'epoch': 0.45} + 45%|████▌ | 236/520 [14:39<17:14, 3.64s/it] 46%|████▌ | 237/520 [14:42<17:13, 3.65s/it] {'loss': 1.2798, 'grad_norm': 0.0011546307586667057, 'learning_rate': 0.02980081514529341, 'epoch': 0.46} + 46%|████▌ | 237/520 [14:42<17:13, 3.65s/it] 46%|████▌ | 238/520 [14:46<17:10, 3.65s/it] {'loss': 1.2174, 'grad_norm': 0.0012364328106963295, 'learning_rate': 0.02964779017907287, 'epoch': 0.46} + 46%|████▌ | 238/520 [14:46<17:10, 3.65s/it] 46%|████▌ | 239/520 [14:50<17:07, 3.65s/it] {'loss': 1.278, 'grad_norm': 0.0011699541946189756, 'learning_rate': 0.02949458462697632, 'epoch': 0.46} + 46%|████▌ | 239/520 [14:50<17:07, 3.65s/it] 46%|████▌ | 240/520 [14:53<17:05, 3.66s/it] {'loss': 1.1177, 'grad_norm': 0.0011218514071010392, 'learning_rate': 0.029341204441673263, 'epoch': 0.46} + 46%|████▌ | 240/520 [14:53<17:05, 3.66s/it] 46%|████▋ | 241/520 [14:57<17:02, 3.66s/it] {'loss': 1.197, 'grad_norm': 0.0012144642649703041, 'learning_rate': 0.02918765558261841, 'epoch': 0.46} + 46%|████▋ | 241/520 [14:57<17:02, 3.66s/it] 47%|████▋ | 242/520 [15:01<16:59, 3.67s/it] {'loss': 1.1945, 'grad_norm': 0.0011212481184333539, 'learning_rate': 0.02903394401582017, 'epoch': 0.47} + 47%|████▋ | 242/520 [15:01<16:59, 3.67s/it] 47%|████▋ | 243/520 [15:04<16:57, 3.67s/it] {'loss': 1.1949, 'grad_norm': 0.0011683590747989935, 'learning_rate': 0.028880075713608788, 'epoch': 0.47} + 47%|████▋ | 243/520 [15:04<16:57, 3.67s/it] 47%|████▋ | 244/520 [15:08<16:55, 3.68s/it] {'loss': 1.2813, 'grad_norm': 0.0011355174564800956, 'learning_rate': 0.02872605665440436, 'epoch': 0.47} + 47%|████▋ | 244/520 [15:08<16:55, 3.68s/it] 47%|████▋ | 245/520 [15:12<16:51, 3.68s/it] {'loss': 1.1781, 'grad_norm': 0.00127223680826851, 'learning_rate': 0.028571892822484502, 'epoch': 0.47} + 47%|████▋ | 245/520 [15:12<16:51, 3.68s/it] 47%|████▋ | 246/520 [15:15<16:49, 3.68s/it] {'loss': 1.2706, 'grad_norm': 0.0011733325248209031, 'learning_rate': 0.028417590207751836, 'epoch': 0.47} + 47%|████▋ | 246/520 [15:15<16:49, 3.68s/it] 48%|████▊ | 247/520 [15:19<16:46, 3.69s/it] {'loss': 1.3463, 'grad_norm': 0.0012188329789452584, 'learning_rate': 0.028263154805501296, 'epoch': 0.47} + 48%|████▊ | 247/520 [15:19<16:46, 3.69s/it] 48%|████▊ | 248/520 [15:23<16:43, 3.69s/it] {'loss': 1.1661, 'grad_norm': 0.0012372292069291787, 'learning_rate': 0.028108592616187134, 'epoch': 0.48} + 48%|████▊ | 248/520 [15:23<16:43, 3.69s/it] 48%|████▊ | 249/520 [15:27<16:37, 3.68s/it] {'loss': 1.2609, 'grad_norm': 0.0012243928330686207, 'learning_rate': 0.027953909645189824, 'epoch': 0.48} + 48%|████▊ | 249/520 [15:27<16:37, 3.68s/it] 48%|████▊ | 250/520 [15:30<16:33, 3.68s/it] {'loss': 1.2169, 'grad_norm': 0.0013215735192743999, 'learning_rate': 0.027799111902582697, 'epoch': 0.48} + 48%|████▊ | 250/520 [15:30<16:33, 3.68s/it] 48%|████▊ | 251/520 [15:34<16:28, 3.67s/it] {'loss': 1.2747, 'grad_norm': 0.0011477248077837185, 'learning_rate': 0.02764420540289845, 'epoch': 0.48} + 48%|████▊ | 251/520 [15:34<16:28, 3.67s/it] 48%|████▊ | 252/520 [15:38<16:25, 3.68s/it] {'loss': 1.18, 'grad_norm': 0.00111534233190339, 'learning_rate': 0.02748919616489542, 'epoch': 0.48} + 48%|████▊ | 252/520 [15:38<16:25, 3.68s/it] 49%|████▊ | 253/520 [15:41<16:22, 3.68s/it] {'loss': 1.2472, 'grad_norm': 0.0012860691213500048, 'learning_rate': 0.027334090211323766, 'epoch': 0.49} + 49%|████▊ | 253/520 [15:41<16:22, 3.68s/it] 49%|████▉ | 254/520 [15:45<16:16, 3.67s/it] {'loss': 1.2003, 'grad_norm': 0.0011826563967914786, 'learning_rate': 0.02717889356869146, 'epoch': 0.49} + 49%|████▉ | 254/520 [15:45<16:16, 3.67s/it] 49%|████▉ | 255/520 [15:49<16:12, 3.67s/it] {'loss': 1.2034, 'grad_norm': 0.0012693812452884833, 'learning_rate': 0.027023612267030084, 'epoch': 0.49} + 49%|████▉ | 255/520 [15:49<16:12, 3.67s/it] 49%|████▉ | 256/520 [15:52<16:09, 3.67s/it] {'loss': 1.267, 'grad_norm': 0.0012839019237175233, 'learning_rate': 0.02686825233966061, 'epoch': 0.49} + 49%|████▉ | 256/520 [15:52<16:09, 3.67s/it] 49%|████▉ | 257/520 [15:56<16:06, 3.67s/it] {'loss': 1.2439, 'grad_norm': 0.0012236633459575409, 'learning_rate': 0.02671281982295892, 'epoch': 0.49} + 49%|████▉ | 257/520 [15:56<16:06, 3.67s/it] 50%|████▉ | 258/520 [16:00<16:02, 3.67s/it] {'loss': 1.2433, 'grad_norm': 0.001028554677829021, 'learning_rate': 0.026557320756121307, 'epoch': 0.5} + 50%|████▉ | 258/520 [16:00<16:02, 3.67s/it] 50%|████▉ | 259/520 [16:03<15:58, 3.67s/it] {'loss': 1.3164, 'grad_norm': 0.0013309069758384005, 'learning_rate': 0.026401761180929795, 'epoch': 0.5} + 50%|████▉ | 259/520 [16:03<15:58, 3.67s/it] 50%|█████ | 260/520 [16:07<15:54, 3.67s/it] {'loss': 1.2233, 'grad_norm': 0.0009401570728703322, 'learning_rate': 0.02624614714151743, 'epoch': 0.5} + 50%|█████ | 260/520 [16:07<15:54, 3.67s/it] 50%|█████ | 261/520 [16:11<15:51, 3.67s/it] {'loss': 1.1722, 'grad_norm': 0.0011213041606284396, 'learning_rate': 0.026090484684133403, 'epoch': 0.5} + 50%|█████ | 261/520 [16:11<15:51, 3.67s/it] 50%|█████ | 262/520 [16:14<15:47, 3.67s/it] {'loss': 1.1916, 'grad_norm': 0.0012410806422166862, 'learning_rate': 0.025934779856908147, 'epoch': 0.5} + 50%|█████ | 262/520 [16:14<15:47, 3.67s/it] 51%|█████ | 263/520 [16:18<15:43, 3.67s/it] {'loss': 1.1995, 'grad_norm': 0.0011599952169027532, 'learning_rate': 0.02577903870961833, 'epoch': 0.51} + 51%|█████ | 263/520 [16:18<15:43, 3.67s/it] 51%|█████ | 264/520 [16:22<15:42, 3.68s/it] {'loss': 1.2685, 'grad_norm': 0.001134083763577374, 'learning_rate': 0.025623267293451824, 'epoch': 0.51} + 51%|█████ | 264/520 [16:22<15:42, 3.68s/it] 51%|█████ | 265/520 [16:25<15:39, 3.68s/it] {'loss': 1.2079, 'grad_norm': 0.0013566469447486762, 'learning_rate': 0.025467471660772557, 'epoch': 0.51} + 51%|█████ | 265/520 [16:25<15:39, 3.68s/it] 51%|█████ | 266/520 [16:29<15:33, 3.68s/it] {'loss': 1.0831, 'grad_norm': 0.001075576941280947, 'learning_rate': 0.02531165786488538, 'epoch': 0.51} + 51%|█████ | 266/520 [16:29<15:33, 3.68s/it] 51%|█████▏ | 267/520 [16:33<15:30, 3.68s/it] {'loss': 1.1971, 'grad_norm': 0.0011828927005603721, 'learning_rate': 0.02515583195980084, 'epoch': 0.51} + 51%|█████▏ | 267/520 [16:33<15:30, 3.68s/it] 52%|█████▏ | 268/520 [16:36<15:27, 3.68s/it] {'loss': 1.2959, 'grad_norm': 0.001175307169790532, 'learning_rate': 0.025, 'epoch': 0.52} + 52%|█████▏ | 268/520 [16:36<15:27, 3.68s/it] 52%|█████▏ | 269/520 [16:40<15:22, 3.68s/it] {'loss': 1.2979, 'grad_norm': 0.0012824628497736757, 'learning_rate': 0.024844168040199163, 'epoch': 0.52} + 52%|█████▏ | 269/520 [16:40<15:22, 3.68s/it] 52%|█████▏ | 270/520 [16:44<15:20, 3.68s/it] {'loss': 1.1362, 'grad_norm': 0.001145661842325793, 'learning_rate': 0.024688342135114626, 'epoch': 0.52} + 52%|█████▏ | 270/520 [16:44<15:20, 3.68s/it] 52%|█████▏ | 271/520 [16:47<15:16, 3.68s/it] {'loss': 1.274, 'grad_norm': 0.001259041133554763, 'learning_rate': 0.02453252833922745, 'epoch': 0.52} + 52%|█████▏ | 271/520 [16:47<15:16, 3.68s/it] 52%|█████▏ | 272/520 [16:51<15:12, 3.68s/it] {'loss': 1.1595, 'grad_norm': 0.001157088393146062, 'learning_rate': 0.024376732706548183, 'epoch': 0.52} + 52%|█████▏ | 272/520 [16:51<15:12, 3.68s/it] 52%|█████▎ | 273/520 [16:55<15:08, 3.68s/it] {'loss': 1.2482, 'grad_norm': 0.001129047244296757, 'learning_rate': 0.024220961290381676, 'epoch': 0.53} + 52%|█████▎ | 273/520 [16:55<15:08, 3.68s/it] 53%|█████▎ | 274/520 [16:58<15:03, 3.67s/it] {'loss': 1.2722, 'grad_norm': 0.001349686067521879, 'learning_rate': 0.024065220143091863, 'epoch': 0.53} + 53%|█████▎ | 274/520 [16:58<15:03, 3.67s/it] 53%|█████▎ | 275/520 [17:02<14:59, 3.67s/it] {'loss': 1.2041, 'grad_norm': 0.0012475967352081803, 'learning_rate': 0.023909515315866603, 'epoch': 0.53} + 53%|█████▎ | 275/520 [17:02<14:59, 3.67s/it] 53%|█████▎ | 276/520 [17:06<14:56, 3.67s/it] {'loss': 1.2598, 'grad_norm': 0.0013128891705927299, 'learning_rate': 0.02375385285848257, 'epoch': 0.53} + 53%|█████▎ | 276/520 [17:06<14:56, 3.67s/it] 53%|█████▎ | 277/520 [17:09<14:53, 3.68s/it] {'loss': 1.2655, 'grad_norm': 0.0011124101115142634, 'learning_rate': 0.023598238819070205, 'epoch': 0.53} + 53%|█████▎ | 277/520 [17:09<14:53, 3.68s/it] 53%|█████▎ | 278/520 [17:13<15:03, 3.73s/it] {'loss': 1.1715, 'grad_norm': 0.0011465479325843413, 'learning_rate': 0.0234426792438787, 'epoch': 0.53} + 53%|█████▎ | 278/520 [17:13<15:03, 3.73s/it] 54%|█████▎ | 279/520 [17:17<15:07, 3.76s/it] {'loss': 1.1475, 'grad_norm': 0.0012707676663334302, 'learning_rate': 0.023287180177041084, 'epoch': 0.54} + 54%|█████▎ | 279/520 [17:17<15:07, 3.76s/it] 54%|█████▍ | 280/520 [17:21<15:11, 3.80s/it] {'loss': 1.1987, 'grad_norm': 0.0012881627868107175, 'learning_rate': 0.023131747660339395, 'epoch': 0.54} + 54%|█████▍ | 280/520 [17:21<15:11, 3.80s/it] 54%|█████▍ | 281/520 [17:25<15:12, 3.82s/it] {'loss': 1.2979, 'grad_norm': 0.0012728954470959664, 'learning_rate': 0.02297638773296992, 'epoch': 0.54} + 54%|█████▍ | 281/520 [17:25<15:12, 3.82s/it] 54%|█████▍ | 282/520 [17:29<15:13, 3.84s/it] {'loss': 1.1801, 'grad_norm': 0.0012030568707613698, 'learning_rate': 0.022821106431308545, 'epoch': 0.54} + 54%|█████▍ | 282/520 [17:29<15:13, 3.84s/it] 54%|█████▍ | 283/520 [17:33<15:07, 3.83s/it] {'loss': 1.3034, 'grad_norm': 0.0013039567528634803, 'learning_rate': 0.022665909788676237, 'epoch': 0.54} + 54%|█████▍ | 283/520 [17:33<15:07, 3.83s/it] 55%|█████▍ | 284/520 [17:36<14:54, 3.79s/it] {'loss': 1.1587, 'grad_norm': 0.0012658114221119933, 'learning_rate': 0.02251080383510459, 'epoch': 0.55} + 55%|█████▍ | 284/520 [17:36<14:54, 3.79s/it] 55%|█████▍ | 285/520 [17:40<14:54, 3.81s/it] {'loss': 1.2021, 'grad_norm': 0.0012273422203669614, 'learning_rate': 0.02235579459710156, 'epoch': 0.55} + 55%|█████▍ | 285/520 [17:40<14:54, 3.81s/it] 55%|█████▌ | 286/520 [17:44<14:54, 3.82s/it] {'loss': 1.0844, 'grad_norm': 0.001225202886323648, 'learning_rate': 0.022200888097417305, 'epoch': 0.55} + 55%|█████▌ | 286/520 [17:44<14:54, 3.82s/it] 55%|█████▌ | 287/520 [17:48<14:53, 3.83s/it] {'loss': 1.3098, 'grad_norm': 0.0011936154602918173, 'learning_rate': 0.022046090354810182, 'epoch': 0.55} + 55%|█████▌ | 287/520 [17:48<14:53, 3.83s/it] 55%|█████▌ | 288/520 [17:52<14:52, 3.84s/it] {'loss': 1.3289, 'grad_norm': 0.0011135000664307107, 'learning_rate': 0.02189140738381288, 'epoch': 0.55} + 55%|█████▌ | 288/520 [17:52<14:52, 3.84s/it] 56%|█████▌ | 289/520 [17:56<14:49, 3.85s/it] {'loss': 1.2257, 'grad_norm': 0.0011701326268426088, 'learning_rate': 0.021736845194498717, 'epoch': 0.56} + 56%|█████▌ | 289/520 [17:56<14:49, 3.85s/it] 56%|█████▌ | 290/520 [17:59<14:34, 3.80s/it] {'loss': 1.145, 'grad_norm': 0.0011211848592975826, 'learning_rate': 0.02158240979224817, 'epoch': 0.56} + 56%|█████▌ | 290/520 [17:59<14:34, 3.80s/it] 56%|█████▌ | 291/520 [18:03<14:22, 3.77s/it] {'loss': 1.1755, 'grad_norm': 0.0011908944061683419, 'learning_rate': 0.021428107177515504, 'epoch': 0.56} + 56%|█████▌ | 291/520 [18:03<14:22, 3.77s/it] 56%|█████▌ | 292/520 [18:07<14:13, 3.74s/it] {'loss': 1.2354, 'grad_norm': 0.001183897342128125, 'learning_rate': 0.02127394334559564, 'epoch': 0.56} + 56%|█████▌ | 292/520 [18:07<14:13, 3.74s/it] 56%|█████▋ | 293/520 [18:10<14:05, 3.72s/it] {'loss': 1.1904, 'grad_norm': 0.0012836286701773385, 'learning_rate': 0.02111992428639121, 'epoch': 0.56} + 56%|█████▋ | 293/520 [18:10<14:05, 3.72s/it] 57%|█████▋ | 294/520 [18:14<13:58, 3.71s/it] {'loss': 1.2108, 'grad_norm': 0.001351191825254117, 'learning_rate': 0.020966055984179832, 'epoch': 0.57} + 57%|█████▋ | 294/520 [18:14<13:58, 3.71s/it] 57%|█████▋ | 295/520 [18:18<13:50, 3.69s/it] {'loss': 1.1752, 'grad_norm': 0.0010572270258436467, 'learning_rate': 0.020812344417381595, 'epoch': 0.57} + 57%|█████▋ | 295/520 [18:18<13:50, 3.69s/it] 57%|█████▋ | 296/520 [18:21<13:45, 3.69s/it] {'loss': 1.1608, 'grad_norm': 0.0012651035089910091, 'learning_rate': 0.020658795558326743, 'epoch': 0.57} + 57%|█████▋ | 296/520 [18:21<13:45, 3.69s/it] 57%|█████▋ | 297/520 [18:25<13:40, 3.68s/it] {'loss': 1.2924, 'grad_norm': 0.0013323323341963244, 'learning_rate': 0.020505415373023684, 'epoch': 0.57} + 57%|█████▋ | 297/520 [18:25<13:40, 3.68s/it] 57%|█████▋ | 298/520 [18:29<13:37, 3.68s/it] {'loss': 1.2457, 'grad_norm': 0.0011447921757343643, 'learning_rate': 0.02035220982092714, 'epoch': 0.57} + 57%|█████▋ | 298/520 [18:29<13:37, 3.68s/it] 57%|█████▊ | 299/520 [18:32<13:32, 3.68s/it] {'loss': 1.2377, 'grad_norm': 0.0011486314545754602, 'learning_rate': 0.0201991848547066, 'epoch': 0.57} + 57%|█████▊ | 299/520 [18:32<13:32, 3.68s/it] 58%|█████▊ | 300/520 [18:36<13:29, 3.68s/it] {'loss': 1.2933, 'grad_norm': 0.0012104899137671617, 'learning_rate': 0.020046346420015067, 'epoch': 0.58} + 58%|█████▊ | 300/520 [18:36<13:29, 3.68s/it] 58%|█████▊ | 301/520 [18:40<13:24, 3.67s/it] {'loss': 1.2911, 'grad_norm': 0.0012207110017398762, 'learning_rate': 0.019893700455257998, 'epoch': 0.58} + 58%|█████▊ | 301/520 [18:40<13:24, 3.67s/it] 58%|█████▊ | 302/520 [18:43<13:21, 3.68s/it] {'loss': 1.2281, 'grad_norm': 0.0011359226110616964, 'learning_rate': 0.01974125289136261, 'epoch': 0.58} + 58%|█████▊ | 302/520 [18:43<13:21, 3.68s/it] 58%|█████▊ | 303/520 [18:47<13:17, 3.67s/it] {'loss': 1.2013, 'grad_norm': 0.001338020137687674, 'learning_rate': 0.01958900965154743, 'epoch': 0.58} + 58%|█████▊ | 303/520 [18:47<13:17, 3.67s/it] 58%|█████▊ | 304/520 [18:51<13:18, 3.70s/it] {'loss': 1.1507, 'grad_norm': 0.001200298343405803, 'learning_rate': 0.019436976651092144, 'epoch': 0.58} + 58%|█████▊ | 304/520 [18:51<13:18, 3.70s/it] 59%|█████▊ | 305/520 [18:54<13:13, 3.69s/it] {'loss': 1.3024, 'grad_norm': 0.001341700455857684, 'learning_rate': 0.019285159797107765, 'epoch': 0.59} + 59%|█████▊ | 305/520 [18:54<13:13, 3.69s/it] 59%|█████▉ | 306/520 [18:58<13:09, 3.69s/it] {'loss': 1.2518, 'grad_norm': 0.001242013840989776, 'learning_rate': 0.019133564988307126, 'epoch': 0.59} + 59%|█████▉ | 306/520 [18:58<13:09, 3.69s/it] 59%|█████▉ | 307/520 [19:02<13:05, 3.69s/it] {'loss': 1.1908, 'grad_norm': 0.0011610951577877744, 'learning_rate': 0.01898219811477568, 'epoch': 0.59} + 59%|█████▉ | 307/520 [19:02<13:05, 3.69s/it] 59%|█████▉ | 308/520 [19:06<13:01, 3.68s/it] {'loss': 1.3114, 'grad_norm': 0.0011367715752735657, 'learning_rate': 0.01883106505774266, 'epoch': 0.59} + 59%|█████▉ | 308/520 [19:06<13:01, 3.68s/it] 59%|█████▉ | 309/520 [19:10<13:17, 3.78s/it] {'loss': 1.206, 'grad_norm': 0.0011623084936764091, 'learning_rate': 0.01868017168935256, 'epoch': 0.59} + 59%|█████▉ | 309/520 [19:10<13:17, 3.78s/it] 60%|█████▉ | 310/520 [19:13<13:06, 3.74s/it] {'loss': 1.1809, 'grad_norm': 0.0011972804327355369, 'learning_rate': 0.01852952387243698, 'epoch': 0.6} + 60%|█████▉ | 310/520 [19:13<13:06, 3.74s/it] 60%|█████▉ | 311/520 [19:17<12:59, 3.73s/it] {'loss': 1.1796, 'grad_norm': 0.0012094559297135307, 'learning_rate': 0.018379127460286818, 'epoch': 0.6} + 60%|█████▉ | 311/520 [19:17<12:59, 3.73s/it] 60%|██████ | 312/520 [19:21<12:52, 3.71s/it] {'loss': 1.1527, 'grad_norm': 0.0012424042683177013, 'learning_rate': 0.018228988296424876, 'epoch': 0.6} + 60%|██████ | 312/520 [19:21<12:52, 3.71s/it] 60%|██████ | 313/520 [19:24<12:48, 3.71s/it] {'loss': 1.1262, 'grad_norm': 0.0011129566546823318, 'learning_rate': 0.01807911221437877, 'epoch': 0.6} + 60%|██████ | 313/520 [19:24<12:48, 3.71s/it] 60%|██████ | 314/520 [19:28<13:11, 3.84s/it] {'loss': 1.1754, 'grad_norm': 0.0011820410464464125, 'learning_rate': 0.017929505037454312, 'epoch': 0.6} + 60%|██████ | 314/520 [19:28<13:11, 3.84s/it] 61%|██████ | 315/520 [19:32<13:00, 3.81s/it] {'loss': 1.1964, 'grad_norm': 0.0012591889837200948, 'learning_rate': 0.017780172578509258, 'epoch': 0.61} + 61%|██████ | 315/520 [19:32<13:00, 3.81s/it] 61%|██████ | 316/520 [19:36<13:17, 3.91s/it] {'loss': 1.1642, 'grad_norm': 0.0012825058762822574, 'learning_rate': 0.017631120639727393, 'epoch': 0.61} + 61%|██████ | 316/520 [19:36<13:17, 3.91s/it] 61%|██████ | 317/520 [19:40<12:58, 3.84s/it] {'loss': 1.1574, 'grad_norm': 0.0011182307651509266, 'learning_rate': 0.017482355012393175, 'epoch': 0.61} + 61%|██████ | 317/520 [19:40<12:58, 3.84s/it] 61%|██████ | 318/520 [19:44<12:47, 3.80s/it] {'loss': 1.273, 'grad_norm': 0.0012775553095494164, 'learning_rate': 0.017333881476666646, 'epoch': 0.61} + 61%|██████ | 318/520 [19:44<12:47, 3.80s/it] 61%|██████▏ | 319/520 [19:48<13:01, 3.89s/it] {'loss': 1.1535, 'grad_norm': 0.0010953725319195, 'learning_rate': 0.01718570580135889, 'epoch': 0.61} + 61%|██████▏ | 319/520 [19:48<13:01, 3.89s/it] 62%|██████▏ | 320/520 [19:51<12:45, 3.83s/it] {'loss': 1.089, 'grad_norm': 0.0012219468700343933, 'learning_rate': 0.017037833743707893, 'epoch': 0.62} + 62%|██████▏ | 320/520 [19:51<12:45, 3.83s/it] 62%|██████▏ | 321/520 [19:55<12:31, 3.78s/it] {'loss': 1.2947, 'grad_norm': 0.001197770976707815, 'learning_rate': 0.016890271049154828, 'epoch': 0.62} + 62%|██████▏ | 321/520 [19:55<12:31, 3.78s/it] 62%|██████▏ | 322/520 [19:59<12:24, 3.76s/it] {'loss': 1.0969, 'grad_norm': 0.0011612763294477665, 'learning_rate': 0.01674302345112083, 'epoch': 0.62} + 62%|██████▏ | 322/520 [19:59<12:24, 3.76s/it] 62%|██████▏ | 323/520 [20:03<12:17, 3.74s/it] {'loss': 1.1761, 'grad_norm': 0.001197087116252218, 'learning_rate': 0.016596096670784235, 'epoch': 0.62} + 62%|██████▏ | 323/520 [20:03<12:17, 3.74s/it] 62%|██████▏ | 324/520 [20:06<12:10, 3.73s/it] {'loss': 1.2391, 'grad_norm': 0.0012550202324498848, 'learning_rate': 0.016449496416858285, 'epoch': 0.62} + 62%|██████▏ | 324/520 [20:06<12:10, 3.73s/it] 62%|██████▎ | 325/520 [20:10<12:03, 3.71s/it] {'loss': 1.2296, 'grad_norm': 0.001304400424933848, 'learning_rate': 0.01630322838536932, 'epoch': 0.62} + 62%|██████▎ | 325/520 [20:10<12:03, 3.71s/it] 63%|██████▎ | 326/520 [20:14<11:58, 3.70s/it] {'loss': 1.233, 'grad_norm': 0.0013029421355094366, 'learning_rate': 0.016157298259435467, 'epoch': 0.63} + 63%|██████▎ | 326/520 [20:14<11:58, 3.70s/it] 63%|██████▎ | 327/520 [20:17<11:52, 3.69s/it] {'loss': 1.2028, 'grad_norm': 0.0012349224870960445, 'learning_rate': 0.016011711709045813, 'epoch': 0.63} + 63%|██████▎ | 327/520 [20:17<11:52, 3.69s/it] 63%|██████▎ | 328/520 [20:21<11:47, 3.68s/it] {'loss': 1.2709, 'grad_norm': 0.0012809048391838158, 'learning_rate': 0.015866474390840126, 'epoch': 0.63} + 63%|██████▎ | 328/520 [20:21<11:47, 3.68s/it] 63%|██████▎ | 329/520 [20:25<11:50, 3.72s/it] {'loss': 1.1562, 'grad_norm': 0.001111657152949265, 'learning_rate': 0.015721591947889052, 'epoch': 0.63} + 63%|██████▎ | 329/520 [20:25<11:50, 3.72s/it] 63%|██████▎ | 330/520 [20:28<11:43, 3.70s/it] {'loss': 1.2374, 'grad_norm': 0.0011719361107123323, 'learning_rate': 0.01557707000947487, 'epoch': 0.63} + 63%|██████▎ | 330/520 [20:28<11:43, 3.70s/it] 64%|██████▎ | 331/520 [20:32<11:38, 3.69s/it] {'loss': 1.19, 'grad_norm': 0.0012974479769271307, 'learning_rate': 0.015432914190872758, 'epoch': 0.64} + 64%|██████▎ | 331/520 [20:32<11:38, 3.69s/it] 64%|██████▍ | 332/520 [20:36<11:38, 3.71s/it] {'loss': 1.2277, 'grad_norm': 0.0011131655875058699, 'learning_rate': 0.015289130093132633, 'epoch': 0.64} + 64%|██████▍ | 332/520 [20:36<11:38, 3.71s/it] 64%|██████▍ | 333/520 [20:39<11:33, 3.71s/it] {'loss': 1.3236, 'grad_norm': 0.0013126326714862455, 'learning_rate': 0.01514572330286152, 'epoch': 0.64} + 64%|██████▍ | 333/520 [20:39<11:33, 3.71s/it] 64%|██████▍ | 334/520 [20:43<11:27, 3.70s/it] {'loss': 1.2411, 'grad_norm': 0.001287878542158269, 'learning_rate': 0.01500269939200648, 'epoch': 0.64} + 64%|██████▍ | 334/520 [20:43<11:27, 3.70s/it] 64%|██████▍ | 335/520 [20:47<11:22, 3.69s/it] {'loss': 1.2391, 'grad_norm': 0.0011825664157503388, 'learning_rate': 0.014860063917638128, 'epoch': 0.64} + 64%|██████▍ | 335/520 [20:47<11:22, 3.69s/it] 65%|██████▍ | 336/520 [20:50<11:17, 3.68s/it] {'loss': 1.1495, 'grad_norm': 0.0013729062985991546, 'learning_rate': 0.014717822421734717, 'epoch': 0.65} + 65%|██████▍ | 336/520 [20:51<11:17, 3.68s/it] 65%|██████▍ | 337/520 [20:54<11:14, 3.68s/it] {'loss': 1.1408, 'grad_norm': 0.0012855533244137884, 'learning_rate': 0.014575980430966806, 'epoch': 0.65} + 65%|██████▍ | 337/520 [20:54<11:14, 3.68s/it] 65%|██████▌ | 338/520 [20:58<11:09, 3.68s/it] {'loss': 1.2484, 'grad_norm': 0.0012479916453413775, 'learning_rate': 0.01443454345648252, 'epoch': 0.65} + 65%|██████▌ | 338/520 [20:58<11:09, 3.68s/it] 65%|██████▌ | 339/520 [21:02<11:06, 3.68s/it] {'loss': 1.186, 'grad_norm': 0.0013107545269988347, 'learning_rate': 0.014293516993693429, 'epoch': 0.65} + 65%|██████▌ | 339/520 [21:02<11:06, 3.68s/it] 65%|██████▌ | 340/520 [21:05<11:04, 3.69s/it] {'loss': 1.1755, 'grad_norm': 0.0012290937259259946, 'learning_rate': 0.014152906522061049, 'epoch': 0.65} + 65%|██████▌ | 340/520 [21:05<11:04, 3.69s/it] 66%|██████▌ | 341/520 [21:09<10:57, 3.68s/it] {'loss': 1.2096, 'grad_norm': 0.0012926570806238295, 'learning_rate': 0.014012717504883873, 'epoch': 0.66} + 66%|██████▌ | 341/520 [21:09<10:57, 3.68s/it] 66%|██████▌ | 342/520 [21:13<10:54, 3.68s/it] {'loss': 1.2147, 'grad_norm': 0.0013825123570147375, 'learning_rate': 0.01387295538908519, 'epoch': 0.66} + 66%|██████▌ | 342/520 [21:13<10:54, 3.68s/it] 66%|██████▌ | 343/520 [21:16<10:49, 3.67s/it] {'loss': 1.1518, 'grad_norm': 0.000978874696829203, 'learning_rate': 0.013733625605001366, 'epoch': 0.66} + 66%|██████▌ | 343/520 [21:16<10:49, 3.67s/it] 66%|██████▌ | 344/520 [21:20<10:45, 3.67s/it] {'loss': 1.1647, 'grad_norm': 0.0011651832118842154, 'learning_rate': 0.013594733566170926, 'epoch': 0.66} + 66%|██████▌ | 344/520 [21:20<10:45, 3.67s/it] 66%|██████▋ | 345/520 [21:24<10:40, 3.66s/it] {'loss': 1.2635, 'grad_norm': 0.0012789622406426757, 'learning_rate': 0.013456284669124158, 'epoch': 0.66} + 66%|██████▋ | 345/520 [21:24<10:40, 3.66s/it] 67%|██████▋ | 346/520 [21:27<10:37, 3.67s/it] {'loss': 1.1771, 'grad_norm': 0.0011482312666976006, 'learning_rate': 0.01331828429317345, 'epoch': 0.67} + 67%|██████▋ | 346/520 [21:27<10:37, 3.67s/it] 67%|██████▋ | 347/520 [21:31<10:32, 3.66s/it] {'loss': 1.1805, 'grad_norm': 0.0011307219599265745, 'learning_rate': 0.01318073780020433, 'epoch': 0.67} + 67%|██████▋ | 347/520 [21:31<10:32, 3.66s/it]Token indices sequence length is longer than the specified maximum sequence length for this model (2778 > 2048). Running this sequence through the model will result in indexing errors + 67%|██████▋ | 348/520 [21:35<10:30, 3.66s/it] {'loss': 1.1381, 'grad_norm': 0.0014448882366062048, 'learning_rate': 0.013043650534467053, 'epoch': 0.67} + 67%|██████▋ | 348/520 [21:35<10:30, 3.66s/it] 67%|██████▋ | 349/520 [21:38<10:25, 3.66s/it] {'loss': 1.1708, 'grad_norm': 0.0012740615334064667, 'learning_rate': 0.012907027822369006, 'epoch': 0.67} + 67%|██████▋ | 349/520 [21:38<10:25, 3.66s/it] 67%|██████▋ | 350/520 [21:42<10:22, 3.66s/it] {'loss': 1.2191, 'grad_norm': 0.0013055719919486128, 'learning_rate': 0.012770874972267776, 'epoch': 0.67} + 67%|██████▋ | 350/520 [21:42<10:22, 3.66s/it] 68%|██████▊ | 351/520 [21:46<10:20, 3.67s/it] {'loss': 1.132, 'grad_norm': 0.001206353486812291, 'learning_rate': 0.012635197274264815, 'epoch': 0.68} + 68%|██████▊ | 351/520 [21:46<10:20, 3.67s/it] 68%|██████▊ | 352/520 [21:49<10:15, 3.67s/it] {'loss': 1.2427, 'grad_norm': 0.0011610146830839615, 'learning_rate': 0.012500000000000006, 'epoch': 0.68} + 68%|██████▊ | 352/520 [21:49<10:15, 3.67s/it] 68%|██████▊ | 353/520 [21:53<10:14, 3.68s/it] {'loss': 1.1544, 'grad_norm': 0.0010065916978334456, 'learning_rate': 0.01236528840244674, 'epoch': 0.68} + 68%|██████▊ | 353/520 [21:53<10:14, 3.68s/it] 68%|██████▊ | 354/520 [21:57<10:09, 3.67s/it] {'loss': 1.2446, 'grad_norm': 0.0011052860634996758, 'learning_rate': 0.012231067715707866, 'epoch': 0.68} + 68%|██████▊ | 354/520 [21:57<10:09, 3.67s/it] 68%|██████▊ | 355/520 [22:00<10:04, 3.66s/it] {'loss': 1.1972, 'grad_norm': 0.00126369983561254, 'learning_rate': 0.012097343154812333, 'epoch': 0.68} + 68%|██████▊ | 355/520 [22:00<10:04, 3.66s/it] 68%|██████▊ | 356/520 [22:04<10:00, 3.66s/it] {'loss': 1.196, 'grad_norm': 0.0012828206086635883, 'learning_rate': 0.01196411991551255, 'epoch': 0.68} + 68%|██████▊ | 356/520 [22:04<10:00, 3.66s/it] 69%|██████▊ | 357/520 [22:08<09:56, 3.66s/it] {'loss': 1.2303, 'grad_norm': 0.0011812491366279278, 'learning_rate': 0.011831403174082482, 'epoch': 0.69} + 69%|██████▊ | 357/520 [22:08<09:56, 3.66s/it] 69%|██████▉ | 358/520 [22:11<09:53, 3.67s/it] {'loss': 1.1595, 'grad_norm': 0.0012373594740002025, 'learning_rate': 0.01169919808711659, 'epoch': 0.69} + 69%|██████▉ | 358/520 [22:11<09:53, 3.67s/it] 69%|██████▉ | 359/520 [22:15<09:50, 3.67s/it] {'loss': 1.1892, 'grad_norm': 0.0012199431292363523, 'learning_rate': 0.011567509791329401, 'epoch': 0.69} + 69%|██████▉ | 359/520 [22:15<09:50, 3.67s/it] 69%|██████▉ | 360/520 [22:19<09:45, 3.66s/it] {'loss': 1.1909, 'grad_norm': 0.0011768604208230306, 'learning_rate': 0.011436343403356017, 'epoch': 0.69} + 69%|██████▉ | 360/520 [22:19<09:45, 3.66s/it] 69%|██████▉ | 361/520 [22:22<09:42, 3.67s/it] {'loss': 1.2133, 'grad_norm': 0.0010614187727858247, 'learning_rate': 0.01130570401955322, 'epoch': 0.69} + 69%|██████▉ | 361/520 [22:22<09:42, 3.67s/it] 70%|██████▉ | 362/520 [22:26<09:38, 3.66s/it] {'loss': 1.2017, 'grad_norm': 0.0013579195473371033, 'learning_rate': 0.011175596715801515, 'epoch': 0.7} + 70%|██████▉ | 362/520 [22:26<09:38, 3.66s/it] 70%|██████▉ | 363/520 [22:29<09:33, 3.65s/it] {'loss': 1.2356, 'grad_norm': 0.00121200867759868, 'learning_rate': 0.011046026547307906, 'epoch': 0.7} + 70%|██████▉ | 363/520 [22:29<09:33, 3.65s/it] 70%|███████ | 364/520 [22:33<09:31, 3.67s/it] {'loss': 1.222, 'grad_norm': 0.0011946601550076538, 'learning_rate': 0.010916998548409449, 'epoch': 0.7} + 70%|███████ | 364/520 [22:33<09:31, 3.67s/it] 70%|███████ | 365/520 [22:37<09:27, 3.66s/it] {'loss': 1.2833, 'grad_norm': 0.0012506662572163618, 'learning_rate': 0.010788517732377696, 'epoch': 0.7} + 70%|███████ | 365/520 [22:37<09:27, 3.66s/it] 70%|███████ | 366/520 [22:40<09:23, 3.66s/it] {'loss': 1.2598, 'grad_norm': 0.0012151741244998597, 'learning_rate': 0.010660589091223855, 'epoch': 0.7} + 70%|███████ | 366/520 [22:40<09:23, 3.66s/it] 71%|███████ | 367/520 [22:44<09:23, 3.68s/it] {'loss': 1.2504, 'grad_norm': 0.0012675374608815042, 'learning_rate': 0.010533217595504858, 'epoch': 0.71} + 71%|███████ | 367/520 [22:44<09:23, 3.68s/it] 71%|███████ | 368/520 [22:48<09:19, 3.68s/it] {'loss': 1.0999, 'grad_norm': 0.0012648588362883224, 'learning_rate': 0.01040640819413026, 'epoch': 0.71} + 71%|███████ | 368/520 [22:48<09:19, 3.68s/it] 71%|███████ | 369/520 [22:52<09:14, 3.67s/it] {'loss': 1.1913, 'grad_norm': 0.0010906198324959339, 'learning_rate': 0.010280165814169885, 'epoch': 0.71} + 71%|███████ | 369/520 [22:52<09:14, 3.67s/it] 71%|███████ | 370/520 [22:55<09:09, 3.67s/it] {'loss': 1.1649, 'grad_norm': 0.0011920430654733369, 'learning_rate': 0.010154495360662465, 'epoch': 0.71} + 71%|███████ | 370/520 [22:55<09:09, 3.67s/it] 71%|███████▏ | 371/520 [22:59<09:05, 3.66s/it] {'loss': 1.1486, 'grad_norm': 0.0013239351892358016, 'learning_rate': 0.010029401716424994, 'epoch': 0.71} + 71%|███████▏ | 371/520 [22:59<09:05, 3.66s/it] 72%|███████▏ | 372/520 [23:03<09:02, 3.67s/it] {'loss': 1.2515, 'grad_norm': 0.0010506535795358203, 'learning_rate': 0.00990488974186306, 'epoch': 0.72} + 72%|███████▏ | 372/520 [23:03<09:02, 3.67s/it] 72%|███████▏ | 373/520 [23:06<08:58, 3.66s/it] {'loss': 1.1461, 'grad_norm': 0.0013064575536395114, 'learning_rate': 0.009780964274781985, 'epoch': 0.72} + 72%|███████▏ | 373/520 [23:06<08:58, 3.66s/it] 72%|███████▏ | 374/520 [23:10<08:54, 3.66s/it] {'loss': 1.2473, 'grad_norm': 0.001294657977639146, 'learning_rate': 0.009657630130198819, 'epoch': 0.72} + 72%|███████▏ | 374/520 [23:10<08:54, 3.66s/it] 72%|███████▏ | 375/520 [23:14<08:55, 3.69s/it] {'loss': 1.1652, 'grad_norm': 0.001282023066254918, 'learning_rate': 0.009534892100155297, 'epoch': 0.72} + 72%|███████▏ | 375/520 [23:14<08:55, 3.69s/it] 72%|███████▏ | 376/520 [23:17<08:50, 3.68s/it] {'loss': 1.2721, 'grad_norm': 0.0011930110283168826, 'learning_rate': 0.009412754953531664, 'epoch': 0.72} + 72%|███████▏ | 376/520 [23:17<08:50, 3.68s/it] 72%|███████▎ | 377/520 [23:21<08:45, 3.68s/it] {'loss': 1.2029, 'grad_norm': 0.0013374749741802948, 'learning_rate': 0.009291223435861319, 'epoch': 0.72} + 72%|███████▎ | 377/520 [23:21<08:45, 3.68s/it] 73%|███████▎ | 378/520 [23:25<08:43, 3.68s/it] {'loss': 1.2663, 'grad_norm': 0.0011680441581739661, 'learning_rate': 0.009170302269146507, 'epoch': 0.73} + 73%|███████▎ | 378/520 [23:25<08:43, 3.68s/it] 73%|███████▎ | 379/520 [23:28<08:38, 3.68s/it] {'loss': 1.2254, 'grad_norm': 0.0011401644383689591, 'learning_rate': 0.009049996151674788, 'epoch': 0.73} + 73%|███████▎ | 379/520 [23:28<08:38, 3.68s/it] 73%|███████▎ | 380/520 [23:32<08:34, 3.67s/it] {'loss': 1.2347, 'grad_norm': 0.0011998962828275398, 'learning_rate': 0.008930309757836517, 'epoch': 0.73} + 73%|███████▎ | 380/520 [23:32<08:34, 3.67s/it] 73%|███████▎ | 381/520 [23:36<08:29, 3.66s/it] {'loss': 1.2424, 'grad_norm': 0.001193116317298423, 'learning_rate': 0.008811247737943242, 'epoch': 0.73} + 73%|███████▎ | 381/520 [23:36<08:29, 3.66s/it] 73%|███████▎ | 382/520 [23:39<08:26, 3.67s/it] {'loss': 1.2092, 'grad_norm': 0.0010982199952817035, 'learning_rate': 0.00869281471804698, 'epoch': 0.73} + 73%|███████▎ | 382/520 [23:39<08:26, 3.67s/it] 74%|███████▎ | 383/520 [23:43<08:23, 3.68s/it] {'loss': 1.085, 'grad_norm': 0.0013562786359610206, 'learning_rate': 0.00857501529976049, 'epoch': 0.74} + 74%|███████▎ | 383/520 [23:43<08:23, 3.68s/it] 74%|███████▍ | 384/520 [23:47<08:23, 3.70s/it] {'loss': 1.217, 'grad_norm': 0.001019041600761386, 'learning_rate': 0.008457854060078521, 'epoch': 0.74} + 74%|███████▍ | 384/520 [23:47<08:23, 3.70s/it] 74%|███████▍ | 385/520 [23:51<08:26, 3.75s/it] {'loss': 1.2345, 'grad_norm': 0.001172801505031085, 'learning_rate': 0.008341335551199902, 'epoch': 0.74} + 74%|███████▍ | 385/520 [23:51<08:26, 3.75s/it] 74%|███████▍ | 386/520 [23:54<08:27, 3.79s/it] {'loss': 1.1781, 'grad_norm': 0.0010729566530116906, 'learning_rate': 0.008225464300350751, 'epoch': 0.74} + 74%|███████▍ | 386/520 [23:54<08:27, 3.79s/it] 74%|███████▍ | 387/520 [23:58<08:26, 3.81s/it] {'loss': 1.2491, 'grad_norm': 0.0011535414225006153, 'learning_rate': 0.008110244809608494, 'epoch': 0.74} + 74%|███████▍ | 387/520 [23:58<08:26, 3.81s/it] 75%|███████▍ | 388/520 [24:02<08:21, 3.80s/it] {'loss': 1.1426, 'grad_norm': 0.001192502853163647, 'learning_rate': 0.00799568155572701, 'epoch': 0.75} + 75%|███████▍ | 388/520 [24:02<08:21, 3.80s/it] 75%|███████▍ | 389/520 [24:06<08:11, 3.75s/it] {'loss': 1.1936, 'grad_norm': 0.0016334533314150128, 'learning_rate': 0.007881778989962663, 'epoch': 0.75} + 75%|███████▍ | 389/520 [24:06<08:11, 3.75s/it] 75%|███████▌ | 390/520 [24:09<08:03, 3.72s/it] {'loss': 1.2559, 'grad_norm': 0.0011614126005776166, 'learning_rate': 0.0077685415379013245, 'epoch': 0.75} + 75%|███████▌ | 390/520 [24:09<08:03, 3.72s/it] 75%|███████▌ | 391/520 [24:13<07:58, 3.71s/it] {'loss': 1.3101, 'grad_norm': 0.0012251148529552414, 'learning_rate': 0.007655973599286459, 'epoch': 0.75} + 75%|███████▌ | 391/520 [24:13<07:58, 3.71s/it] 75%|███████▌ | 392/520 [24:17<07:53, 3.70s/it] {'loss': 1.1419, 'grad_norm': 0.0011940757089515973, 'learning_rate': 0.007544079547848182, 'epoch': 0.75} + 75%|███████▌ | 392/520 [24:17<07:53, 3.70s/it] 76%|███████▌ | 393/520 [24:20<07:48, 3.69s/it] {'loss': 1.1184, 'grad_norm': 0.0010312822216644946, 'learning_rate': 0.007432863731133272, 'epoch': 0.76} + 76%|███████▌ | 393/520 [24:20<07:48, 3.69s/it] 76%|███████▌ | 394/520 [24:24<07:44, 3.69s/it] {'loss': 1.2174, 'grad_norm': 0.001277940661317857, 'learning_rate': 0.0073223304703363135, 'epoch': 0.76} + 76%|███████▌ | 394/520 [24:24<07:44, 3.69s/it] 76%|███████▌ | 395/520 [24:28<07:40, 3.68s/it] {'loss': 1.185, 'grad_norm': 0.0013192578227925757, 'learning_rate': 0.007212484060131752, 'epoch': 0.76} + 76%|███████▌ | 395/520 [24:28<07:40, 3.68s/it] 76%|███████▌ | 396/520 [24:31<07:35, 3.67s/it] {'loss': 1.2511, 'grad_norm': 0.0012807459917905275, 'learning_rate': 0.007103328768507039, 'epoch': 0.76} + 76%|███████▌ | 396/520 [24:31<07:35, 3.67s/it] 76%|███████▋ | 397/520 [24:35<07:32, 3.68s/it] {'loss': 1.2265, 'grad_norm': 0.0012052897945594519, 'learning_rate': 0.006994868836596841, 'epoch': 0.76} + 76%|███████▋ | 397/520 [24:35<07:32, 3.68s/it] 77%|███████▋ | 398/520 [24:39<07:27, 3.67s/it] {'loss': 1.2155, 'grad_norm': 0.0013028661311553663, 'learning_rate': 0.006887108478518184, 'epoch': 0.77} + 77%|███████▋ | 398/520 [24:39<07:27, 3.67s/it] 77%|███████▋ | 399/520 [24:42<07:25, 3.68s/it] {'loss': 1.1432, 'grad_norm': 0.001145851073583165, 'learning_rate': 0.006780051881206792, 'epoch': 0.77} + 77%|███████▋ | 399/520 [24:42<07:25, 3.68s/it] 77%|███████▋ | 400/520 [24:46<07:21, 3.68s/it] {'loss': 1.1815, 'grad_norm': 0.0010906951694823785, 'learning_rate': 0.006673703204254347, 'epoch': 0.77} + 77%|███████▋ | 400/520 [24:46<07:21, 3.68s/it] 77%|███████▋ | 401/520 [24:50<07:17, 3.67s/it] {'loss': 1.0628, 'grad_norm': 0.0013270182777701176, 'learning_rate': 0.006568066579746901, 'epoch': 0.77} + 77%|███████▋ | 401/520 [24:50<07:17, 3.67s/it] 77%|███████▋ | 402/520 [24:53<07:12, 3.67s/it] {'loss': 1.1959, 'grad_norm': 0.001272350705767296, 'learning_rate': 0.006463146112104332, 'epoch': 0.77} + 77%|███████▋ | 402/520 [24:53<07:12, 3.67s/it] 78%|███████▊ | 403/520 [24:57<07:08, 3.66s/it] {'loss': 1.2147, 'grad_norm': 0.001317465503994875, 'learning_rate': 0.0063589458779208605, 'epoch': 0.78} + 78%|███████▊ | 403/520 [24:57<07:08, 3.66s/it] 78%|███████▊ | 404/520 [25:01<07:04, 3.66s/it] {'loss': 1.1338, 'grad_norm': 0.0014739915892693803, 'learning_rate': 0.006255469925806643, 'epoch': 0.78} + 78%|███████▊ | 404/520 [25:01<07:04, 3.66s/it] 78%|███████▊ | 405/520 [25:04<07:01, 3.66s/it] {'loss': 1.1633, 'grad_norm': 0.0011526673633263562, 'learning_rate': 0.006152722276230505, 'epoch': 0.78} + 78%|███████▊ | 405/520 [25:04<07:01, 3.66s/it] 78%|███████▊ | 406/520 [25:08<06:57, 3.66s/it] {'loss': 1.0927, 'grad_norm': 0.0014364717120331175, 'learning_rate': 0.006050706921363672, 'epoch': 0.78} + 78%|███████▊ | 406/520 [25:08<06:57, 3.66s/it] 78%|███████▊ | 407/520 [25:12<06:53, 3.66s/it] {'loss': 1.2866, 'grad_norm': 0.0012424591921075966, 'learning_rate': 0.00594942782492473, 'epoch': 0.78} + 78%|███████▊ | 407/520 [25:12<06:53, 3.66s/it] 78%|███████▊ | 408/520 [25:15<06:49, 3.66s/it] {'loss': 1.2117, 'grad_norm': 0.0013828935804402161, 'learning_rate': 0.005848888922025553, 'epoch': 0.78} + 78%|███████▊ | 408/520 [25:15<06:49, 3.66s/it] 79%|███████▊ | 409/520 [25:19<06:44, 3.65s/it] {'loss': 1.3304, 'grad_norm': 0.0013094404066731934, 'learning_rate': 0.00574909411901843, 'epoch': 0.79} + 79%|███████▊ | 409/520 [25:19<06:44, 3.65s/it] 79%|███████▉ | 410/520 [25:23<06:41, 3.65s/it] {'loss': 1.0648, 'grad_norm': 0.0012680072074427694, 'learning_rate': 0.005650047293344316, 'epoch': 0.79} + 79%|███████▉ | 410/520 [25:23<06:41, 3.65s/it] 79%|███████▉ | 411/520 [25:26<06:37, 3.65s/it] {'loss': 1.3065, 'grad_norm': 0.0013117425510786353, 'learning_rate': 0.005551752293382131, 'epoch': 0.79} + 79%|███████▉ | 411/520 [25:26<06:37, 3.65s/it] 79%|███████▉ | 412/520 [25:30<06:33, 3.65s/it] {'loss': 1.2186, 'grad_norm': 0.001211047918679841, 'learning_rate': 0.005454212938299255, 'epoch': 0.79} + 79%|███████▉ | 412/520 [25:30<06:33, 3.65s/it] 79%|███████▉ | 413/520 [25:34<06:30, 3.65s/it] {'loss': 1.1859, 'grad_norm': 0.0011557636402991464, 'learning_rate': 0.005357433017903163, 'epoch': 0.79} + 79%|███████▉ | 413/520 [25:34<06:30, 3.65s/it] 80%|███████▉ | 414/520 [25:37<06:26, 3.64s/it] {'loss': 0.9916, 'grad_norm': 0.0010269073681710762, 'learning_rate': 0.005261416292494117, 'epoch': 0.8} + 80%|███████▉ | 414/520 [25:37<06:26, 3.64s/it] 80%|███████▉ | 415/520 [25:41<06:22, 3.65s/it] {'loss': 1.1982, 'grad_norm': 0.001181446027514777, 'learning_rate': 0.005166166492719124, 'epoch': 0.8} + 80%|███████▉ | 415/520 [25:41<06:22, 3.65s/it] 80%|████████ | 416/520 [25:45<06:21, 3.67s/it] {'loss': 1.1024, 'grad_norm': 0.001368234552255285, 'learning_rate': 0.005071687319426946, 'epoch': 0.8} + 80%|████████ | 416/520 [25:45<06:21, 3.67s/it] 80%|████████ | 417/520 [25:48<06:22, 3.71s/it] {'loss': 1.2602, 'grad_norm': 0.0012166748354342422, 'learning_rate': 0.004977982443524304, 'epoch': 0.8} + 80%|████████ | 417/520 [25:48<06:22, 3.71s/it] 80%|████████ | 418/520 [25:52<06:22, 3.75s/it] {'loss': 1.2588, 'grad_norm': 0.001151672248796759, 'learning_rate': 0.004885055505833291, 'epoch': 0.8} + 80%|████████ | 418/520 [25:52<06:22, 3.75s/it] 81%|████████ | 419/520 [25:56<06:20, 3.77s/it] {'loss': 1.2568, 'grad_norm': 0.001341763228013401, 'learning_rate': 0.00479291011694987, 'epoch': 0.81} + 81%|████████ | 419/520 [25:56<06:20, 3.77s/it] 81%|████████ | 420/520 [26:00<06:17, 3.78s/it] {'loss': 1.1437, 'grad_norm': 0.0012893022206333733, 'learning_rate': 0.004701549857103588, 'epoch': 0.81} + 81%|████████ | 420/520 [26:00<06:17, 3.78s/it] 81%|████████ | 421/520 [26:04<06:15, 3.80s/it] {'loss': 1.08, 'grad_norm': 0.0013213313304296238, 'learning_rate': 0.004610978276018496, 'epoch': 0.81} + 81%|████████ | 421/520 [26:04<06:15, 3.80s/it] 81%|████████ | 422/520 [26:07<06:12, 3.80s/it] {'loss': 1.2071, 'grad_norm': 0.0013177229557610381, 'learning_rate': 0.0045211988927752024, 'epoch': 0.81} + 81%|████████ | 422/520 [26:07<06:12, 3.80s/it] 81%|████████▏ | 423/520 [26:11<06:09, 3.81s/it] {'loss': 1.1727, 'grad_norm': 0.0013389806083239383, 'learning_rate': 0.004432215195674167, 'epoch': 0.81} + 81%|████████▏ | 423/520 [26:11<06:09, 3.81s/it] 82%|████████▏ | 424/520 [26:15<06:07, 3.82s/it] {'loss': 1.2579, 'grad_norm': 0.0010769408973173853, 'learning_rate': 0.004344030642100133, 'epoch': 0.82} + 82%|████████▏ | 424/520 [26:15<06:07, 3.82s/it] 82%|████████▏ | 425/520 [26:19<06:04, 3.83s/it] {'loss': 1.1806, 'grad_norm': 0.0012072704846982228, 'learning_rate': 0.004256648658387813, 'epoch': 0.82} + 82%|████████▏ | 425/520 [26:19<06:04, 3.83s/it] 82%|████████▏ | 426/520 [26:23<05:59, 3.83s/it] {'loss': 1.2289, 'grad_norm': 0.0016482657408181001, 'learning_rate': 0.00417007263968878, 'epoch': 0.82} + 82%|████████▏ | 426/520 [26:23<05:59, 3.83s/it] 82%|████████▏ | 427/520 [26:27<05:56, 3.83s/it] {'loss': 1.1126, 'grad_norm': 0.0011924219433238578, 'learning_rate': 0.004084305949839506, 'epoch': 0.82} + 82%|████████▏ | 427/520 [26:27<05:56, 3.83s/it] 82%|████████▏ | 428/520 [26:31<05:52, 3.83s/it] {'loss': 1.1127, 'grad_norm': 0.0013497838184938125, 'learning_rate': 0.003999351921230715, 'epoch': 0.82} + 82%|████████▏ | 428/520 [26:31<05:52, 3.83s/it] 82%|████████▎ | 429/520 [26:34<05:48, 3.83s/it] {'loss': 1.2176, 'grad_norm': 0.0012581236589942224, 'learning_rate': 0.003915213854677863, 'epoch': 0.82} + 82%|████████▎ | 429/520 [26:34<05:48, 3.83s/it]Token indices sequence length is longer than the specified maximum sequence length for this model (2076 > 2048). Running this sequence through the model will result in indexing errors + 83%|████████▎ | 430/520 [26:38<05:45, 3.84s/it] {'loss': 1.2138, 'grad_norm': 0.0012035616598302348, 'learning_rate': 0.0038318950192928972, 'epoch': 0.83} + 83%|████████▎ | 430/520 [26:38<05:45, 3.84s/it] 83%|████████▎ | 431/520 [26:42<05:41, 3.84s/it] {'loss': 1.1492, 'grad_norm': 0.001150241949895741, 'learning_rate': 0.003749398652357272, 'epoch': 0.83} + 83%|████████▎ | 431/520 [26:42<05:41, 3.84s/it] 83%|████████▎ | 432/520 [26:46<05:37, 3.84s/it] {'loss': 1.1212, 'grad_norm': 0.0012979143128407416, 'learning_rate': 0.00366772795919611, 'epoch': 0.83} + 83%|████████▎ | 432/520 [26:46<05:37, 3.84s/it] 83%|████████▎ | 433/520 [26:50<05:34, 3.84s/it] {'loss': 1.2536, 'grad_norm': 0.001232445622328954, 'learning_rate': 0.0035868861130537166, 'epoch': 0.83} + 83%|████████▎ | 433/520 [26:50<05:34, 3.84s/it] 83%|████████▎ | 434/520 [26:54<05:30, 3.84s/it] {'loss': 1.0172, 'grad_norm': 0.0013000607415608347, 'learning_rate': 0.0035068762549702428, 'epoch': 0.83} + 83%|████████▎ | 434/520 [26:54<05:30, 3.84s/it] 84%|████████▎ | 435/520 [26:57<05:25, 3.83s/it] {'loss': 1.2876, 'grad_norm': 0.0013391458512797192, 'learning_rate': 0.003427701493659674, 'epoch': 0.84} + 84%|████████▎ | 435/520 [26:57<05:25, 3.83s/it] 84%|████████▍ | 436/520 [27:01<05:21, 3.83s/it] {'loss': 1.0966, 'grad_norm': 0.0013294132280060158, 'learning_rate': 0.0033493649053890325, 'epoch': 0.84} + 84%|████████▍ | 436/520 [27:01<05:21, 3.83s/it] 84%|████████▍ | 437/520 [27:05<05:19, 3.84s/it] {'loss': 1.3081, 'grad_norm': 0.001259615095879299, 'learning_rate': 0.00327186953385884, 'epoch': 0.84} + 84%|████████▍ | 437/520 [27:05<05:19, 3.84s/it] 84%|████████▍ | 438/520 [27:09<05:16, 3.86s/it] {'loss': 1.1265, 'grad_norm': 0.0012995660403431085, 'learning_rate': 0.003195218390084867, 'epoch': 0.84} + 84%|████████▍ | 438/520 [27:09<05:16, 3.86s/it] 84%|████████▍ | 439/520 [27:13<05:13, 3.87s/it] {'loss': 1.1383, 'grad_norm': 0.0009882821292877067, 'learning_rate': 0.003119414452281158, 'epoch': 0.84} + 84%|████████▍ | 439/520 [27:13<05:13, 3.87s/it] 85%|████████▍ | 440/520 [27:17<05:09, 3.86s/it] {'loss': 1.1676, 'grad_norm': 0.0013142690936452774, 'learning_rate': 0.0030444606657442836, 'epoch': 0.85} + 85%|████████▍ | 440/520 [27:17<05:09, 3.86s/it] 85%|████████▍ | 441/520 [27:21<05:06, 3.87s/it] {'loss': 1.1472, 'grad_norm': 0.0011799169339692512, 'learning_rate': 0.00297035994273894, 'epoch': 0.85} + 85%|████████▍ | 441/520 [27:21<05:06, 3.87s/it] 85%|████████▌ | 442/520 [27:24<05:01, 3.86s/it] {'loss': 1.2289, 'grad_norm': 0.001381781579220048, 'learning_rate': 0.002897115162384759, 'epoch': 0.85} + 85%|████████▌ | 442/520 [27:24<05:01, 3.86s/it] 85%|████████▌ | 443/520 [27:28<04:57, 3.86s/it] {'loss': 1.2318, 'grad_norm': 0.0011977250492274172, 'learning_rate': 0.0028247291705444572, 'epoch': 0.85} + 85%|████████▌ | 443/520 [27:28<04:57, 3.86s/it] 85%|████████▌ | 444/520 [27:32<04:49, 3.82s/it] {'loss': 1.1988, 'grad_norm': 0.0011179230018021518, 'learning_rate': 0.0027532047797132865, 'epoch': 0.85} + 85%|████████▌ | 444/520 [27:32<04:49, 3.82s/it] 86%|████████▌ | 445/520 [27:36<04:42, 3.77s/it] {'loss': 1.1239, 'grad_norm': 0.0011801202823325742, 'learning_rate': 0.0026825447689097174, 'epoch': 0.86} + 86%|████████▌ | 445/520 [27:36<04:42, 3.77s/it] 86%|████████▌ | 446/520 [27:39<04:36, 3.73s/it] {'loss': 1.228, 'grad_norm': 0.0010787093611499862, 'learning_rate': 0.002612751883567477, 'epoch': 0.86} + 86%|████████▌ | 446/520 [27:39<04:36, 3.73s/it] 86%|████████▌ | 447/520 [27:43<04:31, 3.72s/it] {'loss': 1.1945, 'grad_norm': 0.0012123189350655833, 'learning_rate': 0.0025438288354288994, 'epoch': 0.86} + 86%|████████▌ | 447/520 [27:43<04:31, 3.72s/it] 86%|████████▌ | 448/520 [27:47<04:25, 3.69s/it] {'loss': 1.1968, 'grad_norm': 0.0013971045743317674, 'learning_rate': 0.0024757783024395245, 'epoch': 0.86} + 86%|████████▌ | 448/520 [27:47<04:25, 3.69s/it] 86%|████████▋ | 449/520 [27:50<04:21, 3.68s/it] {'loss': 1.19, 'grad_norm': 0.0011806489733031421, 'learning_rate': 0.002408602928644088, 'epoch': 0.86} + 86%|████████▋ | 449/520 [27:50<04:21, 3.68s/it] 87%|████████▋ | 450/520 [27:54<04:16, 3.66s/it] {'loss': 1.2244, 'grad_norm': 0.0012881231985615272, 'learning_rate': 0.002342305324083752, 'epoch': 0.87} + 87%|████████▋ | 450/520 [27:54<04:16, 3.66s/it] 87%|████████▋ | 451/520 [27:58<04:12, 3.66s/it] {'loss': 1.2221, 'grad_norm': 0.0012636187292095421, 'learning_rate': 0.0022768880646947265, 'epoch': 0.87} + 87%|████████▋ | 451/520 [27:58<04:12, 3.66s/it] 87%|████████▋ | 452/520 [28:01<04:08, 3.65s/it] {'loss': 1.2361, 'grad_norm': 0.0011390184769379432, 'learning_rate': 0.0022123536922081717, 'epoch': 0.87} + 87%|████████▋ | 452/520 [28:01<04:08, 3.65s/it] 87%|████████▋ | 453/520 [28:05<04:04, 3.65s/it] {'loss': 1.2161, 'grad_norm': 0.0011763543034588554, 'learning_rate': 0.0021487047140514247, 'epoch': 0.87} + 87%|████████▋ | 453/520 [28:05<04:04, 3.65s/it] 87%|████████▋ | 454/520 [28:08<04:00, 3.65s/it] {'loss': 1.1348, 'grad_norm': 0.0012320727120716717, 'learning_rate': 0.002085943603250595, 'epoch': 0.87} + 87%|████████▋ | 454/520 [28:08<04:00, 3.65s/it] 88%|████████▊ | 455/520 [28:12<03:57, 3.66s/it] {'loss': 1.2707, 'grad_norm': 0.0012505688544023708, 'learning_rate': 0.0020240727983344836, 'epoch': 0.88} + 88%|████████▊ | 455/520 [28:12<03:57, 3.66s/it] 88%|████████▊ | 456/520 [28:16<03:53, 3.65s/it] {'loss': 1.2132, 'grad_norm': 0.001312577461889593, 'learning_rate': 0.0019630947032398065, 'epoch': 0.88} + 88%|████████▊ | 456/520 [28:16<03:53, 3.65s/it] 88%|████████▊ | 457/520 [28:19<03:49, 3.65s/it] {'loss': 1.0801, 'grad_norm': 0.0010778110772003168, 'learning_rate': 0.0019030116872178317, 'epoch': 0.88} + 88%|████████▊ | 457/520 [28:19<03:49, 3.65s/it] 88%|████████▊ | 458/520 [28:23<03:48, 3.68s/it] {'loss': 1.3225, 'grad_norm': 0.001318802326599245, 'learning_rate': 0.001843826084742284, 'epoch': 0.88} + 88%|████████▊ | 458/520 [28:23<03:48, 3.68s/it] 88%|████████▊ | 459/520 [28:27<03:44, 3.68s/it] {'loss': 1.2527, 'grad_norm': 0.0012246468500429156, 'learning_rate': 0.0017855401954186612, 'epoch': 0.88} + 88%|████████▊ | 459/520 [28:27<03:44, 3.68s/it] 88%|████████▊ | 460/520 [28:31<03:40, 3.67s/it] {'loss': 1.1467, 'grad_norm': 0.0012533886466015247, 'learning_rate': 0.0017281562838948967, 'epoch': 0.88} + 88%|████████▊ | 460/520 [28:31<03:40, 3.67s/it] 89%|████████▊ | 461/520 [28:34<03:37, 3.68s/it] {'loss': 1.1655, 'grad_norm': 0.000865539746912039, 'learning_rate': 0.0016716765797733374, 'epoch': 0.89} + 89%|████████▊ | 461/520 [28:34<03:37, 3.68s/it] 89%|████████▉ | 462/520 [28:38<03:33, 3.68s/it] {'loss': 1.2729, 'grad_norm': 0.0011554810815691643, 'learning_rate': 0.0016161032775241502, 'epoch': 0.89} + 89%|████████▉ | 462/520 [28:38<03:33, 3.68s/it] 89%|████████▉ | 463/520 [28:42<03:29, 3.68s/it] {'loss': 1.1277, 'grad_norm': 0.0013166121840743595, 'learning_rate': 0.0015614385364000228, 'epoch': 0.89} + 89%|████████▉ | 463/520 [28:42<03:29, 3.68s/it] 89%|████████▉ | 464/520 [28:45<03:26, 3.69s/it] {'loss': 1.2376, 'grad_norm': 0.0012524401788369564, 'learning_rate': 0.0015076844803522923, 'epoch': 0.89} + 89%|████████▉ | 464/520 [28:45<03:26, 3.69s/it] 89%|████████▉ | 465/520 [28:49<03:23, 3.69s/it][E ProcessGroupNCCL.cpp:474] [Rank 6] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=1038882, OpType=_ALLGATHER_BASE, NumelIn=165888, NumelOut=1327104, Timeout(ms)=1800000) ran for 4817222 milliseconds before timing out. +[E ProcessGroupNCCL.cpp:474] [Rank 3] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=1039335, OpType=_ALLGATHER_BASE, NumelIn=16, NumelOut=128, Timeout(ms)=1800000) ran for 4816267 milliseconds before timing out. +[E ProcessGroupNCCL.cpp:474] [Rank 1] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=1039335, OpType=_ALLGATHER_BASE, NumelIn=16, NumelOut=128, Timeout(ms)=1800000) ran for 4816266 milliseconds before timing out. +[E ProcessGroupNCCL.cpp:474] [Rank 5] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=1039231, OpType=_ALLGATHER_BASE, NumelIn=100352, NumelOut=802816, Timeout(ms)=1800000) ran for 4817003 milliseconds before timing out. +[E ProcessGroupNCCL.cpp:474] [Rank 2] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=1039335, OpType=_ALLGATHER_BASE, NumelIn=16, NumelOut=128, Timeout(ms)=1800000) ran for 4816267 milliseconds before timing out. +[E ProcessGroupNCCL.cpp:474] [Rank 7] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=1038887, OpType=_ALLGATHER_BASE, NumelIn=619776, NumelOut=4958208, Timeout(ms)=1800000) ran for 4817222 milliseconds before timing out. +[E ProcessGroupNCCL.cpp:474] [Rank 4] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=1038882, OpType=_ALLGATHER_BASE, NumelIn=165888, NumelOut=1327104, Timeout(ms)=1800000) ran for 4817226 milliseconds before timing out. +[E ProcessGroupNCCL.cpp:474] [Rank 5] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=272379, OpType=_REDUCE_SCATTER_BASE, NumelIn=4358144, NumelOut=544768, Timeout(ms)=1800000) ran for 4816622 milliseconds before timing out. +[E ProcessGroupNCCL.cpp:474] [Rank 1] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=272445, OpType=_REDUCE_SCATTER_BASE, NumelIn=4358144, NumelOut=544768, Timeout(ms)=1800000) ran for 4816440 milliseconds before timing out. +[E ProcessGroupNCCL.cpp:474] [Rank 6] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=272378, OpType=_REDUCE_SCATTER_BASE, NumelIn=802816, NumelOut=100352, Timeout(ms)=1800000) ran for 4816620 milliseconds before timing out. +[E ProcessGroupNCCL.cpp:474] [Rank 4] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=272414, OpType=_REDUCE_SCATTER_BASE, NumelIn=802816, NumelOut=100352, Timeout(ms)=1800000) ran for 4816526 milliseconds before timing out. +[E ProcessGroupNCCL.cpp:474] [Rank 7] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=272429, OpType=_REDUCE_SCATTER_BASE, NumelIn=4358144, NumelOut=544768, Timeout(ms)=1800000) ran for 4816485 milliseconds before timing out. +[E ProcessGroupNCCL.cpp:474] [Rank 2] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=272416, OpType=_REDUCE_SCATTER_BASE, NumelIn=4358144, NumelOut=544768, Timeout(ms)=1800000) ran for 4816524 milliseconds before timing out. +[E ProcessGroupNCCL.cpp:474] [Rank 3] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=272401, OpType=_REDUCE_SCATTER_BASE, NumelIn=229376, NumelOut=28672, Timeout(ms)=1800000) ran for 4816567 milliseconds before timing out. +ywang29-vrdb-test2-worker-0:720430:722079 [2] NCCL INFO [Service thread] Connection closed by localRank 2 +ywang29-vrdb-test2-worker-0:720435:722083 [7] NCCL INFO [Service thread] Connection closed by localRank 7 +ywang29-vrdb-test2-worker-0:720434:722088 [6] NCCL INFO [Service thread] Connection closed by localRank 6 +ywang29-vrdb-test2-worker-0:720433:727150 [5] NCCL INFO [Service thread] Connection closed by localRank 5 +ywang29-vrdb-test2-worker-0:720434:727147 [6] NCCL INFO [Service thread] Connection closed by localRank 6 +ywang29-vrdb-test2-worker-0:720430:727158 [2] NCCL INFO [Service thread] Connection closed by localRank 2 +ywang29-vrdb-test2-worker-0:720435:727157 [7] NCCL INFO [Service thread] Connection closed by localRank 7 +ywang29-vrdb-test2-worker-0:720432:722076 [4] NCCL INFO [Service thread] Connection closed by localRank 4 +ywang29-vrdb-test2-worker-0:720429:727148 [1] NCCL INFO [Service thread] Connection closed by localRank 1 +ywang29-vrdb-test2-worker-0:720433:722075 [5] NCCL INFO [Service thread] Connection closed by localRank 5 +ywang29-vrdb-test2-worker-0:720429:722084 [1] NCCL INFO [Service thread] Connection closed by localRank 1 +ywang29-vrdb-test2-worker-0:720431:722080 [3] NCCL INFO [Service thread] Connection closed by localRank 3 +ywang29-vrdb-test2-worker-0:720432:727153 [4] NCCL INFO [Service thread] Connection closed by localRank 4 +ywang29-vrdb-test2-worker-0:720431:727151 [3] NCCL INFO [Service thread] Connection closed by localRank 3 +ywang29-vrdb-test2-worker-0:720430:721543 [2] NCCL INFO comm 0x56110a1eff40 rank 2 nranks 8 cudaDev 2 busId 201c0 - Abort COMPLETE +[E ProcessGroupNCCL.cpp:488] Some NCCL operations have failed or timed out. Due to the asynchronous nature of CUDA kernels, subsequent GPU operations might run on corrupted/incomplete data. +[E ProcessGroupNCCL.cpp:494] To avoid data inconsistency, we are taking the entire process down. +[E ProcessGroupNCCL.cpp:915] [Rank 2] NCCL watchdog thread terminated with exception: [Rank 2] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=1039335, OpType=_ALLGATHER_BASE, NumelIn=16, NumelOut=128, Timeout(ms)=1800000) ran for 4816267 milliseconds before timing out. +terminate called after throwing an instance of 'std::runtime_error' + what(): [Rank 2] NCCL watchdog thread terminated with exception: [Rank 2] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=1039335, OpType=_ALLGATHER_BASE, NumelIn=16, NumelOut=128, Timeout(ms)=1800000) ran for 4816267 milliseconds before timing out. +ywang29-vrdb-test2-worker-0:720435:722239 [7] NCCL INFO comm 0x7fa86c069af0 rank 7 nranks 8 cudaDev 7 busId a01d0 - Abort COMPLETE +[E ProcessGroupNCCL.cpp:488] Some NCCL operations have failed or timed out. Due to the asynchronous nature of CUDA kernels, subsequent GPU operations might run on corrupted/incomplete data. +[E ProcessGroupNCCL.cpp:494] To avoid data inconsistency, we are taking the entire process down. +[E ProcessGroupNCCL.cpp:915] [Rank 7] NCCL watchdog thread terminated with exception: [Rank 7] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=272429, OpType=_REDUCE_SCATTER_BASE, NumelIn=4358144, NumelOut=544768, Timeout(ms)=1800000) ran for 4816485 milliseconds before timing out. +terminate called after throwing an instance of 'std::runtime_error' + what(): [Rank 7] NCCL watchdog thread terminated with exception: [Rank 7] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=272429, OpType=_REDUCE_SCATTER_BASE, NumelIn=4358144, NumelOut=544768, Timeout(ms)=1800000) ran for 4816485 milliseconds before timing out. +ywang29-vrdb-test2-worker-0:720431:721555 [3] NCCL INFO comm 0x56492cd58b90 rank 3 nranks 8 cudaDev 3 busId 201d0 - Abort COMPLETE +[E ProcessGroupNCCL.cpp:488] Some NCCL operations have failed or timed out. Due to the asynchronous nature of CUDA kernels, subsequent GPU operations might run on corrupted/incomplete data. +[E ProcessGroupNCCL.cpp:494] To avoid data inconsistency, we are taking the entire process down. +[E ProcessGroupNCCL.cpp:915] [Rank 3] NCCL watchdog thread terminated with exception: [Rank 3] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=1039335, OpType=_ALLGATHER_BASE, NumelIn=16, NumelOut=128, Timeout(ms)=1800000) ran for 4816267 milliseconds before timing out. +terminate called after throwing an instance of 'std::runtime_error' + what(): [Rank 3] NCCL watchdog thread terminated with exception: [Rank 3] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=1039335, OpType=_ALLGATHER_BASE, NumelIn=16, NumelOut=128, Timeout(ms)=1800000) ran for 4816267 milliseconds before timing out. +ywang29-vrdb-test2-worker-0:720433:721554 [5] NCCL INFO comm 0x55c945a6d9b0 rank 5 nranks 8 cudaDev 5 busId 901d0 - Abort COMPLETE +[E ProcessGroupNCCL.cpp:488] Some NCCL operations have failed or timed out. Due to the asynchronous nature of CUDA kernels, subsequent GPU operations might run on corrupted/incomplete data. +[E ProcessGroupNCCL.cpp:494] To avoid data inconsistency, we are taking the entire process down. +[E ProcessGroupNCCL.cpp:915] [Rank 5] NCCL watchdog thread terminated with exception: [Rank 5] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=1039231, OpType=_ALLGATHER_BASE, NumelIn=100352, NumelOut=802816, Timeout(ms)=1800000) ran for 4817003 milliseconds before timing out. +terminate called after throwing an instance of 'std::runtime_error' + what(): [Rank 5] NCCL watchdog thread terminated with exception: [Rank 5] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=1039231, OpType=_ALLGATHER_BASE, NumelIn=100352, NumelOut=802816, Timeout(ms)=1800000) ran for 4817003 milliseconds before timing out. +ywang29-vrdb-test2-worker-0:720429:721552 [1] NCCL INFO comm 0x560824237a50 rank 1 nranks 8 cudaDev 1 busId 101d0 - Abort COMPLETE +[E ProcessGroupNCCL.cpp:488] Some NCCL operations have failed or timed out. Due to the asynchronous nature of CUDA kernels, subsequent GPU operations might run on corrupted/incomplete data. +[E ProcessGroupNCCL.cpp:494] To avoid data inconsistency, we are taking the entire process down. +[E ProcessGroupNCCL.cpp:915] [Rank 1] NCCL watchdog thread terminated with exception: [Rank 1] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=1039335, OpType=_ALLGATHER_BASE, NumelIn=16, NumelOut=128, Timeout(ms)=1800000) ran for 4816266 milliseconds before timing out. +terminate called after throwing an instance of 'std::runtime_error' + what(): [Rank 1] NCCL watchdog thread terminated with exception: [Rank 1] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=1039335, OpType=_ALLGATHER_BASE, NumelIn=16, NumelOut=128, Timeout(ms)=1800000) ran for 4816266 milliseconds before timing out. +ywang29-vrdb-test2-worker-0:720432:721542 [4] NCCL INFO comm 0x556b8d5713b0 rank 4 nranks 8 cudaDev 4 busId 901c0 - Abort COMPLETE +[E ProcessGroupNCCL.cpp:488] Some NCCL operations have failed or timed out. Due to the asynchronous nature of CUDA kernels, subsequent GPU operations might run on corrupted/incomplete data. +[E ProcessGroupNCCL.cpp:494] To avoid data inconsistency, we are taking the entire process down. +[E ProcessGroupNCCL.cpp:915] [Rank 4] NCCL watchdog thread terminated with exception: [Rank 4] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=1038882, OpType=_ALLGATHER_BASE, NumelIn=165888, NumelOut=1327104, Timeout(ms)=1800000) ran for 4817226 milliseconds before timing out. +terminate called after throwing an instance of 'std::runtime_error' + what(): [Rank 4] NCCL watchdog thread terminated with exception: [Rank 4] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=1038882, OpType=_ALLGATHER_BASE, NumelIn=165888, NumelOut=1327104, Timeout(ms)=1800000) ran for 4817226 milliseconds before timing out. +ywang29-vrdb-test2-worker-0:720434:721553 [6] NCCL INFO comm 0x5569a6daac30 rank 6 nranks 8 cudaDev 6 busId a01c0 - Abort COMPLETE +[E ProcessGroupNCCL.cpp:488] Some NCCL operations have failed or timed out. Due to the asynchronous nature of CUDA kernels, subsequent GPU operations might run on corrupted/incomplete data. +[E ProcessGroupNCCL.cpp:494] To avoid data inconsistency, we are taking the entire process down. +[E ProcessGroupNCCL.cpp:915] [Rank 6] NCCL watchdog thread terminated with exception: [Rank 6] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=1038882, OpType=_ALLGATHER_BASE, NumelIn=165888, NumelOut=1327104, Timeout(ms)=1800000) ran for 4817222 milliseconds before timing out. +terminate called after throwing an instance of 'std::runtime_error' + what(): [Rank 6] NCCL watchdog thread terminated with exception: [Rank 6] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=1038882, OpType=_ALLGATHER_BASE, NumelIn=165888, NumelOut=1327104, Timeout(ms)=1800000) ran for 4817222 milliseconds before timing out. +[2025-10-12 13:24:32,499] [INFO] [launch.py:316:sigkill_handler] Killing subprocess 720428 +[2025-10-12 13:24:33,397] [INFO] [launch.py:316:sigkill_handler] Killing subprocess 720429 +[2025-10-12 13:24:35,741] [INFO] [launch.py:316:sigkill_handler] Killing subprocess 720430 +[2025-10-12 13:24:35,744] [INFO] [launch.py:316:sigkill_handler] Killing subprocess 720431 +[2025-10-12 13:24:35,745] [INFO] [launch.py:316:sigkill_handler] Killing subprocess 720432 +[2025-10-12 13:24:35,747] [INFO] [launch.py:316:sigkill_handler] Killing subprocess 720433 +[2025-10-12 13:24:35,749] [INFO] [launch.py:316:sigkill_handler] Killing subprocess 720434 +[2025-10-12 13:24:35,749] [INFO] [launch.py:316:sigkill_handler] Killing subprocess 720435 +[2025-10-12 13:24:35,750] [ERROR] [launch.py:322:sigkill_handler] ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=7', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_5e-2_connector-3.0_0.5_5e-2_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '5e-2', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '5e-2', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] exits with return code = -6 +==== EXPERIMENT COMPLETED: qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_5e-2_connector-3.0_0.5_5e-2_ablation ==== +Log File: qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_5e-2_connector-3.0_0.5_5e-2_ablation_20251012_113222.log +Timestamp: 2025-10-12 13:24:36 +===================================== diff --git a/logs_oct11/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_7e-1_connector-3.0_0.5_7e-1_ablation_20251012_071236.log b/logs_oct11/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_7e-1_connector-3.0_0.5_7e-1_ablation_20251012_071236.log new file mode 100644 index 0000000000000000000000000000000000000000..61ba571a4c260141e6cdb20259563bff0789bf61 --- /dev/null +++ b/logs_oct11/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_7e-1_connector-3.0_0.5_7e-1_ablation_20251012_071236.log @@ -0,0 +1,2312 @@ +==== STARTING EXPERIMENT: qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_7e-1_connector-3.0_0.5_7e-1_ablation ==== +Log File: qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_7e-1_connector-3.0_0.5_7e-1_ablation_20251012_071236.log +Timestamp: 2025-10-12 07:12:36 +===================================== +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 07:12:39,373] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 07:12:42,642] [WARNING] [runner.py:202:fetch_hostfile] Unable to find hostfile, will proceed with training with local resources only. +[2025-10-12 07:12:42,644] [INFO] [runner.py:568:main] cmd = /opt/conda/envs/tinyllava/bin/python3.10 -u -m deepspeed.launcher.launch --world_info=eyJsb2NhbGhvc3QiOiBbMCwgMSwgMiwgMywgNCwgNSwgNiwgN119 --master_addr=127.0.0.1 --master_port=29501 --enable_each_rank_log=None tinyllava/train/train.py --deepspeed ./scripts/zero3.json --data_path /s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json --image_folder /s3-code/ywang29/datasets/tinyllava --is_multimodal True --conv_version qwen2_base --model_name_or_path Qwen/Qwen2.5-0.5B --vision_tower google/siglip-so400m-patch14-384 --vision_tower2 --connector_type mlp2x_gelu --mm_vision_select_layer -2 --image_aspect_ratio square --attn_implementation flash_attention_2 --bf16 True --training_recipe common --tune_type_llm full --tune_type_vision_tower frozen --tune_vision_tower_from_layer 0 --tune_type_connector full --group_by_modality_length True --pretrained_model_path /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain --output_dir /nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_7e-1_connector-3.0_0.5_7e-1_ablation --num_train_epochs 1 --per_device_train_batch_size 4 --per_device_eval_batch_size 4 --gradient_accumulation_steps 4 --evaluation_strategy no --learning_rate 7e-1 --weight_decay 0. --warmup_ratio 0.03 --lr_scheduler_type cosine --logging_steps 1 --tf32 False --model_max_length 2048 --gradient_checkpointing True --dataloader_num_workers 8 --lazy_preprocess True --report_to tensorboard --tokenizer_use_fast False --run_name tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune --subnet_mode_text both --subnet_type_text None --mask_type_text soft --init_mean_text 3.0 --temperature_attn_text 0.5 --temperature_mlp_text 0.5 --backward_type_text normal --masked_layers_text all --subnet_mode_vision both --subnet_type_vision None --mask_type_vision soft --init_mean_vision 3.0 --temperature_attn_vision 0.5 --temperature_mlp_vision 0.5 --backward_type_vision normal --masked_layers_vision all --subnet_type_connector global --mask_type_connector soft --init_mean_connector 3.0 --temperature_connector 0.5 --backward_type_connector normal --mm_projector_lr 7e-1 --seed 42 --mask_model llm-connector --save_strategy steps --save_steps 50000 --save_total_limit 1 --train_data_ratio 0.1 +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 07:12:45,288] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 07:12:46,333] [INFO] [launch.py:138:main] 0 NCCL_VERSION=2.21.5 +[2025-10-12 07:12:46,333] [INFO] [launch.py:138:main] 0 NCCL_SOCKET_IFNAME=eth +[2025-10-12 07:12:46,333] [INFO] [launch.py:145:main] WORLD INFO DICT: {'localhost': [0, 1, 2, 3, 4, 5, 6, 7]} +[2025-10-12 07:12:46,333] [INFO] [launch.py:151:main] nnodes=1, num_local_procs=8, node_rank=0 +[2025-10-12 07:12:46,333] [INFO] [launch.py:162:main] global_rank_mapping=defaultdict(, {'localhost': [0, 1, 2, 3, 4, 5, 6, 7]}) +[2025-10-12 07:12:46,333] [INFO] [launch.py:163:main] dist_world_size=8 +[2025-10-12 07:12:46,333] [INFO] [launch.py:165:main] Setting CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 +[2025-10-12 07:12:46,335] [INFO] [launch.py:253:main] process 514115 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=0', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_7e-1_connector-3.0_0.5_7e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '7e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '7e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 07:12:46,337] [INFO] [launch.py:253:main] process 514116 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=1', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_7e-1_connector-3.0_0.5_7e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '7e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '7e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 07:12:46,339] [INFO] [launch.py:253:main] process 514117 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=2', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_7e-1_connector-3.0_0.5_7e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '7e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '7e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 07:12:46,341] [INFO] [launch.py:253:main] process 514118 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=3', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_7e-1_connector-3.0_0.5_7e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '7e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '7e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 07:12:46,343] [INFO] [launch.py:253:main] process 514119 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=4', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_7e-1_connector-3.0_0.5_7e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '7e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '7e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 07:12:46,345] [INFO] [launch.py:253:main] process 514120 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=5', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_7e-1_connector-3.0_0.5_7e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '7e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '7e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 07:12:46,347] [INFO] [launch.py:253:main] process 514121 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=6', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_7e-1_connector-3.0_0.5_7e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '7e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '7e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 07:12:46,349] [INFO] [launch.py:253:main] process 514122 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=7', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_7e-1_connector-3.0_0.5_7e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '7e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '7e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 07:12:53,176] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 07:12:53,194] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 07:12:53,427] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 07:12:53,447] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 07:12:53,453] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 07:12:53,455] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 07:12:53,456] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 07:12:53,461] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 07:12:53,673] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 07:12:53,673] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 07:12:53,823] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 07:12:53,823] [INFO] [comm.py:668:init_distributed] Initializing TorchBackend in DeepSpeed with backend nccl +[2025-10-12 07:12:53,855] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 07:12:53,863] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 07:12:53,865] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 07:12:53,868] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 07:12:53,880] [INFO] [comm.py:637:init_distributed] cdb=None +Apply masks for the following modules: ['llm', 'connector'] +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +{'llm': {'model_name_or_path': 'Qwen/Qwen2.5-0.5B', 'cache_dir': None, 'attn_implementation': 'flash_attention_2', 'subnet_mode': 'both', 'subnet_type': 'None', 'sparsity_attn': None, 'sparsity_mlp': None, 'threshold_attn': None, 'threshold_mlp': None, 'temperature_attn': 0.5, 'temperature_mlp': 0.5, 'masked_layers': 'all', 'mask_type': 'soft', 'backward_type': 'normal'}, 'vision_tower': {'model_name_or_path': 'google/siglip-so400m-patch14-384'}, 'connector': {'connector_type': 'mlp2x_gelu', 'subnet_type': 'global', 'threshold': None, 'sparsity': None, 'temperature': 0.5, 'mask_type': 'soft', 'backward_type': 'normal'}} +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +TinyLlavaConfig { + "backward_type_connector": "normal", + "cache_dir": null, + "connector_type": "mlp2x_gelu", + "hidden_size": 896, + "ignore_index": -100, + "image_aspect_ratio": "square", + "image_token_index": -200, + "llm_model_name_or_path": "Qwen/Qwen2.5-0.5B", + "mask_model": [ + "llm", + "connector" + ], + "mask_type_connector": "soft", + "model_type": "tinyllava", + "num_queries": 128, + "num_resampler_layers": 3, + "pad_token": null, + "resampler_hidden_size": 768, + "sparsity_connector": null, + "subnet_type_connector": "global", + "temperature_connector": 0.5, + "text_config": { + "_name_or_path": "Qwen/Qwen2.5-0.5B", + "architectures": [ + "Qwen2ForCausalLM" + ], + "backward_type": "normal", + "bos_token_id": 151643, + "eos_token_id": 151643, + "hidden_size": 896, + "intermediate_size": 4864, + "mask_type": "soft", + "masked_layers": "all", + "max_position_embeddings": 32768, + "max_window_layers": 24, + "model_type": "qwen2", + "num_attention_heads": 14, + "num_hidden_layers": 24, + "num_key_value_heads": 2, + "rope_theta": 1000000.0, + "sliding_window": 32768, + "subnet_mode": "both", + "subnet_type": "None", + "temperature_attn": 0.5, + "temperature_mlp": 0.5, + "tie_word_embeddings": true, + "torch_dtype": "bfloat16", + "use_mrope": false, + "use_sliding_window": false, + "vocab_size": 151936 + }, + "threshold_connector": null, + "tokenizer_model_max_length": 2048, + "tokenizer_name_or_path": "Qwen/Qwen2.5-0.5B", + "tokenizer_padding_side": "right", + "tokenizer_use_fast": false, + "transformers_version": "4.40.1", + "tune_type_connector": "frozen", + "tune_type_llm": "frozen", + "tune_type_vision_tower": "frozen", + "tune_vision_tower_from_layer": -1, + "use_cache": false, + "vision_config": { + "hidden_act": "gelu_pytorch_tanh", + "hidden_size": 1152, + "image_size": 384, + "intermediate_size": 4304, + "layer_norm_eps": 1e-06, + "model_name_or_path": "google/siglip-so400m-patch14-384", + "model_name_or_path2": "", + "model_type": "siglip_vision_model", + "num_attention_heads": 16, + "num_hidden_layers": 27, + "patch_size": 14 + }, + "vision_feature_layer": -2, + "vision_feature_select_strategy": "patch", + "vision_hidden_size": 1152, + "vision_model_name_or_path": "google/siglip-so400m-patch14-384", + "vision_model_name_or_path2": "", + "vocab_size": 151936 +} + +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test2-worker-0:514115:514115 [0] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:514115:514115 [0] NCCL INFO Bootstrap : Using eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:514115:514115 [0] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test2-worker-0:514115:514115 [0] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test2-worker-0:514115:514115 [0] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test2-worker-0:514115:514115 [0] NCCL INFO cudaDriverVersion 12040 +NCCL version 2.21.5+cuda12.1 +ywang29-vrdb-test2-worker-0:514115:515701 [0] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:514115:515701 [0] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test2-worker-0:514115:515701 [0] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:514115:515701 [0] NCCL INFO NET/Socket : Using [0]eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:514115:515701 [0] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test2-worker-0:514115:515701 [0] NCCL INFO Using network Socket +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test2-worker-0:514117:514117 [2] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test2-worker-0:514117:514117 [2] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:514117:514117 [2] NCCL INFO Bootstrap : Using eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:514117:514117 [2] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test2-worker-0:514117:514117 [2] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test2-worker-0:514117:514117 [2] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test2-worker-0:514122:514122 [7] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test2-worker-0:514122:514122 [7] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:514122:514122 [7] NCCL INFO Bootstrap : Using eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:514122:514122 [7] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test2-worker-0:514122:514122 [7] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test2-worker-0:514122:514122 [7] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test2-worker-0:514120:514120 [5] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test2-worker-0:514120:514120 [5] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:514120:514120 [5] NCCL INFO Bootstrap : Using eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:514120:514120 [5] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test2-worker-0:514120:514120 [5] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test2-worker-0:514120:514120 [5] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test2-worker-0:514116:514116 [1] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test2-worker-0:514116:514116 [1] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:514116:514116 [1] NCCL INFO Bootstrap : Using eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:514116:514116 [1] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test2-worker-0:514116:514116 [1] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test2-worker-0:514116:514116 [1] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test2-worker-0:514121:514121 [6] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test2-worker-0:514121:514121 [6] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:514121:514121 [6] NCCL INFO Bootstrap : Using eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:514119:514119 [4] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test2-worker-0:514121:514121 [6] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test2-worker-0:514121:514121 [6] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test2-worker-0:514121:514121 [6] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test2-worker-0:514119:514119 [4] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:514119:514119 [4] NCCL INFO Bootstrap : Using eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:514118:514118 [3] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test2-worker-0:514118:514118 [3] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:514119:514119 [4] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test2-worker-0:514118:514118 [3] NCCL INFO Bootstrap : Using eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:514119:514119 [4] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test2-worker-0:514119:514119 [4] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test2-worker-0:514118:514118 [3] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test2-worker-0:514118:514118 [3] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test2-worker-0:514118:514118 [3] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test2-worker-0:514117:515719 [2] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:514117:515719 [2] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test2-worker-0:514117:515719 [2] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:514117:515719 [2] NCCL INFO NET/Socket : Using [0]eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:514117:515719 [2] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test2-worker-0:514117:515719 [2] NCCL INFO Using network Socket +ywang29-vrdb-test2-worker-0:514120:515721 [5] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:514120:515721 [5] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test2-worker-0:514120:515721 [5] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:514120:515721 [5] NCCL INFO NET/Socket : Using [0]eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:514120:515721 [5] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test2-worker-0:514120:515721 [5] NCCL INFO Using network Socket +ywang29-vrdb-test2-worker-0:514122:515720 [7] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:514122:515720 [7] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test2-worker-0:514122:515720 [7] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:514122:515720 [7] NCCL INFO NET/Socket : Using [0]eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:514122:515720 [7] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test2-worker-0:514122:515720 [7] NCCL INFO Using network Socket +ywang29-vrdb-test2-worker-0:514118:515724 [3] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:514118:515724 [3] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test2-worker-0:514118:515724 [3] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:514118:515724 [3] NCCL INFO NET/Socket : Using [0]eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:514118:515724 [3] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test2-worker-0:514118:515724 [3] NCCL INFO Using network Socket +ywang29-vrdb-test2-worker-0:514116:515722 [1] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:514119:515725 [4] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:514121:515723 [6] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:514116:515722 [1] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test2-worker-0:514121:515723 [6] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test2-worker-0:514119:515725 [4] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test2-worker-0:514121:515723 [6] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:514119:515725 [4] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:514116:515722 [1] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:514119:515725 [4] NCCL INFO NET/Socket : Using [0]eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:514121:515723 [6] NCCL INFO NET/Socket : Using [0]eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:514116:515722 [1] NCCL INFO NET/Socket : Using [0]eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:514119:515725 [4] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test2-worker-0:514116:515722 [1] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test2-worker-0:514119:515725 [4] NCCL INFO Using network Socket +ywang29-vrdb-test2-worker-0:514116:515722 [1] NCCL INFO Using network Socket +ywang29-vrdb-test2-worker-0:514121:515723 [6] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test2-worker-0:514121:515723 [6] NCCL INFO Using network Socket +ywang29-vrdb-test2-worker-0:514122:515720 [7] NCCL INFO ncclCommInitRank comm 0x55de10eeb450 rank 7 nranks 8 cudaDev 7 nvmlDev 7 busId a01d0 commId 0xf6ae4b96a072379f - Init START +ywang29-vrdb-test2-worker-0:514121:515723 [6] NCCL INFO ncclCommInitRank comm 0x5611899a0a00 rank 6 nranks 8 cudaDev 6 nvmlDev 6 busId a01c0 commId 0xf6ae4b96a072379f - Init START +ywang29-vrdb-test2-worker-0:514120:515721 [5] NCCL INFO ncclCommInitRank comm 0x55d70ee85fd0 rank 5 nranks 8 cudaDev 5 nvmlDev 5 busId 901d0 commId 0xf6ae4b96a072379f - Init START +ywang29-vrdb-test2-worker-0:514119:515725 [4] NCCL INFO ncclCommInitRank comm 0x5638e19a8a70 rank 4 nranks 8 cudaDev 4 nvmlDev 4 busId 901c0 commId 0xf6ae4b96a072379f - Init START +ywang29-vrdb-test2-worker-0:514117:515719 [2] NCCL INFO ncclCommInitRank comm 0x557e86657520 rank 2 nranks 8 cudaDev 2 nvmlDev 2 busId 201c0 commId 0xf6ae4b96a072379f - Init START +ywang29-vrdb-test2-worker-0:514118:515724 [3] NCCL INFO ncclCommInitRank comm 0x561646ab9390 rank 3 nranks 8 cudaDev 3 nvmlDev 3 busId 201d0 commId 0xf6ae4b96a072379f - Init START +ywang29-vrdb-test2-worker-0:514116:515722 [1] NCCL INFO ncclCommInitRank comm 0x558b540cdd80 rank 1 nranks 8 cudaDev 1 nvmlDev 1 busId 101d0 commId 0xf6ae4b96a072379f - Init START +ywang29-vrdb-test2-worker-0:514115:515701 [0] NCCL INFO ncclCommInitRank comm 0x55fddd377540 rank 0 nranks 8 cudaDev 0 nvmlDev 0 busId 101c0 commId 0xf6ae4b96a072379f - Init START +ywang29-vrdb-test2-worker-0:514116:515722 [1] NCCL INFO Setting affinity for GPU 1 to ff,ffff0000,00ffffff +ywang29-vrdb-test2-worker-0:514116:515722 [1] NCCL INFO NVLS multicast support is not available on dev 1 +ywang29-vrdb-test2-worker-0:514115:515701 [0] NCCL INFO Setting affinity for GPU 0 to ff,ffff0000,00ffffff +ywang29-vrdb-test2-worker-0:514115:515701 [0] NCCL INFO NVLS multicast support is not available on dev 0 +ywang29-vrdb-test2-worker-0:514121:515723 [6] NCCL INFO Setting affinity for GPU 6 to ffffff00,0000ffff,ff000000 +ywang29-vrdb-test2-worker-0:514121:515723 [6] NCCL INFO NVLS multicast support is not available on dev 6 +ywang29-vrdb-test2-worker-0:514117:515719 [2] NCCL INFO Setting affinity for GPU 2 to ff,ffff0000,00ffffff +ywang29-vrdb-test2-worker-0:514117:515719 [2] NCCL INFO NVLS multicast support is not available on dev 2 +ywang29-vrdb-test2-worker-0:514118:515724 [3] NCCL INFO Setting affinity for GPU 3 to ff,ffff0000,00ffffff +ywang29-vrdb-test2-worker-0:514118:515724 [3] NCCL INFO NVLS multicast support is not available on dev 3 +ywang29-vrdb-test2-worker-0:514122:515720 [7] NCCL INFO Setting affinity for GPU 7 to ffffff00,0000ffff,ff000000 +ywang29-vrdb-test2-worker-0:514122:515720 [7] NCCL INFO NVLS multicast support is not available on dev 7 +ywang29-vrdb-test2-worker-0:514119:515725 [4] NCCL INFO Setting affinity for GPU 4 to ffffff00,0000ffff,ff000000 +ywang29-vrdb-test2-worker-0:514119:515725 [4] NCCL INFO NVLS multicast support is not available on dev 4 +ywang29-vrdb-test2-worker-0:514120:515721 [5] NCCL INFO Setting affinity for GPU 5 to ffffff00,0000ffff,ff000000 +ywang29-vrdb-test2-worker-0:514120:515721 [5] NCCL INFO NVLS multicast support is not available on dev 5 +ywang29-vrdb-test2-worker-0:514121:515723 [6] NCCL INFO comm 0x5611899a0a00 rank 6 nRanks 8 nNodes 1 localRanks 8 localRank 6 MNNVL 0 +ywang29-vrdb-test2-worker-0:514120:515721 [5] NCCL INFO comm 0x55d70ee85fd0 rank 5 nRanks 8 nNodes 1 localRanks 8 localRank 5 MNNVL 0 +ywang29-vrdb-test2-worker-0:514119:515725 [4] NCCL INFO comm 0x5638e19a8a70 rank 4 nRanks 8 nNodes 1 localRanks 8 localRank 4 MNNVL 0 +ywang29-vrdb-test2-worker-0:514118:515724 [3] NCCL INFO comm 0x561646ab9390 rank 3 nRanks 8 nNodes 1 localRanks 8 localRank 3 MNNVL 0 +ywang29-vrdb-test2-worker-0:514117:515719 [2] NCCL INFO comm 0x557e86657520 rank 2 nRanks 8 nNodes 1 localRanks 8 localRank 2 MNNVL 0 +ywang29-vrdb-test2-worker-0:514116:515722 [1] NCCL INFO comm 0x558b540cdd80 rank 1 nRanks 8 nNodes 1 localRanks 8 localRank 1 MNNVL 0 +ywang29-vrdb-test2-worker-0:514115:515701 [0] NCCL INFO comm 0x55fddd377540 rank 0 nRanks 8 nNodes 1 localRanks 8 localRank 0 MNNVL 0 +ywang29-vrdb-test2-worker-0:514122:515720 [7] NCCL INFO comm 0x55de10eeb450 rank 7 nRanks 8 nNodes 1 localRanks 8 localRank 7 MNNVL 0 +ywang29-vrdb-test2-worker-0:514120:515721 [5] NCCL INFO Trees [0] 6/-1/-1->5->4 [1] 6/-1/-1->5->4 [2] 6/-1/-1->5->4 [3] 6/-1/-1->5->4 [4] 6/-1/-1->5->4 [5] 6/-1/-1->5->4 [6] 6/-1/-1->5->4 [7] 6/-1/-1->5->4 [8] 6/-1/-1->5->4 [9] 6/-1/-1->5->4 [10] 6/-1/-1->5->4 [11] 6/-1/-1->5->4 [12] 6/-1/-1->5->4 [13] 6/-1/-1->5->4 [14] 6/-1/-1->5->4 [15] 6/-1/-1->5->4 [16] 6/-1/-1->5->4 [17] 6/-1/-1->5->4 [18] 6/-1/-1->5->4 [19] 6/-1/-1->5->4 [20] 6/-1/-1->5->4 [21] 6/-1/-1->5->4 [22] 6/-1/-1->5->4 [23] 6/-1/-1->5->4 +ywang29-vrdb-test2-worker-0:514121:515723 [6] NCCL INFO Trees [0] 7/-1/-1->6->5 [1] 7/-1/-1->6->5 [2] 7/-1/-1->6->5 [3] 7/-1/-1->6->5 [4] 7/-1/-1->6->5 [5] 7/-1/-1->6->5 [6] 7/-1/-1->6->5 [7] 7/-1/-1->6->5 [8] 7/-1/-1->6->5 [9] 7/-1/-1->6->5 [10] 7/-1/-1->6->5 [11] 7/-1/-1->6->5 [12] 7/-1/-1->6->5 [13] 7/-1/-1->6->5 [14] 7/-1/-1->6->5 [15] 7/-1/-1->6->5 [16] 7/-1/-1->6->5 [17] 7/-1/-1->6->5 [18] 7/-1/-1->6->5 [19] 7/-1/-1->6->5 [20] 7/-1/-1->6->5 [21] 7/-1/-1->6->5 [22] 7/-1/-1->6->5 [23] 7/-1/-1->6->5 +ywang29-vrdb-test2-worker-0:514120:515721 [5] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:514115:515701 [0] NCCL INFO Channel 00/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:514115:515701 [0] NCCL INFO Channel 01/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:514118:515724 [3] NCCL INFO Trees [0] 4/-1/-1->3->2 [1] 4/-1/-1->3->2 [2] 4/-1/-1->3->2 [3] 4/-1/-1->3->2 [4] 4/-1/-1->3->2 [5] 4/-1/-1->3->2 [6] 4/-1/-1->3->2 [7] 4/-1/-1->3->2 [8] 4/-1/-1->3->2 [9] 4/-1/-1->3->2 [10] 4/-1/-1->3->2 [11] 4/-1/-1->3->2 [12] 4/-1/-1->3->2 [13] 4/-1/-1->3->2 [14] 4/-1/-1->3->2 [15] 4/-1/-1->3->2 [16] 4/-1/-1->3->2 [17] 4/-1/-1->3->2 [18] 4/-1/-1->3->2 [19] 4/-1/-1->3->2 [20] 4/-1/-1->3->2 [21] 4/-1/-1->3->2 [22] 4/-1/-1->3->2 [23] 4/-1/-1->3->2 +ywang29-vrdb-test2-worker-0:514121:515723 [6] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:514115:515701 [0] NCCL INFO Channel 02/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:514115:515701 [0] NCCL INFO Channel 03/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:514118:515724 [3] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:514115:515701 [0] NCCL INFO Channel 04/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:514115:515701 [0] NCCL INFO Channel 05/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:514115:515701 [0] NCCL INFO Channel 06/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:514117:515719 [2] NCCL INFO Trees [0] 3/-1/-1->2->1 [1] 3/-1/-1->2->1 [2] 3/-1/-1->2->1 [3] 3/-1/-1->2->1 [4] 3/-1/-1->2->1 [5] 3/-1/-1->2->1 [6] 3/-1/-1->2->1 [7] 3/-1/-1->2->1 [8] 3/-1/-1->2->1 [9] 3/-1/-1->2->1 [10] 3/-1/-1->2->1 [11] 3/-1/-1->2->1 [12] 3/-1/-1->2->1 [13] 3/-1/-1->2->1 [14] 3/-1/-1->2->1 [15] 3/-1/-1->2->1 [16] 3/-1/-1->2->1 [17] 3/-1/-1->2->1 [18] 3/-1/-1->2->1 [19] 3/-1/-1->2->1 [20] 3/-1/-1->2->1 [21] 3/-1/-1->2->1 [22] 3/-1/-1->2->1 [23] 3/-1/-1->2->1 +ywang29-vrdb-test2-worker-0:514115:515701 [0] NCCL INFO Channel 07/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:514116:515722 [1] NCCL INFO Trees [0] 2/-1/-1->1->0 [1] 2/-1/-1->1->0 [2] 2/-1/-1->1->0 [3] 2/-1/-1->1->0 [4] 2/-1/-1->1->0 [5] 2/-1/-1->1->0 [6] 2/-1/-1->1->0 [7] 2/-1/-1->1->0 [8] 2/-1/-1->1->0 [9] 2/-1/-1->1->0 [10] 2/-1/-1->1->0 [11] 2/-1/-1->1->0 [12] 2/-1/-1->1->0 [13] 2/-1/-1->1->0 [14] 2/-1/-1->1->0 [15] 2/-1/-1->1->0 [16] 2/-1/-1->1->0 [17] 2/-1/-1->1->0 [18] 2/-1/-1->1->0 [19] 2/-1/-1->1->0 [20] 2/-1/-1->1->0 [21] 2/-1/-1->1->0 [22] 2/-1/-1->1->0 [23] 2/-1/-1->1->0 +ywang29-vrdb-test2-worker-0:514119:515725 [4] NCCL INFO Trees [0] 5/-1/-1->4->3 [1] 5/-1/-1->4->3 [2] 5/-1/-1->4->3 [3] 5/-1/-1->4->3 [4] 5/-1/-1->4->3 [5] 5/-1/-1->4->3 [6] 5/-1/-1->4->3 [7] 5/-1/-1->4->3 [8] 5/-1/-1->4->3 [9] 5/-1/-1->4->3 [10] 5/-1/-1->4->3 [11] 5/-1/-1->4->3 [12] 5/-1/-1->4->3 [13] 5/-1/-1->4->3 [14] 5/-1/-1->4->3 [15] 5/-1/-1->4->3 [16] 5/-1/-1->4->3 [17] 5/-1/-1->4->3 [18] 5/-1/-1->4->3 [19] 5/-1/-1->4->3 [20] 5/-1/-1->4->3 [21] 5/-1/-1->4->3 [22] 5/-1/-1->4->3 [23] 5/-1/-1->4->3 +ywang29-vrdb-test2-worker-0:514115:515701 [0] NCCL INFO Channel 08/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:514117:515719 [2] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:514115:515701 [0] NCCL INFO Channel 09/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:514116:515722 [1] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:514115:515701 [0] NCCL INFO Channel 10/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:514119:515725 [4] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:514115:515701 [0] NCCL INFO Channel 11/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:514115:515701 [0] NCCL INFO Channel 12/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:514115:515701 [0] NCCL INFO Channel 13/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:514115:515701 [0] NCCL INFO Channel 14/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:514115:515701 [0] NCCL INFO Channel 15/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:514115:515701 [0] NCCL INFO Channel 16/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:514115:515701 [0] NCCL INFO Channel 17/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:514115:515701 [0] NCCL INFO Channel 18/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:514115:515701 [0] NCCL INFO Channel 19/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:514122:515720 [7] NCCL INFO Trees [0] -1/-1/-1->7->6 [1] -1/-1/-1->7->6 [2] -1/-1/-1->7->6 [3] -1/-1/-1->7->6 [4] -1/-1/-1->7->6 [5] -1/-1/-1->7->6 [6] -1/-1/-1->7->6 [7] -1/-1/-1->7->6 [8] -1/-1/-1->7->6 [9] -1/-1/-1->7->6 [10] -1/-1/-1->7->6 [11] -1/-1/-1->7->6 [12] -1/-1/-1->7->6 [13] -1/-1/-1->7->6 [14] -1/-1/-1->7->6 [15] -1/-1/-1->7->6 [16] -1/-1/-1->7->6 [17] -1/-1/-1->7->6 [18] -1/-1/-1->7->6 [19] -1/-1/-1->7->6 [20] -1/-1/-1->7->6 [21] -1/-1/-1->7->6 [22] -1/-1/-1->7->6 [23] -1/-1/-1->7->6 +ywang29-vrdb-test2-worker-0:514115:515701 [0] NCCL INFO Channel 20/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:514115:515701 [0] NCCL INFO Channel 21/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:514115:515701 [0] NCCL INFO Channel 22/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:514115:515701 [0] NCCL INFO Channel 23/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:514122:515720 [7] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:514115:515701 [0] NCCL INFO Trees [0] 1/-1/-1->0->-1 [1] 1/-1/-1->0->-1 [2] 1/-1/-1->0->-1 [3] 1/-1/-1->0->-1 [4] 1/-1/-1->0->-1 [5] 1/-1/-1->0->-1 [6] 1/-1/-1->0->-1 [7] 1/-1/-1->0->-1 [8] 1/-1/-1->0->-1 [9] 1/-1/-1->0->-1 [10] 1/-1/-1->0->-1 [11] 1/-1/-1->0->-1 [12] 1/-1/-1->0->-1 [13] 1/-1/-1->0->-1 [14] 1/-1/-1->0->-1 [15] 1/-1/-1->0->-1 [16] 1/-1/-1->0->-1 [17] 1/-1/-1->0->-1 [18] 1/-1/-1->0->-1 [19] 1/-1/-1->0->-1 [20] 1/-1/-1->0->-1 [21] 1/-1/-1->0->-1 [22] 1/-1/-1->0->-1 [23] 1/-1/-1->0->-1 +ywang29-vrdb-test2-worker-0:514115:515701 [0] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:514121:515723 [6] NCCL INFO Channel 00/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:515724 [3] NCCL INFO Channel 00/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:515719 [2] NCCL INFO Channel 00/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:515721 [5] NCCL INFO Channel 00/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:515723 [6] NCCL INFO Channel 01/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:515722 [1] NCCL INFO Channel 00/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:515724 [3] NCCL INFO Channel 01/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514115:515701 [0] NCCL INFO Channel 00/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:515719 [2] NCCL INFO Channel 01/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:515721 [5] NCCL INFO Channel 01/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:515723 [6] NCCL INFO Channel 02/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:515722 [1] NCCL INFO Channel 01/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:515724 [3] NCCL INFO Channel 02/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514115:515701 [0] NCCL INFO Channel 01/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:515720 [7] NCCL INFO Channel 00/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:515719 [2] NCCL INFO Channel 02/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:515725 [4] NCCL INFO Channel 00/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:515721 [5] NCCL INFO Channel 02/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:515723 [6] NCCL INFO Channel 03/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:515722 [1] NCCL INFO Channel 02/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:515724 [3] NCCL INFO Channel 03/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514115:515701 [0] NCCL INFO Channel 02/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:515720 [7] NCCL INFO Channel 01/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:515719 [2] NCCL INFO Channel 03/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:515721 [5] NCCL INFO Channel 03/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:515723 [6] NCCL INFO Channel 04/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:515725 [4] NCCL INFO Channel 01/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:515722 [1] NCCL INFO Channel 03/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:515724 [3] NCCL INFO Channel 04/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514115:515701 [0] NCCL INFO Channel 03/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:515720 [7] NCCL INFO Channel 02/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:515719 [2] NCCL INFO Channel 04/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:515721 [5] NCCL INFO Channel 04/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:515725 [4] NCCL INFO Channel 02/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:515722 [1] NCCL INFO Channel 04/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:515723 [6] NCCL INFO Channel 05/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:515724 [3] NCCL INFO Channel 05/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514115:515701 [0] NCCL INFO Channel 04/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:515720 [7] NCCL INFO Channel 03/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:515719 [2] NCCL INFO Channel 05/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:515721 [5] NCCL INFO Channel 05/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:515725 [4] NCCL INFO Channel 03/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:515722 [1] NCCL INFO Channel 05/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:515723 [6] NCCL INFO Channel 06/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:515724 [3] NCCL INFO Channel 06/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514115:515701 [0] NCCL INFO Channel 05/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:515720 [7] NCCL INFO Channel 04/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:515719 [2] NCCL INFO Channel 06/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:515721 [5] NCCL INFO Channel 06/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:515725 [4] NCCL INFO Channel 04/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:515722 [1] NCCL INFO Channel 06/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:515723 [6] NCCL INFO Channel 07/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:515724 [3] NCCL INFO Channel 07/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514115:515701 [0] NCCL INFO Channel 06/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:515720 [7] NCCL INFO Channel 05/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:515719 [2] NCCL INFO Channel 07/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:515721 [5] NCCL INFO Channel 07/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:515724 [3] NCCL INFO Channel 08/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:515725 [4] NCCL INFO Channel 05/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:515722 [1] NCCL INFO Channel 07/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:515723 [6] NCCL INFO Channel 08/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514115:515701 [0] NCCL INFO Channel 07/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:515719 [2] NCCL INFO Channel 08/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:515720 [7] NCCL INFO Channel 06/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:515721 [5] NCCL INFO Channel 08/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:515724 [3] NCCL INFO Channel 09/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:515722 [1] NCCL INFO Channel 08/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:515723 [6] NCCL INFO Channel 09/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:515725 [4] NCCL INFO Channel 06/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514115:515701 [0] NCCL INFO Channel 08/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:515719 [2] NCCL INFO Channel 09/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:515720 [7] NCCL INFO Channel 07/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:515721 [5] NCCL INFO Channel 09/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:515724 [3] NCCL INFO Channel 10/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:515722 [1] NCCL INFO Channel 09/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:515723 [6] NCCL INFO Channel 10/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:515725 [4] NCCL INFO Channel 07/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514115:515701 [0] NCCL INFO Channel 09/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:515719 [2] NCCL INFO Channel 10/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:515720 [7] NCCL INFO Channel 08/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:515721 [5] NCCL INFO Channel 10/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:515724 [3] NCCL INFO Channel 11/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:515722 [1] NCCL INFO Channel 10/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:515723 [6] NCCL INFO Channel 11/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514115:515701 [0] NCCL INFO Channel 10/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:515725 [4] NCCL INFO Channel 08/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:515719 [2] NCCL INFO Channel 11/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:515720 [7] NCCL INFO Channel 09/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:515721 [5] NCCL INFO Channel 11/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:515724 [3] NCCL INFO Channel 12/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:515722 [1] NCCL INFO Channel 11/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:515723 [6] NCCL INFO Channel 12/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514115:515701 [0] NCCL INFO Channel 11/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:515725 [4] NCCL INFO Channel 09/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:515719 [2] NCCL INFO Channel 12/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:515720 [7] NCCL INFO Channel 10/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:515721 [5] NCCL INFO Channel 12/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:515724 [3] NCCL INFO Channel 13/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:515722 [1] NCCL INFO Channel 12/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:515723 [6] NCCL INFO Channel 13/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514115:515701 [0] NCCL INFO Channel 12/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:515725 [4] NCCL INFO Channel 10/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:515719 [2] NCCL INFO Channel 13/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:515720 [7] NCCL INFO Channel 11/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:515721 [5] NCCL INFO Channel 13/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:515724 [3] NCCL INFO Channel 14/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:515722 [1] NCCL INFO Channel 13/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:515723 [6] NCCL INFO Channel 14/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514115:515701 [0] NCCL INFO Channel 13/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:515725 [4] NCCL INFO Channel 11/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:515719 [2] NCCL INFO Channel 14/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:515721 [5] NCCL INFO Channel 14/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:515720 [7] NCCL INFO Channel 12/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:515724 [3] NCCL INFO Channel 15/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:515722 [1] NCCL INFO Channel 14/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:515723 [6] NCCL INFO Channel 15/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514115:515701 [0] NCCL INFO Channel 14/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:515725 [4] NCCL INFO Channel 12/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:515719 [2] NCCL INFO Channel 15/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:515721 [5] NCCL INFO Channel 15/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:515720 [7] NCCL INFO Channel 13/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:515724 [3] NCCL INFO Channel 16/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:515722 [1] NCCL INFO Channel 15/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:515723 [6] NCCL INFO Channel 16/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514115:515701 [0] NCCL INFO Channel 15/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:515719 [2] NCCL INFO Channel 16/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:515725 [4] NCCL INFO Channel 13/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:515721 [5] NCCL INFO Channel 16/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:515720 [7] NCCL INFO Channel 14/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:515724 [3] NCCL INFO Channel 17/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:515722 [1] NCCL INFO Channel 16/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:515723 [6] NCCL INFO Channel 17/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514115:515701 [0] NCCL INFO Channel 16/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:515719 [2] NCCL INFO Channel 17/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:515725 [4] NCCL INFO Channel 14/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:515721 [5] NCCL INFO Channel 17/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:515724 [3] NCCL INFO Channel 18/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:515720 [7] NCCL INFO Channel 15/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:515722 [1] NCCL INFO Channel 17/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:515723 [6] NCCL INFO Channel 18/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514115:515701 [0] NCCL INFO Channel 17/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:515719 [2] NCCL INFO Channel 18/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:515725 [4] NCCL INFO Channel 15/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:515721 [5] NCCL INFO Channel 18/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:515724 [3] NCCL INFO Channel 19/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:515720 [7] NCCL INFO Channel 16/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:515722 [1] NCCL INFO Channel 18/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:515723 [6] NCCL INFO Channel 19/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514115:515701 [0] NCCL INFO Channel 18/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:515719 [2] NCCL INFO Channel 19/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:515721 [5] NCCL INFO Channel 19/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:515725 [4] NCCL INFO Channel 16/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:515724 [3] NCCL INFO Channel 20/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:515722 [1] NCCL INFO Channel 19/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:515720 [7] NCCL INFO Channel 17/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:515723 [6] NCCL INFO Channel 20/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514115:515701 [0] NCCL INFO Channel 19/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:515719 [2] NCCL INFO Channel 20/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:515721 [5] NCCL INFO Channel 20/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:515724 [3] NCCL INFO Channel 21/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:515725 [4] NCCL INFO Channel 17/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:515722 [1] NCCL INFO Channel 20/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514115:515701 [0] NCCL INFO Channel 20/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:515723 [6] NCCL INFO Channel 21/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:515720 [7] NCCL INFO Channel 18/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:515719 [2] NCCL INFO Channel 21/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:515721 [5] NCCL INFO Channel 21/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:515724 [3] NCCL INFO Channel 22/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:515725 [4] NCCL INFO Channel 18/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:515722 [1] NCCL INFO Channel 21/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514115:515701 [0] NCCL INFO Channel 21/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:515723 [6] NCCL INFO Channel 22/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:515720 [7] NCCL INFO Channel 19/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:515721 [5] NCCL INFO Channel 22/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:515719 [2] NCCL INFO Channel 22/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:515724 [3] NCCL INFO Channel 23/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:515725 [4] NCCL INFO Channel 19/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:515722 [1] NCCL INFO Channel 22/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514115:515701 [0] NCCL INFO Channel 22/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:515723 [6] NCCL INFO Channel 23/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:515720 [7] NCCL INFO Channel 20/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:515721 [5] NCCL INFO Channel 23/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:515719 [2] NCCL INFO Channel 23/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:515725 [4] NCCL INFO Channel 20/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:515722 [1] NCCL INFO Channel 23/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514115:515701 [0] NCCL INFO Channel 23/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:515720 [7] NCCL INFO Channel 21/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:515725 [4] NCCL INFO Channel 21/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:515720 [7] NCCL INFO Channel 22/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:515725 [4] NCCL INFO Channel 22/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:515725 [4] NCCL INFO Channel 23/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:515720 [7] NCCL INFO Channel 23/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:515719 [2] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:514118:515724 [3] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:514116:515722 [1] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:514115:515701 [0] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:514119:515725 [4] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:514117:515719 [2] NCCL INFO Channel 00/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:515719 [2] NCCL INFO Channel 01/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:515724 [3] NCCL INFO Channel 00/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:515719 [2] NCCL INFO Channel 02/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:515724 [3] NCCL INFO Channel 01/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:515719 [2] NCCL INFO Channel 03/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:515724 [3] NCCL INFO Channel 02/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:515719 [2] NCCL INFO Channel 04/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:515724 [3] NCCL INFO Channel 03/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:515719 [2] NCCL INFO Channel 05/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:515724 [3] NCCL INFO Channel 04/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:515720 [7] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:514122:515720 [7] NCCL INFO Channel 00/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:515721 [5] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:514121:515723 [6] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:514118:515724 [3] NCCL INFO Channel 05/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:515719 [2] NCCL INFO Channel 06/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:515720 [7] NCCL INFO Channel 01/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:515724 [3] NCCL INFO Channel 06/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:515719 [2] NCCL INFO Channel 07/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:515720 [7] NCCL INFO Channel 02/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:515724 [3] NCCL INFO Channel 07/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:515719 [2] NCCL INFO Channel 08/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:515724 [3] NCCL INFO Channel 08/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:515719 [2] NCCL INFO Channel 09/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:515724 [3] NCCL INFO Channel 09/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:515719 [2] NCCL INFO Channel 10/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:515724 [3] NCCL INFO Channel 10/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:515719 [2] NCCL INFO Channel 11/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:515724 [3] NCCL INFO Channel 11/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:515720 [7] NCCL INFO Channel 03/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:515719 [2] NCCL INFO Channel 12/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:515724 [3] NCCL INFO Channel 12/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:515720 [7] NCCL INFO Channel 04/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:515719 [2] NCCL INFO Channel 13/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:515724 [3] NCCL INFO Channel 13/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:515720 [7] NCCL INFO Channel 05/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:515719 [2] NCCL INFO Channel 14/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:515724 [3] NCCL INFO Channel 14/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:515720 [7] NCCL INFO Channel 06/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:515719 [2] NCCL INFO Channel 15/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:515724 [3] NCCL INFO Channel 15/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:515720 [7] NCCL INFO Channel 07/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:515719 [2] NCCL INFO Channel 16/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:515724 [3] NCCL INFO Channel 16/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:515720 [7] NCCL INFO Channel 08/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:515719 [2] NCCL INFO Channel 17/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:515724 [3] NCCL INFO Channel 17/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:515720 [7] NCCL INFO Channel 09/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:515719 [2] NCCL INFO Channel 18/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:515724 [3] NCCL INFO Channel 18/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:515720 [7] NCCL INFO Channel 10/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:515719 [2] NCCL INFO Channel 19/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:515724 [3] NCCL INFO Channel 19/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:515720 [7] NCCL INFO Channel 11/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:515719 [2] NCCL INFO Channel 20/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:515724 [3] NCCL INFO Channel 20/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:515720 [7] NCCL INFO Channel 12/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:515719 [2] NCCL INFO Channel 21/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:515724 [3] NCCL INFO Channel 21/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:515720 [7] NCCL INFO Channel 13/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:515719 [2] NCCL INFO Channel 22/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:515724 [3] NCCL INFO Channel 22/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:515719 [2] NCCL INFO Channel 23/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:515724 [3] NCCL INFO Channel 23/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:515725 [4] NCCL INFO Channel 00/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:515725 [4] NCCL INFO Channel 01/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:515725 [4] NCCL INFO Channel 02/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:515725 [4] NCCL INFO Channel 03/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:515725 [4] NCCL INFO Channel 04/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:515725 [4] NCCL INFO Channel 05/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:515723 [6] NCCL INFO Channel 00/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:515721 [5] NCCL INFO Channel 00/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:515725 [4] NCCL INFO Channel 06/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:515723 [6] NCCL INFO Channel 01/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:515721 [5] NCCL INFO Channel 01/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:515725 [4] NCCL INFO Channel 07/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:515723 [6] NCCL INFO Channel 02/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:515721 [5] NCCL INFO Channel 02/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:515725 [4] NCCL INFO Channel 08/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:515723 [6] NCCL INFO Channel 03/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:515721 [5] NCCL INFO Channel 03/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:515725 [4] NCCL INFO Channel 09/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:515723 [6] NCCL INFO Channel 04/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:515721 [5] NCCL INFO Channel 04/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:515725 [4] NCCL INFO Channel 10/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:515722 [1] NCCL INFO Channel 00/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:515723 [6] NCCL INFO Channel 05/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:515721 [5] NCCL INFO Channel 05/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:515722 [1] NCCL INFO Channel 01/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:515723 [6] NCCL INFO Channel 06/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:515721 [5] NCCL INFO Channel 06/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:515722 [1] NCCL INFO Channel 02/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:515723 [6] NCCL INFO Channel 07/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:515721 [5] NCCL INFO Channel 07/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:515722 [1] NCCL INFO Channel 03/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:515723 [6] NCCL INFO Channel 08/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:515721 [5] NCCL INFO Channel 08/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:515722 [1] NCCL INFO Channel 04/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:515723 [6] NCCL INFO Channel 09/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:515721 [5] NCCL INFO Channel 09/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:515723 [6] NCCL INFO Channel 10/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:515722 [1] NCCL INFO Channel 05/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:515721 [5] NCCL INFO Channel 10/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:515723 [6] NCCL INFO Channel 11/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:515722 [1] NCCL INFO Channel 06/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:515721 [5] NCCL INFO Channel 11/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:515723 [6] NCCL INFO Channel 12/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:515721 [5] NCCL INFO Channel 12/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:515722 [1] NCCL INFO Channel 07/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:515723 [6] NCCL INFO Channel 13/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:515720 [7] NCCL INFO Channel 14/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:515721 [5] NCCL INFO Channel 13/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:515722 [1] NCCL INFO Channel 08/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:515723 [6] NCCL INFO Channel 14/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:515720 [7] NCCL INFO Channel 15/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:515721 [5] NCCL INFO Channel 14/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:515722 [1] NCCL INFO Channel 09/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:515723 [6] NCCL INFO Channel 15/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:515720 [7] NCCL INFO Channel 16/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:515721 [5] NCCL INFO Channel 15/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:515723 [6] NCCL INFO Channel 16/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:515722 [1] NCCL INFO Channel 10/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:515720 [7] NCCL INFO Channel 17/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:515721 [5] NCCL INFO Channel 16/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:515723 [6] NCCL INFO Channel 17/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:515720 [7] NCCL INFO Channel 18/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:515722 [1] NCCL INFO Channel 11/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:515721 [5] NCCL INFO Channel 17/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:515723 [6] NCCL INFO Channel 18/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:515720 [7] NCCL INFO Channel 19/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:515722 [1] NCCL INFO Channel 12/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:515723 [6] NCCL INFO Channel 19/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:515720 [7] NCCL INFO Channel 20/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:515722 [1] NCCL INFO Channel 13/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:515723 [6] NCCL INFO Channel 20/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:515720 [7] NCCL INFO Channel 21/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:515725 [4] NCCL INFO Channel 11/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:515723 [6] NCCL INFO Channel 21/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:515722 [1] NCCL INFO Channel 14/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:515720 [7] NCCL INFO Channel 22/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:515725 [4] NCCL INFO Channel 12/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:515723 [6] NCCL INFO Channel 22/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:515722 [1] NCCL INFO Channel 15/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:515720 [7] NCCL INFO Channel 23/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:515725 [4] NCCL INFO Channel 13/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:515723 [6] NCCL INFO Channel 23/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:515722 [1] NCCL INFO Channel 16/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:515725 [4] NCCL INFO Channel 14/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:515722 [1] NCCL INFO Channel 17/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:515725 [4] NCCL INFO Channel 15/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:515722 [1] NCCL INFO Channel 18/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:515725 [4] NCCL INFO Channel 16/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:515725 [4] NCCL INFO Channel 17/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:515722 [1] NCCL INFO Channel 19/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:515725 [4] NCCL INFO Channel 18/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:515722 [1] NCCL INFO Channel 20/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:515725 [4] NCCL INFO Channel 19/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:515722 [1] NCCL INFO Channel 21/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:515725 [4] NCCL INFO Channel 20/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:515721 [5] NCCL INFO Channel 18/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:515725 [4] NCCL INFO Channel 21/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:515722 [1] NCCL INFO Channel 22/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:515721 [5] NCCL INFO Channel 19/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:515725 [4] NCCL INFO Channel 22/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:515721 [5] NCCL INFO Channel 20/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:515722 [1] NCCL INFO Channel 23/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:515725 [4] NCCL INFO Channel 23/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:515721 [5] NCCL INFO Channel 21/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:515721 [5] NCCL INFO Channel 22/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:515721 [5] NCCL INFO Channel 23/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514115:515701 [0] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:514115:515701 [0] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:514115:515701 [0] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:514116:515722 [1] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:514116:515722 [1] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:514116:515722 [1] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:514117:515719 [2] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:514117:515719 [2] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:514117:515719 [2] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:514118:515724 [3] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:514118:515724 [3] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:514118:515724 [3] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:514122:515720 [7] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:514122:515720 [7] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:514122:515720 [7] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:514121:515723 [6] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:514120:515721 [5] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:514121:515723 [6] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:514121:515723 [6] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:514120:515721 [5] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:514120:515721 [5] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:514119:515725 [4] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:514119:515725 [4] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:514119:515725 [4] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:514117:515719 [2] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test2-worker-0:514118:515724 [3] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test2-worker-0:514118:515724 [3] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test2-worker-0:514118:515724 [3] NCCL INFO ncclCommInitRank comm 0x561646ab9390 rank 3 nranks 8 cudaDev 3 nvmlDev 3 busId 201d0 commId 0xf6ae4b96a072379f - Init COMPLETE +ywang29-vrdb-test2-worker-0:514117:515719 [2] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test2-worker-0:514119:515725 [4] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test2-worker-0:514121:515723 [6] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test2-worker-0:514122:515720 [7] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test2-worker-0:514120:515721 [5] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test2-worker-0:514117:515719 [2] NCCL INFO ncclCommInitRank comm 0x557e86657520 rank 2 nranks 8 cudaDev 2 nvmlDev 2 busId 201c0 commId 0xf6ae4b96a072379f - Init COMPLETE +ywang29-vrdb-test2-worker-0:514119:515725 [4] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test2-worker-0:514121:515723 [6] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test2-worker-0:514122:515720 [7] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test2-worker-0:514120:515721 [5] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test2-worker-0:514119:515725 [4] NCCL INFO ncclCommInitRank comm 0x5638e19a8a70 rank 4 nranks 8 cudaDev 4 nvmlDev 4 busId 901c0 commId 0xf6ae4b96a072379f - Init COMPLETE +ywang29-vrdb-test2-worker-0:514122:515720 [7] NCCL INFO ncclCommInitRank comm 0x55de10eeb450 rank 7 nranks 8 cudaDev 7 nvmlDev 7 busId a01d0 commId 0xf6ae4b96a072379f - Init COMPLETE +ywang29-vrdb-test2-worker-0:514121:515723 [6] NCCL INFO ncclCommInitRank comm 0x5611899a0a00 rank 6 nranks 8 cudaDev 6 nvmlDev 6 busId a01c0 commId 0xf6ae4b96a072379f - Init COMPLETE +ywang29-vrdb-test2-worker-0:514120:515721 [5] NCCL INFO ncclCommInitRank comm 0x55d70ee85fd0 rank 5 nranks 8 cudaDev 5 nvmlDev 5 busId 901d0 commId 0xf6ae4b96a072379f - Init COMPLETE +ywang29-vrdb-test2-worker-0:514115:515701 [0] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test2-worker-0:514115:515701 [0] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test2-worker-0:514115:515701 [0] NCCL INFO ncclCommInitRank comm 0x55fddd377540 rank 0 nranks 8 cudaDev 0 nvmlDev 0 busId 101c0 commId 0xf6ae4b96a072379f - Init COMPLETE +ywang29-vrdb-test2-worker-0:514116:515722 [1] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test2-worker-0:514116:515722 [1] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test2-worker-0:514116:515722 [1] NCCL INFO ncclCommInitRank comm 0x558b540cdd80 rank 1 nranks 8 cudaDev 1 nvmlDev 1 busId 101d0 commId 0xf6ae4b96a072379f - Init COMPLETE +[2025-10-12 07:13:42,010] [INFO] [partition_parameters.py:343:__exit__] finished initializing model - num_params = 459, num_elems = 0.99B +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +[2025-10-12 07:13:45,486] [INFO] [partition_parameters.py:343:__exit__] finished initializing model - num_params = 907, num_elems = 1.42B +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +TinyLlavaForConditionalGeneration( + (language_model): Qwen2ForCausalLM( + (model): Qwen2Model( + (embed_tokens): Embedding(151936, 896) + (layers): ModuleList( + (0-23): 24 x Qwen2DecoderLayer( + (self_attn): Qwen2FlashAttention2( + (q_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=True) + (k_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=128, bias=True) + (v_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=128, bias=True) + (o_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=False) + (rotary_emb): Qwen2RotaryEmbedding() + ) + (mlp): Qwen2MLP( + (gate_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=4864, bias=False) + (up_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=4864, bias=False) + (down_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=4864, out_features=896, bias=False) + (act_fn): SiLU() + ) + (input_layernorm): Qwen2RMSNorm() + (post_attention_layernorm): Qwen2RMSNorm() + ) + ) + (norm): Qwen2RMSNorm() + ) + (lm_head): Linear(in_features=896, out_features=151936, bias=False) + ) + (vision_tower): SIGLIPVisionTower( + (_vision_tower): SiglipVisionModel( + (vision_model): SiglipVisionTransformer( + (embeddings): SiglipVisionEmbeddings( + (patch_embedding): Conv2d(3, 1152, kernel_size=(14, 14), stride=(14, 14), padding=valid) + (position_embedding): Embedding(729, 1152) + ) + (encoder): SiglipEncoder( + (layers): ModuleList( + (0-26): 27 x SiglipEncoderLayer( + (self_attn): SiglipAttention( + (k_proj): Linear(in_features=1152, out_features=1152, bias=True) + (v_proj): Linear(in_features=1152, out_features=1152, bias=True) + (q_proj): Linear(in_features=1152, out_features=1152, bias=True) + (out_proj): Linear(in_features=1152, out_features=1152, bias=True) + ) + (layer_norm1): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + (layer_norm2): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + ) + ) + ) + (post_layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (head): SiglipMultiheadAttentionPoolingHead( + (attention): MultiheadAttention( + (out_proj): NonDynamicallyQuantizableLinear(in_features=1152, out_features=1152, bias=True) + ) + (layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + ) + ) + ) + ) + (connector): MLPConnector( + (_connector): Sequential( + (0): SupermaskLinearSparsity_SoftForward_Normal(in_features=1152, out_features=896, bias=True) + (1): GELU(approximate='none') + (2): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=True) + ) + ) +) +Pre-training init language_model.model.layers.0.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.0.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.0.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.0.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.0.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.0.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.0.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.mlp.down_proj.scores: Mean=3.000000 +Pre-training init connector._connector.0.scores: Mean=3.000005 +Pre-training init connector._connector.2.scores: Mean=2.999970 +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +2025-10-12 07:32:50,819 | INFO: Total Parameters: 1283756736, Total Trainable Parameters: 359661568 +2025-10-12 07:32:50,823 | INFO: Trainable Parameters: +language_model.model.layers.0.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.0.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.0.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.0.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.0.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.0.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.0.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.1.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.1.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.1.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.1.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.1.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.1.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.1.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.2.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.2.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.2.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.2.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.2.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.2.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.2.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.3.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.3.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.3.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.3.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.3.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.3.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.3.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.4.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.4.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.4.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.4.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.4.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.4.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.4.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.5.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.5.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.5.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.5.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.5.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.5.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.5.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.6.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.6.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.6.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.6.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.6.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.6.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.6.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.7.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.7.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.7.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.7.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.7.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.7.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.7.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.8.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.8.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.8.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.8.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.8.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.8.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.8.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.9.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.9.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.9.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.9.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.9.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.9.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.9.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.10.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.10.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.10.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.10.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.10.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.10.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.10.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.11.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.11.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.11.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.11.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.11.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.11.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.11.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.12.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.12.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.12.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.12.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.12.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.12.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.12.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.13.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.13.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.13.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.13.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.13.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.13.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.13.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.14.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.14.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.14.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.14.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.14.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.14.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.14.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.15.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.15.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.15.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.15.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.15.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.15.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.15.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.16.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.16.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.16.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.16.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.16.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.16.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.16.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.17.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.17.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.17.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.17.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.17.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.17.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.17.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.18.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.18.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.18.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.18.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.18.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.18.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.18.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.19.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.19.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.19.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.19.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.19.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.19.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.19.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.20.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.20.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.20.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.20.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.20.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.20.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.20.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.21.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.21.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.21.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.21.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.21.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.21.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.21.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.22.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.22.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.22.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.22.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.22.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.22.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.22.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.23.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.23.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.23.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.23.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.23.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.23.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.23.mlp.down_proj.scores: 4358144 parameters +connector._connector.0.scores: 1032192 parameters +connector._connector.2.scores: 802816 parameters +Parameter Offload: Total persistent parameters: 486464 in 403 params + 0%| | 0/520 [00:007->6 [1] -1/-1/-1->7->6 [2] -1/-1/-1->7->6 [3] -1/-1/-1->7->6 [4] -1/-1/-1->7->6 [5] -1/-1/-1->7->6 [6] -1/-1/-1->7->6 [7] -1/-1/-1->7->6 [8] -1/-1/-1->7->6 [9] -1/-1/-1->7->6 [10] -1/-1/-1->7->6 [11] -1/-1/-1->7->6 [12] -1/-1/-1->7->6 [13] -1/-1/-1->7->6 [14] -1/-1/-1->7->6 [15] -1/-1/-1->7->6 [16] -1/-1/-1->7->6 [17] -1/-1/-1->7->6 [18] -1/-1/-1->7->6 [19] -1/-1/-1->7->6 [20] -1/-1/-1->7->6 [21] -1/-1/-1->7->6 [22] -1/-1/-1->7->6 [23] -1/-1/-1->7->6 +ywang29-vrdb-test2-worker-0:514115:520999 [0] NCCL INFO Channel 07/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:514122:521006 [7] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:514121:521004 [6] NCCL INFO Trees [0] 7/-1/-1->6->5 [1] 7/-1/-1->6->5 [2] 7/-1/-1->6->5 [3] 7/-1/-1->6->5 [4] 7/-1/-1->6->5 [5] 7/-1/-1->6->5 [6] 7/-1/-1->6->5 [7] 7/-1/-1->6->5 [8] 7/-1/-1->6->5 [9] 7/-1/-1->6->5 [10] 7/-1/-1->6->5 [11] 7/-1/-1->6->5 [12] 7/-1/-1->6->5 [13] 7/-1/-1->6->5 [14] 7/-1/-1->6->5 [15] 7/-1/-1->6->5 [16] 7/-1/-1->6->5 [17] 7/-1/-1->6->5 [18] 7/-1/-1->6->5 [19] 7/-1/-1->6->5 [20] 7/-1/-1->6->5 [21] 7/-1/-1->6->5 [22] 7/-1/-1->6->5 [23] 7/-1/-1->6->5 +ywang29-vrdb-test2-worker-0:514121:521004 [6] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:514115:520999 [0] NCCL INFO Channel 08/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:514120:521005 [5] NCCL INFO Trees [0] 6/-1/-1->5->4 [1] 6/-1/-1->5->4 [2] 6/-1/-1->5->4 [3] 6/-1/-1->5->4 [4] 6/-1/-1->5->4 [5] 6/-1/-1->5->4 [6] 6/-1/-1->5->4 [7] 6/-1/-1->5->4 [8] 6/-1/-1->5->4 [9] 6/-1/-1->5->4 [10] 6/-1/-1->5->4 [11] 6/-1/-1->5->4 [12] 6/-1/-1->5->4 [13] 6/-1/-1->5->4 [14] 6/-1/-1->5->4 [15] 6/-1/-1->5->4 [16] 6/-1/-1->5->4 [17] 6/-1/-1->5->4 [18] 6/-1/-1->5->4 [19] 6/-1/-1->5->4 [20] 6/-1/-1->5->4 [21] 6/-1/-1->5->4 [22] 6/-1/-1->5->4 [23] 6/-1/-1->5->4 +ywang29-vrdb-test2-worker-0:514119:521001 [4] NCCL INFO Trees [0] 5/-1/-1->4->3 [1] 5/-1/-1->4->3 [2] 5/-1/-1->4->3 [3] 5/-1/-1->4->3 [4] 5/-1/-1->4->3 [5] 5/-1/-1->4->3 [6] 5/-1/-1->4->3 [7] 5/-1/-1->4->3 [8] 5/-1/-1->4->3 [9] 5/-1/-1->4->3 [10] 5/-1/-1->4->3 [11] 5/-1/-1->4->3 [12] 5/-1/-1->4->3 [13] 5/-1/-1->4->3 [14] 5/-1/-1->4->3 [15] 5/-1/-1->4->3 [16] 5/-1/-1->4->3 [17] 5/-1/-1->4->3 [18] 5/-1/-1->4->3 [19] 5/-1/-1->4->3 [20] 5/-1/-1->4->3 [21] 5/-1/-1->4->3 [22] 5/-1/-1->4->3 [23] 5/-1/-1->4->3 +ywang29-vrdb-test2-worker-0:514115:520999 [0] NCCL INFO Channel 09/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:514120:521005 [5] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:514119:521001 [4] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:514118:521000 [3] NCCL INFO Trees [0] 4/-1/-1->3->2 [1] 4/-1/-1->3->2 [2] 4/-1/-1->3->2 [3] 4/-1/-1->3->2 [4] 4/-1/-1->3->2 [5] 4/-1/-1->3->2 [6] 4/-1/-1->3->2 [7] 4/-1/-1->3->2 [8] 4/-1/-1->3->2 [9] 4/-1/-1->3->2 [10] 4/-1/-1->3->2 [11] 4/-1/-1->3->2 [12] 4/-1/-1->3->2 [13] 4/-1/-1->3->2 [14] 4/-1/-1->3->2 [15] 4/-1/-1->3->2 [16] 4/-1/-1->3->2 [17] 4/-1/-1->3->2 [18] 4/-1/-1->3->2 [19] 4/-1/-1->3->2 [20] 4/-1/-1->3->2 [21] 4/-1/-1->3->2 [22] 4/-1/-1->3->2 [23] 4/-1/-1->3->2 +ywang29-vrdb-test2-worker-0:514115:520999 [0] NCCL INFO Channel 10/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:514118:521000 [3] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:514115:520999 [0] NCCL INFO Channel 11/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:514115:520999 [0] NCCL INFO Channel 12/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:514115:520999 [0] NCCL INFO Channel 13/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:514115:520999 [0] NCCL INFO Channel 14/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:514115:520999 [0] NCCL INFO Channel 15/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:514116:521002 [1] NCCL INFO Trees [0] 2/-1/-1->1->0 [1] 2/-1/-1->1->0 [2] 2/-1/-1->1->0 [3] 2/-1/-1->1->0 [4] 2/-1/-1->1->0 [5] 2/-1/-1->1->0 [6] 2/-1/-1->1->0 [7] 2/-1/-1->1->0 [8] 2/-1/-1->1->0 [9] 2/-1/-1->1->0 [10] 2/-1/-1->1->0 [11] 2/-1/-1->1->0 [12] 2/-1/-1->1->0 [13] 2/-1/-1->1->0 [14] 2/-1/-1->1->0 [15] 2/-1/-1->1->0 [16] 2/-1/-1->1->0 [17] 2/-1/-1->1->0 [18] 2/-1/-1->1->0 [19] 2/-1/-1->1->0 [20] 2/-1/-1->1->0 [21] 2/-1/-1->1->0 [22] 2/-1/-1->1->0 [23] 2/-1/-1->1->0 +ywang29-vrdb-test2-worker-0:514115:520999 [0] NCCL INFO Channel 16/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:514117:521003 [2] NCCL INFO Trees [0] 3/-1/-1->2->1 [1] 3/-1/-1->2->1 [2] 3/-1/-1->2->1 [3] 3/-1/-1->2->1 [4] 3/-1/-1->2->1 [5] 3/-1/-1->2->1 [6] 3/-1/-1->2->1 [7] 3/-1/-1->2->1 [8] 3/-1/-1->2->1 [9] 3/-1/-1->2->1 [10] 3/-1/-1->2->1 [11] 3/-1/-1->2->1 [12] 3/-1/-1->2->1 [13] 3/-1/-1->2->1 [14] 3/-1/-1->2->1 [15] 3/-1/-1->2->1 [16] 3/-1/-1->2->1 [17] 3/-1/-1->2->1 [18] 3/-1/-1->2->1 [19] 3/-1/-1->2->1 [20] 3/-1/-1->2->1 [21] 3/-1/-1->2->1 [22] 3/-1/-1->2->1 [23] 3/-1/-1->2->1 +ywang29-vrdb-test2-worker-0:514115:520999 [0] NCCL INFO Channel 17/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:514116:521002 [1] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:514115:520999 [0] NCCL INFO Channel 18/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:514117:521003 [2] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:514115:520999 [0] NCCL INFO Channel 19/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:514115:520999 [0] NCCL INFO Channel 20/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:514115:520999 [0] NCCL INFO Channel 21/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:514115:520999 [0] NCCL INFO Channel 22/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:514115:520999 [0] NCCL INFO Channel 23/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:514115:520999 [0] NCCL INFO Trees [0] 1/-1/-1->0->-1 [1] 1/-1/-1->0->-1 [2] 1/-1/-1->0->-1 [3] 1/-1/-1->0->-1 [4] 1/-1/-1->0->-1 [5] 1/-1/-1->0->-1 [6] 1/-1/-1->0->-1 [7] 1/-1/-1->0->-1 [8] 1/-1/-1->0->-1 [9] 1/-1/-1->0->-1 [10] 1/-1/-1->0->-1 [11] 1/-1/-1->0->-1 [12] 1/-1/-1->0->-1 [13] 1/-1/-1->0->-1 [14] 1/-1/-1->0->-1 [15] 1/-1/-1->0->-1 [16] 1/-1/-1->0->-1 [17] 1/-1/-1->0->-1 [18] 1/-1/-1->0->-1 [19] 1/-1/-1->0->-1 [20] 1/-1/-1->0->-1 [21] 1/-1/-1->0->-1 [22] 1/-1/-1->0->-1 [23] 1/-1/-1->0->-1 +ywang29-vrdb-test2-worker-0:514115:520999 [0] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:514122:521006 [7] NCCL INFO Channel 00/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:521001 [4] NCCL INFO Channel 00/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:521004 [6] NCCL INFO Channel 00/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:521006 [7] NCCL INFO Channel 01/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:521005 [5] NCCL INFO Channel 00/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514115:520999 [0] NCCL INFO Channel 00/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:521001 [4] NCCL INFO Channel 01/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:521004 [6] NCCL INFO Channel 01/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:521006 [7] NCCL INFO Channel 02/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:521005 [5] NCCL INFO Channel 01/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514115:520999 [0] NCCL INFO Channel 01/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:521001 [4] NCCL INFO Channel 02/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:521004 [6] NCCL INFO Channel 02/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:521006 [7] NCCL INFO Channel 03/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:521005 [5] NCCL INFO Channel 02/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514115:520999 [0] NCCL INFO Channel 02/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:521001 [4] NCCL INFO Channel 03/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:521003 [2] NCCL INFO Channel 00/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:521004 [6] NCCL INFO Channel 03/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:521002 [1] NCCL INFO Channel 00/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:521006 [7] NCCL INFO Channel 04/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:521005 [5] NCCL INFO Channel 03/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:521001 [4] NCCL INFO Channel 04/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:521003 [2] NCCL INFO Channel 01/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:521004 [6] NCCL INFO Channel 04/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:521006 [7] NCCL INFO Channel 05/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:521005 [5] NCCL INFO Channel 04/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:521002 [1] NCCL INFO Channel 01/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:521001 [4] NCCL INFO Channel 05/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:521003 [2] NCCL INFO Channel 02/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:521004 [6] NCCL INFO Channel 05/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:521006 [7] NCCL INFO Channel 06/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:521005 [5] NCCL INFO Channel 05/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:521002 [1] NCCL INFO Channel 02/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:521001 [4] NCCL INFO Channel 06/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:521004 [6] NCCL INFO Channel 06/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:521003 [2] NCCL INFO Channel 03/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:521006 [7] NCCL INFO Channel 07/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:521005 [5] NCCL INFO Channel 06/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:521001 [4] NCCL INFO Channel 07/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:521002 [1] NCCL INFO Channel 03/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:521004 [6] NCCL INFO Channel 07/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:521003 [2] NCCL INFO Channel 04/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:521005 [5] NCCL INFO Channel 07/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:521006 [7] NCCL INFO Channel 08/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:521001 [4] NCCL INFO Channel 08/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:521002 [1] NCCL INFO Channel 04/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:521004 [6] NCCL INFO Channel 08/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:521003 [2] NCCL INFO Channel 05/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:521005 [5] NCCL INFO Channel 08/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:521001 [4] NCCL INFO Channel 09/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:521006 [7] NCCL INFO Channel 09/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:521002 [1] NCCL INFO Channel 05/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:521004 [6] NCCL INFO Channel 09/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:521003 [2] NCCL INFO Channel 06/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:521005 [5] NCCL INFO Channel 09/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:521001 [4] NCCL INFO Channel 10/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:521006 [7] NCCL INFO Channel 10/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:521002 [1] NCCL INFO Channel 06/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:521004 [6] NCCL INFO Channel 10/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:521003 [2] NCCL INFO Channel 07/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:521005 [5] NCCL INFO Channel 10/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:521001 [4] NCCL INFO Channel 11/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:521006 [7] NCCL INFO Channel 11/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:521002 [1] NCCL INFO Channel 07/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:521004 [6] NCCL INFO Channel 11/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:521005 [5] NCCL INFO Channel 11/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:521003 [2] NCCL INFO Channel 08/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:521001 [4] NCCL INFO Channel 12/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:521006 [7] NCCL INFO Channel 12/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:521002 [1] NCCL INFO Channel 08/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:521004 [6] NCCL INFO Channel 12/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:521005 [5] NCCL INFO Channel 12/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:521001 [4] NCCL INFO Channel 13/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:521003 [2] NCCL INFO Channel 09/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:521006 [7] NCCL INFO Channel 13/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:521000 [3] NCCL INFO Channel 00/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:521002 [1] NCCL INFO Channel 09/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:521004 [6] NCCL INFO Channel 13/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:521005 [5] NCCL INFO Channel 13/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:521001 [4] NCCL INFO Channel 14/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:521003 [2] NCCL INFO Channel 10/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:521006 [7] NCCL INFO Channel 14/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:521000 [3] NCCL INFO Channel 01/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:521004 [6] NCCL INFO Channel 14/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:521002 [1] NCCL INFO Channel 10/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:521005 [5] NCCL INFO Channel 14/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:521001 [4] NCCL INFO Channel 15/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:521003 [2] NCCL INFO Channel 11/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:521006 [7] NCCL INFO Channel 15/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:521000 [3] NCCL INFO Channel 02/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:521004 [6] NCCL INFO Channel 15/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:521002 [1] NCCL INFO Channel 11/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:521005 [5] NCCL INFO Channel 15/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:521001 [4] NCCL INFO Channel 16/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:521003 [2] NCCL INFO Channel 12/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:521006 [7] NCCL INFO Channel 16/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:521000 [3] NCCL INFO Channel 03/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:521004 [6] NCCL INFO Channel 16/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:521002 [1] NCCL INFO Channel 12/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:521001 [4] NCCL INFO Channel 17/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514115:520999 [0] NCCL INFO Channel 03/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:521006 [7] NCCL INFO Channel 17/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:521005 [5] NCCL INFO Channel 16/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:521003 [2] NCCL INFO Channel 13/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:521000 [3] NCCL INFO Channel 04/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:521004 [6] NCCL INFO Channel 17/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:521002 [1] NCCL INFO Channel 13/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:521001 [4] NCCL INFO Channel 18/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514115:520999 [0] NCCL INFO Channel 04/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:521006 [7] NCCL INFO Channel 18/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:521005 [5] NCCL INFO Channel 17/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:521003 [2] NCCL INFO Channel 14/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:521000 [3] NCCL INFO Channel 05/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:521004 [6] NCCL INFO Channel 18/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:521002 [1] NCCL INFO Channel 14/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:521001 [4] NCCL INFO Channel 19/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514115:520999 [0] NCCL INFO Channel 05/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:521006 [7] NCCL INFO Channel 19/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:521005 [5] NCCL INFO Channel 18/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:521004 [6] NCCL INFO Channel 19/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:521003 [2] NCCL INFO Channel 15/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:521000 [3] NCCL INFO Channel 06/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:521002 [1] NCCL INFO Channel 15/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:521001 [4] NCCL INFO Channel 20/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514115:520999 [0] NCCL INFO Channel 06/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:521006 [7] NCCL INFO Channel 20/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:521005 [5] NCCL INFO Channel 19/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:521004 [6] NCCL INFO Channel 20/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:521000 [3] NCCL INFO Channel 07/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:521003 [2] NCCL INFO Channel 16/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:521002 [1] NCCL INFO Channel 16/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:521001 [4] NCCL INFO Channel 21/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514115:520999 [0] NCCL INFO Channel 07/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:521006 [7] NCCL INFO Channel 21/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:521005 [5] NCCL INFO Channel 20/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:521004 [6] NCCL INFO Channel 21/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:521000 [3] NCCL INFO Channel 08/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:521003 [2] NCCL INFO Channel 17/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:521001 [4] NCCL INFO Channel 22/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:521002 [1] NCCL INFO Channel 17/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514115:520999 [0] NCCL INFO Channel 08/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:521006 [7] NCCL INFO Channel 22/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:521005 [5] NCCL INFO Channel 21/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:521000 [3] NCCL INFO Channel 09/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:521003 [2] NCCL INFO Channel 18/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:521004 [6] NCCL INFO Channel 22/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:521001 [4] NCCL INFO Channel 23/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514115:520999 [0] NCCL INFO Channel 09/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:521002 [1] NCCL INFO Channel 18/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:521006 [7] NCCL INFO Channel 23/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:521005 [5] NCCL INFO Channel 22/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:521000 [3] NCCL INFO Channel 10/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:521003 [2] NCCL INFO Channel 19/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:521004 [6] NCCL INFO Channel 23/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514115:520999 [0] NCCL INFO Channel 10/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:521005 [5] NCCL INFO Channel 23/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:521002 [1] NCCL INFO Channel 19/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:521000 [3] NCCL INFO Channel 11/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:521003 [2] NCCL INFO Channel 20/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514115:520999 [0] NCCL INFO Channel 11/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:521002 [1] NCCL INFO Channel 20/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:521000 [3] NCCL INFO Channel 12/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:521003 [2] NCCL INFO Channel 21/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514115:520999 [0] NCCL INFO Channel 12/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:521000 [3] NCCL INFO Channel 13/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:521003 [2] NCCL INFO Channel 22/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514115:520999 [0] NCCL INFO Channel 13/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:521000 [3] NCCL INFO Channel 14/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:521002 [1] NCCL INFO Channel 21/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514115:520999 [0] NCCL INFO Channel 14/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:521003 [2] NCCL INFO Channel 23/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:521000 [3] NCCL INFO Channel 15/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:521002 [1] NCCL INFO Channel 22/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514115:520999 [0] NCCL INFO Channel 15/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:521000 [3] NCCL INFO Channel 16/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:521002 [1] NCCL INFO Channel 23/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514115:520999 [0] NCCL INFO Channel 16/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:521000 [3] NCCL INFO Channel 17/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514115:520999 [0] NCCL INFO Channel 17/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:521000 [3] NCCL INFO Channel 18/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514115:520999 [0] NCCL INFO Channel 18/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:521000 [3] NCCL INFO Channel 19/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514115:520999 [0] NCCL INFO Channel 19/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514115:520999 [0] NCCL INFO Channel 20/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:521000 [3] NCCL INFO Channel 20/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514115:520999 [0] NCCL INFO Channel 21/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:521000 [3] NCCL INFO Channel 21/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514115:520999 [0] NCCL INFO Channel 22/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:521000 [3] NCCL INFO Channel 22/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514115:520999 [0] NCCL INFO Channel 23/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:521000 [3] NCCL INFO Channel 23/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:521003 [2] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:514118:521000 [3] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:514119:521001 [4] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:514116:521002 [1] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:514115:520999 [0] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:514122:521006 [7] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:514122:521006 [7] NCCL INFO Channel 00/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:521005 [5] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:514121:521004 [6] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:514122:521006 [7] NCCL INFO Channel 01/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:521006 [7] NCCL INFO Channel 02/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:521006 [7] NCCL INFO Channel 03/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:521006 [7] NCCL INFO Channel 04/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:521006 [7] NCCL INFO Channel 05/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:521006 [7] NCCL INFO Channel 06/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:521006 [7] NCCL INFO Channel 07/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:521006 [7] NCCL INFO Channel 08/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:521003 [2] NCCL INFO Channel 00/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:521006 [7] NCCL INFO Channel 09/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:521003 [2] NCCL INFO Channel 01/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:521006 [7] NCCL INFO Channel 10/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:521003 [2] NCCL INFO Channel 02/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:521006 [7] NCCL INFO Channel 11/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:521006 [7] NCCL INFO Channel 12/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:521003 [2] NCCL INFO Channel 03/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:521006 [7] NCCL INFO Channel 13/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:521003 [2] NCCL INFO Channel 04/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:521006 [7] NCCL INFO Channel 14/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:521003 [2] NCCL INFO Channel 05/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:521006 [7] NCCL INFO Channel 15/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:521003 [2] NCCL INFO Channel 06/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:521006 [7] NCCL INFO Channel 16/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:521003 [2] NCCL INFO Channel 07/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:521006 [7] NCCL INFO Channel 17/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:521003 [2] NCCL INFO Channel 08/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:521006 [7] NCCL INFO Channel 18/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:521003 [2] NCCL INFO Channel 09/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:521006 [7] NCCL INFO Channel 19/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:521003 [2] NCCL INFO Channel 10/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:521006 [7] NCCL INFO Channel 20/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:521003 [2] NCCL INFO Channel 11/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:521006 [7] NCCL INFO Channel 21/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:521003 [2] NCCL INFO Channel 12/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:521006 [7] NCCL INFO Channel 22/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:521003 [2] NCCL INFO Channel 13/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:521001 [4] NCCL INFO Channel 00/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514122:521006 [7] NCCL INFO Channel 23/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:521003 [2] NCCL INFO Channel 14/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:521002 [1] NCCL INFO Channel 00/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:521001 [4] NCCL INFO Channel 01/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:521003 [2] NCCL INFO Channel 15/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:521004 [6] NCCL INFO Channel 00/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:521002 [1] NCCL INFO Channel 01/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:521001 [4] NCCL INFO Channel 02/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:521003 [2] NCCL INFO Channel 16/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:521002 [1] NCCL INFO Channel 02/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:521004 [6] NCCL INFO Channel 01/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:521001 [4] NCCL INFO Channel 03/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:521005 [5] NCCL INFO Channel 00/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:521003 [2] NCCL INFO Channel 17/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:521002 [1] NCCL INFO Channel 03/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:521004 [6] NCCL INFO Channel 02/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:521001 [4] NCCL INFO Channel 04/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:521005 [5] NCCL INFO Channel 01/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:521003 [2] NCCL INFO Channel 18/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:521002 [1] NCCL INFO Channel 04/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:521004 [6] NCCL INFO Channel 03/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:521001 [4] NCCL INFO Channel 05/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:521005 [5] NCCL INFO Channel 02/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:521003 [2] NCCL INFO Channel 19/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:521002 [1] NCCL INFO Channel 05/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:521004 [6] NCCL INFO Channel 04/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:521001 [4] NCCL INFO Channel 06/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:521005 [5] NCCL INFO Channel 03/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:521003 [2] NCCL INFO Channel 20/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:521002 [1] NCCL INFO Channel 06/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:521001 [4] NCCL INFO Channel 07/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:521004 [6] NCCL INFO Channel 05/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:521005 [5] NCCL INFO Channel 04/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:521003 [2] NCCL INFO Channel 21/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:521002 [1] NCCL INFO Channel 07/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:521001 [4] NCCL INFO Channel 08/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:521005 [5] NCCL INFO Channel 05/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:521004 [6] NCCL INFO Channel 06/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:521003 [2] NCCL INFO Channel 22/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:521002 [1] NCCL INFO Channel 08/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:521001 [4] NCCL INFO Channel 09/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:521005 [5] NCCL INFO Channel 06/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:521004 [6] NCCL INFO Channel 07/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514117:521003 [2] NCCL INFO Channel 23/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:521000 [3] NCCL INFO Channel 00/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:521002 [1] NCCL INFO Channel 09/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:521001 [4] NCCL INFO Channel 10/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:521005 [5] NCCL INFO Channel 07/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:521004 [6] NCCL INFO Channel 08/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:521000 [3] NCCL INFO Channel 01/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:521002 [1] NCCL INFO Channel 10/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:521001 [4] NCCL INFO Channel 11/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:521005 [5] NCCL INFO Channel 08/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:521004 [6] NCCL INFO Channel 09/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:521000 [3] NCCL INFO Channel 02/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:521002 [1] NCCL INFO Channel 11/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:521001 [4] NCCL INFO Channel 12/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:521005 [5] NCCL INFO Channel 09/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:521004 [6] NCCL INFO Channel 10/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:521000 [3] NCCL INFO Channel 03/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:521002 [1] NCCL INFO Channel 12/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:521001 [4] NCCL INFO Channel 13/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:521005 [5] NCCL INFO Channel 10/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:521004 [6] NCCL INFO Channel 11/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:521000 [3] NCCL INFO Channel 04/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:521002 [1] NCCL INFO Channel 13/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:521001 [4] NCCL INFO Channel 14/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:521005 [5] NCCL INFO Channel 11/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:521004 [6] NCCL INFO Channel 12/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:521000 [3] NCCL INFO Channel 05/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:521002 [1] NCCL INFO Channel 14/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:521001 [4] NCCL INFO Channel 15/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:521005 [5] NCCL INFO Channel 12/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:521004 [6] NCCL INFO Channel 13/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:521000 [3] NCCL INFO Channel 06/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:521002 [1] NCCL INFO Channel 15/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:521001 [4] NCCL INFO Channel 16/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:521005 [5] NCCL INFO Channel 13/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:521004 [6] NCCL INFO Channel 14/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:521000 [3] NCCL INFO Channel 07/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:521002 [1] NCCL INFO Channel 16/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:521001 [4] NCCL INFO Channel 17/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:521005 [5] NCCL INFO Channel 14/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:521004 [6] NCCL INFO Channel 15/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:521000 [3] NCCL INFO Channel 08/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:521002 [1] NCCL INFO Channel 17/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:521001 [4] NCCL INFO Channel 18/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:521005 [5] NCCL INFO Channel 15/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:521004 [6] NCCL INFO Channel 16/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:521000 [3] NCCL INFO Channel 09/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:521002 [1] NCCL INFO Channel 18/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:521001 [4] NCCL INFO Channel 19/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:521005 [5] NCCL INFO Channel 16/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:521004 [6] NCCL INFO Channel 17/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:521000 [3] NCCL INFO Channel 10/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:521002 [1] NCCL INFO Channel 19/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:521001 [4] NCCL INFO Channel 20/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:521005 [5] NCCL INFO Channel 17/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:521004 [6] NCCL INFO Channel 18/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:521000 [3] NCCL INFO Channel 11/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:521002 [1] NCCL INFO Channel 20/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:521001 [4] NCCL INFO Channel 21/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:521005 [5] NCCL INFO Channel 18/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:521004 [6] NCCL INFO Channel 19/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:521000 [3] NCCL INFO Channel 12/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:521002 [1] NCCL INFO Channel 21/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:521001 [4] NCCL INFO Channel 22/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:521005 [5] NCCL INFO Channel 19/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:521004 [6] NCCL INFO Channel 20/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:521000 [3] NCCL INFO Channel 13/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:521002 [1] NCCL INFO Channel 22/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514119:521001 [4] NCCL INFO Channel 23/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:521005 [5] NCCL INFO Channel 20/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:521004 [6] NCCL INFO Channel 21/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:521000 [3] NCCL INFO Channel 14/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514116:521002 [1] NCCL INFO Channel 23/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:521005 [5] NCCL INFO Channel 21/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:521004 [6] NCCL INFO Channel 22/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:521000 [3] NCCL INFO Channel 15/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:521005 [5] NCCL INFO Channel 22/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:521000 [3] NCCL INFO Channel 16/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514121:521004 [6] NCCL INFO Channel 23/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514120:521005 [5] NCCL INFO Channel 23/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:521000 [3] NCCL INFO Channel 17/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:521000 [3] NCCL INFO Channel 18/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:521000 [3] NCCL INFO Channel 19/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:521000 [3] NCCL INFO Channel 20/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:521000 [3] NCCL INFO Channel 21/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:521000 [3] NCCL INFO Channel 22/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514118:521000 [3] NCCL INFO Channel 23/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:514115:520999 [0] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:514115:520999 [0] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:514115:520999 [0] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:514119:521001 [4] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:514119:521001 [4] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:514119:521001 [4] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:514120:521005 [5] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:514116:521002 [1] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:514120:521005 [5] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:514116:521002 [1] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:514116:521002 [1] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:514120:521005 [5] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:514118:521000 [3] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:514117:521003 [2] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:514122:521006 [7] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:514118:521000 [3] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:514122:521006 [7] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:514117:521003 [2] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:514122:521006 [7] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:514117:521003 [2] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:514121:521004 [6] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:514118:521000 [3] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:514121:521004 [6] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:514121:521004 [6] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:514121:521004 [6] NCCL INFO ncclCommInitRank comm 0x7f724c06b420 rank 6 nranks 8 cudaDev 6 nvmlDev 6 busId a01c0 commId 0x3a5ef83485165fb9 - Init COMPLETE +ywang29-vrdb-test2-worker-0:514119:521001 [4] NCCL INFO ncclCommInitRank comm 0x7fbdbc06aef0 rank 4 nranks 8 cudaDev 4 nvmlDev 4 busId 901c0 commId 0x3a5ef83485165fb9 - Init COMPLETE +ywang29-vrdb-test2-worker-0:514117:521003 [2] NCCL INFO ncclCommInitRank comm 0x7f5a1006a5a0 rank 2 nranks 8 cudaDev 2 nvmlDev 2 busId 201c0 commId 0x3a5ef83485165fb9 - Init COMPLETE +ywang29-vrdb-test2-worker-0:514115:520999 [0] NCCL INFO ncclCommInitRank comm 0x7feddc06b3c0 rank 0 nranks 8 cudaDev 0 nvmlDev 0 busId 101c0 commId 0x3a5ef83485165fb9 - Init COMPLETE +ywang29-vrdb-test2-worker-0:514120:521005 [5] NCCL INFO ncclCommInitRank comm 0x7f608406aff0 rank 5 nranks 8 cudaDev 5 nvmlDev 5 busId 901d0 commId 0x3a5ef83485165fb9 - Init COMPLETE +ywang29-vrdb-test2-worker-0:514116:521002 [1] NCCL INFO ncclCommInitRank comm 0x7f943c06af40 rank 1 nranks 8 cudaDev 1 nvmlDev 1 busId 101d0 commId 0x3a5ef83485165fb9 - Init COMPLETE +ywang29-vrdb-test2-worker-0:514118:521000 [3] NCCL INFO ncclCommInitRank comm 0x7f90fc06aac0 rank 3 nranks 8 cudaDev 3 nvmlDev 3 busId 201d0 commId 0x3a5ef83485165fb9 - Init COMPLETE +ywang29-vrdb-test2-worker-0:514122:521006 [7] NCCL INFO ncclCommInitRank comm 0x7fc11006a6a0 rank 7 nranks 8 cudaDev 7 nvmlDev 7 busId a01d0 commId 0x3a5ef83485165fb9 - Init COMPLETE + 0%| | 1/520 [00:14<2:03:52, 14.32s/it] {'loss': 2.0453, 'grad_norm': 0.004834119298472289, 'learning_rate': 0.04375, 'epoch': 0.0} + 0%| | 1/520 [00:14<2:03:52, 14.32s/it] 0%| | 2/520 [00:17<1:09:19, 8.03s/it] {'loss': 2.0549, 'grad_norm': 0.005249197783401624, 'learning_rate': 0.0875, 'epoch': 0.0} + 0%| | 2/520 [00:17<1:09:19, 8.03s/it] 1%| | 3/520 [00:21<51:54, 6.02s/it] {'loss': 2.1899, 'grad_norm': 0.006005730102190992, 'learning_rate': 0.13124999999999998, 'epoch': 0.01} + 1%| | 3/520 [00:21<51:54, 6.02s/it] 1%| | 4/520 [00:25<44:10, 5.14s/it] {'loss': 1.6533, 'grad_norm': 0.0017239983987748268, 'learning_rate': 0.175, 'epoch': 0.01} + 1%| | 4/520 [00:25<44:10, 5.14s/it] 1%| | 5/520 [00:29<40:04, 4.67s/it] {'loss': 1.6652, 'grad_norm': 0.0009328233595336485, 'learning_rate': 0.21875, 'epoch': 0.01} + 1%| | 5/520 [00:29<40:04, 4.67s/it] 1%| | 6/520 [00:32<37:09, 4.34s/it] {'loss': 1.3884, 'grad_norm': 0.0004411662172364917, 'learning_rate': 0.26249999999999996, 'epoch': 0.01} + 1%| | 6/520 [00:32<37:09, 4.34s/it] 1%|▏ | 7/520 [00:36<35:06, 4.11s/it] {'loss': 1.4337, 'grad_norm': 0.0006278316959182495, 'learning_rate': 0.30624999999999997, 'epoch': 0.01} + 1%|▏ | 7/520 [00:36<35:06, 4.11s/it] 2%|▏ | 8/520 [00:40<35:39, 4.18s/it] {'loss': 1.4601, 'grad_norm': 0.0006504599498420875, 'learning_rate': 0.35, 'epoch': 0.02} + 2%|▏ | 8/520 [00:40<35:39, 4.18s/it] 2%|▏ | 9/520 [00:45<35:34, 4.18s/it] {'loss': 1.5258, 'grad_norm': 0.0008331624432439663, 'learning_rate': 0.39375, 'epoch': 0.02} + 2%|▏ | 9/520 [00:45<35:34, 4.18s/it] 2%|▏ | 10/520 [00:48<33:57, 3.99s/it] {'loss': 1.3587, 'grad_norm': 0.001166785346404256, 'learning_rate': 0.4375, 'epoch': 0.02} + 2%|▏ | 10/520 [00:48<33:57, 3.99s/it] 2%|▏ | 11/520 [00:52<34:13, 4.03s/it] {'loss': 1.4424, 'grad_norm': 0.002320169791739115, 'learning_rate': 0.48124999999999996, 'epoch': 0.02} + 2%|▏ | 11/520 [00:52<34:13, 4.03s/it] 2%|▏ | 12/520 [00:56<33:11, 3.92s/it] {'loss': 1.3576, 'grad_norm': 0.0020981722498399575, 'learning_rate': 0.5249999999999999, 'epoch': 0.02} + 2%|▏ | 12/520 [00:56<33:11, 3.92s/it][2025-10-12 07:33:56,118] [WARNING] [stage3.py:2069:step] 1 pytorch allocator cache flushes since last step. this happens when there is high memory pressure and is detrimental to performance. if this is happening frequently consider adjusting settings to reduce memory consumption. If you are unable to make the cache flushes go away consider adding get_accelerator().empty_cache() calls in your training loop to ensure that all ranks flush their caches at the same time + 2%|▎ | 13/520 [01:00<34:00, 4.02s/it] {'loss': 1.4394, 'grad_norm': 0.0030643013489280425, 'learning_rate': 0.56875, 'epoch': 0.03} + 2%|▎ | 13/520 [01:00<34:00, 4.02s/it] 3%|▎ | 14/520 [01:04<33:03, 3.92s/it] {'loss': 1.5417, 'grad_norm': 0.007835424361061636, 'learning_rate': 0.6124999999999999, 'epoch': 0.03} + 3%|▎ | 14/520 [01:04<33:03, 3.92s/it] 3%|▎ | 15/520 [01:07<32:17, 3.84s/it] {'loss': 1.5147, 'grad_norm': 0.0036441803466819556, 'learning_rate': 0.65625, 'epoch': 0.03} + 3%|▎ | 15/520 [01:07<32:17, 3.84s/it] 3%|▎ | 16/520 [01:11<31:38, 3.77s/it] {'loss': 1.4881, 'grad_norm': 0.003779873533171586, 'learning_rate': 0.7, 'epoch': 0.03} + 3%|▎ | 16/520 [01:11<31:38, 3.77s/it] 3%|▎ | 17/520 [01:15<31:22, 3.74s/it] {'loss': 1.8851, 'grad_norm': 0.022125510580164123, 'learning_rate': 0.6999932005260383, 'epoch': 0.03} + 3%|▎ | 17/520 [01:15<31:22, 3.74s/it] 3%|▎ | 18/520 [01:18<31:02, 3.71s/it] {'loss': 1.8623, 'grad_norm': 0.014327280256340384, 'learning_rate': 0.6999728023683411, 'epoch': 0.03} + 3%|▎ | 18/520 [01:18<31:02, 3.71s/it] 4%|▎ | 19/520 [01:22<30:51, 3.70s/it] {'loss': 2.6522, 'grad_norm': 0.02362921366436859, 'learning_rate': 0.6999388063194613, 'epoch': 0.04} + 4%|▎ | 19/520 [01:22<30:51, 3.70s/it] 4%|▍ | 20/520 [01:26<30:42, 3.68s/it] {'loss': 3.0321, 'grad_norm': 0.03515626640393358, 'learning_rate': 0.6998912137002856, 'epoch': 0.04} + 4%|▍ | 20/520 [01:26<30:42, 3.68s/it] 4%|▍ | 21/520 [01:29<30:38, 3.68s/it] {'loss': 2.3443, 'grad_norm': 0.021316512960373672, 'learning_rate': 0.6998300263599846, 'epoch': 0.04} + 4%|▍ | 21/520 [01:29<30:38, 3.68s/it] 4%|▍ | 22/520 [01:33<30:31, 3.68s/it] {'loss': 2.1684, 'grad_norm': 0.010272681955064152, 'learning_rate': 0.6997552466759395, 'epoch': 0.04} + 4%|▍ | 22/520 [01:33<30:31, 3.68s/it] 4%|▍ | 23/520 [01:37<30:27, 3.68s/it] {'loss': 2.2453, 'grad_norm': 0.013925191492810908, 'learning_rate': 0.6996668775536502, 'epoch': 0.04} + 4%|▍ | 23/520 [01:37<30:27, 3.68s/it] 5%|▍ | 24/520 [01:40<30:21, 3.67s/it] {'loss': 3.4139, 'grad_norm': 0.07271626954434203, 'learning_rate': 0.6995649224266227, 'epoch': 0.05} + 5%|▍ | 24/520 [01:40<30:21, 3.67s/it] 5%|▍ | 25/520 [01:44<30:11, 3.66s/it] {'loss': 3.0489, 'grad_norm': 0.4755405848922491, 'learning_rate': 0.6994493852562358, 'epoch': 0.05} + 5%|▍ | 25/520 [01:44<30:11, 3.66s/it] 5%|▌ | 26/520 [01:48<30:06, 3.66s/it] {'loss': 2.3395, 'grad_norm': 0.00836219358418084, 'learning_rate': 0.6993202705315862, 'epoch': 0.05} + 5%|▌ | 26/520 [01:48<30:06, 3.66s/it] 5%|▌ | 27/520 [01:51<30:02, 3.66s/it] {'loss': 2.0267, 'grad_norm': 0.00525670321963642, 'learning_rate': 0.6991775832693151, 'epoch': 0.05} + 5%|▌ | 27/520 [01:51<30:02, 3.66s/it] 5%|▌ | 28/520 [01:55<30:19, 3.70s/it] {'loss': 1.9417, 'grad_norm': 0.005787903348505464, 'learning_rate': 0.699021329013413, 'epoch': 0.05} + 5%|▌ | 28/520 [01:55<30:19, 3.70s/it] 6%|▌ | 29/520 [01:59<30:41, 3.75s/it] {'loss': 1.9092, 'grad_norm': 0.005201359930012625, 'learning_rate': 0.6988515138350042, 'epoch': 0.06} + 6%|▌ | 29/520 [01:59<30:41, 3.75s/it] 6%|▌ | 30/520 [02:03<30:52, 3.78s/it] {'loss': 2.7658, 'grad_norm': 0.0283890372251381, 'learning_rate': 0.6986681443321109, 'epoch': 0.06} + 6%|▌ | 30/520 [02:03<30:52, 3.78s/it] 6%|▌ | 31/520 [02:07<31:00, 3.80s/it] {'loss': 1.9734, 'grad_norm': 0.025930754018769334, 'learning_rate': 0.6984712276293967, 'epoch': 0.06} + 6%|▌ | 31/520 [02:07<31:00, 3.80s/it] 6%|▌ | 32/520 [02:11<31:05, 3.82s/it] {'loss': 3.51, 'grad_norm': 0.2157270319120493, 'learning_rate': 0.6982607713778904, 'epoch': 0.06} + 6%|▌ | 32/520 [02:11<31:05, 3.82s/it] 6%|▋ | 33/520 [02:14<31:02, 3.82s/it] {'loss': 1.9933, 'grad_norm': 0.030784876399002657, 'learning_rate': 0.6980367837546879, 'epoch': 0.06} + 6%|▋ | 33/520 [02:14<31:02, 3.82s/it] 7%|▋ | 34/520 [02:18<30:55, 3.82s/it] {'loss': 1.851, 'grad_norm': 0.004646739539796258, 'learning_rate': 0.6977992734626349, 'epoch': 0.07} + 7%|▋ | 34/520 [02:18<30:55, 3.82s/it] 7%|▋ | 35/520 [02:22<30:52, 3.82s/it] {'loss': 1.8706, 'grad_norm': 0.004716408256065537, 'learning_rate': 0.6975482497299887, 'epoch': 0.07} + 7%|▋ | 35/520 [02:22<30:52, 3.82s/it] 7%|▋ | 36/520 [02:26<30:57, 3.84s/it] {'loss': 1.9406, 'grad_norm': 0.003028564857480359, 'learning_rate': 0.6972837223100602, 'epoch': 0.07} + 7%|▋ | 36/520 [02:26<30:57, 3.84s/it] 7%|▋ | 37/520 [02:30<30:52, 3.83s/it] {'loss': 2.358, 'grad_norm': 0.006952931711047297, 'learning_rate': 0.6970057014808336, 'epoch': 0.07} + 7%|▋ | 37/520 [02:30<30:52, 3.83s/it] 7%|▋ | 38/520 [02:33<30:22, 3.78s/it] {'loss': 1.9904, 'grad_norm': 0.002905062545813216, 'learning_rate': 0.6967141980445684, 'epoch': 0.07} + 7%|▋ | 38/520 [02:33<30:22, 3.78s/it] 8%|▊ | 39/520 [02:37<30:02, 3.75s/it] {'loss': 1.7655, 'grad_norm': 0.0034195499947714384, 'learning_rate': 0.6964092233273791, 'epoch': 0.07} + 8%|▊ | 39/520 [02:37<30:02, 3.75s/it] 8%|▊ | 40/520 [02:41<29:52, 3.73s/it] {'loss': 1.7961, 'grad_norm': 0.0023926055124263938, 'learning_rate': 0.6960907891787949, 'epoch': 0.08} + 8%|▊ | 40/520 [02:41<29:52, 3.73s/it] 8%|▊ | 41/520 [02:44<29:38, 3.71s/it] {'loss': 1.7349, 'grad_norm': 0.0022436107645030868, 'learning_rate': 0.6957589079713, 'epoch': 0.08} + 8%|▊ | 41/520 [02:44<29:38, 3.71s/it] 8%|▊ | 42/520 [02:48<29:19, 3.68s/it] {'loss': 1.7792, 'grad_norm': 0.002448713963537461, 'learning_rate': 0.6954135925998524, 'epoch': 0.08} + 8%|▊ | 42/520 [02:48<29:19, 3.68s/it] 8%|▊ | 43/520 [02:52<29:14, 3.68s/it] {'loss': 2.0335, 'grad_norm': 0.006745401641962259, 'learning_rate': 0.6950548564813824, 'epoch': 0.08} + 8%|▊ | 43/520 [02:52<29:14, 3.68s/it] 8%|▊ | 44/520 [02:56<29:27, 3.71s/it] {'loss': 2.095, 'grad_norm': 0.0035674293728654625, 'learning_rate': 0.6946827135542728, 'epoch': 0.08} + 8%|▊ | 44/520 [02:56<29:27, 3.71s/it] 9%|▊ | 45/520 [02:59<29:16, 3.70s/it] {'loss': 1.7222, 'grad_norm': 0.0017526203387228734, 'learning_rate': 0.6942971782778155, 'epoch': 0.09} + 9%|▊ | 45/520 [02:59<29:16, 3.70s/it] 9%|▉ | 46/520 [03:03<29:23, 3.72s/it] {'loss': 2.1275, 'grad_norm': 0.0025029536768020437, 'learning_rate': 0.6938982656316509, 'epoch': 0.09} + 9%|▉ | 46/520 [03:03<29:23, 3.72s/it] 9%|▉ | 47/520 [03:07<29:17, 3.71s/it] {'loss': 1.709, 'grad_norm': 0.001846424150599567, 'learning_rate': 0.6934859911151857, 'epoch': 0.09} + 9%|▉ | 47/520 [03:07<29:17, 3.71s/it] 9%|▉ | 48/520 [03:10<29:08, 3.71s/it] {'loss': 1.6806, 'grad_norm': 0.0017567162728664024, 'learning_rate': 0.6930603707469903, 'epoch': 0.09} + 9%|▉ | 48/520 [03:10<29:08, 3.71s/it] 9%|▉ | 49/520 [03:14<29:20, 3.74s/it] {'loss': 1.6833, 'grad_norm': 0.0015193556004590898, 'learning_rate': 0.6926214210641769, 'epoch': 0.09} + 9%|▉ | 49/520 [03:14<29:20, 3.74s/it] 10%|▉ | 50/520 [03:18<29:23, 3.75s/it] {'loss': 1.6634, 'grad_norm': 0.0013326711692247808, 'learning_rate': 0.6921691591217566, 'epoch': 0.1} + 10%|▉ | 50/520 [03:18<29:23, 3.75s/it] 10%|▉ | 51/520 [03:22<29:29, 3.77s/it] {'loss': 1.5808, 'grad_norm': 0.0014872831915028395, 'learning_rate': 0.6917036024919766, 'epoch': 0.1} + 10%|▉ | 51/520 [03:22<29:29, 3.77s/it] 10%|█ | 52/520 [03:26<29:30, 3.78s/it] {'loss': 1.7164, 'grad_norm': 0.0015444438854221834, 'learning_rate': 0.6912247692636382, 'epoch': 0.1} + 10%|█ | 52/520 [03:26<29:30, 3.78s/it] 10%|█ | 53/520 [03:29<29:28, 3.79s/it] {'loss': 1.7179, 'grad_norm': 0.0013667238530343127, 'learning_rate': 0.690732678041393, 'epoch': 0.1} + 10%|█ | 53/520 [03:29<29:28, 3.79s/it] 10%|█ | 54/520 [03:33<29:26, 3.79s/it] {'loss': 1.5752, 'grad_norm': 0.0011774999402535216, 'learning_rate': 0.690227347945021, 'epoch': 0.1} + 10%|█ | 54/520 [03:33<29:26, 3.79s/it] 11%|█ | 55/520 [03:37<29:29, 3.81s/it] {'loss': 1.5752, 'grad_norm': 0.0013265303745747705, 'learning_rate': 0.6897087986086868, 'epoch': 0.11} + 11%|█ | 55/520 [03:37<29:29, 3.81s/it] 11%|█ | 56/520 [03:41<29:31, 3.82s/it] {'loss': 1.7112, 'grad_norm': 0.0013371291766145178, 'learning_rate': 0.6891770501801773, 'epoch': 0.11} + 11%|█ | 56/520 [03:41<29:31, 3.82s/it] 11%|█ | 57/520 [03:45<29:29, 3.82s/it] {'loss': 1.5509, 'grad_norm': 0.0011693085467470435, 'learning_rate': 0.6886321233201187, 'epoch': 0.11} + 11%|█ | 57/520 [03:45<29:29, 3.82s/it] 11%|█ | 58/520 [03:48<29:25, 3.82s/it] {'loss': 1.7179, 'grad_norm': 0.0011230767209257538, 'learning_rate': 0.6880740392011738, 'epoch': 0.11} + 11%|█ | 58/520 [03:49<29:25, 3.82s/it] 11%|█▏ | 59/520 [03:52<29:26, 3.83s/it] {'loss': 1.7177, 'grad_norm': 0.0014154258355809054, 'learning_rate': 0.6875028195072197, 'epoch': 0.11} + 11%|█▏ | 59/520 [03:52<29:26, 3.83s/it] 12%|█▏ | 60/520 [03:56<29:22, 3.83s/it] {'loss': 1.6111, 'grad_norm': 0.0011245642069607933, 'learning_rate': 0.6869184864325041, 'epoch': 0.12} + 12%|█▏ | 60/520 [03:56<29:22, 3.83s/it] 12%|█▏ | 61/520 [04:00<29:21, 3.84s/it] {'loss': 1.9091, 'grad_norm': 0.001791630890975574, 'learning_rate': 0.6863210626807849, 'epoch': 0.12} + 12%|█▏ | 61/520 [04:00<29:21, 3.84s/it] 12%|█▏ | 62/520 [04:04<29:16, 3.84s/it] {'loss': 1.5869, 'grad_norm': 0.001154596485872782, 'learning_rate': 0.6857105714644457, 'epoch': 0.12} + 12%|█▏ | 62/520 [04:04<29:16, 3.84s/it] 12%|█▏ | 63/520 [04:08<29:12, 3.83s/it] {'loss': 1.6059, 'grad_norm': 0.0011634324535370964, 'learning_rate': 0.6850870365035963, 'epoch': 0.12} + 12%|█▏ | 63/520 [04:08<29:12, 3.83s/it] 12%|█▏ | 64/520 [04:11<28:58, 3.81s/it] {'loss': 1.6074, 'grad_norm': 0.0013164835359053158, 'learning_rate': 0.6844504820251492, 'epoch': 0.12} + 12%|█▏ | 64/520 [04:11<28:58, 3.81s/it] 12%|█▎ | 65/520 [04:15<28:36, 3.77s/it] {'loss': 1.6156, 'grad_norm': 0.0012066736255389565, 'learning_rate': 0.6838009327618794, 'epoch': 0.12} + 12%|█▎ | 65/520 [04:15<28:36, 3.77s/it] 13%|█▎ | 66/520 [04:19<28:16, 3.74s/it] {'loss': 1.5893, 'grad_norm': 0.0010519431506425495, 'learning_rate': 0.6831384139514628, 'epoch': 0.13} + 13%|█▎ | 66/520 [04:19<28:16, 3.74s/it] 13%|█▎ | 67/520 [04:23<28:27, 3.77s/it] {'loss': 1.4411, 'grad_norm': 0.0009748155975201621, 'learning_rate': 0.6824629513354961, 'epoch': 0.13} + 13%|█▎ | 67/520 [04:23<28:27, 3.77s/it] 13%|█▎ | 68/520 [04:27<28:55, 3.84s/it] {'loss': 1.4953, 'grad_norm': 0.0019515769258250033, 'learning_rate': 0.681774571158496, 'epoch': 0.13} + 13%|█▎ | 68/520 [04:27<28:55, 3.84s/it] 13%|█▎ | 69/520 [04:31<29:29, 3.92s/it] {'loss': 1.4637, 'grad_norm': 0.001014435939070233, 'learning_rate': 0.6810733001668805, 'epoch': 0.13} + 13%|█▎ | 69/520 [04:31<29:29, 3.92s/it] 13%|█▎ | 70/520 [04:35<29:50, 3.98s/it] {'loss': 1.5542, 'grad_norm': 0.0015266374003019216, 'learning_rate': 0.6803591656079286, 'epoch': 0.13} + 13%|█▎ | 70/520 [04:35<29:50, 3.98s/it] 14%|█▎ | 71/520 [04:39<30:00, 4.01s/it] {'loss': 1.4421, 'grad_norm': 0.0008746724166023978, 'learning_rate': 0.6796321952287222, 'epoch': 0.14} + 14%|█▎ | 71/520 [04:39<30:00, 4.01s/it] 14%|█▍ | 72/520 [04:43<30:07, 4.03s/it] {'loss': 1.6012, 'grad_norm': 0.0010221437692451655, 'learning_rate': 0.6788924172750679, 'epoch': 0.14} + 14%|█▍ | 72/520 [04:43<30:07, 4.03s/it] 14%|█▍ | 73/520 [04:47<30:12, 4.05s/it] {'loss': 1.4275, 'grad_norm': 0.0009317469942965107, 'learning_rate': 0.6781398604903998, 'epoch': 0.14} + 14%|█▍ | 73/520 [04:47<30:12, 4.05s/it] 14%|█▍ | 74/520 [04:51<30:13, 4.07s/it] {'loss': 1.5438, 'grad_norm': 0.0011142084490660925, 'learning_rate': 0.6773745541146619, 'epoch': 0.14} + 14%|█▍ | 74/520 [04:51<30:13, 4.07s/it] 14%|█▍ | 75/520 [04:55<30:14, 4.08s/it] {'loss': 1.4497, 'grad_norm': 0.0008604815503671251, 'learning_rate': 0.6765965278831731, 'epoch': 0.14} + 14%|█▍ | 75/520 [04:55<30:14, 4.08s/it] 15%|█▍ | 76/520 [04:59<30:14, 4.09s/it] {'loss': 1.8599, 'grad_norm': 0.0010413327619381715, 'learning_rate': 0.6758058120254714, 'epoch': 0.15} + 15%|█▍ | 76/520 [04:59<30:14, 4.09s/it] 15%|█▍ | 77/520 [05:04<30:13, 4.09s/it] {'loss': 1.3658, 'grad_norm': 0.0010820410632276527, 'learning_rate': 0.6750024372641387, 'epoch': 0.15} + 15%|█▍ | 77/520 [05:04<30:13, 4.09s/it] 15%|█▌ | 78/520 [05:08<30:02, 4.08s/it] {'loss': 1.4902, 'grad_norm': 0.0008293036571383674, 'learning_rate': 0.674186434813608, 'epoch': 0.15} + 15%|█▌ | 78/520 [05:08<30:02, 4.08s/it] 15%|█▌ | 79/520 [05:11<29:19, 3.99s/it] {'loss': 1.4663, 'grad_norm': 0.0007646256543301868, 'learning_rate': 0.6733578363789503, 'epoch': 0.15} + 15%|█▌ | 79/520 [05:11<29:19, 3.99s/it] 15%|█▌ | 80/520 [05:15<28:39, 3.91s/it] {'loss': 1.8786, 'grad_norm': 0.0010588074468903291, 'learning_rate': 0.6725166741546427, 'epoch': 0.15} + 15%|█▌ | 80/520 [05:15<28:39, 3.91s/it] 16%|█▌ | 81/520 [05:19<28:03, 3.83s/it] {'loss': 1.6527, 'grad_norm': 0.0009753566289456692, 'learning_rate': 0.6716629808233171, 'epoch': 0.16} + 16%|█▌ | 81/520 [05:19<28:03, 3.83s/it] 16%|█▌ | 82/520 [05:22<27:44, 3.80s/it] {'loss': 1.5603, 'grad_norm': 0.0008567864805183087, 'learning_rate': 0.6707967895544916, 'epoch': 0.16} + 16%|█▌ | 82/520 [05:22<27:44, 3.80s/it] 16%|█▌ | 83/520 [05:26<27:21, 3.76s/it] {'loss': 1.5917, 'grad_norm': 0.0008655239236940441, 'learning_rate': 0.66991813400328, 'epoch': 0.16} + 16%|█▌ | 83/520 [05:26<27:21, 3.76s/it] 16%|█▌ | 84/520 [05:30<27:01, 3.72s/it] {'loss': 1.5713, 'grad_norm': 0.000934848289148945, 'learning_rate': 0.6690270483090855, 'epoch': 0.16} + 16%|█▌ | 84/520 [05:30<27:01, 3.72s/it] 16%|█▋ | 85/520 [05:33<26:45, 3.69s/it] {'loss': 1.5752, 'grad_norm': 0.0007664422023455304, 'learning_rate': 0.6681235670942738, 'epoch': 0.16} + 16%|█▋ | 85/520 [05:33<26:45, 3.69s/it] 17%|█▋ | 86/520 [05:37<26:36, 3.68s/it] {'loss': 1.6322, 'grad_norm': 0.0007966614647669055, 'learning_rate': 0.6672077254628275, 'epoch': 0.17} + 17%|█▋ | 86/520 [05:37<26:36, 3.68s/it] 17%|█▋ | 87/520 [05:41<26:36, 3.69s/it] {'loss': 1.8058, 'grad_norm': 0.0011794550709510292, 'learning_rate': 0.6662795589989828, 'epoch': 0.17} + 17%|█▋ | 87/520 [05:41<26:36, 3.69s/it] 17%|█▋ | 88/520 [05:44<26:29, 3.68s/it] {'loss': 1.9279, 'grad_norm': 0.000981566098284031, 'learning_rate': 0.6653391037658466, 'epoch': 0.17} + 17%|█▋ | 88/520 [05:44<26:29, 3.68s/it] 17%|█▋ | 89/520 [05:48<26:24, 3.68s/it] {'loss': 1.5423, 'grad_norm': 0.0007628940782252337, 'learning_rate': 0.6643863963039954, 'epoch': 0.17} + 17%|█▋ | 89/520 [05:48<26:24, 3.68s/it] 17%|█▋ | 90/520 [05:52<26:17, 3.67s/it] {'loss': 1.4689, 'grad_norm': 0.0009303855314684361, 'learning_rate': 0.6634214736300553, 'epoch': 0.17} + 17%|█▋ | 90/520 [05:52<26:17, 3.67s/it] 18%|█▊ | 91/520 [05:55<26:11, 3.66s/it] {'loss': 1.5592, 'grad_norm': 0.0006408022084401333, 'learning_rate': 0.662444373235264, 'epoch': 0.17} + 18%|█▊ | 91/520 [05:55<26:11, 3.66s/it] 18%|█▊ | 92/520 [05:59<26:04, 3.66s/it] {'loss': 1.4966, 'grad_norm': 0.0007823357569881217, 'learning_rate': 0.661455133084014, 'epoch': 0.18} + 18%|█▊ | 92/520 [05:59<26:04, 3.66s/it] 18%|█▊ | 93/520 [06:03<26:06, 3.67s/it] {'loss': 1.4778, 'grad_norm': 0.0007154691339146993, 'learning_rate': 0.6604537916123775, 'epoch': 0.18} + 18%|█▊ | 93/520 [06:03<26:06, 3.67s/it] 18%|█▊ | 94/520 [06:06<26:00, 3.66s/it] {'loss': 1.6028, 'grad_norm': 0.0006969988941028545, 'learning_rate': 0.6594403877266134, 'epoch': 0.18} + 18%|█▊ | 94/520 [06:06<26:00, 3.66s/it] 18%|█▊ | 95/520 [06:10<25:56, 3.66s/it] {'loss': 1.4681, 'grad_norm': 0.0008382656583375377, 'learning_rate': 0.6584149608016547, 'epoch': 0.18} + 18%|█▊ | 95/520 [06:10<25:56, 3.66s/it] 18%|█▊ | 96/520 [06:14<25:53, 3.66s/it] {'loss': 1.4752, 'grad_norm': 0.0007159093886481745, 'learning_rate': 0.6573775506795799, 'epoch': 0.18} + 18%|█▊ | 96/520 [06:14<25:53, 3.66s/it] 19%|█▊ | 97/520 [06:17<25:48, 3.66s/it] {'loss': 1.4365, 'grad_norm': 0.0007072081375628331, 'learning_rate': 0.6563281976680638, 'epoch': 0.19} + 19%|█▊ | 97/520 [06:17<25:48, 3.66s/it] 19%|█▉ | 98/520 [06:21<25:46, 3.66s/it] {'loss': 1.4579, 'grad_norm': 0.0006148776812247525, 'learning_rate': 0.6552669425388119, 'epoch': 0.19} + 19%|█▉ | 98/520 [06:21<25:46, 3.66s/it] 19%|█▉ | 99/520 [06:25<25:39, 3.66s/it] {'loss': 1.4735, 'grad_norm': 0.0006643697273457525, 'learning_rate': 0.6541938265259762, 'epoch': 0.19} + 19%|█▉ | 99/520 [06:25<25:39, 3.66s/it] 19%|█▉ | 100/520 [06:28<25:29, 3.64s/it] {'loss': 1.6606, 'grad_norm': 0.0007650304272852298, 'learning_rate': 0.6531088913245535, 'epoch': 0.19} + 19%|█▉ | 100/520 [06:28<25:29, 3.64s/it] 19%|█▉ | 101/520 [06:32<25:22, 3.63s/it] {'loss': 1.471, 'grad_norm': 0.0007232482654513256, 'learning_rate': 0.6520121790887645, 'epoch': 0.19} + 19%|█▉ | 101/520 [06:32<25:22, 3.63s/it] 20%|█▉ | 102/520 [06:36<25:17, 3.63s/it] {'loss': 1.4607, 'grad_norm': 0.0006974733927668036, 'learning_rate': 0.6509037324304165, 'epoch': 0.2} + 20%|█▉ | 102/520 [06:36<25:17, 3.63s/it] 20%|█▉ | 103/520 [06:39<25:07, 3.62s/it] {'loss': 1.4211, 'grad_norm': 0.0006638113766793617, 'learning_rate': 0.649783594417248, 'epoch': 0.2} + 20%|█▉ | 103/520 [06:39<25:07, 3.62s/it] 20%|██ | 104/520 [06:43<25:06, 3.62s/it] {'loss': 1.4911, 'grad_norm': 0.0007003744833459835, 'learning_rate': 0.6486518085712544, 'epoch': 0.2} + 20%|██ | 104/520 [06:43<25:06, 3.62s/it] 20%|██ | 105/520 [06:46<25:03, 3.62s/it] {'loss': 1.4783, 'grad_norm': 0.0006157989636452896, 'learning_rate': 0.6475084188669982, 'epoch': 0.2} + 20%|██ | 105/520 [06:46<25:03, 3.62s/it] 20%|██ | 106/520 [06:50<25:02, 3.63s/it] {'loss': 1.67, 'grad_norm': 0.0009876499589925177, 'learning_rate': 0.6463534697298994, 'epoch': 0.2} + 20%|██ | 106/520 [06:50<25:02, 3.63s/it] 21%|██ | 107/520 [06:54<24:56, 3.62s/it] {'loss': 1.6496, 'grad_norm': 0.0011060075644756453, 'learning_rate': 0.64518700603451, 'epoch': 0.21} + 21%|██ | 107/520 [06:54<24:56, 3.62s/it] 21%|██ | 108/520 [06:57<24:58, 3.64s/it] {'loss': 1.4287, 'grad_norm': 0.0007229411661780069, 'learning_rate': 0.64400907310277, 'epoch': 0.21} + 21%|██ | 108/520 [06:57<24:58, 3.64s/it] 21%|██ | 109/520 [07:01<25:04, 3.66s/it] {'loss': 1.645, 'grad_norm': 0.0006658000246786751, 'learning_rate': 0.6428197167022469, 'epoch': 0.21} + 21%|██ | 109/520 [07:01<25:04, 3.66s/it] 21%|██ | 110/520 [07:05<24:54, 3.64s/it] {'loss': 1.6179, 'grad_norm': 0.0006484630329079996, 'learning_rate': 0.641618983044357, 'epoch': 0.21} + 21%|██ | 110/520 [07:05<24:54, 3.64s/it] 21%|██▏ | 111/520 [07:08<24:48, 3.64s/it] {'loss': 1.6286, 'grad_norm': 0.0007117011200700489, 'learning_rate': 0.6404069187825706, 'epoch': 0.21} + 21%|██▏ | 111/520 [07:08<24:48, 3.64s/it] 22%|██▏ | 112/520 [07:12<24:44, 3.64s/it] {'loss': 1.5076, 'grad_norm': 0.0006616274319277269, 'learning_rate': 0.6391835710105982, 'epoch': 0.22} + 22%|██▏ | 112/520 [07:12<24:44, 3.64s/it] 22%|██▏ | 113/520 [07:15<24:39, 3.63s/it] {'loss': 1.3572, 'grad_norm': 0.0008014235391020023, 'learning_rate': 0.6379489872605616, 'epoch': 0.22} + 22%|██▏ | 113/520 [07:15<24:39, 3.63s/it] 22%|██▏ | 114/520 [07:19<24:39, 3.64s/it] {'loss': 1.4838, 'grad_norm': 0.0006354177975344596, 'learning_rate': 0.6367032155011472, 'epoch': 0.22} + 22%|██▏ | 114/520 [07:19<24:39, 3.64s/it] 22%|██▏ | 115/520 [07:23<24:37, 3.65s/it] {'loss': 1.6065, 'grad_norm': 0.000637753891106319, 'learning_rate': 0.635446304135741, 'epoch': 0.22} + 22%|██▏ | 115/520 [07:23<24:37, 3.65s/it] 22%|██▏ | 116/520 [07:26<24:32, 3.65s/it] {'loss': 1.5918, 'grad_norm': 0.0005357507058592157, 'learning_rate': 0.6341783020005498, 'epoch': 0.22} + 22%|██▏ | 116/520 [07:26<24:32, 3.65s/it] 22%|██▎ | 117/520 [07:30<24:28, 3.64s/it] {'loss': 1.5728, 'grad_norm': 0.0006727728874238503, 'learning_rate': 0.6328992583627018, 'epoch': 0.23} + 22%|██▎ | 117/520 [07:30<24:28, 3.64s/it] 23%|██▎ | 118/520 [07:34<24:20, 3.63s/it] {'loss': 1.4541, 'grad_norm': 0.000745140634354666, 'learning_rate': 0.6316092229183339, 'epoch': 0.23} + 23%|██▎ | 118/520 [07:34<24:20, 3.63s/it] 23%|██▎ | 119/520 [07:37<24:15, 3.63s/it] {'loss': 1.4094, 'grad_norm': 0.0006039812059090084, 'learning_rate': 0.6303082457906598, 'epoch': 0.23} + 23%|██▎ | 119/520 [07:37<24:15, 3.63s/it] 23%|██▎ | 120/520 [07:41<24:16, 3.64s/it] {'loss': 1.4448, 'grad_norm': 0.0009828602320977313, 'learning_rate': 0.6289963775280228, 'epoch': 0.23} + 23%|██▎ | 120/520 [07:41<24:16, 3.64s/it] 23%|██▎ | 121/520 [07:45<24:11, 3.64s/it] {'loss': 1.5033, 'grad_norm': 0.0008287504466378255, 'learning_rate': 0.6276736691019322, 'epoch': 0.23} + 23%|██▎ | 121/520 [07:45<24:11, 3.64s/it] 23%|██▎ | 122/520 [07:48<24:09, 3.64s/it] {'loss': 1.3557, 'grad_norm': 0.0006466719095919816, 'learning_rate': 0.6263401719050824, 'epoch': 0.23} + 23%|██▎ | 122/520 [07:48<24:09, 3.64s/it] 24%|██▎ | 123/520 [07:52<24:12, 3.66s/it] {'loss': 1.7187, 'grad_norm': 0.0009894682973807514, 'learning_rate': 0.6249959377493558, 'epoch': 0.24} + 24%|██▎ | 123/520 [07:52<24:12, 3.66s/it] 24%|██▍ | 124/520 [07:56<24:09, 3.66s/it] {'loss': 1.4622, 'grad_norm': 0.000900232015000808, 'learning_rate': 0.6236410188638104, 'epoch': 0.24} + 24%|██▍ | 124/520 [07:56<24:09, 3.66s/it] 24%|██▍ | 125/520 [07:59<24:10, 3.67s/it] {'loss': 1.4333, 'grad_norm': 0.0007622828598853199, 'learning_rate': 0.6222754678926502, 'epoch': 0.24} + 24%|██▍ | 125/520 [07:59<24:10, 3.67s/it] 24%|██▍ | 126/520 [08:04<25:36, 3.90s/it] {'loss': 1.5999, 'grad_norm': 0.0005864575966040374, 'learning_rate': 0.6208993378931796, 'epoch': 0.24} + 24%|██▍ | 126/520 [08:04<25:36, 3.90s/it] 24%|██▍ | 127/520 [08:08<25:23, 3.88s/it] {'loss': 1.4211, 'grad_norm': 0.000603466405513231, 'learning_rate': 0.619512682333742, 'epoch': 0.24} + 24%|██▍ | 127/520 [08:08<25:23, 3.88s/it] 25%|██▍ | 128/520 [08:11<25:12, 3.86s/it] {'loss': 1.4839, 'grad_norm': 0.0006066369550124095, 'learning_rate': 0.6181155550916422, 'epoch': 0.25} + 25%|██▍ | 128/520 [08:11<25:12, 3.86s/it] 25%|██▍ | 129/520 [08:15<25:08, 3.86s/it] {'loss': 1.4011, 'grad_norm': 0.00048733066807110875, 'learning_rate': 0.6167080104510537, 'epoch': 0.25} + 25%|██▍ | 129/520 [08:15<25:08, 3.86s/it] 25%|██▌ | 130/520 [08:19<24:58, 3.84s/it] {'loss': 1.4739, 'grad_norm': 0.0005736438676574286, 'learning_rate': 0.6152901031009085, 'epoch': 0.25} + 25%|██▌ | 130/520 [08:19<24:58, 3.84s/it] 25%|██▌ | 131/520 [08:23<24:58, 3.85s/it] {'loss': 1.5921, 'grad_norm': 0.0005634277801526212, 'learning_rate': 0.6138618881327729, 'epoch': 0.25} + 25%|██▌ | 131/520 [08:23<24:58, 3.85s/it] 25%|██▌ | 132/520 [08:27<24:49, 3.84s/it] {'loss': 1.5209, 'grad_norm': 0.0005998201333842215, 'learning_rate': 0.612423421038707, 'epoch': 0.25} + 25%|██▌ | 132/520 [08:27<24:49, 3.84s/it] 26%|██▌ | 133/520 [08:31<24:50, 3.85s/it] {'loss': 1.429, 'grad_norm': 0.0006442189447042366, 'learning_rate': 0.6109747577091079, 'epoch': 0.26} + 26%|██▌ | 133/520 [08:31<24:50, 3.85s/it] 26%|██▌ | 134/520 [08:34<24:24, 3.79s/it] {'loss': 1.515, 'grad_norm': 0.0005472727135819568, 'learning_rate': 0.6095159544305393, 'epoch': 0.26} + 26%|██▌ | 134/520 [08:34<24:24, 3.79s/it] 26%|██▌ | 135/520 [08:38<24:01, 3.75s/it] {'loss': 1.5842, 'grad_norm': 0.0007064687438524041, 'learning_rate': 0.6080470678835433, 'epoch': 0.26} + 26%|██▌ | 135/520 [08:38<24:01, 3.75s/it] 26%|██▌ | 136/520 [08:42<23:47, 3.72s/it] {'loss': 1.5093, 'grad_norm': 0.0005449026487825841, 'learning_rate': 0.6065681551404392, 'epoch': 0.26} + 26%|██▌ | 136/520 [08:42<23:47, 3.72s/it] 26%|██▋ | 137/520 [08:45<23:35, 3.70s/it] {'loss': 1.4217, 'grad_norm': 0.0007098452485919402, 'learning_rate': 0.6050792736631049, 'epoch': 0.26} + 26%|██▋ | 137/520 [08:45<23:35, 3.70s/it] 27%|██▋ | 138/520 [08:49<23:26, 3.68s/it] {'loss': 1.433, 'grad_norm': 0.0005814124381775528, 'learning_rate': 0.6035804813007454, 'epoch': 0.27} + 27%|██▋ | 138/520 [08:49<23:26, 3.68s/it] 27%|██▋ | 139/520 [08:53<23:20, 3.68s/it] {'loss': 1.4862, 'grad_norm': 0.0005699814421196888, 'learning_rate': 0.6020718362876443, 'epoch': 0.27} + 27%|██▋ | 139/520 [08:53<23:20, 3.68s/it] 27%|██▋ | 140/520 [08:56<23:33, 3.72s/it] {'loss': 1.6264, 'grad_norm': 0.0005873187481539899, 'learning_rate': 0.6005533972409014, 'epoch': 0.27} + 27%|██▋ | 140/520 [08:56<23:33, 3.72s/it] 27%|██▋ | 141/520 [09:00<23:50, 3.77s/it] {'loss': 1.5455, 'grad_norm': 0.000514542379032307, 'learning_rate': 0.5990252231581555, 'epoch': 0.27} + 27%|██▋ | 141/520 [09:00<23:50, 3.77s/it] 27%|██▋ | 142/520 [09:04<23:52, 3.79s/it] {'loss': 1.6738, 'grad_norm': 0.0007194801510053238, 'learning_rate': 0.5974873734152916, 'epoch': 0.27} + 27%|██▋ | 142/520 [09:04<23:52, 3.79s/it] 28%|██▊ | 143/520 [09:08<23:53, 3.80s/it] {'loss': 1.4771, 'grad_norm': 0.0006635802067226929, 'learning_rate': 0.5959399077641342, 'epoch': 0.28} + 28%|██▊ | 143/520 [09:08<23:53, 3.80s/it] 28%|██▊ | 144/520 [09:12<23:53, 3.81s/it] {'loss': 1.4083, 'grad_norm': 0.0006200409608566874, 'learning_rate': 0.5943828863301254, 'epoch': 0.28} + 28%|██▊ | 144/520 [09:12<23:53, 3.81s/it] 28%|██▊ | 145/520 [09:16<23:56, 3.83s/it] {'loss': 1.3431, 'grad_norm': 0.0005898855213142247, 'learning_rate': 0.5928163696099895, 'epoch': 0.28} + 28%|██▊ | 145/520 [09:16<23:56, 3.83s/it] 28%|██▊ | 146/520 [09:19<23:52, 3.83s/it] {'loss': 1.7107, 'grad_norm': 0.0007573004473444756, 'learning_rate': 0.5912404184693815, 'epoch': 0.28} + 28%|██▊ | 146/520 [09:19<23:52, 3.83s/it] 28%|██▊ | 147/520 [09:23<23:51, 3.84s/it] {'loss': 1.3772, 'grad_norm': 0.0005100191133244284, 'learning_rate': 0.5896550941405226, 'epoch': 0.28} + 28%|██▊ | 147/520 [09:23<23:51, 3.84s/it] 28%|██▊ | 148/520 [09:27<23:47, 3.84s/it] {'loss': 1.4328, 'grad_norm': 0.0004895391903476123, 'learning_rate': 0.5880604582198217, 'epoch': 0.28} + 28%|██▊ | 148/520 [09:27<23:47, 3.84s/it] 29%|██▊ | 149/520 [09:31<23:27, 3.79s/it] {'loss': 1.3685, 'grad_norm': 0.0005278815642087895, 'learning_rate': 0.5864565726654811, 'epoch': 0.29} + 29%|██▊ | 149/520 [09:31<23:27, 3.79s/it] 29%|██▉ | 150/520 [09:34<23:09, 3.75s/it] {'loss': 1.5935, 'grad_norm': 0.0005440338314705325, 'learning_rate': 0.5848434997950894, 'epoch': 0.29} + 29%|██▉ | 150/520 [09:34<23:09, 3.75s/it] 29%|██▉ | 151/520 [09:38<22:55, 3.73s/it] {'loss': 1.4087, 'grad_norm': 0.0006968758577512192, 'learning_rate': 0.5832213022832013, 'epoch': 0.29} + 29%|██▉ | 151/520 [09:38<22:55, 3.73s/it] 29%|██▉ | 152/520 [09:42<22:49, 3.72s/it] {'loss': 1.373, 'grad_norm': 0.0007145273011159097, 'learning_rate': 0.5815900431589008, 'epoch': 0.29} + 29%|██▉ | 152/520 [09:42<22:49, 3.72s/it] 29%|██▉ | 153/520 [09:46<22:38, 3.70s/it] {'loss': 1.4078, 'grad_norm': 0.0006967752721003978, 'learning_rate': 0.5799497858033532, 'epoch': 0.29} + 29%|██▉ | 153/520 [09:46<22:38, 3.70s/it] 30%|██▉ | 154/520 [09:49<22:31, 3.69s/it] {'loss': 1.5131, 'grad_norm': 0.0005804147205481287, 'learning_rate': 0.5783005939473423, 'epoch': 0.3} + 30%|██▉ | 154/520 [09:49<22:31, 3.69s/it] 30%|██▉ | 155/520 [09:53<22:25, 3.69s/it] {'loss': 1.3973, 'grad_norm': 0.0007801696982415267, 'learning_rate': 0.5766425316687946, 'epoch': 0.3} + 30%|██▉ | 155/520 [09:53<22:25, 3.69s/it] 30%|███ | 156/520 [09:57<22:24, 3.69s/it] {'loss': 1.4406, 'grad_norm': 0.0011395069424135961, 'learning_rate': 0.5749756633902887, 'epoch': 0.3} + 30%|███ | 156/520 [09:57<22:24, 3.69s/it] 30%|███ | 157/520 [10:00<22:37, 3.74s/it] {'loss': 1.737, 'grad_norm': 0.0007118624060299276, 'learning_rate': 0.573300053876553, 'epoch': 0.3} + 30%|███ | 157/520 [10:00<22:37, 3.74s/it] 30%|███ | 158/520 [10:04<22:25, 3.72s/it] {'loss': 1.4088, 'grad_norm': 0.0005753409342636853, 'learning_rate': 0.5716157682319489, 'epoch': 0.3} + 30%|███ | 158/520 [10:04<22:25, 3.72s/it] 31%|███ | 159/520 [10:08<22:27, 3.73s/it] {'loss': 1.4402, 'grad_norm': 0.0008124897108817514, 'learning_rate': 0.5699228718979416, 'epoch': 0.31} + 31%|███ | 159/520 [10:08<22:27, 3.73s/it] 31%|███ | 160/520 [10:12<22:42, 3.78s/it] {'loss': 1.4786, 'grad_norm': 0.00046851902862068425, 'learning_rate': 0.5682214306505567, 'epoch': 0.31} + 31%|███ | 160/520 [10:12<22:42, 3.78s/it] 31%|███ | 161/520 [10:16<22:49, 3.82s/it] {'loss': 1.4527, 'grad_norm': 0.0005370710454975423, 'learning_rate': 0.5665115105978258, 'epoch': 0.31} + 31%|███ | 161/520 [10:16<22:49, 3.82s/it] 31%|███ | 162/520 [10:20<22:52, 3.83s/it] {'loss': 1.6282, 'grad_norm': 0.0005627306183021376, 'learning_rate': 0.5647931781772165, 'epoch': 0.31} + 31%|███ | 162/520 [10:20<22:52, 3.83s/it] 31%|███▏ | 163/520 [10:23<22:48, 3.83s/it] {'loss': 1.3532, 'grad_norm': 0.0006370766195367352, 'learning_rate': 0.5630665001530522, 'epoch': 0.31} + 31%|███▏ | 163/520 [10:23<22:48, 3.83s/it] 32%|███▏ | 164/520 [10:27<22:50, 3.85s/it] {'loss': 1.2875, 'grad_norm': 0.000537605856143653, 'learning_rate': 0.561331543613917, 'epoch': 0.32} + 32%|███▏ | 164/520 [10:27<22:50, 3.85s/it] 32%|███▏ | 165/520 [10:31<22:28, 3.80s/it] {'loss': 1.4452, 'grad_norm': 0.0005395408707593313, 'learning_rate': 0.5595883759700501, 'epoch': 0.32} + 32%|███▏ | 165/520 [10:31<22:28, 3.80s/it] 32%|███▏ | 166/520 [10:35<22:05, 3.75s/it] {'loss': 1.4225, 'grad_norm': 0.000597390955591206, 'learning_rate': 0.5578370649507255, 'epoch': 0.32} + 32%|███▏ | 166/520 [10:35<22:05, 3.75s/it] 32%|███▏ | 167/520 [10:38<21:50, 3.71s/it] {'loss': 1.4178, 'grad_norm': 0.0005563813744603428, 'learning_rate': 0.5560776786016216, 'epoch': 0.32} + 32%|███▏ | 167/520 [10:38<21:50, 3.71s/it] 32%|███▏ | 168/520 [10:42<21:39, 3.69s/it] {'loss': 1.3757, 'grad_norm': 0.0005226932069450797, 'learning_rate': 0.5543102852821764, 'epoch': 0.32} + 32%|███▏ | 168/520 [10:42<21:39, 3.69s/it] 32%|███▎ | 169/520 [10:45<21:32, 3.68s/it] {'loss': 1.4361, 'grad_norm': 0.0005613889599130086, 'learning_rate': 0.552534953662932, 'epoch': 0.33} + 32%|███▎ | 169/520 [10:45<21:32, 3.68s/it] 33%|███▎ | 170/520 [10:49<21:26, 3.67s/it] {'loss': 1.5402, 'grad_norm': 0.0005760648528987027, 'learning_rate': 0.550751752722866, 'epoch': 0.33} + 33%|███▎ | 170/520 [10:49<21:26, 3.67s/it] 33%|███▎ | 171/520 [10:53<21:17, 3.66s/it] {'loss': 1.3689, 'grad_norm': 0.0005445214290717133, 'learning_rate': 0.5489607517467123, 'epoch': 0.33} + 33%|███▎ | 171/520 [10:53<21:17, 3.66s/it] 33%|███▎ | 172/520 [10:56<21:11, 3.65s/it] {'loss': 1.4561, 'grad_norm': 0.0005087974939266256, 'learning_rate': 0.5471620203222677, 'epoch': 0.33} + 33%|███▎ | 172/520 [10:56<21:11, 3.65s/it] 33%|███▎ | 173/520 [11:00<21:06, 3.65s/it] {'loss': 1.3701, 'grad_norm': 0.0006052196861888979, 'learning_rate': 0.5453556283376894, 'epoch': 0.33} + 33%|███▎ | 173/520 [11:00<21:06, 3.65s/it] 33%|███▎ | 174/520 [11:04<21:04, 3.65s/it] {'loss': 1.4638, 'grad_norm': 0.0004479786834865571, 'learning_rate': 0.5435416459787787, 'epoch': 0.33} + 33%|███▎ | 174/520 [11:04<21:04, 3.65s/it] 34%|███▎ | 175/520 [11:07<21:11, 3.69s/it] {'loss': 1.3563, 'grad_norm': 0.0005889825523495954, 'learning_rate': 0.541720143726255, 'epoch': 0.34} + 34%|███▎ | 175/520 [11:07<21:11, 3.69s/it] 34%|███▍ | 176/520 [11:11<21:28, 3.74s/it] {'loss': 1.6445, 'grad_norm': 0.0005591738871644112, 'learning_rate': 0.5398911923530157, 'epoch': 0.34} + 34%|███▍ | 176/520 [11:11<21:28, 3.74s/it] 34%|███▍ | 177/520 [11:15<21:39, 3.79s/it] {'loss': 1.5162, 'grad_norm': 0.0005364053289956426, 'learning_rate': 0.5380548629213884, 'epoch': 0.34} + 34%|███▍ | 177/520 [11:15<21:39, 3.79s/it] 34%|███▍ | 178/520 [11:19<21:49, 3.83s/it] {'loss': 1.4266, 'grad_norm': 0.0005559706367287986, 'learning_rate': 0.5362112267803678, 'epoch': 0.34} + 34%|███▍ | 178/520 [11:19<21:49, 3.83s/it] 34%|███▍ | 179/520 [11:23<21:49, 3.84s/it] {'loss': 1.5259, 'grad_norm': 0.0004937755852452003, 'learning_rate': 0.5343603555628452, 'epoch': 0.34} + 34%|███▍ | 179/520 [11:23<21:49, 3.84s/it] 35%|███▍ | 180/520 [11:27<21:53, 3.86s/it] {'loss': 1.4132, 'grad_norm': 0.0005019582255165848, 'learning_rate': 0.5325023211828243, 'epoch': 0.35} + 35%|███▍ | 180/520 [11:27<21:53, 3.86s/it] 35%|███▍ | 181/520 [11:31<21:51, 3.87s/it] {'loss': 1.4106, 'grad_norm': 0.0004215112903813852, 'learning_rate': 0.5306371958326273, 'epoch': 0.35} + 35%|███▍ | 181/520 [11:31<21:51, 3.87s/it] 35%|███▌ | 182/520 [11:35<21:47, 3.87s/it] {'loss': 1.4239, 'grad_norm': 0.0005073291704009966, 'learning_rate': 0.5287650519800899, 'epoch': 0.35} + 35%|███▌ | 182/520 [11:35<21:47, 3.87s/it] 35%|███▌ | 183/520 [11:39<21:45, 3.87s/it] {'loss': 1.4446, 'grad_norm': 0.0005274632295537882, 'learning_rate': 0.5268859623657458, 'epoch': 0.35} + 35%|███▌ | 183/520 [11:39<21:45, 3.87s/it] 35%|███▌ | 184/520 [11:42<21:42, 3.88s/it] {'loss': 1.3733, 'grad_norm': 0.0005016766111064163, 'learning_rate': 0.5249999999999999, 'epoch': 0.35} + 35%|███▌ | 184/520 [11:42<21:42, 3.88s/it] 36%|███▌ | 185/520 [11:46<21:37, 3.87s/it] {'loss': 1.5273, 'grad_norm': 0.0004773132711516344, 'learning_rate': 0.5231072381602926, 'epoch': 0.36} + 36%|███▌ | 185/520 [11:46<21:37, 3.87s/it] 36%|███▌ | 186/520 [11:50<21:31, 3.87s/it] {'loss': 1.3785, 'grad_norm': 0.000493076585925075, 'learning_rate': 0.5212077503882513, 'epoch': 0.36} + 36%|███▌ | 186/520 [11:50<21:31, 3.87s/it] 36%|███▌ | 187/520 [11:54<21:27, 3.87s/it] {'loss': 1.3897, 'grad_norm': 0.0005952517943863045, 'learning_rate': 0.5193016104868339, 'epoch': 0.36} + 36%|███▌ | 187/520 [11:54<21:27, 3.87s/it] 36%|███▌ | 188/520 [11:58<21:22, 3.86s/it] {'loss': 1.4648, 'grad_norm': 0.0005584586664628785, 'learning_rate': 0.5173888925174613, 'epoch': 0.36} + 36%|███▌ | 188/520 [11:58<21:22, 3.86s/it] 36%|███▋ | 189/520 [12:02<21:25, 3.88s/it] {'loss': 1.491, 'grad_norm': 0.0007078855714489971, 'learning_rate': 0.5154696707971395, 'epoch': 0.36} + 36%|███▋ | 189/520 [12:02<21:25, 3.88s/it] 37%|███▋ | 190/520 [12:06<21:35, 3.93s/it] {'loss': 1.3786, 'grad_norm': 0.0005302286700833257, 'learning_rate': 0.5135440198955716, 'epoch': 0.37} + 37%|███▋ | 190/520 [12:06<21:35, 3.93s/it] 37%|███▋ | 191/520 [12:10<21:45, 3.97s/it] {'loss': 1.3396, 'grad_norm': 0.0004902621772198421, 'learning_rate': 0.5116120146322619, 'epoch': 0.37} + 37%|███▋ | 191/520 [12:10<21:45, 3.97s/it] 37%|███▋ | 192/520 [12:14<21:32, 3.94s/it] {'loss': 1.4382, 'grad_norm': 0.0004910743187699697, 'learning_rate': 0.5096737300736071, 'epoch': 0.37} + 37%|███▋ | 192/520 [12:14<21:32, 3.94s/it] 37%|███▋ | 193/520 [12:18<21:13, 3.90s/it] {'loss': 1.5842, 'grad_norm': 0.00050940008983597, 'learning_rate': 0.5077292415299809, 'epoch': 0.37} + 37%|███▋ | 193/520 [12:18<21:13, 3.90s/it] 37%|███▋ | 194/520 [12:21<21:02, 3.87s/it] {'loss': 1.4675, 'grad_norm': 0.0004937878484443576, 'learning_rate': 0.5057786245528072, 'epoch': 0.37} + 37%|███▋ | 194/520 [12:21<21:02, 3.87s/it] 38%|███▊ | 195/520 [12:25<20:53, 3.86s/it] {'loss': 1.4423, 'grad_norm': 0.0005354750462462966, 'learning_rate': 0.5038219549316257, 'epoch': 0.38} + 38%|███▊ | 195/520 [12:25<20:53, 3.86s/it] 38%|███▊ | 196/520 [12:29<20:43, 3.84s/it] {'loss': 1.4341, 'grad_norm': 0.0014118118701630872, 'learning_rate': 0.5018593086911453, 'epoch': 0.38} + 38%|███▊ | 196/520 [12:29<20:43, 3.84s/it] 38%|███▊ | 197/520 [12:33<20:38, 3.83s/it] {'loss': 1.3761, 'grad_norm': 0.0006739091088300933, 'learning_rate': 0.4998907620882919, 'epoch': 0.38} + 38%|███▊ | 197/520 [12:33<20:38, 3.83s/it] 38%|███▊ | 198/520 [12:37<20:20, 3.79s/it] {'loss': 1.4622, 'grad_norm': 0.0006287234982014448, 'learning_rate': 0.4979163916092448, 'epoch': 0.38} + 38%|███▊ | 198/520 [12:37<20:20, 3.79s/it] 38%|███▊ | 199/520 [12:40<20:04, 3.75s/it] {'loss': 1.3693, 'grad_norm': 0.0005445026129088481, 'learning_rate': 0.49593627396646484, 'epoch': 0.38} + 38%|███▊ | 199/520 [12:40<20:04, 3.75s/it] 38%|███▊ | 200/520 [12:44<19:51, 3.72s/it] {'loss': 1.5108, 'grad_norm': 0.0007192802133818344, 'learning_rate': 0.49395048609571407, 'epoch': 0.38} + 38%|███▊ | 200/520 [12:44<19:51, 3.72s/it] 39%|███▊ | 201/520 [12:48<19:44, 3.71s/it] {'loss': 1.5378, 'grad_norm': 0.000508520631174224, 'learning_rate': 0.49195910515306623, 'epoch': 0.39} + 39%|███▊ | 201/520 [12:48<19:44, 3.71s/it] 39%|███▉ | 202/520 [12:51<19:35, 3.70s/it] {'loss': 1.3633, 'grad_norm': 0.00047119814025430934, 'learning_rate': 0.48996220851190925, 'epoch': 0.39} + 39%|███▉ | 202/520 [12:51<19:35, 3.70s/it] 39%|███▉ | 203/520 [12:55<19:29, 3.69s/it] {'loss': 1.419, 'grad_norm': 0.00044747233042833923, 'learning_rate': 0.48795987375993877, 'epoch': 0.39} + 39%|███▉ | 203/520 [12:55<19:29, 3.69s/it] 39%|███▉ | 204/520 [12:59<19:22, 3.68s/it] {'loss': 1.4594, 'grad_norm': 0.000609450803870691, 'learning_rate': 0.48595217869614316, 'epoch': 0.39} + 39%|███▉ | 204/520 [12:59<19:22, 3.68s/it] 39%|███▉ | 205/520 [13:02<19:18, 3.68s/it] {'loss': 1.5431, 'grad_norm': 0.0005218375928045123, 'learning_rate': 0.4839392013277814, 'epoch': 0.39} + 39%|███▉ | 205/520 [13:02<19:18, 3.68s/it] 40%|███▉ | 206/520 [13:06<19:12, 3.67s/it] {'loss': 1.4898, 'grad_norm': 0.0006999399387297532, 'learning_rate': 0.4819210198673518, 'epoch': 0.4} + 40%|███▉ | 206/520 [13:06<19:12, 3.67s/it] 40%|███▉ | 207/520 [13:10<19:08, 3.67s/it] {'loss': 1.5175, 'grad_norm': 0.0004926100492845071, 'learning_rate': 0.4798977127295532, 'epoch': 0.4} + 40%|███▉ | 207/520 [13:10<19:08, 3.67s/it] 40%|████ | 208/520 [13:13<19:06, 3.67s/it] {'loss': 1.4406, 'grad_norm': 0.00047202095592093716, 'learning_rate': 0.4778693585282383, 'epoch': 0.4} + 40%|████ | 208/520 [13:13<19:06, 3.67s/it] 40%|████ | 209/520 [13:17<19:01, 3.67s/it] {'loss': 1.3916, 'grad_norm': 0.00045095065297616076, 'learning_rate': 0.4758360360733586, 'epoch': 0.4} + 40%|████ | 209/520 [13:17<19:01, 3.67s/it] 40%|████ | 210/520 [13:21<18:56, 3.67s/it] {'loss': 1.4461, 'grad_norm': 0.00047677699142035173, 'learning_rate': 0.47379782436790346, 'epoch': 0.4} + 40%|████ | 210/520 [13:21<18:56, 3.67s/it] 41%|████ | 211/520 [13:24<18:53, 3.67s/it] {'loss': 1.4642, 'grad_norm': 0.00040522927913855093, 'learning_rate': 0.47175480260482944, 'epoch': 0.41} + 41%|████ | 211/520 [13:24<18:53, 3.67s/it] 41%|████ | 212/520 [13:28<18:50, 3.67s/it] {'loss': 1.443, 'grad_norm': 0.000511713838600061, 'learning_rate': 0.46970705016398406, 'epoch': 0.41} + 41%|████ | 212/520 [13:28<18:50, 3.67s/it] 41%|████ | 213/520 [13:32<18:47, 3.67s/it] {'loss': 1.3909, 'grad_norm': 0.0006885164073973452, 'learning_rate': 0.4676546466090208, 'epoch': 0.41} + 41%|████ | 213/520 [13:32<18:47, 3.67s/it] 41%|████ | 214/520 [13:35<18:41, 3.67s/it] {'loss': 1.4164, 'grad_norm': 0.0004838883814349136, 'learning_rate': 0.4655976716843085, 'epoch': 0.41} + 41%|████ | 214/520 [13:35<18:41, 3.67s/it] 41%|████▏ | 215/520 [13:39<18:38, 3.67s/it] {'loss': 1.4486, 'grad_norm': 0.0007277298981247313, 'learning_rate': 0.4635362053118325, 'epoch': 0.41} + 41%|████▏ | 215/520 [13:39<18:38, 3.67s/it] 42%|████▏ | 216/520 [13:43<18:36, 3.67s/it] {'loss': 1.3005, 'grad_norm': 0.0005443431038782748, 'learning_rate': 0.46147032758808954, 'epoch': 0.42} + 42%|████▏ | 216/520 [13:43<18:36, 3.67s/it] 42%|████▏ | 217/520 [13:46<18:34, 3.68s/it] {'loss': 1.4386, 'grad_norm': 0.00048508464894913086, 'learning_rate': 0.4594001187809756, 'epoch': 0.42} + 42%|████▏ | 217/520 [13:46<18:34, 3.68s/it] 42%|████▏ | 218/520 [13:50<18:34, 3.69s/it] {'loss': 1.4483, 'grad_norm': 0.0005782121789331677, 'learning_rate': 0.457325659326667, 'epoch': 0.42} + 42%|████▏ | 218/520 [13:50<18:34, 3.69s/it] 42%|████▏ | 219/520 [13:54<18:28, 3.68s/it] {'loss': 1.4145, 'grad_norm': 0.0004949470427176764, 'learning_rate': 0.4552470298264955, 'epoch': 0.42} + 42%|████▏ | 219/520 [13:54<18:28, 3.68s/it] 42%|████▏ | 220/520 [13:57<18:24, 3.68s/it] {'loss': 1.5177, 'grad_norm': 0.0005645595029189905, 'learning_rate': 0.45316431104381644, 'epoch': 0.42} + 42%|████▏ | 220/520 [13:57<18:24, 3.68s/it] 42%|████▎ | 221/520 [14:01<18:19, 3.68s/it] {'loss': 1.4366, 'grad_norm': 0.00045271291433035966, 'learning_rate': 0.45107758390087044, 'epoch': 0.42} + 42%|████▎ | 221/520 [14:01<18:19, 3.68s/it] 43%|████▎ | 222/520 [14:05<18:17, 3.68s/it] {'loss': 1.3377, 'grad_norm': 0.00044366144039816724, 'learning_rate': 0.44898692947563956, 'epoch': 0.43} + 43%|████▎ | 222/520 [14:05<18:17, 3.68s/it] 43%|████▎ | 223/520 [14:08<18:15, 3.69s/it] {'loss': 1.3452, 'grad_norm': 0.00047777903768998515, 'learning_rate': 0.44689242899869724, 'epoch': 0.43} + 43%|████▎ | 223/520 [14:08<18:15, 3.69s/it] 43%|████▎ | 224/520 [14:12<18:14, 3.70s/it] {'loss': 1.7937, 'grad_norm': 0.0006667040502422344, 'learning_rate': 0.44479416385005177, 'epoch': 0.43} + 43%|████▎ | 224/520 [14:12<18:14, 3.70s/it] 43%|████▎ | 225/520 [14:16<18:11, 3.70s/it] {'loss': 1.357, 'grad_norm': 0.0004966627016949919, 'learning_rate': 0.4426922155559845, 'epoch': 0.43} + 43%|████▎ | 225/520 [14:16<18:11, 3.70s/it] 43%|████▎ | 226/520 [14:19<18:05, 3.69s/it] {'loss': 1.4702, 'grad_norm': 0.0004027859500428354, 'learning_rate': 0.44058666578588224, 'epoch': 0.43} + 43%|████▎ | 226/520 [14:19<18:05, 3.69s/it] 44%|████▎ | 227/520 [14:23<18:02, 3.69s/it] {'loss': 1.44, 'grad_norm': 0.0004519779371994806, 'learning_rate': 0.4384775963490641, 'epoch': 0.44} + 44%|████▎ | 227/520 [14:23<18:02, 3.69s/it] 44%|████▍ | 228/520 [14:27<18:01, 3.71s/it] {'loss': 1.6639, 'grad_norm': 0.0004937825272232031, 'learning_rate': 0.4363650891916027, 'epoch': 0.44} + 44%|████▍ | 228/520 [14:27<18:01, 3.71s/it] 44%|████▍ | 229/520 [14:31<17:57, 3.70s/it] {'loss': 1.4444, 'grad_norm': 0.0006408941458044617, 'learning_rate': 0.43424922639314056, 'epoch': 0.44} + 44%|████▍ | 229/520 [14:31<17:57, 3.70s/it] 44%|████▍ | 230/520 [14:34<17:53, 3.70s/it] {'loss': 1.3065, 'grad_norm': 0.0005321473369931823, 'learning_rate': 0.43213009016370035, 'epoch': 0.44} + 44%|████▍ | 230/520 [14:34<17:53, 3.70s/it] 44%|████▍ | 231/520 [14:38<17:48, 3.70s/it] {'loss': 1.3678, 'grad_norm': 0.00047751458204912847, 'learning_rate': 0.43000776284049136, 'epoch': 0.44} + 44%|████▍ | 231/520 [14:38<17:48, 3.70s/it] 45%|████▍ | 232/520 [14:42<17:45, 3.70s/it] {'loss': 1.6981, 'grad_norm': 0.0005949267102636048, 'learning_rate': 0.42788232688471, 'epoch': 0.45} + 45%|████▍ | 232/520 [14:42<17:45, 3.70s/it] 45%|████▍ | 233/520 [14:45<17:37, 3.68s/it] {'loss': 1.562, 'grad_norm': 0.0007753838706676592, 'learning_rate': 0.425753864878336, 'epoch': 0.45} + 45%|████▍ | 233/520 [14:45<17:37, 3.68s/it] 45%|████▌ | 234/520 [14:49<17:32, 3.68s/it] {'loss': 1.3358, 'grad_norm': 0.0006242876209572992, 'learning_rate': 0.42362245952092353, 'epoch': 0.45} + 45%|████▌ | 234/520 [14:49<17:32, 3.68s/it] 45%|████▌ | 235/520 [14:53<17:29, 3.68s/it] {'loss': 1.3714, 'grad_norm': 0.0005292556397796142, 'learning_rate': 0.4214881936263881, 'epoch': 0.45} + 45%|████▌ | 235/520 [14:53<17:29, 3.68s/it] 45%|████▌ | 236/520 [14:56<17:23, 3.67s/it] {'loss': 1.5004, 'grad_norm': 0.0005739280866983194, 'learning_rate': 0.41935115011978913, 'epoch': 0.45} + 45%|████▌ | 236/520 [14:56<17:23, 3.67s/it] 46%|████▌ | 237/520 [15:00<17:20, 3.68s/it] {'loss': 1.4404, 'grad_norm': 0.0004633178492778161, 'learning_rate': 0.41721141203410766, 'epoch': 0.46} + 46%|████▌ | 237/520 [15:00<17:20, 3.68s/it] 46%|████▌ | 238/520 [15:04<17:16, 3.68s/it] {'loss': 1.4074, 'grad_norm': 0.0006685660596674742, 'learning_rate': 0.41506906250702014, 'epoch': 0.46} + 46%|████▌ | 238/520 [15:04<17:16, 3.68s/it] 46%|████▌ | 239/520 [15:07<17:13, 3.68s/it] {'loss': 1.4994, 'grad_norm': 0.0007436378219388383, 'learning_rate': 0.4129241847776684, 'epoch': 0.46} + 46%|████▌ | 239/520 [15:07<17:13, 3.68s/it] 46%|████▌ | 240/520 [15:11<17:09, 3.68s/it] {'loss': 1.2465, 'grad_norm': 0.0004019714842638116, 'learning_rate': 0.4107768621834257, 'epoch': 0.46} + 46%|████▌ | 240/520 [15:11<17:09, 3.68s/it] 46%|████▋ | 241/520 [15:15<17:06, 3.68s/it] {'loss': 1.3368, 'grad_norm': 0.0004935765963706099, 'learning_rate': 0.4086271781566577, 'epoch': 0.46} + 46%|████▋ | 241/520 [15:15<17:06, 3.68s/it] 47%|████▋ | 242/520 [15:18<17:02, 3.68s/it] {'loss': 1.3702, 'grad_norm': 0.0003936859732779889, 'learning_rate': 0.4064752162214823, 'epoch': 0.47} + 47%|████▋ | 242/520 [15:18<17:02, 3.68s/it] 47%|████▋ | 243/520 [15:22<16:58, 3.68s/it] {'loss': 1.3556, 'grad_norm': 0.0006246341361028007, 'learning_rate': 0.40432105999052304, 'epoch': 0.47} + 47%|████▋ | 243/520 [15:22<16:58, 3.68s/it] 47%|████▋ | 244/520 [15:26<16:55, 3.68s/it] {'loss': 1.499, 'grad_norm': 0.0008506910742661841, 'learning_rate': 0.402164793161661, 'epoch': 0.47} + 47%|████▋ | 244/520 [15:26<16:55, 3.68s/it] 47%|████▋ | 245/520 [15:30<16:55, 3.69s/it] {'loss': 1.3575, 'grad_norm': 0.0004414492536338264, 'learning_rate': 0.400006499514783, 'epoch': 0.47} + 47%|████▋ | 245/520 [15:30<16:55, 3.69s/it] 47%|████▋ | 246/520 [15:33<17:07, 3.75s/it] {'loss': 1.6686, 'grad_norm': 0.000527417623746539, 'learning_rate': 0.3978462629085257, 'epoch': 0.47} + 47%|████▋ | 246/520 [15:33<17:07, 3.75s/it] 48%|████▊ | 247/520 [15:37<17:14, 3.79s/it] {'loss': 1.5228, 'grad_norm': 0.001031222160705322, 'learning_rate': 0.3956841672770181, 'epoch': 0.47} + 48%|████▊ | 247/520 [15:37<17:14, 3.79s/it] 48%|████▊ | 248/520 [15:41<17:16, 3.81s/it] {'loss': 1.3348, 'grad_norm': 0.00048083856774258974, 'learning_rate': 0.39352029662661986, 'epoch': 0.48} + 48%|████▊ | 248/520 [15:41<17:16, 3.81s/it] 48%|████▊ | 249/520 [15:45<17:17, 3.83s/it] {'loss': 1.4683, 'grad_norm': 0.0007067319315677308, 'learning_rate': 0.3913547350326575, 'epoch': 0.48} + 48%|████▊ | 249/520 [15:45<17:17, 3.83s/it] 48%|████▊ | 250/520 [15:49<17:17, 3.84s/it] {'loss': 1.4149, 'grad_norm': 0.0006028972888859568, 'learning_rate': 0.3891875666361577, 'epoch': 0.48} + 48%|████▊ | 250/520 [15:49<17:17, 3.84s/it] 48%|████▊ | 251/520 [15:53<17:14, 3.85s/it] {'loss': 1.4549, 'grad_norm': 0.0005564872919105603, 'learning_rate': 0.38701887564057824, 'epoch': 0.48} + 48%|████▊ | 251/520 [15:53<17:14, 3.85s/it] 48%|████▊ | 252/520 [15:57<17:09, 3.84s/it] {'loss': 1.5254, 'grad_norm': 0.0007367557104246526, 'learning_rate': 0.3848487463085358, 'epoch': 0.48} + 48%|████▊ | 252/520 [15:57<17:09, 3.84s/it] 49%|████▊ | 253/520 [16:00<17:10, 3.86s/it] {'loss': 1.4502, 'grad_norm': 0.0004409573139459021, 'learning_rate': 0.38267726295853266, 'epoch': 0.49} + 49%|████▊ | 253/520 [16:00<17:10, 3.86s/it] 49%|████▉ | 254/520 [16:04<17:07, 3.86s/it] {'loss': 1.3567, 'grad_norm': 0.0005134683404206887, 'learning_rate': 0.3805045099616804, 'epoch': 0.49} + 49%|████▉ | 254/520 [16:04<17:07, 3.86s/it] 49%|████▉ | 255/520 [16:08<17:10, 3.89s/it] {'loss': 1.4014, 'grad_norm': 0.0004488946388570298, 'learning_rate': 0.37833057173842116, 'epoch': 0.49} + 49%|████▉ | 255/520 [16:08<17:10, 3.89s/it] 49%|████▉ | 256/520 [16:12<17:20, 3.94s/it] {'loss': 1.44, 'grad_norm': 0.0005921280219045587, 'learning_rate': 0.3761555327552485, 'epoch': 0.49} + 49%|████▉ | 256/520 [16:12<17:20, 3.94s/it] 49%|████▉ | 257/520 [16:16<17:25, 3.98s/it] {'loss': 1.4411, 'grad_norm': 0.0004756461008158528, 'learning_rate': 0.3739794775214248, 'epoch': 0.49} + 49%|████▉ | 257/520 [16:16<17:25, 3.98s/it] 50%|████▉ | 258/520 [16:21<17:30, 4.01s/it] {'loss': 1.4598, 'grad_norm': 0.0004548725985363074, 'learning_rate': 0.37180249058569825, 'epoch': 0.5} + 50%|████▉ | 258/520 [16:21<17:30, 4.01s/it] 50%|████▉ | 259/520 [16:25<17:35, 4.05s/it] {'loss': 1.5156, 'grad_norm': 0.0004849515093266967, 'learning_rate': 0.3696246565330171, 'epoch': 0.5} + 50%|████▉ | 259/520 [16:25<17:35, 4.05s/it] 50%|█████ | 260/520 [16:29<17:35, 4.06s/it] {'loss': 1.6596, 'grad_norm': 0.00046420824660183154, 'learning_rate': 0.36744605998124397, 'epoch': 0.5} + 50%|█████ | 260/520 [16:29<17:35, 4.06s/it] 50%|█████ | 261/520 [16:33<17:33, 4.07s/it] {'loss': 1.5642, 'grad_norm': 0.0006073728943042922, 'learning_rate': 0.3652667855778676, 'epoch': 0.5} + 50%|█████ | 261/520 [16:33<17:33, 4.07s/it] 50%|█████ | 262/520 [16:37<17:25, 4.05s/it] {'loss': 1.3729, 'grad_norm': 0.0004426115060500364, 'learning_rate': 0.36308691799671405, 'epoch': 0.5} + 50%|█████ | 262/520 [16:37<17:25, 4.05s/it] 51%|█████ | 263/520 [16:41<17:05, 3.99s/it] {'loss': 1.5689, 'grad_norm': 0.00046311642241989167, 'learning_rate': 0.3609065419346566, 'epoch': 0.51} + 51%|█████ | 263/520 [16:41<17:05, 3.99s/it] 51%|█████ | 264/520 [16:45<16:51, 3.95s/it] {'loss': 1.4686, 'grad_norm': 0.00042042358084380985, 'learning_rate': 0.35872574210832553, 'epoch': 0.51} + 51%|█████ | 264/520 [16:45<16:51, 3.95s/it] 51%|█████ | 265/520 [16:48<16:29, 3.88s/it] {'loss': 1.3681, 'grad_norm': 0.000728168722772418, 'learning_rate': 0.3565446032508158, 'epoch': 0.51} + 51%|█████ | 265/520 [16:48<16:29, 3.88s/it] 51%|█████ | 266/520 [16:52<16:09, 3.82s/it] {'loss': 1.2229, 'grad_norm': 0.00040218103668159926, 'learning_rate': 0.3543632101083953, 'epoch': 0.51} + 51%|█████ | 266/520 [16:52<16:09, 3.82s/it] 51%|█████▏ | 267/520 [16:56<15:55, 3.78s/it] {'loss': 1.3818, 'grad_norm': 0.0005851886162969596, 'learning_rate': 0.35218164743721175, 'epoch': 0.51} + 51%|█████▏ | 267/520 [16:56<15:55, 3.78s/it] 52%|█████▏ | 268/520 [16:59<15:45, 3.75s/it] {'loss': 1.7226, 'grad_norm': 0.0012600896400938029, 'learning_rate': 0.35, 'epoch': 0.52} + 52%|█████▏ | 268/520 [16:59<15:45, 3.75s/it] 52%|█████▏ | 269/520 [17:03<15:34, 3.72s/it] {'loss': 1.4756, 'grad_norm': 0.0008570549946830843, 'learning_rate': 0.3478183525627882, 'epoch': 0.52} + 52%|█████▏ | 269/520 [17:03<15:34, 3.72s/it] 52%|█████▏ | 270/520 [17:07<15:29, 3.72s/it] {'loss': 1.4895, 'grad_norm': 0.000467640785831139, 'learning_rate': 0.34563678989160473, 'epoch': 0.52} + 52%|█████▏ | 270/520 [17:07<15:29, 3.72s/it] 52%|█████▏ | 271/520 [17:10<15:26, 3.72s/it] {'loss': 1.4715, 'grad_norm': 0.0005797515037310036, 'learning_rate': 0.3434553967491843, 'epoch': 0.52} + 52%|█████▏ | 271/520 [17:10<15:26, 3.72s/it] 52%|█████▏ | 272/520 [17:14<15:26, 3.73s/it] {'loss': 1.5069, 'grad_norm': 0.0006919700394875857, 'learning_rate': 0.34127425789167454, 'epoch': 0.52} + 52%|█████▏ | 272/520 [17:14<15:26, 3.73s/it] 52%|█████▎ | 273/520 [17:18<15:24, 3.74s/it] {'loss': 1.6287, 'grad_norm': 0.0006310176470933784, 'learning_rate': 0.33909345806534347, 'epoch': 0.53} + 52%|█████▎ | 273/520 [17:18<15:24, 3.74s/it] 53%|█████▎ | 274/520 [17:22<15:13, 3.71s/it] {'loss': 1.4257, 'grad_norm': 0.0007910185807010805, 'learning_rate': 0.33691308200328607, 'epoch': 0.53} + 53%|█████▎ | 274/520 [17:22<15:13, 3.71s/it] 53%|█████▎ | 275/520 [17:25<15:05, 3.69s/it] {'loss': 1.3605, 'grad_norm': 0.0004148835283617035, 'learning_rate': 0.33473321442213244, 'epoch': 0.53} + 53%|█████▎ | 275/520 [17:25<15:05, 3.69s/it] 53%|█████▎ | 276/520 [17:29<14:58, 3.68s/it] {'loss': 1.4479, 'grad_norm': 0.0006772982175486665, 'learning_rate': 0.33255394001875593, 'epoch': 0.53} + 53%|█████▎ | 276/520 [17:29<14:58, 3.68s/it] 53%|█████▎ | 277/520 [17:33<14:54, 3.68s/it] {'loss': 1.6187, 'grad_norm': 0.0005782856642007809, 'learning_rate': 0.33037534346698283, 'epoch': 0.53} + 53%|█████▎ | 277/520 [17:33<14:54, 3.68s/it] 53%|█████▎ | 278/520 [17:36<14:48, 3.67s/it] {'loss': 1.3276, 'grad_norm': 0.00043273914936000686, 'learning_rate': 0.32819750941430176, 'epoch': 0.53} + 53%|█████▎ | 278/520 [17:36<14:48, 3.67s/it] 54%|█████▎ | 279/520 [17:40<14:44, 3.67s/it] {'loss': 1.5397, 'grad_norm': 0.000574210996811887, 'learning_rate': 0.32602052247857516, 'epoch': 0.54} + 54%|█████▎ | 279/520 [17:40<14:44, 3.67s/it] 54%|█████▍ | 280/520 [17:44<14:40, 3.67s/it] {'loss': 1.3641, 'grad_norm': 0.0007770164964782644, 'learning_rate': 0.3238444672447515, 'epoch': 0.54} + 54%|█████▍ | 280/520 [17:44<14:40, 3.67s/it] 54%|█████▍ | 281/520 [17:47<14:35, 3.66s/it] {'loss': 1.481, 'grad_norm': 0.00047090926448574156, 'learning_rate': 0.3216694282615788, 'epoch': 0.54} + 54%|█████▍ | 281/520 [17:47<14:35, 3.66s/it] 54%|█████▍ | 282/520 [17:51<14:34, 3.67s/it] {'loss': 1.336, 'grad_norm': 0.0005187274831335425, 'learning_rate': 0.3194954900383196, 'epoch': 0.54} + 54%|█████▍ | 282/520 [17:51<14:34, 3.67s/it] 54%|█████▍ | 283/520 [17:55<14:29, 3.67s/it] {'loss': 1.5256, 'grad_norm': 0.0005582393486024636, 'learning_rate': 0.3173227370414673, 'epoch': 0.54} + 54%|█████▍ | 283/520 [17:55<14:29, 3.67s/it] 55%|█████▍ | 284/520 [17:58<14:25, 3.67s/it] {'loss': 1.4968, 'grad_norm': 0.0006581601778386872, 'learning_rate': 0.3151512536914642, 'epoch': 0.55} + 55%|█████▍ | 284/520 [17:58<14:25, 3.67s/it] 55%|█████▍ | 285/520 [18:02<14:25, 3.68s/it] {'loss': 1.3681, 'grad_norm': 0.0004820916192752337, 'learning_rate': 0.31298112435942177, 'epoch': 0.55} + 55%|█████▍ | 285/520 [18:02<14:25, 3.68s/it] 55%|█████▌ | 286/520 [18:06<14:20, 3.68s/it] {'loss': 1.227, 'grad_norm': 0.00045608697121557343, 'learning_rate': 0.31081243336384223, 'epoch': 0.55} + 55%|█████▌ | 286/520 [18:06<14:20, 3.68s/it] 55%|█████▌ | 287/520 [18:09<14:17, 3.68s/it] {'loss': 1.4975, 'grad_norm': 0.0004596464734418638, 'learning_rate': 0.3086452649673425, 'epoch': 0.55} + 55%|█████▌ | 287/520 [18:09<14:17, 3.68s/it] 55%|█████▌ | 288/520 [18:13<14:15, 3.69s/it] {'loss': 1.5518, 'grad_norm': 0.0006847115573073452, 'learning_rate': 0.30647970337338026, 'epoch': 0.55} + 55%|█████▌ | 288/520 [18:13<14:15, 3.69s/it] 56%|█████▌ | 289/520 [18:17<14:11, 3.68s/it] {'loss': 1.3926, 'grad_norm': 0.00042983308331124084, 'learning_rate': 0.304315832722982, 'epoch': 0.56} + 56%|█████▌ | 289/520 [18:17<14:11, 3.68s/it] 56%|█████▌ | 290/520 [18:20<14:08, 3.69s/it] {'loss': 1.3203, 'grad_norm': 0.00043471658865542766, 'learning_rate': 0.30215373709147436, 'epoch': 0.56} + 56%|█████▌ | 290/520 [18:20<14:08, 3.69s/it] 56%|█████▌ | 291/520 [18:24<14:05, 3.69s/it] {'loss': 1.355, 'grad_norm': 0.0008798135360790489, 'learning_rate': 0.29999350048521706, 'epoch': 0.56} + 56%|█████▌ | 291/520 [18:24<14:05, 3.69s/it] 56%|█████▌ | 292/520 [18:28<14:00, 3.68s/it] {'loss': 1.4292, 'grad_norm': 0.00073439568572247, 'learning_rate': 0.29783520683833886, 'epoch': 0.56} + 56%|█████▌ | 292/520 [18:28<14:00, 3.68s/it] 56%|█████▋ | 293/520 [18:31<13:55, 3.68s/it] {'loss': 1.346, 'grad_norm': 0.0007967789466070812, 'learning_rate': 0.2956789400094769, 'epoch': 0.56} + 56%|█████▋ | 293/520 [18:31<13:55, 3.68s/it] 57%|█████▋ | 294/520 [18:35<13:53, 3.69s/it] {'loss': 1.3843, 'grad_norm': 0.0005425668105609267, 'learning_rate': 0.29352478377851765, 'epoch': 0.57} + 57%|█████▋ | 294/520 [18:35<13:53, 3.69s/it] 57%|█████▋ | 295/520 [18:39<13:48, 3.68s/it] {'loss': 1.6106, 'grad_norm': 0.0004442535075531298, 'learning_rate': 0.2913728218433423, 'epoch': 0.57} + 57%|█████▋ | 295/520 [18:39<13:48, 3.68s/it] 57%|█████▋ | 296/520 [18:42<13:44, 3.68s/it] {'loss': 1.3316, 'grad_norm': 0.0004341767226049842, 'learning_rate': 0.2892231378165744, 'epoch': 0.57} + 57%|█████▋ | 296/520 [18:42<13:44, 3.68s/it] 57%|█████▋ | 297/520 [18:46<13:38, 3.67s/it] {'loss': 1.4761, 'grad_norm': 0.00046719504732749625, 'learning_rate': 0.28707581522233155, 'epoch': 0.57} + 57%|█████▋ | 297/520 [18:46<13:38, 3.67s/it] 57%|█████▋ | 298/520 [18:50<13:33, 3.67s/it] {'loss': 1.4384, 'grad_norm': 0.00036832195901276234, 'learning_rate': 0.2849309374929799, 'epoch': 0.57} + 57%|█████▋ | 298/520 [18:50<13:33, 3.67s/it] 57%|█████▊ | 299/520 [18:53<13:30, 3.67s/it] {'loss': 1.5928, 'grad_norm': 0.0005927090189535422, 'learning_rate': 0.28278858796589235, 'epoch': 0.57} + 57%|█████▊ | 299/520 [18:53<13:30, 3.67s/it] 58%|█████▊ | 300/520 [18:57<13:27, 3.67s/it] {'loss': 1.4885, 'grad_norm': 0.00040692872864545244, 'learning_rate': 0.2806488498802109, 'epoch': 0.58} + 58%|█████▊ | 300/520 [18:57<13:27, 3.67s/it] 58%|█████▊ | 301/520 [19:01<13:24, 3.67s/it] {'loss': 1.4866, 'grad_norm': 0.0004065959640385841, 'learning_rate': 0.2785118063736119, 'epoch': 0.58} + 58%|█████▊ | 301/520 [19:01<13:24, 3.67s/it] 58%|█████▊ | 302/520 [19:04<13:20, 3.67s/it] {'loss': 1.611, 'grad_norm': 0.0005338349629624692, 'learning_rate': 0.27637754047907653, 'epoch': 0.58} + 58%|█████▊ | 302/520 [19:04<13:20, 3.67s/it] 58%|█████▊ | 303/520 [19:08<13:15, 3.67s/it] {'loss': 1.3801, 'grad_norm': 0.0009551245960220929, 'learning_rate': 0.27424613512166396, 'epoch': 0.58} + 58%|█████▊ | 303/520 [19:08<13:15, 3.67s/it] 58%|█████▊ | 304/520 [19:12<13:13, 3.67s/it] {'loss': 1.4897, 'grad_norm': 0.00042153060324356733, 'learning_rate': 0.27211767311528995, 'epoch': 0.58} + 58%|█████▊ | 304/520 [19:12<13:13, 3.67s/it] 59%|█████▊ | 305/520 [19:15<13:10, 3.68s/it] {'loss': 1.5509, 'grad_norm': 0.0007873107885300344, 'learning_rate': 0.2699922371595087, 'epoch': 0.59} + 59%|█████▊ | 305/520 [19:15<13:10, 3.68s/it] 59%|█████▉ | 306/520 [19:19<13:07, 3.68s/it] {'loss': 1.4411, 'grad_norm': 0.00044965897082381215, 'learning_rate': 0.2678699098362997, 'epoch': 0.59} + 59%|█████▉ | 306/520 [19:19<13:07, 3.68s/it] 59%|█████▉ | 307/520 [19:23<13:33, 3.82s/it] {'loss': 1.394, 'grad_norm': 0.0004917139508718828, 'learning_rate': 0.2657507736068595, 'epoch': 0.59} + 59%|█████▉ | 307/520 [19:23<13:33, 3.82s/it] 59%|█████▉ | 308/520 [19:27<13:25, 3.80s/it] {'loss': 1.4987, 'grad_norm': 0.00046799788282528895, 'learning_rate': 0.2636349108083972, 'epoch': 0.59} + 59%|█████▉ | 308/520 [19:27<13:25, 3.80s/it] 59%|█████▉ | 309/520 [19:31<13:16, 3.77s/it] {'loss': 1.3816, 'grad_norm': 0.000649308464276775, 'learning_rate': 0.2615224036509358, 'epoch': 0.59} + 59%|█████▉ | 309/520 [19:31<13:16, 3.77s/it] 60%|█████▉ | 310/520 [19:34<13:08, 3.76s/it] {'loss': 1.3536, 'grad_norm': 0.00046152374581168675, 'learning_rate': 0.25941333421411766, 'epoch': 0.6} + 60%|█████▉ | 310/520 [19:34<13:08, 3.76s/it] 60%|█████▉ | 311/520 [19:38<13:00, 3.74s/it] {'loss': 1.3371, 'grad_norm': 0.0007521969236863106, 'learning_rate': 0.2573077844440154, 'epoch': 0.6} + 60%|█████▉ | 311/520 [19:38<13:00, 3.74s/it] 60%|██████ | 312/520 [19:42<12:56, 3.73s/it] {'loss': 1.3119, 'grad_norm': 0.0005026667305737267, 'learning_rate': 0.25520583614994824, 'epoch': 0.6} + 60%|██████ | 312/520 [19:42<12:56, 3.73s/it] 60%|██████ | 313/520 [19:46<12:49, 3.72s/it] {'loss': 1.2859, 'grad_norm': 0.0005483482284725593, 'learning_rate': 0.2531075710013027, 'epoch': 0.6} + 60%|██████ | 313/520 [19:46<12:49, 3.72s/it] 60%|██████ | 314/520 [19:50<13:08, 3.83s/it] {'loss': 1.3354, 'grad_norm': 0.00042522861636782184, 'learning_rate': 0.2510130705243604, 'epoch': 0.6} + 60%|██████ | 314/520 [19:50<13:08, 3.83s/it] 61%|██████ | 315/520 [19:53<12:52, 3.77s/it] {'loss': 1.6125, 'grad_norm': 0.00057737924291805, 'learning_rate': 0.24892241609912957, 'epoch': 0.61} + 61%|██████ | 315/520 [19:53<12:52, 3.77s/it] 61%|██████ | 316/520 [19:57<13:04, 3.84s/it] {'loss': 1.3228, 'grad_norm': 0.0005526693989898164, 'learning_rate': 0.2468356889561835, 'epoch': 0.61} + 61%|██████ | 316/520 [19:57<13:04, 3.84s/it] 61%|██████ | 317/520 [20:01<12:48, 3.79s/it] {'loss': 1.33, 'grad_norm': 0.0006871701136070948, 'learning_rate': 0.24475297017350442, 'epoch': 0.61} + 61%|██████ | 317/520 [20:01<12:48, 3.79s/it] 61%|██████ | 318/520 [20:05<12:35, 3.74s/it] {'loss': 1.4733, 'grad_norm': 0.00048535043608663046, 'learning_rate': 0.24267434067333302, 'epoch': 0.61} + 61%|██████ | 318/520 [20:05<12:35, 3.74s/it] 61%|██████▏ | 319/520 [20:09<13:13, 3.95s/it] {'loss': 1.3255, 'grad_norm': 0.0005401226242561812, 'learning_rate': 0.24059988121902445, 'epoch': 0.61} + 61%|██████▏ | 319/520 [20:09<13:13, 3.95s/it] 62%|██████▏ | 320/520 [20:13<12:50, 3.85s/it] {'loss': 1.2595, 'grad_norm': 0.000554619571284987, 'learning_rate': 0.23852967241191048, 'epoch': 0.62} + 62%|██████▏ | 320/520 [20:13<12:50, 3.85s/it] 62%|██████▏ | 321/520 [20:16<12:33, 3.79s/it] {'loss': 1.4957, 'grad_norm': 0.0006837852536401415, 'learning_rate': 0.23646379468816756, 'epoch': 0.62} + 62%|██████▏ | 321/520 [20:16<12:33, 3.79s/it] 62%|██████▏ | 322/520 [20:20<12:21, 3.74s/it] {'loss': 1.446, 'grad_norm': 0.0009106462286312261, 'learning_rate': 0.2344023283156916, 'epoch': 0.62} + 62%|██████▏ | 322/520 [20:20<12:21, 3.74s/it] 62%|██████▏ | 323/520 [20:24<12:10, 3.71s/it] {'loss': 1.5259, 'grad_norm': 0.0009057738351010193, 'learning_rate': 0.23234535339097928, 'epoch': 0.62} + 62%|██████▏ | 323/520 [20:24<12:10, 3.71s/it] 62%|██████▏ | 324/520 [20:27<12:01, 3.68s/it] {'loss': 1.419, 'grad_norm': 0.0004923256522568081, 'learning_rate': 0.23029294983601595, 'epoch': 0.62} + 62%|██████▏ | 324/520 [20:27<12:01, 3.68s/it] 62%|██████▎ | 325/520 [20:31<11:59, 3.69s/it] {'loss': 1.4186, 'grad_norm': 0.0004462460045837693, 'learning_rate': 0.22824519739517043, 'epoch': 0.62} + 62%|██████▎ | 325/520 [20:31<11:59, 3.69s/it] 63%|██████▎ | 326/520 [20:35<12:04, 3.74s/it] {'loss': 1.4086, 'grad_norm': 0.00048088477720213755, 'learning_rate': 0.2262021756320965, 'epoch': 0.63} + 63%|██████▎ | 326/520 [20:35<12:04, 3.74s/it] 63%|██████▎ | 327/520 [20:39<12:06, 3.77s/it] {'loss': 1.609, 'grad_norm': 0.0005911557236376249, 'learning_rate': 0.22416396392664134, 'epoch': 0.63} + 63%|██████▎ | 327/520 [20:39<12:06, 3.77s/it] 63%|██████▎ | 328/520 [20:42<12:07, 3.79s/it] {'loss': 1.4857, 'grad_norm': 0.0005058894881647668, 'learning_rate': 0.22213064147176173, 'epoch': 0.63} + 63%|██████▎ | 328/520 [20:42<12:07, 3.79s/it] 63%|██████▎ | 329/520 [20:46<12:10, 3.82s/it] {'loss': 1.3134, 'grad_norm': 0.0003827829114063173, 'learning_rate': 0.2201022872704467, 'epoch': 0.63} + 63%|██████▎ | 329/520 [20:46<12:10, 3.82s/it] 63%|██████▎ | 330/520 [20:50<12:09, 3.84s/it] {'loss': 1.414, 'grad_norm': 0.0008163565220124956, 'learning_rate': 0.21807898013264815, 'epoch': 0.63} + 63%|██████▎ | 330/520 [20:50<12:09, 3.84s/it] 64%|██████▎ | 331/520 [20:54<11:58, 3.80s/it] {'loss': 1.3577, 'grad_norm': 0.0005335114724372876, 'learning_rate': 0.21606079867221858, 'epoch': 0.64} + 64%|██████▎ | 331/520 [20:54<11:58, 3.80s/it] 64%|██████▍ | 332/520 [20:58<11:45, 3.75s/it] {'loss': 1.6024, 'grad_norm': 0.000538983181562698, 'learning_rate': 0.21404782130385686, 'epoch': 0.64} + 64%|██████▍ | 332/520 [20:58<11:45, 3.75s/it] 64%|██████▍ | 333/520 [21:01<11:38, 3.74s/it] {'loss': 1.5436, 'grad_norm': 0.0005769283398758064, 'learning_rate': 0.21204012624006124, 'epoch': 0.64} + 64%|██████▍ | 333/520 [21:01<11:38, 3.74s/it] 64%|██████▍ | 334/520 [21:05<11:29, 3.71s/it] {'loss': 1.4097, 'grad_norm': 0.0008353147473279478, 'learning_rate': 0.2100377914880907, 'epoch': 0.64} + 64%|██████▍ | 334/520 [21:05<11:29, 3.71s/it] 64%|██████▍ | 335/520 [21:09<11:23, 3.69s/it] {'loss': 1.425, 'grad_norm': 0.0008513198648191213, 'learning_rate': 0.20804089484693378, 'epoch': 0.64} + 64%|██████▍ | 335/520 [21:09<11:23, 3.69s/it] 65%|██████▍ | 336/520 [21:12<11:17, 3.68s/it] {'loss': 1.2981, 'grad_norm': 0.0006988631226925639, 'learning_rate': 0.20604951390428602, 'epoch': 0.65} + 65%|██████▍ | 336/520 [21:12<11:17, 3.68s/it] 65%|██████▍ | 337/520 [21:16<11:12, 3.67s/it] {'loss': 1.3025, 'grad_norm': 0.0006334537917409819, 'learning_rate': 0.2040637260335353, 'epoch': 0.65} + 65%|██████▍ | 337/520 [21:16<11:12, 3.67s/it] 65%|██████▌ | 338/520 [21:20<11:08, 3.67s/it] {'loss': 1.4364, 'grad_norm': 0.0009100085748195875, 'learning_rate': 0.20208360839075523, 'epoch': 0.65} + 65%|██████▌ | 338/520 [21:20<11:08, 3.67s/it] 65%|██████▌ | 339/520 [21:23<11:05, 3.67s/it] {'loss': 1.3791, 'grad_norm': 0.0006274407662642824, 'learning_rate': 0.20010923791170798, 'epoch': 0.65} + 65%|██████▌ | 339/520 [21:23<11:05, 3.67s/it] 65%|██████▌ | 340/520 [21:27<11:00, 3.67s/it] {'loss': 1.3388, 'grad_norm': 0.0007730226247947219, 'learning_rate': 0.19814069130885467, 'epoch': 0.65} + 65%|██████▌ | 340/520 [21:27<11:00, 3.67s/it] 66%|██████▌ | 341/520 [21:30<10:56, 3.66s/it] {'loss': 1.3752, 'grad_norm': 0.0004231389112698688, 'learning_rate': 0.19617804506837422, 'epoch': 0.66} + 66%|██████▌ | 341/520 [21:30<10:56, 3.66s/it] 66%|██████▌ | 342/520 [21:34<10:56, 3.69s/it] {'loss': 1.5879, 'grad_norm': 0.0011851839053319399, 'learning_rate': 0.19422137544719265, 'epoch': 0.66} + 66%|██████▌ | 342/520 [21:34<10:56, 3.69s/it] 66%|██████▌ | 343/520 [21:38<10:56, 3.71s/it] {'loss': 1.5464, 'grad_norm': 0.000616598266099426, 'learning_rate': 0.1922707584700191, 'epoch': 0.66} + 66%|██████▌ | 343/520 [21:38<10:56, 3.71s/it] 66%|██████▌ | 344/520 [21:42<10:55, 3.72s/it] {'loss': 1.3226, 'grad_norm': 0.0006992291467756126, 'learning_rate': 0.19032626992639293, 'epoch': 0.66} + 66%|██████▌ | 344/520 [21:42<10:55, 3.72s/it] 66%|██████▋ | 345/520 [21:46<10:53, 3.74s/it] {'loss': 1.4447, 'grad_norm': 0.000637078129754577, 'learning_rate': 0.1883879853677382, 'epoch': 0.66} + 66%|██████▋ | 345/520 [21:46<10:53, 3.74s/it] 67%|██████▋ | 346/520 [21:49<10:46, 3.71s/it] {'loss': 1.5354, 'grad_norm': 0.0005655392412468906, 'learning_rate': 0.18645598010442826, 'epoch': 0.67} + 67%|██████▋ | 346/520 [21:49<10:46, 3.71s/it] 67%|██████▋ | 347/520 [21:53<10:39, 3.69s/it] {'loss': 1.3451, 'grad_norm': 0.00041436128749199166, 'learning_rate': 0.18453032920286058, 'epoch': 0.67} + 67%|██████▋ | 347/520 [21:53<10:39, 3.69s/it]Token indices sequence length is longer than the specified maximum sequence length for this model (2778 > 2048). Running this sequence through the model will result in indexing errors + 67%|██████▋ | 348/520 [21:57<10:38, 3.71s/it] {'loss': 1.3313, 'grad_norm': 0.0005071418127664326, 'learning_rate': 0.18261110748253873, 'epoch': 0.67} + 67%|██████▋ | 348/520 [21:57<10:38, 3.71s/it] 67%|██████▋ | 349/520 [22:00<10:40, 3.75s/it] {'loss': 1.3545, 'grad_norm': 0.000706685310575106, 'learning_rate': 0.18069838951316605, 'epoch': 0.67} + 67%|██████▋ | 349/520 [22:00<10:40, 3.75s/it] 67%|██████▋ | 350/520 [22:04<10:39, 3.76s/it] {'loss': 1.3806, 'grad_norm': 0.0004209647631267898, 'learning_rate': 0.17879224961174886, 'epoch': 0.67} + 67%|██████▋ | 350/520 [22:04<10:39, 3.76s/it] 68%|██████▊ | 351/520 [22:08<10:34, 3.75s/it] {'loss': 1.293, 'grad_norm': 0.0004892573391290484, 'learning_rate': 0.1768927618397074, 'epoch': 0.68} + 68%|██████▊ | 351/520 [22:08<10:34, 3.75s/it] 68%|██████▊ | 352/520 [22:12<10:29, 3.75s/it] {'loss': 1.43, 'grad_norm': 0.000624125717680862, 'learning_rate': 0.17500000000000007, 'epoch': 0.68} + 68%|██████▊ | 352/520 [22:12<10:29, 3.75s/it] 68%|██████▊ | 353/520 [22:16<10:32, 3.79s/it] {'loss': 1.4904, 'grad_norm': 0.0007215089465314275, 'learning_rate': 0.17311403763425434, 'epoch': 0.68} + 68%|██████▊ | 353/520 [22:16<10:32, 3.79s/it] 68%|██████▊ | 354/520 [22:19<10:32, 3.81s/it] {'loss': 1.6277, 'grad_norm': 0.00043419106128871017, 'learning_rate': 0.17123494801991013, 'epoch': 0.68} + 68%|██████▊ | 354/520 [22:19<10:32, 3.81s/it] 68%|██████▊ | 355/520 [22:23<10:31, 3.83s/it] {'loss': 1.365, 'grad_norm': 0.0006064344348524903, 'learning_rate': 0.16936280416737262, 'epoch': 0.68} + 68%|██████▊ | 355/520 [22:23<10:31, 3.83s/it] 68%|██████▊ | 356/520 [22:27<10:28, 3.83s/it] {'loss': 1.366, 'grad_norm': 0.0006448844135000268, 'learning_rate': 0.1674976788171757, 'epoch': 0.68} + 68%|██████▊ | 356/520 [22:27<10:28, 3.83s/it] 69%|██████▊ | 357/520 [22:31<10:25, 3.84s/it] {'loss': 1.3704, 'grad_norm': 0.0005044782618341049, 'learning_rate': 0.1656396444371547, 'epoch': 0.69} + 69%|██████▊ | 357/520 [22:31<10:25, 3.84s/it] 69%|██████▉ | 358/520 [22:35<10:23, 3.85s/it] {'loss': 1.3192, 'grad_norm': 0.0008468759038088787, 'learning_rate': 0.16378877321963223, 'epoch': 0.69} + 69%|██████▉ | 358/520 [22:35<10:23, 3.85s/it] 69%|██████▉ | 359/520 [22:39<10:20, 3.86s/it] {'loss': 1.558, 'grad_norm': 0.0004947322139095163, 'learning_rate': 0.16194513707861158, 'epoch': 0.69} + 69%|██████▉ | 359/520 [22:39<10:20, 3.86s/it] 69%|██████▉ | 360/520 [22:43<10:17, 3.86s/it] {'loss': 1.5607, 'grad_norm': 0.0007273285127078566, 'learning_rate': 0.16010880764698424, 'epoch': 0.69} + 69%|██████▉ | 360/520 [22:43<10:17, 3.86s/it] 69%|██████▉ | 361/520 [22:46<10:14, 3.86s/it] {'loss': 1.5687, 'grad_norm': 0.0005125388098145123, 'learning_rate': 0.15827985627374508, 'epoch': 0.69} + 69%|██████▉ | 361/520 [22:46<10:14, 3.86s/it] 70%|██████▉ | 362/520 [22:50<10:10, 3.86s/it] {'loss': 1.3545, 'grad_norm': 0.0005803769637025345, 'learning_rate': 0.15645835402122119, 'epoch': 0.7} + 70%|██████▉ | 362/520 [22:50<10:10, 3.86s/it] 70%|██████▉ | 363/520 [22:54<10:06, 3.86s/it] {'loss': 1.4453, 'grad_norm': 0.0005621722311024822, 'learning_rate': 0.15464437166231065, 'epoch': 0.7} + 70%|██████▉ | 363/520 [22:54<10:06, 3.86s/it] 70%|███████ | 364/520 [22:58<10:02, 3.87s/it] {'loss': 1.5798, 'grad_norm': 0.0006662642042184853, 'learning_rate': 0.15283797967773227, 'epoch': 0.7} + 70%|███████ | 364/520 [22:58<10:02, 3.87s/it] 70%|███████ | 365/520 [23:02<09:58, 3.86s/it] {'loss': 1.4884, 'grad_norm': 0.0005421268127118339, 'learning_rate': 0.1510392482532877, 'epoch': 0.7} + 70%|███████ | 365/520 [23:02<09:58, 3.86s/it] 70%|███████ | 366/520 [23:06<09:50, 3.83s/it] {'loss': 1.4399, 'grad_norm': 0.00043524259579742, 'learning_rate': 0.14924824727713396, 'epoch': 0.7} + 70%|███████ | 366/520 [23:06<09:50, 3.83s/it] 71%|███████ | 367/520 [23:09<09:38, 3.78s/it] {'loss': 1.4406, 'grad_norm': 0.0004655059825379772, 'learning_rate': 0.14746504633706797, 'epoch': 0.71} + 71%|███████ | 367/520 [23:09<09:38, 3.78s/it] 71%|███████ | 368/520 [23:13<09:30, 3.75s/it] {'loss': 1.2895, 'grad_norm': 0.0004726047295466064, 'learning_rate': 0.14568971471782363, 'epoch': 0.71} + 71%|███████ | 368/520 [23:13<09:30, 3.75s/it] 71%|███████ | 369/520 [23:17<09:23, 3.73s/it] {'loss': 1.5308, 'grad_norm': 0.0004334368971359381, 'learning_rate': 0.14392232139837835, 'epoch': 0.71} + 71%|███████ | 369/520 [23:17<09:23, 3.73s/it] 71%|███████ | 370/520 [23:20<09:18, 3.72s/it] {'loss': 1.3311, 'grad_norm': 0.0004905674495765672, 'learning_rate': 0.14216293504927446, 'epoch': 0.71} + 71%|███████ | 370/520 [23:20<09:18, 3.72s/it] 71%|███████▏ | 371/520 [23:24<09:12, 3.71s/it] {'loss': 1.3288, 'grad_norm': 0.00044259137408038764, 'learning_rate': 0.1404116240299499, 'epoch': 0.71} + 71%|███████▏ | 371/520 [23:24<09:12, 3.71s/it] 72%|███████▏ | 372/520 [23:28<09:06, 3.70s/it] {'loss': 1.6401, 'grad_norm': 0.00046405817344021233, 'learning_rate': 0.13866845638608283, 'epoch': 0.72} + 72%|███████▏ | 372/520 [23:28<09:06, 3.70s/it] 72%|███████▏ | 373/520 [23:31<09:03, 3.70s/it] {'loss': 1.5158, 'grad_norm': 0.0008797587113858141, 'learning_rate': 0.13693349984694775, 'epoch': 0.72} + 72%|███████▏ | 373/520 [23:31<09:03, 3.70s/it] 72%|███████▏ | 374/520 [23:35<09:02, 3.71s/it] {'loss': 1.433, 'grad_norm': 0.0005970235295699794, 'learning_rate': 0.13520682182278346, 'epoch': 0.72} + 72%|███████▏ | 374/520 [23:35<09:02, 3.71s/it] 72%|███████▏ | 375/520 [23:39<08:59, 3.72s/it] {'loss': 1.3293, 'grad_norm': 0.00042833856839616315, 'learning_rate': 0.13348848940217412, 'epoch': 0.72} + 72%|███████▏ | 375/520 [23:39<08:59, 3.72s/it] 72%|███████▏ | 376/520 [23:43<08:55, 3.72s/it] {'loss': 1.4544, 'grad_norm': 0.00048788514109980656, 'learning_rate': 0.13177856934944326, 'epoch': 0.72} + 72%|███████▏ | 376/520 [23:43<08:55, 3.72s/it] 72%|███████▎ | 377/520 [23:46<08:51, 3.72s/it] {'loss': 1.3849, 'grad_norm': 0.000522208457607011, 'learning_rate': 0.13007712810205843, 'epoch': 0.72} + 72%|███████▎ | 377/520 [23:46<08:51, 3.72s/it] 73%|███████▎ | 378/520 [23:50<08:48, 3.72s/it] {'loss': 1.4369, 'grad_norm': 0.0004045589341096242, 'learning_rate': 0.1283842317680511, 'epoch': 0.73} + 73%|███████▎ | 378/520 [23:50<08:48, 3.72s/it] 73%|███████▎ | 379/520 [23:54<08:43, 3.72s/it] {'loss': 1.4337, 'grad_norm': 0.0006243091562218784, 'learning_rate': 0.12669994612344704, 'epoch': 0.73} + 73%|███████▎ | 379/520 [23:54<08:43, 3.72s/it] 73%|███████▎ | 380/520 [23:57<08:38, 3.70s/it] {'loss': 1.6227, 'grad_norm': 0.0005160749553491518, 'learning_rate': 0.1250243366097112, 'epoch': 0.73} + 73%|███████▎ | 380/520 [23:57<08:38, 3.70s/it] 73%|███████▎ | 381/520 [24:01<08:36, 3.71s/it] {'loss': 1.4309, 'grad_norm': 0.0004664205586687409, 'learning_rate': 0.12335746833120538, 'epoch': 0.73} + 73%|███████▎ | 381/520 [24:01<08:36, 3.71s/it] 73%|███████▎ | 382/520 [24:05<08:32, 3.71s/it] {'loss': 1.5511, 'grad_norm': 0.0009837460049954899, 'learning_rate': 0.1216994060526577, 'epoch': 0.73} + 73%|███████▎ | 382/520 [24:05<08:32, 3.71s/it] 74%|███████▎ | 383/520 [24:09<08:27, 3.70s/it] {'loss': 1.2623, 'grad_norm': 0.0006212879307759977, 'learning_rate': 0.12005021419664687, 'epoch': 0.74} + 74%|███████▎ | 383/520 [24:09<08:27, 3.70s/it] 74%|███████▍ | 384/520 [24:12<08:26, 3.72s/it] {'loss': 1.7482, 'grad_norm': 0.0004933572738709953, 'learning_rate': 0.11840995684109928, 'epoch': 0.74} + 74%|███████▍ | 384/520 [24:12<08:26, 3.72s/it] 74%|███████▍ | 385/520 [24:16<08:26, 3.75s/it] {'loss': 1.4099, 'grad_norm': 0.0003802377930364608, 'learning_rate': 0.11677869771679862, 'epoch': 0.74} + 74%|███████▍ | 385/520 [24:16<08:26, 3.75s/it] 74%|███████▍ | 386/520 [24:20<08:26, 3.78s/it] {'loss': 1.3386, 'grad_norm': 0.000376909901022902, 'learning_rate': 0.11515650020491051, 'epoch': 0.74} + 74%|███████▍ | 386/520 [24:20<08:26, 3.78s/it] 74%|███████▍ | 387/520 [24:24<08:23, 3.78s/it] {'loss': 1.657, 'grad_norm': 0.000571769277560168, 'learning_rate': 0.1135434273345189, 'epoch': 0.74} + 74%|███████▍ | 387/520 [24:24<08:23, 3.78s/it] 75%|███████▍ | 388/520 [24:28<08:20, 3.79s/it] {'loss': 1.3107, 'grad_norm': 0.0006381614418471515, 'learning_rate': 0.11193954178017813, 'epoch': 0.75} + 75%|███████▍ | 388/520 [24:28<08:20, 3.79s/it] 75%|███████▍ | 389/520 [24:31<08:15, 3.79s/it] {'loss': 1.3778, 'grad_norm': 0.0005536717639952049, 'learning_rate': 0.11034490585947726, 'epoch': 0.75} + 75%|███████▍ | 389/520 [24:31<08:15, 3.79s/it] 75%|███████▌ | 390/520 [24:35<08:08, 3.76s/it] {'loss': 1.4356, 'grad_norm': 0.0004012425820506373, 'learning_rate': 0.10875958153061854, 'epoch': 0.75} + 75%|███████▌ | 390/520 [24:35<08:08, 3.76s/it] 75%|███████▌ | 391/520 [24:39<08:01, 3.74s/it] {'loss': 1.5256, 'grad_norm': 0.0005754836910141735, 'learning_rate': 0.10718363039001041, 'epoch': 0.75} + 75%|███████▌ | 391/520 [24:39<08:01, 3.74s/it] 75%|███████▌ | 392/520 [24:42<07:55, 3.72s/it] {'loss': 1.3236, 'grad_norm': 0.0004551993289015618, 'learning_rate': 0.10561711366987453, 'epoch': 0.75} + 75%|███████▌ | 392/520 [24:42<07:55, 3.72s/it] 76%|███████▌ | 393/520 [24:46<07:49, 3.70s/it] {'loss': 1.4248, 'grad_norm': 0.0004767492980022474, 'learning_rate': 0.10406009223586579, 'epoch': 0.76} + 76%|███████▌ | 393/520 [24:46<07:49, 3.70s/it] 76%|███████▌ | 394/520 [24:50<07:44, 3.69s/it] {'loss': 1.3936, 'grad_norm': 0.00047169489344610983, 'learning_rate': 0.10251262658470839, 'epoch': 0.76} + 76%|███████▌ | 394/520 [24:50<07:44, 3.69s/it] 76%|███████▌ | 395/520 [24:53<07:40, 3.68s/it] {'loss': 1.3479, 'grad_norm': 0.00044968092226383777, 'learning_rate': 0.10097477684184453, 'epoch': 0.76} + 76%|███████▌ | 395/520 [24:53<07:40, 3.68s/it] 76%|███████▌ | 396/520 [24:57<07:34, 3.66s/it] {'loss': 1.4441, 'grad_norm': 0.00046381794815676864, 'learning_rate': 0.09944660275909854, 'epoch': 0.76} + 76%|███████▌ | 396/520 [24:57<07:34, 3.66s/it] 76%|███████▋ | 397/520 [25:01<07:30, 3.67s/it] {'loss': 1.4094, 'grad_norm': 0.0006571326349834386, 'learning_rate': 0.09792816371235576, 'epoch': 0.76} + 76%|███████▋ | 397/520 [25:01<07:30, 3.67s/it] 77%|███████▋ | 398/520 [25:04<07:29, 3.69s/it] {'loss': 1.387, 'grad_norm': 0.0004414139262237814, 'learning_rate': 0.09641951869925457, 'epoch': 0.77} + 77%|███████▋ | 398/520 [25:04<07:29, 3.69s/it] 77%|███████▋ | 399/520 [25:08<07:33, 3.74s/it] {'loss': 1.492, 'grad_norm': 0.0006255544478600451, 'learning_rate': 0.09492072633689508, 'epoch': 0.77} + 77%|███████▋ | 399/520 [33:40<07:33, 3.74s/it] 77%|███████▋ | 400/520 [33:44<5:14:48, 157.40s/it] {'loss': 1.5516, 'grad_norm': 0.00047504196717009634, 'learning_rate': 0.09343184485956085, 'epoch': 0.77} + 77%|███████▋ | 400/520 [33:44<5:14:48, 157.40s/it] 77%|███████▋ | 401/520 [33:48<3:40:50, 111.35s/it] {'loss': 1.2095, 'grad_norm': 0.0005508953910039906, 'learning_rate': 0.09195293211645661, 'epoch': 0.77} + 77%|███████▋ | 401/520 [33:48<3:40:50, 111.35s/it] 77%|███████▋ | 402/520 [33:52<2:35:33, 79.10s/it] {'loss': 1.344, 'grad_norm': 0.0007300584654390279, 'learning_rate': 0.09048404556946063, 'epoch': 0.77} + 77%|███████▋ | 402/520 [33:52<2:35:33, 79.10s/it] 78%|███████▊ | 403/520 [33:56<1:50:14, 56.53s/it] {'loss': 1.4035, 'grad_norm': 0.0006467560370387071, 'learning_rate': 0.08902524229089204, 'epoch': 0.78} + 78%|███████▊ | 403/520 [33:56<1:50:14, 56.53s/it] 78%|███████▊ | 404/520 [34:00<1:18:45, 40.73s/it] {'loss': 1.2788, 'grad_norm': 0.0005711970619089951, 'learning_rate': 0.08757657896129298, 'epoch': 0.78} + 78%|███████▊ | 404/520 [34:00<1:18:45, 40.73s/it] 78%|███████▊ | 405/520 [34:04<56:52, 29.67s/it] {'loss': 1.4877, 'grad_norm': 0.0005461426104572795, 'learning_rate': 0.08613811186722706, 'epoch': 0.78} + 78%|███████▊ | 405/520 [34:04<56:52, 29.67s/it] 78%|███████▊ | 406/520 [34:07<41:36, 21.90s/it] {'loss': 1.4396, 'grad_norm': 0.0005188125390726163, 'learning_rate': 0.08470989689909139, 'epoch': 0.78} + 78%|███████▊ | 406/520 [34:07<41:36, 21.90s/it] 78%|███████▊ | 407/520 [34:11<30:55, 16.42s/it] {'loss': 1.493, 'grad_norm': 0.0004896600659209933, 'learning_rate': 0.08329198954894622, 'epoch': 0.78} + 78%|███████▊ | 407/520 [34:11<30:55, 16.42s/it] 78%|███████▊ | 408/520 [34:15<23:28, 12.58s/it] {'loss': 1.3765, 'grad_norm': 0.0004631414605342586, 'learning_rate': 0.08188444490835772, 'epoch': 0.78} + 78%|███████▊ | 408/520 [34:15<23:28, 12.58s/it] 79%|███████▊ | 409/520 [34:18<18:18, 9.90s/it] {'loss': 1.5139, 'grad_norm': 0.0009276698878223071, 'learning_rate': 0.08048731766625802, 'epoch': 0.79} + 79%|███████▊ | 409/520 [34:18<18:18, 9.90s/it] 79%|███████▉ | 410/520 [34:22<14:43, 8.03s/it] {'loss': 1.2101, 'grad_norm': 0.0006883500272071328, 'learning_rate': 0.0791006621068204, 'epoch': 0.79} + 79%|███████▉ | 410/520 [34:22<14:43, 8.03s/it] 79%|███████▉ | 411/520 [34:26<12:11, 6.71s/it] {'loss': 1.4765, 'grad_norm': 0.00044245656001224603, 'learning_rate': 0.07772453210734984, 'epoch': 0.79} + 79%|███████▉ | 411/520 [34:26<12:11, 6.71s/it] 79%|███████▉ | 412/520 [34:29<10:26, 5.80s/it] {'loss': 1.3939, 'grad_norm': 0.0006231033592381162, 'learning_rate': 0.07635898113618957, 'epoch': 0.79} + 79%|███████▉ | 412/520 [34:29<10:26, 5.80s/it] 79%|███████▉ | 413/520 [34:33<09:12, 5.16s/it] {'loss': 1.5628, 'grad_norm': 0.0004823621165461344, 'learning_rate': 0.07500406225064427, 'epoch': 0.79} + 79%|███████▉ | 413/520 [34:33<09:12, 5.16s/it] 80%|███████▉ | 414/520 [34:37<08:19, 4.71s/it] {'loss': 1.3089, 'grad_norm': 0.0004974326188841984, 'learning_rate': 0.07365982809491764, 'epoch': 0.8} + 80%|███████▉ | 414/520 [34:37<08:19, 4.71s/it] 80%|███████▉ | 415/520 [34:40<07:41, 4.40s/it] {'loss': 1.3622, 'grad_norm': 0.0005468912124976609, 'learning_rate': 0.07232633089806773, 'epoch': 0.8} + 80%|███████▉ | 415/520 [34:40<07:41, 4.40s/it] 80%|████████ | 416/520 [34:44<07:14, 4.18s/it] {'loss': 1.2678, 'grad_norm': 0.0006797684506496729, 'learning_rate': 0.07100362247197724, 'epoch': 0.8} + 80%|████████ | 416/520 [34:44<07:14, 4.18s/it] 80%|████████ | 417/520 [34:48<06:54, 4.02s/it] {'loss': 1.4469, 'grad_norm': 0.0004861842297662198, 'learning_rate': 0.06969175420934025, 'epoch': 0.8} + 80%|████████ | 417/520 [34:48<06:54, 4.02s/it] 80%|████████ | 418/520 [34:51<06:38, 3.91s/it] {'loss': 1.4403, 'grad_norm': 0.00046428042893602154, 'learning_rate': 0.06839077708166608, 'epoch': 0.8} + 80%|████████ | 418/520 [34:51<06:38, 3.91s/it] 81%|████████ | 419/520 [34:55<06:31, 3.88s/it] {'loss': 1.4197, 'grad_norm': 0.0006260586757207416, 'learning_rate': 0.06710074163729816, 'epoch': 0.81} + 81%|████████ | 419/520 [34:55<06:31, 3.88s/it] 81%|████████ | 420/520 [34:59<06:23, 3.83s/it] {'loss': 1.2951, 'grad_norm': 0.0006004416741162088, 'learning_rate': 0.06582169799945022, 'epoch': 0.81} + 81%|████████ | 420/520 [34:59<06:23, 3.83s/it] 81%|████████ | 421/520 [35:02<06:15, 3.79s/it] {'loss': 1.224, 'grad_norm': 0.0009330757152992532, 'learning_rate': 0.06455369586425894, 'epoch': 0.81} + 81%|████████ | 421/520 [35:02<06:15, 3.79s/it] 81%|████████ | 422/520 [35:06<06:07, 3.75s/it] {'loss': 1.3573, 'grad_norm': 0.0004497605902411217, 'learning_rate': 0.06329678449885283, 'epoch': 0.81} + 81%|████████ | 422/520 [35:06<06:07, 3.75s/it] 81%|████████▏ | 423/520 [35:10<06:02, 3.73s/it] {'loss': 1.3628, 'grad_norm': 0.0008392555508169044, 'learning_rate': 0.062051012739438326, 'epoch': 0.81} + 81%|████████▏ | 423/520 [35:10<06:02, 3.73s/it] 82%|████████▏ | 424/520 [35:14<05:57, 3.72s/it] {'loss': 1.644, 'grad_norm': 0.0008294862042125481, 'learning_rate': 0.06081642898940186, 'epoch': 0.82} + 82%|████████▏ | 424/520 [35:14<05:57, 3.72s/it] 82%|████████▏ | 425/520 [35:17<05:51, 3.70s/it] {'loss': 1.3464, 'grad_norm': 0.0006309071541981985, 'learning_rate': 0.05959308121742937, 'epoch': 0.82} + 82%|████████▏ | 425/520 [35:17<05:51, 3.70s/it] 82%|████████▏ | 426/520 [35:21<05:46, 3.68s/it] {'loss': 1.4113, 'grad_norm': 0.0009252268831703649, 'learning_rate': 0.058381016955642906, 'epoch': 0.82} + 82%|████████▏ | 426/520 [35:21<05:46, 3.68s/it] 82%|████████▏ | 427/520 [35:25<05:42, 3.69s/it] {'loss': 1.2847, 'grad_norm': 0.0005057545596016187, 'learning_rate': 0.057180283297753084, 'epoch': 0.82} + 82%|████████▏ | 427/520 [35:25<05:42, 3.69s/it] 82%|████████▏ | 428/520 [35:28<05:38, 3.68s/it] {'loss': 1.2524, 'grad_norm': 0.0004853205705312531, 'learning_rate': 0.05599092689723, 'epoch': 0.82} + 82%|████████▏ | 428/520 [35:28<05:38, 3.68s/it] 82%|████████▎ | 429/520 [35:32<05:35, 3.69s/it] {'loss': 1.3839, 'grad_norm': 0.00042236308463973455, 'learning_rate': 0.054812993965490074, 'epoch': 0.82} + 82%|████████▎ | 429/520 [35:32<05:35, 3.69s/it]Token indices sequence length is longer than the specified maximum sequence length for this model (2076 > 2048). Running this sequence through the model will result in indexing errors + 83%|████████▎ | 430/520 [35:36<05:31, 3.68s/it] {'loss': 1.3651, 'grad_norm': 0.0004902507044142381, 'learning_rate': 0.05364653027010055, 'epoch': 0.83} + 83%|████████▎ | 430/520 [35:36<05:31, 3.68s/it] 83%|████████▎ | 431/520 [35:39<05:26, 3.67s/it] {'loss': 1.5239, 'grad_norm': 0.0005433268936914266, 'learning_rate': 0.052491581133001806, 'epoch': 0.83} + 83%|████████▎ | 431/520 [35:39<05:26, 3.67s/it] 83%|████████▎ | 432/520 [35:43<05:22, 3.66s/it] {'loss': 1.2768, 'grad_norm': 0.0005608453428077928, 'learning_rate': 0.051348191428745533, 'epoch': 0.83} + 83%|████████▎ | 432/520 [35:43<05:22, 3.66s/it] 83%|████████▎ | 433/520 [35:46<05:17, 3.65s/it] {'loss': 1.4338, 'grad_norm': 0.0004938324809790384, 'learning_rate': 0.05021640558275203, 'epoch': 0.83} + 83%|████████▎ | 433/520 [35:46<05:17, 3.65s/it] 83%|████████▎ | 434/520 [35:50<05:14, 3.65s/it] {'loss': 1.1515, 'grad_norm': 0.000842368887861486, 'learning_rate': 0.04909626756958339, 'epoch': 0.83} + 83%|████████▎ | 434/520 [35:50<05:14, 3.65s/it] 84%|████████▎ | 435/520 [35:54<05:10, 3.65s/it] {'loss': 1.473, 'grad_norm': 0.00046239028479031755, 'learning_rate': 0.047987820911235435, 'epoch': 0.84} + 84%|████████▎ | 435/520 [35:54<05:10, 3.65s/it] 84%|████████▍ | 436/520 [35:57<05:06, 3.64s/it] {'loss': 1.244, 'grad_norm': 0.0004919410490109671, 'learning_rate': 0.04689110867544645, 'epoch': 0.84} + 84%|████████▍ | 436/520 [35:57<05:06, 3.64s/it] 84%|████████▍ | 437/520 [36:01<05:03, 3.65s/it] {'loss': 1.5006, 'grad_norm': 0.0007613165343860034, 'learning_rate': 0.045806173474023756, 'epoch': 0.84} + 84%|████████▍ | 437/520 [36:01<05:03, 3.65s/it] 84%|████████▍ | 438/520 [36:05<04:59, 3.66s/it] {'loss': 1.2757, 'grad_norm': 0.00046169689730777056, 'learning_rate': 0.044733057461188136, 'epoch': 0.84} + 84%|████████▍ | 438/520 [36:05<04:59, 3.66s/it] 84%|████████▍ | 439/520 [36:08<04:55, 3.65s/it] {'loss': 1.4701, 'grad_norm': 0.0005164481217123496, 'learning_rate': 0.04367180233193621, 'epoch': 0.84} + 84%|████████▍ | 439/520 [36:08<04:55, 3.65s/it] 85%|████████▍ | 440/520 [36:12<04:52, 3.66s/it] {'loss': 1.3489, 'grad_norm': 0.0007602406093843402, 'learning_rate': 0.04262244932041997, 'epoch': 0.85} + 85%|████████▍ | 440/520 [36:12<04:52, 3.66s/it] 85%|████████▍ | 441/520 [36:16<04:49, 3.67s/it] {'loss': 1.5266, 'grad_norm': 0.0006322485916927767, 'learning_rate': 0.04158503919834516, 'epoch': 0.85} + 85%|████████▍ | 441/520 [36:16<04:49, 3.67s/it] 85%|████████▌ | 442/520 [36:19<04:47, 3.68s/it] {'loss': 1.3943, 'grad_norm': 0.0007271565606816162, 'learning_rate': 0.040559612273386614, 'epoch': 0.85} + 85%|████████▌ | 442/520 [36:19<04:47, 3.68s/it] 85%|████████▌ | 443/520 [36:23<04:43, 3.69s/it] {'loss': 1.4079, 'grad_norm': 0.000447313276883113, 'learning_rate': 0.0395462083876224, 'epoch': 0.85} + 85%|████████▌ | 443/520 [36:23<04:43, 3.69s/it] 85%|████████▌ | 444/520 [36:27<04:39, 3.68s/it] {'loss': 1.3816, 'grad_norm': 0.0005460987950060438, 'learning_rate': 0.038544866915986006, 'epoch': 0.85} + 85%|████████▌ | 444/520 [36:27<04:39, 3.68s/it] 86%|████████▌ | 445/520 [36:30<04:35, 3.67s/it] {'loss': 1.2899, 'grad_norm': 0.0005114542956912539, 'learning_rate': 0.03755562676473603, 'epoch': 0.86} + 86%|████████▌ | 445/520 [36:30<04:35, 3.67s/it] 86%|████████▌ | 446/520 [36:34<04:31, 3.67s/it] {'loss': 1.5777, 'grad_norm': 0.0004899011296453307, 'learning_rate': 0.03657852636994467, 'epoch': 0.86} + 86%|████████▌ | 446/520 [36:34<04:31, 3.67s/it] 86%|████████▌ | 447/520 [36:38<04:27, 3.67s/it] {'loss': 1.3904, 'grad_norm': 0.0006066323604686835, 'learning_rate': 0.03561360369600458, 'epoch': 0.86} + 86%|████████▌ | 447/520 [36:38<04:27, 3.67s/it] 86%|████████▌ | 448/520 [36:41<04:23, 3.66s/it] {'loss': 1.3613, 'grad_norm': 0.0006381269756532971, 'learning_rate': 0.03466089623415333, 'epoch': 0.86} + 86%|████████▌ | 448/520 [36:41<04:23, 3.66s/it] 86%|████████▋ | 449/520 [36:45<04:19, 3.65s/it] {'loss': 1.5588, 'grad_norm': 0.0007413933638629671, 'learning_rate': 0.03372044100101723, 'epoch': 0.86} + 86%|████████▋ | 449/520 [36:45<04:19, 3.65s/it] 87%|████████▋ | 450/520 [36:49<04:15, 3.65s/it] {'loss': 1.4181, 'grad_norm': 0.000690966554542421, 'learning_rate': 0.03279227453717252, 'epoch': 0.87} + 87%|████████▋ | 450/520 [36:49<04:15, 3.65s/it] 87%|████████▋ | 451/520 [36:52<04:12, 3.66s/it] {'loss': 1.4154, 'grad_norm': 0.000577763423165936, 'learning_rate': 0.03187643290572617, 'epoch': 0.87} + 87%|████████▋ | 451/520 [36:52<04:12, 3.66s/it] 87%|████████▋ | 452/520 [36:56<04:08, 3.65s/it] {'loss': 1.5938, 'grad_norm': 0.0005300111480442249, 'learning_rate': 0.0309729516909144, 'epoch': 0.87} + 87%|████████▋ | 452/520 [36:56<04:08, 3.65s/it] 87%|████████▋ | 453/520 [37:00<04:05, 3.66s/it] {'loss': 1.5638, 'grad_norm': 0.0009325414777686971, 'learning_rate': 0.030081865996719945, 'epoch': 0.87} + 87%|████████▋ | 453/520 [37:00<04:05, 3.66s/it] 87%|████████▋ | 454/520 [37:03<04:03, 3.69s/it] {'loss': 1.3064, 'grad_norm': 0.0005106513150346246, 'learning_rate': 0.02920321044550833, 'epoch': 0.87} + 87%|████████▋ | 454/520 [37:03<04:03, 3.69s/it] 88%|████████▊ | 455/520 [37:07<04:01, 3.71s/it] {'loss': 1.4645, 'grad_norm': 0.000545209608599765, 'learning_rate': 0.028337019176682767, 'epoch': 0.88} + 88%|████████▊ | 455/520 [37:07<04:01, 3.71s/it] 88%|████████▊ | 456/520 [37:11<03:58, 3.73s/it] {'loss': 1.3727, 'grad_norm': 0.0005141848923108656, 'learning_rate': 0.02748332584535729, 'epoch': 0.88} + 88%|████████▊ | 456/520 [37:11<03:58, 3.73s/it] 88%|████████▊ | 457/520 [37:15<03:55, 3.74s/it] {'loss': 1.6159, 'grad_norm': 0.0005035076433900273, 'learning_rate': 0.02664216362104964, 'epoch': 0.88} + 88%|████████▊ | 457/520 [37:15<03:55, 3.74s/it] 88%|████████▊ | 458/520 [37:19<03:52, 3.75s/it] {'loss': 1.5385, 'grad_norm': 0.000616959860088459, 'learning_rate': 0.02581356518639197, 'epoch': 0.88} + 88%|████████▊ | 458/520 [37:19<03:52, 3.75s/it] 88%|████████▊ | 459/520 [37:22<03:48, 3.74s/it] {'loss': 1.4411, 'grad_norm': 0.00044459442386341375, 'learning_rate': 0.024997562735861255, 'epoch': 0.88} + 88%|████████▊ | 459/520 [37:22<03:48, 3.74s/it] 88%|████████▊ | 460/520 [37:26<03:43, 3.72s/it] {'loss': 1.3058, 'grad_norm': 0.0004554402724094227, 'learning_rate': 0.02419418797452855, 'epoch': 0.88} + 88%|████████▊ | 460/520 [37:26<03:43, 3.72s/it] 89%|████████▊ | 461/520 [37:30<03:39, 3.72s/it] {'loss': 1.6793, 'grad_norm': 0.0005801647237270231, 'learning_rate': 0.023403472116826723, 'epoch': 0.89} + 89%|████████▊ | 461/520 [37:30<03:39, 3.72s/it] 89%|████████▉ | 462/520 [37:33<03:34, 3.71s/it] {'loss': 1.6422, 'grad_norm': 0.0006342016035687171, 'learning_rate': 0.022625445885338102, 'epoch': 0.89} + 89%|████████▉ | 462/520 [37:33<03:34, 3.71s/it] 89%|████████▉ | 463/520 [37:37<03:30, 3.70s/it] {'loss': 1.2739, 'grad_norm': 0.0008082338936726696, 'learning_rate': 0.021860139509600317, 'epoch': 0.89} + 89%|████████▉ | 463/520 [37:37<03:30, 3.70s/it] 89%|████████▉ | 464/520 [37:41<03:26, 3.70s/it] {'loss': 1.4423, 'grad_norm': 0.0005263997372310292, 'learning_rate': 0.021107582724932086, 'epoch': 0.89} + 89%|████████▉ | 464/520 [37:41<03:26, 3.70s/it] 89%|████████▉ | 465/520 [37:44<03:23, 3.70s/it] {'loss': 1.5404, 'grad_norm': 0.0005185281754449685, 'learning_rate': 0.020367804771277784, 'epoch': 0.89} + 89%|████████▉ | 465/520 [37:44<03:23, 3.70s/it] 90%|████████▉ | 466/520 [37:48<03:19, 3.70s/it] {'loss': 1.4318, 'grad_norm': 0.0006033906321761081, 'learning_rate': 0.019640834392071347, 'epoch': 0.9} + 90%|████████▉ | 466/520 [37:48<03:19, 3.70s/it] 90%|████████▉ | 467/520 [37:52<03:15, 3.69s/it] {'loss': 1.5209, 'grad_norm': 0.0004644889472600972, 'learning_rate': 0.01892669983311939, 'epoch': 0.9} + 90%|████████▉ | 467/520 [37:52<03:15, 3.69s/it] 90%|█████████ | 468/520 [37:55<03:12, 3.69s/it] {'loss': 1.4146, 'grad_norm': 0.0005847219006089949, 'learning_rate': 0.018225428841503903, 'epoch': 0.9} + 90%|█████████ | 468/520 [37:55<03:12, 3.69s/it] 90%|█████████ | 469/520 [37:59<03:08, 3.70s/it] {'loss': 1.4678, 'grad_norm': 0.000544506956605576, 'learning_rate': 0.0175370486645039, 'epoch': 0.9} + 90%|█████████ | 469/520 [37:59<03:08, 3.70s/it] 90%|█████████ | 470/520 [38:03<03:04, 3.70s/it] {'loss': 1.3085, 'grad_norm': 0.0004756594741380657, 'learning_rate': 0.016861586048537172, 'epoch': 0.9} + 90%|█████████ | 470/520 [38:03<03:04, 3.70s/it] 91%|█████████ | 471/520 [38:07<03:00, 3.68s/it] {'loss': 1.3586, 'grad_norm': 0.0004842919481660303, 'learning_rate': 0.01619906723812061, 'epoch': 0.91} + 91%|█████████ | 471/520 [38:07<03:00, 3.68s/it] 91%|█████████ | 472/520 [38:10<02:57, 3.69s/it] {'loss': 1.3136, 'grad_norm': 0.0006404605338396665, 'learning_rate': 0.015549517974850723, 'epoch': 0.91} + 91%|█████████ | 472/520 [38:10<02:57, 3.69s/it] 91%|█████████ | 473/520 [38:14<02:53, 3.69s/it] {'loss': 1.3939, 'grad_norm': 0.00048352822831155103, 'learning_rate': 0.014912963496403675, 'epoch': 0.91} + 91%|█████████ | 473/520 [38:14<02:53, 3.69s/it] 91%|█████████ | 474/520 [38:18<02:49, 3.69s/it] {'loss': 1.5447, 'grad_norm': 0.0005198724526149538, 'learning_rate': 0.014289428535554281, 'epoch': 0.91} + 91%|█████████ | 474/520 [38:18<02:49, 3.69s/it] 91%|█████████▏| 475/520 [38:21<02:45, 3.68s/it] {'loss': 1.4382, 'grad_norm': 0.0005101509352067302, 'learning_rate': 0.013678937319215177, 'epoch': 0.91} + 91%|█████████▏| 475/520 [38:21<02:45, 3.68s/it] 92%|█████████▏| 476/520 [38:25<02:41, 3.68s/it] {'loss': 1.3576, 'grad_norm': 0.0005313935690868042, 'learning_rate': 0.01308151356749579, 'epoch': 0.92} + 92%|█████████▏| 476/520 [38:25<02:41, 3.68s/it] 92%|█████████▏| 477/520 [38:29<02:38, 3.68s/it] {'loss': 1.3509, 'grad_norm': 0.0004773330793836459, 'learning_rate': 0.012497180492780318, 'epoch': 0.92} + 92%|█████████▏| 477/520 [38:29<02:38, 3.68s/it] 92%|█████████▏| 478/520 [38:32<02:34, 3.68s/it] {'loss': 1.2953, 'grad_norm': 0.0006159375683860342, 'learning_rate': 0.01192596079882613, 'epoch': 0.92} + 92%|█████████▏| 478/520 [38:32<02:34, 3.68s/it] 92%|█████████▏| 479/520 [38:36<02:30, 3.68s/it] {'loss': 1.5649, 'grad_norm': 0.0005536815751808903, 'learning_rate': 0.01136787667988136, 'epoch': 0.92} + 92%|█████████▏| 479/520 [38:36<02:30, 3.68s/it] 92%|█████████▏| 480/520 [38:40<02:27, 3.69s/it] {'loss': 1.5308, 'grad_norm': 0.00042313056370172323, 'learning_rate': 0.010822949819822751, 'epoch': 0.92} + 92%|█████████▏| 480/520 [38:40<02:27, 3.69s/it] 92%|█████████▎| 481/520 [38:43<02:24, 3.70s/it] {'loss': 1.5724, 'grad_norm': 0.000480328303506591, 'learning_rate': 0.010291201391313164, 'epoch': 0.93} + 92%|█████████▎| 481/520 [38:43<02:24, 3.70s/it] 93%|█████████▎| 482/520 [38:47<02:20, 3.70s/it] {'loss': 1.575, 'grad_norm': 0.0005344482379991292, 'learning_rate': 0.009772652054978924, 'epoch': 0.93} + 93%|█████████▎| 482/520 [38:47<02:20, 3.70s/it] 93%|█████████▎| 483/520 [38:51<02:16, 3.69s/it] {'loss': 1.4045, 'grad_norm': 0.0005514668379636528, 'learning_rate': 0.009267321958606828, 'epoch': 0.93} + 93%|█████████▎| 483/520 [38:51<02:16, 3.69s/it] 93%|█████████▎| 484/520 [38:54<02:12, 3.69s/it] {'loss': 1.3915, 'grad_norm': 0.00047366516324147834, 'learning_rate': 0.008775230736361732, 'epoch': 0.93} + 93%|█████████▎| 484/520 [38:54<02:12, 3.69s/it] 93%|█████████▎| 485/520 [38:58<02:08, 3.68s/it] {'loss': 1.3417, 'grad_norm': 0.00046000957870219927, 'learning_rate': 0.008296397508023322, 'epoch': 0.93} + 93%|█████████▎| 485/520 [38:58<02:08, 3.68s/it] 93%|█████████▎| 486/520 [39:02<02:04, 3.67s/it] {'loss': 1.4695, 'grad_norm': 0.0006503558583941112, 'learning_rate': 0.00783084087824341, 'epoch': 0.93} + 93%|█████████▎| 486/520 [39:02<02:04, 3.67s/it] 94%|█████████▎| 487/520 [39:05<02:01, 3.67s/it] {'loss': 1.3077, 'grad_norm': 0.0004126979634007093, 'learning_rate': 0.007378578935823071, 'epoch': 0.94} + 94%|█████████▎| 487/520 [39:05<02:01, 3.67s/it] 94%|█████████▍| 488/520 [39:09<01:57, 3.67s/it] {'loss': 1.2494, 'grad_norm': 0.0005591801210997227, 'learning_rate': 0.00693962925300966, 'epoch': 0.94} + 94%|█████████▍| 488/520 [39:09<01:57, 3.67s/it] 94%|█████████▍| 489/520 [39:13<01:53, 3.67s/it] {'loss': 1.5334, 'grad_norm': 0.0004016860381781927, 'learning_rate': 0.006514008884814321, 'epoch': 0.94} + 94%|█████████▍| 489/520 [39:13<01:53, 3.67s/it] 94%|█████████▍| 490/520 [39:16<01:50, 3.67s/it] {'loss': 1.392, 'grad_norm': 0.00047355322093949135, 'learning_rate': 0.006101734368349104, 'epoch': 0.94} + 94%|█████████▍| 490/520 [39:16<01:50, 3.67s/it] 94%|█████████▍| 491/520 [39:20<01:46, 3.67s/it] {'loss': 1.3343, 'grad_norm': 0.0008261250321301893, 'learning_rate': 0.005702821722184536, 'epoch': 0.94} + 94%|█████████▍| 491/520 [39:20<01:46, 3.67s/it] 95%|█████████▍| 492/520 [39:24<01:42, 3.67s/it] {'loss': 1.477, 'grad_norm': 0.0006631493644162045, 'learning_rate': 0.005317286445727192, 'epoch': 0.95} + 95%|█████████▍| 492/520 [39:24<01:42, 3.67s/it] 95%|█████████▍| 493/520 [39:27<01:39, 3.67s/it] {'loss': 1.6584, 'grad_norm': 0.0008069019844161083, 'learning_rate': 0.00494514351861744, 'epoch': 0.95} + 95%|█████████▍| 493/520 [39:27<01:39, 3.67s/it] 95%|█████████▌| 494/520 [39:31<01:35, 3.68s/it] {'loss': 1.3847, 'grad_norm': 0.0004151800217984771, 'learning_rate': 0.004586407400147618, 'epoch': 0.95} + 95%|█████████▌| 494/520 [39:31<01:35, 3.68s/it] 95%|█████████▌| 495/520 [39:35<01:31, 3.67s/it] {'loss': 1.3584, 'grad_norm': 0.0005804193908891454, 'learning_rate': 0.00424109202869985, 'epoch': 0.95} + 95%|█████████▌| 495/520 [39:35<01:31, 3.67s/it] 95%|█████████▌| 496/520 [39:39<01:28, 3.68s/it] {'loss': 1.2907, 'grad_norm': 0.0006093707123534557, 'learning_rate': 0.003909210821205017, 'epoch': 0.95} + 95%|█████████▌| 496/520 [39:39<01:28, 3.68s/it] 96%|█████████▌| 497/520 [39:42<01:24, 3.68s/it] {'loss': 1.4869, 'grad_norm': 0.0006563481291270557, 'learning_rate': 0.0035907766726209042, 'epoch': 0.96} + 96%|█████████▌| 497/520 [39:42<01:24, 3.68s/it] 96%|█████████▌| 498/520 [39:46<01:20, 3.68s/it] {'loss': 1.3667, 'grad_norm': 0.0005907159943919531, 'learning_rate': 0.0032858019554315165, 'epoch': 0.96} + 96%|█████████▌| 498/520 [39:46<01:20, 3.68s/it] 96%|█████████▌| 499/520 [39:50<01:17, 3.69s/it] {'loss': 1.6277, 'grad_norm': 0.0004985755558740698, 'learning_rate': 0.002994298519166366, 'epoch': 0.96} + 96%|█████████▌| 499/520 [39:50<01:17, 3.69s/it] 96%|█████████▌| 500/520 [39:53<01:13, 3.68s/it] {'loss': 1.499, 'grad_norm': 0.0007359842782696027, 'learning_rate': 0.0027162776899397776, 'epoch': 0.96} + 96%|█████████▌| 500/520 [39:53<01:13, 3.68s/it] 96%|█████████▋| 501/520 [39:57<01:09, 3.68s/it] {'loss': 1.5442, 'grad_norm': 0.0009097259888447756, 'learning_rate': 0.0024517502700111323, 'epoch': 0.96} + 96%|█████████▋| 501/520 [39:57<01:09, 3.68s/it] 97%|█████████▋| 502/520 [40:01<01:06, 3.69s/it] {'loss': 1.3884, 'grad_norm': 0.0004176022849866897, 'learning_rate': 0.0022007265373650883, 'epoch': 0.97} + 97%|█████████▋| 502/520 [40:01<01:06, 3.69s/it] 97%|█████████▋| 503/520 [40:04<01:02, 3.69s/it] {'loss': 1.4984, 'grad_norm': 0.0005116319725163506, 'learning_rate': 0.0019632162453120825, 'epoch': 0.97} + 97%|█████████▋| 503/520 [40:04<01:02, 3.69s/it] 97%|█████████▋| 504/520 [40:08<00:58, 3.68s/it] {'loss': 1.3932, 'grad_norm': 0.0006813366714817493, 'learning_rate': 0.0017392286221095065, 'epoch': 0.97} + 97%|█████████▋| 504/520 [40:08<00:58, 3.68s/it] 97%|█████████▋| 505/520 [40:12<00:55, 3.68s/it] {'loss': 1.4463, 'grad_norm': 0.0005585886240907945, 'learning_rate': 0.0015287723706031652, 'epoch': 0.97} + 97%|█████████▋| 505/520 [40:12<00:55, 3.68s/it] 97%|█████████▋| 506/520 [40:15<00:51, 3.67s/it] {'loss': 1.3406, 'grad_norm': 0.0004737431336297393, 'learning_rate': 0.001331855667889059, 'epoch': 0.97} + 97%|█████████▋| 506/520 [40:15<00:51, 3.67s/it] 98%|█████████▊| 507/520 [40:19<00:47, 3.67s/it] {'loss': 1.6681, 'grad_norm': 0.00045585192852172875, 'learning_rate': 0.001148486164995721, 'epoch': 0.97} + 98%|█████████▊| 507/520 [40:19<00:47, 3.67s/it] 98%|█████████▊| 508/520 [40:23<00:44, 3.67s/it] {'loss': 1.4676, 'grad_norm': 0.0006140963029914364, 'learning_rate': 0.0009786709865869548, 'epoch': 0.98} + 98%|█████████▊| 508/520 [40:23<00:44, 3.67s/it] 98%|█████████▊| 509/520 [40:26<00:40, 3.67s/it] {'loss': 1.4364, 'grad_norm': 0.0005903664576302505, 'learning_rate': 0.000822416730684894, 'epoch': 0.98} + 98%|█████████▊| 509/520 [40:26<00:40, 3.67s/it] 98%|█████████▊| 510/520 [40:30<00:36, 3.68s/it] {'loss': 1.391, 'grad_norm': 0.0004997823866364712, 'learning_rate': 0.0006797294684138533, 'epoch': 0.98} + 98%|█████████▊| 510/520 [40:30<00:36, 3.68s/it] 98%|█████████▊| 511/520 [40:34<00:33, 3.68s/it] {'loss': 1.3516, 'grad_norm': 0.0007581413063475167, 'learning_rate': 0.0005506147437641883, 'epoch': 0.98} + 98%|█████████▊| 511/520 [40:34<00:33, 3.68s/it] 98%|█████████▊| 512/520 [40:37<00:29, 3.68s/it] {'loss': 1.2351, 'grad_norm': 0.0005150032340413121, 'learning_rate': 0.0004350775733771794, 'epoch': 0.98} + 98%|█████████▊| 512/520 [40:37<00:29, 3.68s/it] 99%|█████████▊| 513/520 [40:41<00:25, 3.68s/it] {'loss': 1.4503, 'grad_norm': 0.0005645563741492992, 'learning_rate': 0.0003331224463497706, 'epoch': 0.99} + 99%|█████████▊| 513/520 [40:41<00:25, 3.68s/it] 99%|█████████▉| 514/520 [40:45<00:22, 3.68s/it] {'loss': 1.4263, 'grad_norm': 0.0004426245052607797, 'learning_rate': 0.0002447533240604871, 'epoch': 0.99} + 99%|█████████▉| 514/520 [40:45<00:22, 3.68s/it] 99%|█████████▉| 515/520 [40:48<00:18, 3.66s/it] {'loss': 1.507, 'grad_norm': 0.000622910264981831, 'learning_rate': 0.0001699736400153251, 'epoch': 0.99} + 99%|█████████▉| 515/520 [40:48<00:18, 3.66s/it] 99%|█████████▉| 516/520 [40:52<00:14, 3.65s/it] {'loss': 1.3567, 'grad_norm': 0.0006442848598972872, 'learning_rate': 0.00010878629971431407, 'epoch': 0.99} + 99%|█████████▉| 516/520 [40:52<00:14, 3.65s/it] 99%|█████████▉| 517/520 [40:56<00:10, 3.63s/it] {'loss': 1.5864, 'grad_norm': 0.0004466872193015139, 'learning_rate': 6.11936805387514e-05, 'epoch': 0.99} + 99%|█████████▉| 517/520 [40:56<00:10, 3.63s/it] 100%|█████████▉| 518/520 [40:59<00:07, 3.63s/it] {'loss': 1.3966, 'grad_norm': 0.0004804797079679704, 'learning_rate': 2.7197631658798513e-05, 'epoch': 1.0} + 100%|█████████▉| 518/520 [40:59<00:07, 3.63s/it] 100%|█████████▉| 519/520 [41:03<00:03, 3.63s/it] {'loss': 1.5376, 'grad_norm': 0.000821605620310897, 'learning_rate': 6.799473961632829e-06, 'epoch': 1.0} + 100%|█████████▉| 519/520 [41:03<00:03, 3.63s/it] 100%|██████████| 520/520 [41:07<00:00, 3.88s/it] {'loss': 1.6136, 'grad_norm': 0.0005847375722189914, 'learning_rate': 0.0, 'epoch': 1.0} + 100%|██████████| 520/520 [41:07<00:00, 3.88s/it] {'train_runtime': 2467.8512, 'train_samples_per_second': 26.958, 'train_steps_per_second': 0.211, 'train_loss': 1.5078876571013378, 'epoch': 1.0} + 100%|██████████| 520/520 [41:07<00:00, 3.88s/it] 100%|██████████| 520/520 [41:07<00:00, 4.75s/it] +[2025-10-12 08:14:09,255] [INFO] [launch.py:348:main] Process 514122 exits successfully. +[2025-10-12 08:14:09,256] [INFO] [launch.py:348:main] Process 514119 exits successfully. +[2025-10-12 08:14:10,257] [INFO] [launch.py:348:main] Process 514121 exits successfully. +[2025-10-12 08:14:10,258] [INFO] [launch.py:348:main] Process 514118 exits successfully. +[2025-10-12 08:14:10,258] [INFO] [launch.py:348:main] Process 514116 exits successfully. +[2025-10-12 08:14:10,258] [INFO] [launch.py:348:main] Process 514120 exits successfully. +[2025-10-12 08:14:10,259] [INFO] [launch.py:348:main] Process 514117 exits successfully. +[2025-10-12 08:14:14,263] [INFO] [launch.py:348:main] Process 514115 exits successfully. +==== EXPERIMENT COMPLETED: qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_7e-1_connector-3.0_0.5_7e-1_ablation ==== +Log File: qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_7e-1_connector-3.0_0.5_7e-1_ablation_20251012_071236.log +Timestamp: 2025-10-12 08:14:16 +===================================== diff --git a/logs_oct11/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_7e-2_connector-3.0_0.5_7e-2_ablation_20251012_132440.log b/logs_oct11/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_7e-2_connector-3.0_0.5_7e-2_ablation_20251012_132440.log new file mode 100644 index 0000000000000000000000000000000000000000..31c2ef9a1776d8b224841600a52bb1fb5989557f --- /dev/null +++ b/logs_oct11/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_7e-2_connector-3.0_0.5_7e-2_ablation_20251012_132440.log @@ -0,0 +1,2312 @@ +==== STARTING EXPERIMENT: qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_7e-2_connector-3.0_0.5_7e-2_ablation ==== +Log File: qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_7e-2_connector-3.0_0.5_7e-2_ablation_20251012_132440.log +Timestamp: 2025-10-12 13:24:40 +===================================== +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 13:24:43,644] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 13:24:47,142] [WARNING] [runner.py:202:fetch_hostfile] Unable to find hostfile, will proceed with training with local resources only. +[2025-10-12 13:24:47,143] [INFO] [runner.py:568:main] cmd = /opt/conda/envs/tinyllava/bin/python3.10 -u -m deepspeed.launcher.launch --world_info=eyJsb2NhbGhvc3QiOiBbMCwgMSwgMiwgMywgNCwgNSwgNiwgN119 --master_addr=127.0.0.1 --master_port=29501 --enable_each_rank_log=None tinyllava/train/train.py --deepspeed ./scripts/zero3.json --data_path /s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json --image_folder /s3-code/ywang29/datasets/tinyllava --is_multimodal True --conv_version qwen2_base --model_name_or_path Qwen/Qwen2.5-0.5B --vision_tower google/siglip-so400m-patch14-384 --vision_tower2 --connector_type mlp2x_gelu --mm_vision_select_layer -2 --image_aspect_ratio square --attn_implementation flash_attention_2 --bf16 True --training_recipe common --tune_type_llm full --tune_type_vision_tower frozen --tune_vision_tower_from_layer 0 --tune_type_connector full --group_by_modality_length True --pretrained_model_path /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain --output_dir /nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_7e-2_connector-3.0_0.5_7e-2_ablation --num_train_epochs 1 --per_device_train_batch_size 4 --per_device_eval_batch_size 4 --gradient_accumulation_steps 4 --evaluation_strategy no --learning_rate 7e-2 --weight_decay 0. --warmup_ratio 0.03 --lr_scheduler_type cosine --logging_steps 1 --tf32 False --model_max_length 2048 --gradient_checkpointing True --dataloader_num_workers 8 --lazy_preprocess True --report_to tensorboard --tokenizer_use_fast False --run_name tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune --subnet_mode_text both --subnet_type_text None --mask_type_text soft --init_mean_text 3.0 --temperature_attn_text 0.5 --temperature_mlp_text 0.5 --backward_type_text normal --masked_layers_text all --subnet_mode_vision both --subnet_type_vision None --mask_type_vision soft --init_mean_vision 3.0 --temperature_attn_vision 0.5 --temperature_mlp_vision 0.5 --backward_type_vision normal --masked_layers_vision all --subnet_type_connector global --mask_type_connector soft --init_mean_connector 3.0 --temperature_connector 0.5 --backward_type_connector normal --mm_projector_lr 7e-2 --seed 42 --mask_model llm-connector --save_strategy steps --save_steps 50000 --save_total_limit 1 --train_data_ratio 0.1 +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 13:24:49,729] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 13:24:50,790] [INFO] [launch.py:138:main] 0 NCCL_VERSION=2.21.5 +[2025-10-12 13:24:50,790] [INFO] [launch.py:138:main] 0 NCCL_SOCKET_IFNAME=eth +[2025-10-12 13:24:50,790] [INFO] [launch.py:145:main] WORLD INFO DICT: {'localhost': [0, 1, 2, 3, 4, 5, 6, 7]} +[2025-10-12 13:24:50,790] [INFO] [launch.py:151:main] nnodes=1, num_local_procs=8, node_rank=0 +[2025-10-12 13:24:50,790] [INFO] [launch.py:162:main] global_rank_mapping=defaultdict(, {'localhost': [0, 1, 2, 3, 4, 5, 6, 7]}) +[2025-10-12 13:24:50,790] [INFO] [launch.py:163:main] dist_world_size=8 +[2025-10-12 13:24:50,790] [INFO] [launch.py:165:main] Setting CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 +[2025-10-12 13:24:50,792] [INFO] [launch.py:253:main] process 732982 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=0', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_7e-2_connector-3.0_0.5_7e-2_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '7e-2', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '7e-2', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 13:24:50,794] [INFO] [launch.py:253:main] process 732983 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=1', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_7e-2_connector-3.0_0.5_7e-2_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '7e-2', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '7e-2', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 13:24:50,796] [INFO] [launch.py:253:main] process 732984 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=2', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_7e-2_connector-3.0_0.5_7e-2_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '7e-2', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '7e-2', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 13:24:50,798] [INFO] [launch.py:253:main] process 732985 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=3', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_7e-2_connector-3.0_0.5_7e-2_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '7e-2', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '7e-2', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 13:24:50,799] [INFO] [launch.py:253:main] process 732986 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=4', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_7e-2_connector-3.0_0.5_7e-2_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '7e-2', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '7e-2', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 13:24:50,801] [INFO] [launch.py:253:main] process 732987 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=5', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_7e-2_connector-3.0_0.5_7e-2_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '7e-2', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '7e-2', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 13:24:50,803] [INFO] [launch.py:253:main] process 732988 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=6', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_7e-2_connector-3.0_0.5_7e-2_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '7e-2', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '7e-2', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 13:24:50,805] [INFO] [launch.py:253:main] process 732989 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=7', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_7e-2_connector-3.0_0.5_7e-2_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '7e-2', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '7e-2', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 13:24:57,723] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 13:24:57,723] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 13:24:57,788] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 13:24:57,788] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 13:24:57,791] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 13:24:57,794] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 13:24:57,794] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 13:24:57,816] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 13:24:58,298] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 13:24:58,298] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 13:24:58,298] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 13:24:58,298] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 13:24:58,298] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 13:24:58,298] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 13:24:58,298] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 13:24:58,298] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 13:24:58,298] [INFO] [comm.py:668:init_distributed] Initializing TorchBackend in DeepSpeed with backend nccl +Apply masks for the following modules: ['llm', 'connector'] +{'llm': {'model_name_or_path': 'Qwen/Qwen2.5-0.5B', 'cache_dir': None, 'attn_implementation': 'flash_attention_2', 'subnet_mode': 'both', 'subnet_type': 'None', 'sparsity_attn': None, 'sparsity_mlp': None, 'threshold_attn': None, 'threshold_mlp': None, 'temperature_attn': 0.5, 'temperature_mlp': 0.5, 'masked_layers': 'all', 'mask_type': 'soft', 'backward_type': 'normal'}, 'vision_tower': {'model_name_or_path': 'google/siglip-so400m-patch14-384'}, 'connector': {'connector_type': 'mlp2x_gelu', 'subnet_type': 'global', 'threshold': None, 'sparsity': None, 'temperature': 0.5, 'mask_type': 'soft', 'backward_type': 'normal'}} +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +TinyLlavaConfig { + "backward_type_connector": "normal", + "cache_dir": null, + "connector_type": "mlp2x_gelu", + "hidden_size": 896, + "ignore_index": -100, + "image_aspect_ratio": "square", + "image_token_index": -200, + "llm_model_name_or_path": "Qwen/Qwen2.5-0.5B", + "mask_model": [ + "llm", + "connector" + ], + "mask_type_connector": "soft", + "model_type": "tinyllava", + "num_queries": 128, + "num_resampler_layers": 3, + "pad_token": null, + "resampler_hidden_size": 768, + "sparsity_connector": null, + "subnet_type_connector": "global", + "temperature_connector": 0.5, + "text_config": { + "_name_or_path": "Qwen/Qwen2.5-0.5B", + "architectures": [ + "Qwen2ForCausalLM" + ], + "backward_type": "normal", + "bos_token_id": 151643, + "eos_token_id": 151643, + "hidden_size": 896, + "intermediate_size": 4864, + "mask_type": "soft", + "masked_layers": "all", + "max_position_embeddings": 32768, + "max_window_layers": 24, + "model_type": "qwen2", + "num_attention_heads": 14, + "num_hidden_layers": 24, + "num_key_value_heads": 2, + "rope_theta": 1000000.0, + "sliding_window": 32768, + "subnet_mode": "both", + "subnet_type": "None", + "temperature_attn": 0.5, + "temperature_mlp": 0.5, + "tie_word_embeddings": true, + "torch_dtype": "bfloat16", + "use_mrope": false, + "use_sliding_window": false, + "vocab_size": 151936 + }, + "threshold_connector": null, + "tokenizer_model_max_length": 2048, + "tokenizer_name_or_path": "Qwen/Qwen2.5-0.5B", + "tokenizer_padding_side": "right", + "tokenizer_use_fast": false, + "transformers_version": "4.40.1", + "tune_type_connector": "frozen", + "tune_type_llm": "frozen", + "tune_type_vision_tower": "frozen", + "tune_vision_tower_from_layer": -1, + "use_cache": false, + "vision_config": { + "hidden_act": "gelu_pytorch_tanh", + "hidden_size": 1152, + "image_size": 384, + "intermediate_size": 4304, + "layer_norm_eps": 1e-06, + "model_name_or_path": "google/siglip-so400m-patch14-384", + "model_name_or_path2": "", + "model_type": "siglip_vision_model", + "num_attention_heads": 16, + "num_hidden_layers": 27, + "patch_size": 14 + }, + "vision_feature_layer": -2, + "vision_feature_select_strategy": "patch", + "vision_hidden_size": 1152, + "vision_model_name_or_path": "google/siglip-so400m-patch14-384", + "vision_model_name_or_path2": "", + "vocab_size": 151936 +} + +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test2-worker-0:732982:732982 [0] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:732982:732982 [0] NCCL INFO Bootstrap : Using eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:732982:732982 [0] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test2-worker-0:732982:732982 [0] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test2-worker-0:732982:732982 [0] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test2-worker-0:732982:732982 [0] NCCL INFO cudaDriverVersion 12040 +NCCL version 2.21.5+cuda12.1 +ywang29-vrdb-test2-worker-0:732982:734577 [0] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:732982:734577 [0] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test2-worker-0:732982:734577 [0] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:732982:734577 [0] NCCL INFO NET/Socket : Using [0]eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:732982:734577 [0] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test2-worker-0:732982:734577 [0] NCCL INFO Using network Socket +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test2-worker-0:732987:732987 [5] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test2-worker-0:732987:732987 [5] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:732987:732987 [5] NCCL INFO Bootstrap : Using eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:732987:732987 [5] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test2-worker-0:732987:732987 [5] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test2-worker-0:732987:732987 [5] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test2-worker-0:732989:732989 [7] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test2-worker-0:732989:732989 [7] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:732989:732989 [7] NCCL INFO Bootstrap : Using eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:732989:732989 [7] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test2-worker-0:732989:732989 [7] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test2-worker-0:732989:732989 [7] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test2-worker-0:732983:732983 [1] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test2-worker-0:732983:732983 [1] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:732983:732983 [1] NCCL INFO Bootstrap : Using eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:732983:732983 [1] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test2-worker-0:732983:732983 [1] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test2-worker-0:732983:732983 [1] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test2-worker-0:732985:732985 [3] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test2-worker-0:732985:732985 [3] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:732985:732985 [3] NCCL INFO Bootstrap : Using eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:732985:732985 [3] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test2-worker-0:732985:732985 [3] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test2-worker-0:732985:732985 [3] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test2-worker-0:732988:732988 [6] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test2-worker-0:732988:732988 [6] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:732988:732988 [6] NCCL INFO Bootstrap : Using eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:732988:732988 [6] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test2-worker-0:732988:732988 [6] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test2-worker-0:732988:732988 [6] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test2-worker-0:732986:732986 [4] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test2-worker-0:732986:732986 [4] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:732986:732986 [4] NCCL INFO Bootstrap : Using eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:732986:732986 [4] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test2-worker-0:732986:732986 [4] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test2-worker-0:732986:732986 [4] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test2-worker-0:732984:732984 [2] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test2-worker-0:732984:732984 [2] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:732984:732984 [2] NCCL INFO Bootstrap : Using eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:732984:732984 [2] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test2-worker-0:732984:732984 [2] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test2-worker-0:732984:732984 [2] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test2-worker-0:732989:734595 [7] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:732989:734595 [7] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test2-worker-0:732989:734595 [7] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:732989:734595 [7] NCCL INFO NET/Socket : Using [0]eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:732989:734595 [7] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test2-worker-0:732989:734595 [7] NCCL INFO Using network Socket +ywang29-vrdb-test2-worker-0:732988:734599 [6] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:732988:734599 [6] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test2-worker-0:732988:734599 [6] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:732988:734599 [6] NCCL INFO NET/Socket : Using [0]eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:732988:734599 [6] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test2-worker-0:732988:734599 [6] NCCL INFO Using network Socket +ywang29-vrdb-test2-worker-0:732983:734597 [1] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:732983:734597 [1] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test2-worker-0:732983:734597 [1] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:732983:734597 [1] NCCL INFO NET/Socket : Using [0]eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:732983:734597 [1] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test2-worker-0:732983:734597 [1] NCCL INFO Using network Socket +ywang29-vrdb-test2-worker-0:732987:734596 [5] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:732987:734596 [5] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test2-worker-0:732987:734596 [5] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:732987:734596 [5] NCCL INFO NET/Socket : Using [0]eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:732987:734596 [5] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test2-worker-0:732987:734596 [5] NCCL INFO Using network Socket +ywang29-vrdb-test2-worker-0:732985:734598 [3] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:732985:734598 [3] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test2-worker-0:732985:734598 [3] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:732985:734598 [3] NCCL INFO NET/Socket : Using [0]eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:732985:734598 [3] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test2-worker-0:732985:734598 [3] NCCL INFO Using network Socket +ywang29-vrdb-test2-worker-0:732984:734600 [2] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:732986:734601 [4] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:732986:734601 [4] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test2-worker-0:732984:734600 [2] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test2-worker-0:732984:734600 [2] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:732986:734601 [4] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:732984:734600 [2] NCCL INFO NET/Socket : Using [0]eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:732986:734601 [4] NCCL INFO NET/Socket : Using [0]eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:732986:734601 [4] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test2-worker-0:732984:734600 [2] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test2-worker-0:732986:734601 [4] NCCL INFO Using network Socket +ywang29-vrdb-test2-worker-0:732984:734600 [2] NCCL INFO Using network Socket +ywang29-vrdb-test2-worker-0:732988:734599 [6] NCCL INFO ncclCommInitRank comm 0x555b762d1e00 rank 6 nranks 8 cudaDev 6 nvmlDev 6 busId a01c0 commId 0x6344ea7ffa2ad409 - Init START +ywang29-vrdb-test2-worker-0:732986:734601 [4] NCCL INFO ncclCommInitRank comm 0x55759d9095e0 rank 4 nranks 8 cudaDev 4 nvmlDev 4 busId 901c0 commId 0x6344ea7ffa2ad409 - Init START +ywang29-vrdb-test2-worker-0:732989:734595 [7] NCCL INFO ncclCommInitRank comm 0x5627125f0780 rank 7 nranks 8 cudaDev 7 nvmlDev 7 busId a01d0 commId 0x6344ea7ffa2ad409 - Init START +ywang29-vrdb-test2-worker-0:732987:734596 [5] NCCL INFO ncclCommInitRank comm 0x55739f323440 rank 5 nranks 8 cudaDev 5 nvmlDev 5 busId 901d0 commId 0x6344ea7ffa2ad409 - Init START +ywang29-vrdb-test2-worker-0:732985:734598 [3] NCCL INFO ncclCommInitRank comm 0x560c62423240 rank 3 nranks 8 cudaDev 3 nvmlDev 3 busId 201d0 commId 0x6344ea7ffa2ad409 - Init START +ywang29-vrdb-test2-worker-0:732984:734600 [2] NCCL INFO ncclCommInitRank comm 0x55d3b6137120 rank 2 nranks 8 cudaDev 2 nvmlDev 2 busId 201c0 commId 0x6344ea7ffa2ad409 - Init START +ywang29-vrdb-test2-worker-0:732983:734597 [1] NCCL INFO ncclCommInitRank comm 0x55630c90eec0 rank 1 nranks 8 cudaDev 1 nvmlDev 1 busId 101d0 commId 0x6344ea7ffa2ad409 - Init START +ywang29-vrdb-test2-worker-0:732982:734577 [0] NCCL INFO ncclCommInitRank comm 0x557e15b8f500 rank 0 nranks 8 cudaDev 0 nvmlDev 0 busId 101c0 commId 0x6344ea7ffa2ad409 - Init START +ywang29-vrdb-test2-worker-0:732982:734577 [0] NCCL INFO Setting affinity for GPU 0 to ff,ffff0000,00ffffff +ywang29-vrdb-test2-worker-0:732982:734577 [0] NCCL INFO NVLS multicast support is not available on dev 0 +ywang29-vrdb-test2-worker-0:732989:734595 [7] NCCL INFO Setting affinity for GPU 7 to ffffff00,0000ffff,ff000000 +ywang29-vrdb-test2-worker-0:732989:734595 [7] NCCL INFO NVLS multicast support is not available on dev 7 +ywang29-vrdb-test2-worker-0:732988:734599 [6] NCCL INFO Setting affinity for GPU 6 to ffffff00,0000ffff,ff000000 +ywang29-vrdb-test2-worker-0:732988:734599 [6] NCCL INFO NVLS multicast support is not available on dev 6 +ywang29-vrdb-test2-worker-0:732984:734600 [2] NCCL INFO Setting affinity for GPU 2 to ff,ffff0000,00ffffff +ywang29-vrdb-test2-worker-0:732984:734600 [2] NCCL INFO NVLS multicast support is not available on dev 2 +ywang29-vrdb-test2-worker-0:732987:734596 [5] NCCL INFO Setting affinity for GPU 5 to ffffff00,0000ffff,ff000000 +ywang29-vrdb-test2-worker-0:732987:734596 [5] NCCL INFO NVLS multicast support is not available on dev 5 +ywang29-vrdb-test2-worker-0:732986:734601 [4] NCCL INFO Setting affinity for GPU 4 to ffffff00,0000ffff,ff000000 +ywang29-vrdb-test2-worker-0:732986:734601 [4] NCCL INFO NVLS multicast support is not available on dev 4 +ywang29-vrdb-test2-worker-0:732985:734598 [3] NCCL INFO Setting affinity for GPU 3 to ff,ffff0000,00ffffff +ywang29-vrdb-test2-worker-0:732985:734598 [3] NCCL INFO NVLS multicast support is not available on dev 3 +ywang29-vrdb-test2-worker-0:732983:734597 [1] NCCL INFO Setting affinity for GPU 1 to ff,ffff0000,00ffffff +ywang29-vrdb-test2-worker-0:732983:734597 [1] NCCL INFO NVLS multicast support is not available on dev 1 +ywang29-vrdb-test2-worker-0:732989:734595 [7] NCCL INFO comm 0x5627125f0780 rank 7 nRanks 8 nNodes 1 localRanks 8 localRank 7 MNNVL 0 +ywang29-vrdb-test2-worker-0:732983:734597 [1] NCCL INFO comm 0x55630c90eec0 rank 1 nRanks 8 nNodes 1 localRanks 8 localRank 1 MNNVL 0 +ywang29-vrdb-test2-worker-0:732984:734600 [2] NCCL INFO comm 0x55d3b6137120 rank 2 nRanks 8 nNodes 1 localRanks 8 localRank 2 MNNVL 0 +ywang29-vrdb-test2-worker-0:732982:734577 [0] NCCL INFO comm 0x557e15b8f500 rank 0 nRanks 8 nNodes 1 localRanks 8 localRank 0 MNNVL 0 +ywang29-vrdb-test2-worker-0:732987:734596 [5] NCCL INFO comm 0x55739f323440 rank 5 nRanks 8 nNodes 1 localRanks 8 localRank 5 MNNVL 0 +ywang29-vrdb-test2-worker-0:732988:734599 [6] NCCL INFO comm 0x555b762d1e00 rank 6 nRanks 8 nNodes 1 localRanks 8 localRank 6 MNNVL 0 +ywang29-vrdb-test2-worker-0:732986:734601 [4] NCCL INFO comm 0x55759d9095e0 rank 4 nRanks 8 nNodes 1 localRanks 8 localRank 4 MNNVL 0 +ywang29-vrdb-test2-worker-0:732985:734598 [3] NCCL INFO comm 0x560c62423240 rank 3 nRanks 8 nNodes 1 localRanks 8 localRank 3 MNNVL 0 +ywang29-vrdb-test2-worker-0:732982:734577 [0] NCCL INFO Channel 00/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:732982:734577 [0] NCCL INFO Channel 01/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:732982:734577 [0] NCCL INFO Channel 02/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:732982:734577 [0] NCCL INFO Channel 03/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:732982:734577 [0] NCCL INFO Channel 04/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:732982:734577 [0] NCCL INFO Channel 05/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:732982:734577 [0] NCCL INFO Channel 06/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:732983:734597 [1] NCCL INFO Trees [0] 2/-1/-1->1->0 [1] 2/-1/-1->1->0 [2] 2/-1/-1->1->0 [3] 2/-1/-1->1->0 [4] 2/-1/-1->1->0 [5] 2/-1/-1->1->0 [6] 2/-1/-1->1->0 [7] 2/-1/-1->1->0 [8] 2/-1/-1->1->0 [9] 2/-1/-1->1->0 [10] 2/-1/-1->1->0 [11] 2/-1/-1->1->0 [12] 2/-1/-1->1->0 [13] 2/-1/-1->1->0 [14] 2/-1/-1->1->0 [15] 2/-1/-1->1->0 [16] 2/-1/-1->1->0 [17] 2/-1/-1->1->0 [18] 2/-1/-1->1->0 [19] 2/-1/-1->1->0 [20] 2/-1/-1->1->0 [21] 2/-1/-1->1->0 [22] 2/-1/-1->1->0 [23] 2/-1/-1->1->0 +ywang29-vrdb-test2-worker-0:732982:734577 [0] NCCL INFO Channel 07/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:732982:734577 [0] NCCL INFO Channel 08/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:732984:734600 [2] NCCL INFO Trees [0] 3/-1/-1->2->1 [1] 3/-1/-1->2->1 [2] 3/-1/-1->2->1 [3] 3/-1/-1->2->1 [4] 3/-1/-1->2->1 [5] 3/-1/-1->2->1 [6] 3/-1/-1->2->1 [7] 3/-1/-1->2->1 [8] 3/-1/-1->2->1 [9] 3/-1/-1->2->1 [10] 3/-1/-1->2->1 [11] 3/-1/-1->2->1 [12] 3/-1/-1->2->1 [13] 3/-1/-1->2->1 [14] 3/-1/-1->2->1 [15] 3/-1/-1->2->1 [16] 3/-1/-1->2->1 [17] 3/-1/-1->2->1 [18] 3/-1/-1->2->1 [19] 3/-1/-1->2->1 [20] 3/-1/-1->2->1 [21] 3/-1/-1->2->1 [22] 3/-1/-1->2->1 [23] 3/-1/-1->2->1 +ywang29-vrdb-test2-worker-0:732983:734597 [1] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:732982:734577 [0] NCCL INFO Channel 09/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:732989:734595 [7] NCCL INFO Trees [0] -1/-1/-1->7->6 [1] -1/-1/-1->7->6 [2] -1/-1/-1->7->6 [3] -1/-1/-1->7->6 [4] -1/-1/-1->7->6 [5] -1/-1/-1->7->6 [6] -1/-1/-1->7->6 [7] -1/-1/-1->7->6 [8] -1/-1/-1->7->6 [9] -1/-1/-1->7->6 [10] -1/-1/-1->7->6 [11] -1/-1/-1->7->6 [12] -1/-1/-1->7->6 [13] -1/-1/-1->7->6 [14] -1/-1/-1->7->6 [15] -1/-1/-1->7->6 [16] -1/-1/-1->7->6 [17] -1/-1/-1->7->6 [18] -1/-1/-1->7->6 [19] -1/-1/-1->7->6 [20] -1/-1/-1->7->6 [21] -1/-1/-1->7->6 [22] -1/-1/-1->7->6 [23] -1/-1/-1->7->6 +ywang29-vrdb-test2-worker-0:732987:734596 [5] NCCL INFO Trees [0] 6/-1/-1->5->4 [1] 6/-1/-1->5->4 [2] 6/-1/-1->5->4 [3] 6/-1/-1->5->4 [4] 6/-1/-1->5->4 [5] 6/-1/-1->5->4 [6] 6/-1/-1->5->4 [7] 6/-1/-1->5->4 [8] 6/-1/-1->5->4 [9] 6/-1/-1->5->4 [10] 6/-1/-1->5->4 [11] 6/-1/-1->5->4 [12] 6/-1/-1->5->4 [13] 6/-1/-1->5->4 [14] 6/-1/-1->5->4 [15] 6/-1/-1->5->4 [16] 6/-1/-1->5->4 [17] 6/-1/-1->5->4 [18] 6/-1/-1->5->4 [19] 6/-1/-1->5->4 [20] 6/-1/-1->5->4 [21] 6/-1/-1->5->4 [22] 6/-1/-1->5->4 [23] 6/-1/-1->5->4 +ywang29-vrdb-test2-worker-0:732986:734601 [4] NCCL INFO Trees [0] 5/-1/-1->4->3 [1] 5/-1/-1->4->3 [2] 5/-1/-1->4->3 [3] 5/-1/-1->4->3 [4] 5/-1/-1->4->3 [5] 5/-1/-1->4->3 [6] 5/-1/-1->4->3 [7] 5/-1/-1->4->3 [8] 5/-1/-1->4->3 [9] 5/-1/-1->4->3 [10] 5/-1/-1->4->3 [11] 5/-1/-1->4->3 [12] 5/-1/-1->4->3 [13] 5/-1/-1->4->3 [14] 5/-1/-1->4->3 [15] 5/-1/-1->4->3 [16] 5/-1/-1->4->3 [17] 5/-1/-1->4->3 [18] 5/-1/-1->4->3 [19] 5/-1/-1->4->3 [20] 5/-1/-1->4->3 [21] 5/-1/-1->4->3 [22] 5/-1/-1->4->3 [23] 5/-1/-1->4->3 +ywang29-vrdb-test2-worker-0:732982:734577 [0] NCCL INFO Channel 10/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:732988:734599 [6] NCCL INFO Trees [0] 7/-1/-1->6->5 [1] 7/-1/-1->6->5 [2] 7/-1/-1->6->5 [3] 7/-1/-1->6->5 [4] 7/-1/-1->6->5 [5] 7/-1/-1->6->5 [6] 7/-1/-1->6->5 [7] 7/-1/-1->6->5 [8] 7/-1/-1->6->5 [9] 7/-1/-1->6->5 [10] 7/-1/-1->6->5 [11] 7/-1/-1->6->5 [12] 7/-1/-1->6->5 [13] 7/-1/-1->6->5 [14] 7/-1/-1->6->5 [15] 7/-1/-1->6->5 [16] 7/-1/-1->6->5 [17] 7/-1/-1->6->5 [18] 7/-1/-1->6->5 [19] 7/-1/-1->6->5 [20] 7/-1/-1->6->5 [21] 7/-1/-1->6->5 [22] 7/-1/-1->6->5 [23] 7/-1/-1->6->5 +ywang29-vrdb-test2-worker-0:732984:734600 [2] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:732985:734598 [3] NCCL INFO Trees [0] 4/-1/-1->3->2 [1] 4/-1/-1->3->2 [2] 4/-1/-1->3->2 [3] 4/-1/-1->3->2 [4] 4/-1/-1->3->2 [5] 4/-1/-1->3->2 [6] 4/-1/-1->3->2 [7] 4/-1/-1->3->2 [8] 4/-1/-1->3->2 [9] 4/-1/-1->3->2 [10] 4/-1/-1->3->2 [11] 4/-1/-1->3->2 [12] 4/-1/-1->3->2 [13] 4/-1/-1->3->2 [14] 4/-1/-1->3->2 [15] 4/-1/-1->3->2 [16] 4/-1/-1->3->2 [17] 4/-1/-1->3->2 [18] 4/-1/-1->3->2 [19] 4/-1/-1->3->2 [20] 4/-1/-1->3->2 [21] 4/-1/-1->3->2 [22] 4/-1/-1->3->2 [23] 4/-1/-1->3->2 +ywang29-vrdb-test2-worker-0:732982:734577 [0] NCCL INFO Channel 11/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:732989:734595 [7] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:732986:734601 [4] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:732987:734596 [5] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:732988:734599 [6] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:732982:734577 [0] NCCL INFO Channel 12/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:732985:734598 [3] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:732982:734577 [0] NCCL INFO Channel 13/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:732982:734577 [0] NCCL INFO Channel 14/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:732982:734577 [0] NCCL INFO Channel 15/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:732982:734577 [0] NCCL INFO Channel 16/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:732982:734577 [0] NCCL INFO Channel 17/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:732982:734577 [0] NCCL INFO Channel 18/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:732982:734577 [0] NCCL INFO Channel 19/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:732982:734577 [0] NCCL INFO Channel 20/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:732982:734577 [0] NCCL INFO Channel 21/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:732982:734577 [0] NCCL INFO Channel 22/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:732982:734577 [0] NCCL INFO Channel 23/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:732982:734577 [0] NCCL INFO Trees [0] 1/-1/-1->0->-1 [1] 1/-1/-1->0->-1 [2] 1/-1/-1->0->-1 [3] 1/-1/-1->0->-1 [4] 1/-1/-1->0->-1 [5] 1/-1/-1->0->-1 [6] 1/-1/-1->0->-1 [7] 1/-1/-1->0->-1 [8] 1/-1/-1->0->-1 [9] 1/-1/-1->0->-1 [10] 1/-1/-1->0->-1 [11] 1/-1/-1->0->-1 [12] 1/-1/-1->0->-1 [13] 1/-1/-1->0->-1 [14] 1/-1/-1->0->-1 [15] 1/-1/-1->0->-1 [16] 1/-1/-1->0->-1 [17] 1/-1/-1->0->-1 [18] 1/-1/-1->0->-1 [19] 1/-1/-1->0->-1 [20] 1/-1/-1->0->-1 [21] 1/-1/-1->0->-1 [22] 1/-1/-1->0->-1 [23] 1/-1/-1->0->-1 +ywang29-vrdb-test2-worker-0:732982:734577 [0] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:732984:734600 [2] NCCL INFO Channel 00/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:734601 [4] NCCL INFO Channel 00/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:734598 [3] NCCL INFO Channel 00/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:734599 [6] NCCL INFO Channel 00/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:734596 [5] NCCL INFO Channel 00/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:734595 [7] NCCL INFO Channel 00/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:734600 [2] NCCL INFO Channel 01/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:734601 [4] NCCL INFO Channel 01/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:734598 [3] NCCL INFO Channel 01/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:734599 [6] NCCL INFO Channel 01/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:734596 [5] NCCL INFO Channel 01/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:734595 [7] NCCL INFO Channel 01/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:734600 [2] NCCL INFO Channel 02/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:734601 [4] NCCL INFO Channel 02/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:734598 [3] NCCL INFO Channel 02/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:734599 [6] NCCL INFO Channel 02/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:734596 [5] NCCL INFO Channel 02/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:734595 [7] NCCL INFO Channel 02/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:734600 [2] NCCL INFO Channel 03/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:734601 [4] NCCL INFO Channel 03/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:734598 [3] NCCL INFO Channel 03/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:734599 [6] NCCL INFO Channel 03/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:734596 [5] NCCL INFO Channel 03/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:734595 [7] NCCL INFO Channel 03/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:734600 [2] NCCL INFO Channel 04/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:734599 [6] NCCL INFO Channel 04/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:734601 [4] NCCL INFO Channel 04/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:734598 [3] NCCL INFO Channel 04/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:734596 [5] NCCL INFO Channel 04/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:734595 [7] NCCL INFO Channel 04/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:734600 [2] NCCL INFO Channel 05/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:734599 [6] NCCL INFO Channel 05/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:734601 [4] NCCL INFO Channel 05/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:734598 [3] NCCL INFO Channel 05/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:734596 [5] NCCL INFO Channel 05/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:734595 [7] NCCL INFO Channel 05/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:734600 [2] NCCL INFO Channel 06/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:734599 [6] NCCL INFO Channel 06/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:734601 [4] NCCL INFO Channel 06/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:734598 [3] NCCL INFO Channel 06/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:734596 [5] NCCL INFO Channel 06/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:734595 [7] NCCL INFO Channel 06/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:734600 [2] NCCL INFO Channel 07/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:734599 [6] NCCL INFO Channel 07/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:734601 [4] NCCL INFO Channel 07/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:734598 [3] NCCL INFO Channel 07/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:734596 [5] NCCL INFO Channel 07/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:734595 [7] NCCL INFO Channel 07/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:734600 [2] NCCL INFO Channel 08/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:734599 [6] NCCL INFO Channel 08/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:734601 [4] NCCL INFO Channel 08/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:734598 [3] NCCL INFO Channel 08/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:734596 [5] NCCL INFO Channel 08/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:734595 [7] NCCL INFO Channel 08/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:734600 [2] NCCL INFO Channel 09/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:734599 [6] NCCL INFO Channel 09/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:734601 [4] NCCL INFO Channel 09/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:734598 [3] NCCL INFO Channel 09/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:734596 [5] NCCL INFO Channel 09/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:734595 [7] NCCL INFO Channel 09/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:734600 [2] NCCL INFO Channel 10/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:734599 [6] NCCL INFO Channel 10/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:734601 [4] NCCL INFO Channel 10/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:734598 [3] NCCL INFO Channel 10/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:734596 [5] NCCL INFO Channel 10/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:734595 [7] NCCL INFO Channel 10/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:734600 [2] NCCL INFO Channel 11/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:734599 [6] NCCL INFO Channel 11/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:734601 [4] NCCL INFO Channel 11/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:734598 [3] NCCL INFO Channel 11/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:734595 [7] NCCL INFO Channel 11/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:734596 [5] NCCL INFO Channel 11/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:734600 [2] NCCL INFO Channel 12/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:734599 [6] NCCL INFO Channel 12/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:734601 [4] NCCL INFO Channel 12/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:734598 [3] NCCL INFO Channel 12/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:734595 [7] NCCL INFO Channel 12/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:734596 [5] NCCL INFO Channel 12/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:734597 [1] NCCL INFO Channel 00/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732982:734577 [0] NCCL INFO Channel 00/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:734600 [2] NCCL INFO Channel 13/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:734599 [6] NCCL INFO Channel 13/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:734601 [4] NCCL INFO Channel 13/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:734598 [3] NCCL INFO Channel 13/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:734595 [7] NCCL INFO Channel 13/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:734596 [5] NCCL INFO Channel 13/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:734597 [1] NCCL INFO Channel 01/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732982:734577 [0] NCCL INFO Channel 01/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:734600 [2] NCCL INFO Channel 14/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:734599 [6] NCCL INFO Channel 14/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:734601 [4] NCCL INFO Channel 14/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:734598 [3] NCCL INFO Channel 14/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:734595 [7] NCCL INFO Channel 14/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:734596 [5] NCCL INFO Channel 14/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:734597 [1] NCCL INFO Channel 02/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732982:734577 [0] NCCL INFO Channel 02/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:734600 [2] NCCL INFO Channel 15/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:734599 [6] NCCL INFO Channel 15/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:734601 [4] NCCL INFO Channel 15/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:734598 [3] NCCL INFO Channel 15/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:734595 [7] NCCL INFO Channel 15/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:734596 [5] NCCL INFO Channel 15/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:734597 [1] NCCL INFO Channel 03/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732982:734577 [0] NCCL INFO Channel 03/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:734600 [2] NCCL INFO Channel 16/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:734599 [6] NCCL INFO Channel 16/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:734601 [4] NCCL INFO Channel 16/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:734598 [3] NCCL INFO Channel 16/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:734597 [1] NCCL INFO Channel 04/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:734595 [7] NCCL INFO Channel 16/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:734596 [5] NCCL INFO Channel 16/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732982:734577 [0] NCCL INFO Channel 04/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:734600 [2] NCCL INFO Channel 17/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:734599 [6] NCCL INFO Channel 17/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:734601 [4] NCCL INFO Channel 17/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:734598 [3] NCCL INFO Channel 17/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:734595 [7] NCCL INFO Channel 17/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:734597 [1] NCCL INFO Channel 05/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:734596 [5] NCCL INFO Channel 17/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732982:734577 [0] NCCL INFO Channel 05/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:734600 [2] NCCL INFO Channel 18/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:734599 [6] NCCL INFO Channel 18/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:734601 [4] NCCL INFO Channel 18/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:734598 [3] NCCL INFO Channel 18/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:734595 [7] NCCL INFO Channel 18/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:734597 [1] NCCL INFO Channel 06/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:734596 [5] NCCL INFO Channel 18/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732982:734577 [0] NCCL INFO Channel 06/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:734600 [2] NCCL INFO Channel 19/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:734599 [6] NCCL INFO Channel 19/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:734601 [4] NCCL INFO Channel 19/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:734596 [5] NCCL INFO Channel 19/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:734598 [3] NCCL INFO Channel 19/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:734595 [7] NCCL INFO Channel 19/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:734597 [1] NCCL INFO Channel 07/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732982:734577 [0] NCCL INFO Channel 07/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:734600 [2] NCCL INFO Channel 20/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:734599 [6] NCCL INFO Channel 20/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:734601 [4] NCCL INFO Channel 20/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:734596 [5] NCCL INFO Channel 20/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:734598 [3] NCCL INFO Channel 20/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:734595 [7] NCCL INFO Channel 20/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:734597 [1] NCCL INFO Channel 08/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732982:734577 [0] NCCL INFO Channel 08/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:734600 [2] NCCL INFO Channel 21/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:734599 [6] NCCL INFO Channel 21/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:734601 [4] NCCL INFO Channel 21/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:734596 [5] NCCL INFO Channel 21/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:734598 [3] NCCL INFO Channel 21/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:734595 [7] NCCL INFO Channel 21/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:734597 [1] NCCL INFO Channel 09/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732982:734577 [0] NCCL INFO Channel 09/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:734600 [2] NCCL INFO Channel 22/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:734599 [6] NCCL INFO Channel 22/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:734601 [4] NCCL INFO Channel 22/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:734596 [5] NCCL INFO Channel 22/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:734598 [3] NCCL INFO Channel 22/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:734595 [7] NCCL INFO Channel 22/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:734597 [1] NCCL INFO Channel 10/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732982:734577 [0] NCCL INFO Channel 10/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:734600 [2] NCCL INFO Channel 23/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:734599 [6] NCCL INFO Channel 23/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:734601 [4] NCCL INFO Channel 23/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:734596 [5] NCCL INFO Channel 23/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:734598 [3] NCCL INFO Channel 23/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:734595 [7] NCCL INFO Channel 23/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:734597 [1] NCCL INFO Channel 11/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732982:734577 [0] NCCL INFO Channel 11/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:734597 [1] NCCL INFO Channel 12/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732982:734577 [0] NCCL INFO Channel 12/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:734597 [1] NCCL INFO Channel 13/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732982:734577 [0] NCCL INFO Channel 13/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:734597 [1] NCCL INFO Channel 14/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732982:734577 [0] NCCL INFO Channel 14/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:734597 [1] NCCL INFO Channel 15/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732982:734577 [0] NCCL INFO Channel 15/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:734597 [1] NCCL INFO Channel 16/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732982:734577 [0] NCCL INFO Channel 16/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732982:734577 [0] NCCL INFO Channel 17/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:734597 [1] NCCL INFO Channel 17/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732982:734577 [0] NCCL INFO Channel 18/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:734597 [1] NCCL INFO Channel 18/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732982:734577 [0] NCCL INFO Channel 19/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:734597 [1] NCCL INFO Channel 19/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732982:734577 [0] NCCL INFO Channel 20/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:734597 [1] NCCL INFO Channel 20/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732982:734577 [0] NCCL INFO Channel 21/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:734597 [1] NCCL INFO Channel 21/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732982:734577 [0] NCCL INFO Channel 22/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732982:734577 [0] NCCL INFO Channel 23/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:734597 [1] NCCL INFO Channel 22/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:734597 [1] NCCL INFO Channel 23/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:734600 [2] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:732985:734598 [3] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:732983:734597 [1] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:732989:734595 [7] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:732982:734577 [0] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:732989:734595 [7] NCCL INFO Channel 00/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:734595 [7] NCCL INFO Channel 01/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:734595 [7] NCCL INFO Channel 02/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:734595 [7] NCCL INFO Channel 03/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:734595 [7] NCCL INFO Channel 04/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:734599 [6] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:732986:734601 [4] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:732987:734596 [5] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:732985:734598 [3] NCCL INFO Channel 00/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:734598 [3] NCCL INFO Channel 01/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:734595 [7] NCCL INFO Channel 05/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:734598 [3] NCCL INFO Channel 02/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:734595 [7] NCCL INFO Channel 06/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:734598 [3] NCCL INFO Channel 03/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:734595 [7] NCCL INFO Channel 07/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:734598 [3] NCCL INFO Channel 04/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:734595 [7] NCCL INFO Channel 08/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:734598 [3] NCCL INFO Channel 05/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:734595 [7] NCCL INFO Channel 09/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:734598 [3] NCCL INFO Channel 06/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:734595 [7] NCCL INFO Channel 10/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:734598 [3] NCCL INFO Channel 07/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:734600 [2] NCCL INFO Channel 00/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:734595 [7] NCCL INFO Channel 11/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:734598 [3] NCCL INFO Channel 08/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:734600 [2] NCCL INFO Channel 01/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:734598 [3] NCCL INFO Channel 09/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:734600 [2] NCCL INFO Channel 02/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:734598 [3] NCCL INFO Channel 10/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:734600 [2] NCCL INFO Channel 03/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:734598 [3] NCCL INFO Channel 11/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:734595 [7] NCCL INFO Channel 12/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:734600 [2] NCCL INFO Channel 04/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:734598 [3] NCCL INFO Channel 12/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:734595 [7] NCCL INFO Channel 13/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:734600 [2] NCCL INFO Channel 05/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:734598 [3] NCCL INFO Channel 13/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:734595 [7] NCCL INFO Channel 14/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:734600 [2] NCCL INFO Channel 06/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:734598 [3] NCCL INFO Channel 14/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:734595 [7] NCCL INFO Channel 15/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:734600 [2] NCCL INFO Channel 07/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:734598 [3] NCCL INFO Channel 15/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:734595 [7] NCCL INFO Channel 16/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:734600 [2] NCCL INFO Channel 08/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:734598 [3] NCCL INFO Channel 16/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:734595 [7] NCCL INFO Channel 17/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:734600 [2] NCCL INFO Channel 09/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:734598 [3] NCCL INFO Channel 17/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:734595 [7] NCCL INFO Channel 18/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:734601 [4] NCCL INFO Channel 00/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:734600 [2] NCCL INFO Channel 10/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:734598 [3] NCCL INFO Channel 18/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:734595 [7] NCCL INFO Channel 19/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:734596 [5] NCCL INFO Channel 00/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:734601 [4] NCCL INFO Channel 01/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:734600 [2] NCCL INFO Channel 11/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:734598 [3] NCCL INFO Channel 19/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:734595 [7] NCCL INFO Channel 20/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:734596 [5] NCCL INFO Channel 01/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:734601 [4] NCCL INFO Channel 02/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:734600 [2] NCCL INFO Channel 12/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:734598 [3] NCCL INFO Channel 20/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:734595 [7] NCCL INFO Channel 21/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:734596 [5] NCCL INFO Channel 02/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:734601 [4] NCCL INFO Channel 03/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:734600 [2] NCCL INFO Channel 13/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:734598 [3] NCCL INFO Channel 21/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:734595 [7] NCCL INFO Channel 22/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:734596 [5] NCCL INFO Channel 03/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:734601 [4] NCCL INFO Channel 04/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:734600 [2] NCCL INFO Channel 14/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:734598 [3] NCCL INFO Channel 22/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:734595 [7] NCCL INFO Channel 23/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:734596 [5] NCCL INFO Channel 04/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:734599 [6] NCCL INFO Channel 00/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:734601 [4] NCCL INFO Channel 05/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:734600 [2] NCCL INFO Channel 15/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:734598 [3] NCCL INFO Channel 23/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:734596 [5] NCCL INFO Channel 05/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:734599 [6] NCCL INFO Channel 01/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:734601 [4] NCCL INFO Channel 06/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:734600 [2] NCCL INFO Channel 16/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:734599 [6] NCCL INFO Channel 02/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:734596 [5] NCCL INFO Channel 06/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:734601 [4] NCCL INFO Channel 07/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:734600 [2] NCCL INFO Channel 17/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:734599 [6] NCCL INFO Channel 03/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:734596 [5] NCCL INFO Channel 07/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:734601 [4] NCCL INFO Channel 08/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:734600 [2] NCCL INFO Channel 18/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:734599 [6] NCCL INFO Channel 04/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:734596 [5] NCCL INFO Channel 08/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:734601 [4] NCCL INFO Channel 09/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:734600 [2] NCCL INFO Channel 19/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:734599 [6] NCCL INFO Channel 05/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:734596 [5] NCCL INFO Channel 09/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:734601 [4] NCCL INFO Channel 10/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:734600 [2] NCCL INFO Channel 20/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:734599 [6] NCCL INFO Channel 06/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:734596 [5] NCCL INFO Channel 10/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:734601 [4] NCCL INFO Channel 11/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:734600 [2] NCCL INFO Channel 21/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:734599 [6] NCCL INFO Channel 07/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:734596 [5] NCCL INFO Channel 11/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:734601 [4] NCCL INFO Channel 12/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:734600 [2] NCCL INFO Channel 22/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:734599 [6] NCCL INFO Channel 08/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:734596 [5] NCCL INFO Channel 12/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:734601 [4] NCCL INFO Channel 13/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:734600 [2] NCCL INFO Channel 23/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:734599 [6] NCCL INFO Channel 09/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:734596 [5] NCCL INFO Channel 13/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:734601 [4] NCCL INFO Channel 14/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:734599 [6] NCCL INFO Channel 10/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:734596 [5] NCCL INFO Channel 14/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:734601 [4] NCCL INFO Channel 15/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:734599 [6] NCCL INFO Channel 11/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:734596 [5] NCCL INFO Channel 15/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:734601 [4] NCCL INFO Channel 16/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:734599 [6] NCCL INFO Channel 12/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:734596 [5] NCCL INFO Channel 16/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:734601 [4] NCCL INFO Channel 17/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:734599 [6] NCCL INFO Channel 13/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:734596 [5] NCCL INFO Channel 17/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:734601 [4] NCCL INFO Channel 18/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:734599 [6] NCCL INFO Channel 14/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:734596 [5] NCCL INFO Channel 18/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:734601 [4] NCCL INFO Channel 19/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:734599 [6] NCCL INFO Channel 15/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:734596 [5] NCCL INFO Channel 19/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:734601 [4] NCCL INFO Channel 20/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:734599 [6] NCCL INFO Channel 16/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:734596 [5] NCCL INFO Channel 20/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:734601 [4] NCCL INFO Channel 21/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:734599 [6] NCCL INFO Channel 17/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:734596 [5] NCCL INFO Channel 21/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:734601 [4] NCCL INFO Channel 22/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:734599 [6] NCCL INFO Channel 18/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:734596 [5] NCCL INFO Channel 22/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:734601 [4] NCCL INFO Channel 23/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:734599 [6] NCCL INFO Channel 19/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:734596 [5] NCCL INFO Channel 23/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:734599 [6] NCCL INFO Channel 20/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:734599 [6] NCCL INFO Channel 21/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:734599 [6] NCCL INFO Channel 22/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:734599 [6] NCCL INFO Channel 23/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:734597 [1] NCCL INFO Channel 00/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:734597 [1] NCCL INFO Channel 01/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:734597 [1] NCCL INFO Channel 02/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:734597 [1] NCCL INFO Channel 03/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:734597 [1] NCCL INFO Channel 04/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:734597 [1] NCCL INFO Channel 05/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:734597 [1] NCCL INFO Channel 06/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:734597 [1] NCCL INFO Channel 07/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:734597 [1] NCCL INFO Channel 08/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:734597 [1] NCCL INFO Channel 09/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:734597 [1] NCCL INFO Channel 10/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:734597 [1] NCCL INFO Channel 11/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:734597 [1] NCCL INFO Channel 12/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:734597 [1] NCCL INFO Channel 13/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:734597 [1] NCCL INFO Channel 14/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:734597 [1] NCCL INFO Channel 15/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:734597 [1] NCCL INFO Channel 16/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:734597 [1] NCCL INFO Channel 17/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:734597 [1] NCCL INFO Channel 18/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:734597 [1] NCCL INFO Channel 19/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:734597 [1] NCCL INFO Channel 20/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:734597 [1] NCCL INFO Channel 21/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:734597 [1] NCCL INFO Channel 22/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:734597 [1] NCCL INFO Channel 23/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732982:734577 [0] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:732982:734577 [0] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:732982:734577 [0] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:732986:734601 [4] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:732986:734601 [4] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:732986:734601 [4] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:732987:734596 [5] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:732987:734596 [5] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:732987:734596 [5] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:732989:734595 [7] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:732989:734595 [7] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:732988:734599 [6] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:732989:734595 [7] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:732988:734599 [6] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:732988:734599 [6] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:732983:734597 [1] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:732983:734597 [1] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:732983:734597 [1] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:732985:734598 [3] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:732984:734600 [2] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:732984:734600 [2] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:732985:734598 [3] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:732985:734598 [3] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:732984:734600 [2] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:732988:734599 [6] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test2-worker-0:732987:734596 [5] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test2-worker-0:732986:734601 [4] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test2-worker-0:732989:734595 [7] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test2-worker-0:732987:734596 [5] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test2-worker-0:732986:734601 [4] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test2-worker-0:732989:734595 [7] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test2-worker-0:732986:734601 [4] NCCL INFO ncclCommInitRank comm 0x55759d9095e0 rank 4 nranks 8 cudaDev 4 nvmlDev 4 busId 901c0 commId 0x6344ea7ffa2ad409 - Init COMPLETE +ywang29-vrdb-test2-worker-0:732987:734596 [5] NCCL INFO ncclCommInitRank comm 0x55739f323440 rank 5 nranks 8 cudaDev 5 nvmlDev 5 busId 901d0 commId 0x6344ea7ffa2ad409 - Init COMPLETE +ywang29-vrdb-test2-worker-0:732989:734595 [7] NCCL INFO ncclCommInitRank comm 0x5627125f0780 rank 7 nranks 8 cudaDev 7 nvmlDev 7 busId a01d0 commId 0x6344ea7ffa2ad409 - Init COMPLETE +ywang29-vrdb-test2-worker-0:732988:734599 [6] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test2-worker-0:732988:734599 [6] NCCL INFO ncclCommInitRank comm 0x555b762d1e00 rank 6 nranks 8 cudaDev 6 nvmlDev 6 busId a01c0 commId 0x6344ea7ffa2ad409 - Init COMPLETE +ywang29-vrdb-test2-worker-0:732983:734597 [1] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test2-worker-0:732982:734577 [0] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test2-worker-0:732985:734598 [3] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test2-worker-0:732983:734597 [1] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test2-worker-0:732984:734600 [2] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test2-worker-0:732982:734577 [0] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test2-worker-0:732985:734598 [3] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test2-worker-0:732984:734600 [2] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test2-worker-0:732983:734597 [1] NCCL INFO ncclCommInitRank comm 0x55630c90eec0 rank 1 nranks 8 cudaDev 1 nvmlDev 1 busId 101d0 commId 0x6344ea7ffa2ad409 - Init COMPLETE +ywang29-vrdb-test2-worker-0:732982:734577 [0] NCCL INFO ncclCommInitRank comm 0x557e15b8f500 rank 0 nranks 8 cudaDev 0 nvmlDev 0 busId 101c0 commId 0x6344ea7ffa2ad409 - Init COMPLETE +ywang29-vrdb-test2-worker-0:732985:734598 [3] NCCL INFO ncclCommInitRank comm 0x560c62423240 rank 3 nranks 8 cudaDev 3 nvmlDev 3 busId 201d0 commId 0x6344ea7ffa2ad409 - Init COMPLETE +ywang29-vrdb-test2-worker-0:732984:734600 [2] NCCL INFO ncclCommInitRank comm 0x55d3b6137120 rank 2 nranks 8 cudaDev 2 nvmlDev 2 busId 201c0 commId 0x6344ea7ffa2ad409 - Init COMPLETE +[2025-10-12 13:25:48,580] [INFO] [partition_parameters.py:343:__exit__] finished initializing model - num_params = 459, num_elems = 0.99B +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model + +loading language model from loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_modelloading language model from + /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_modelloading language model from + /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model + /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +[2025-10-12 13:36:31,741] [INFO] [partition_parameters.py:343:__exit__] finished initializing model - num_params = 907, num_elems = 1.42B +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +TinyLlavaForConditionalGeneration( + (language_model): Qwen2ForCausalLM( + (model): Qwen2Model( + (embed_tokens): Embedding(151936, 896) + (layers): ModuleList( + (0-23): 24 x Qwen2DecoderLayer( + (self_attn): Qwen2FlashAttention2( + (q_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=True) + (k_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=128, bias=True) + (v_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=128, bias=True) + (o_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=False) + (rotary_emb): Qwen2RotaryEmbedding() + ) + (mlp): Qwen2MLP( + (gate_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=4864, bias=False) + (up_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=4864, bias=False) + (down_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=4864, out_features=896, bias=False) + (act_fn): SiLU() + ) + (input_layernorm): Qwen2RMSNorm() + (post_attention_layernorm): Qwen2RMSNorm() + ) + ) + (norm): Qwen2RMSNorm() + ) + (lm_head): Linear(in_features=896, out_features=151936, bias=False) + ) + (vision_tower): SIGLIPVisionTower( + (_vision_tower): SiglipVisionModel( + (vision_model): SiglipVisionTransformer( + (embeddings): SiglipVisionEmbeddings( + (patch_embedding): Conv2d(3, 1152, kernel_size=(14, 14), stride=(14, 14), padding=valid) + (position_embedding): Embedding(729, 1152) + ) + (encoder): SiglipEncoder( + (layers): ModuleList( + (0-26): 27 x SiglipEncoderLayer( + (self_attn): SiglipAttention( + (k_proj): Linear(in_features=1152, out_features=1152, bias=True) + (v_proj): Linear(in_features=1152, out_features=1152, bias=True) + (q_proj): Linear(in_features=1152, out_features=1152, bias=True) + (out_proj): Linear(in_features=1152, out_features=1152, bias=True) + ) + (layer_norm1): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + (layer_norm2): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + ) + ) + ) + (post_layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (head): SiglipMultiheadAttentionPoolingHead( + (attention): MultiheadAttention( + (out_proj): NonDynamicallyQuantizableLinear(in_features=1152, out_features=1152, bias=True) + ) + (layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + ) + ) + ) + ) + (connector): MLPConnector( + (_connector): Sequential( + (0): SupermaskLinearSparsity_SoftForward_Normal(in_features=1152, out_features=896, bias=True) + (1): GELU(approximate='none') + (2): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=True) + ) + ) +) +Pre-training init language_model.model.layers.0.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.0.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.0.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.0.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.0.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.0.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.0.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.mlp.down_proj.scores: Mean=3.000000 +Pre-training init connector._connector.0.scores: Mean=3.000005 +Pre-training init connector._connector.2.scores: Mean=2.999970 +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +2025-10-12 13:36:49,697 | INFO: Total Parameters: 1283756736, Total Trainable Parameters: 359661568 +2025-10-12 13:36:49,709 | INFO: Trainable Parameters: +language_model.model.layers.0.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.0.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.0.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.0.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.0.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.0.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.0.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.1.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.1.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.1.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.1.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.1.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.1.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.1.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.2.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.2.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.2.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.2.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.2.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.2.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.2.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.3.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.3.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.3.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.3.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.3.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.3.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.3.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.4.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.4.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.4.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.4.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.4.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.4.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.4.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.5.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.5.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.5.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.5.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.5.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.5.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.5.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.6.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.6.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.6.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.6.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.6.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.6.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.6.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.7.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.7.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.7.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.7.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.7.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.7.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.7.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.8.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.8.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.8.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.8.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.8.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.8.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.8.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.9.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.9.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.9.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.9.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.9.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.9.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.9.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.10.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.10.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.10.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.10.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.10.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.10.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.10.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.11.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.11.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.11.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.11.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.11.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.11.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.11.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.12.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.12.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.12.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.12.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.12.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.12.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.12.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.13.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.13.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.13.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.13.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.13.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.13.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.13.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.14.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.14.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.14.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.14.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.14.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.14.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.14.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.15.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.15.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.15.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.15.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.15.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.15.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.15.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.16.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.16.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.16.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.16.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.16.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.16.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.16.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.17.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.17.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.17.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.17.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.17.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.17.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.17.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.18.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.18.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.18.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.18.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.18.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.18.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.18.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.19.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.19.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.19.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.19.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.19.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.19.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.19.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.20.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.20.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.20.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.20.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.20.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.20.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.20.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.21.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.21.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.21.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.21.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.21.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.21.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.21.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.22.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.22.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.22.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.22.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.22.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.22.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.22.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.23.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.23.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.23.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.23.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.23.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.23.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.23.mlp.down_proj.scores: 4358144 parameters +connector._connector.0.scores: 1032192 parameters +connector._connector.2.scores: 802816 parameters +Parameter Offload: Total persistent parameters: 486464 in 403 params + 0%| | 0/520 [00:003->2 [1] 4/-1/-1->3->2 [2] 4/-1/-1->3->2 [3] 4/-1/-1->3->2 [4] 4/-1/-1->3->2 [5] 4/-1/-1->3->2 [6] 4/-1/-1->3->2 [7] 4/-1/-1->3->2 [8] 4/-1/-1->3->2 [9] 4/-1/-1->3->2 [10] 4/-1/-1->3->2 [11] 4/-1/-1->3->2 [12] 4/-1/-1->3->2 [13] 4/-1/-1->3->2 [14] 4/-1/-1->3->2 [15] 4/-1/-1->3->2 [16] 4/-1/-1->3->2 [17] 4/-1/-1->3->2 [18] 4/-1/-1->3->2 [19] 4/-1/-1->3->2 [20] 4/-1/-1->3->2 [21] 4/-1/-1->3->2 [22] 4/-1/-1->3->2 [23] 4/-1/-1->3->2 +ywang29-vrdb-test2-worker-0:732982:740019 [0] NCCL INFO Channel 07/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:732989:740023 [7] NCCL INFO Trees [0] -1/-1/-1->7->6 [1] -1/-1/-1->7->6 [2] -1/-1/-1->7->6 [3] -1/-1/-1->7->6 [4] -1/-1/-1->7->6 [5] -1/-1/-1->7->6 [6] -1/-1/-1->7->6 [7] -1/-1/-1->7->6 [8] -1/-1/-1->7->6 [9] -1/-1/-1->7->6 [10] -1/-1/-1->7->6 [11] -1/-1/-1->7->6 [12] -1/-1/-1->7->6 [13] -1/-1/-1->7->6 [14] -1/-1/-1->7->6 [15] -1/-1/-1->7->6 [16] -1/-1/-1->7->6 [17] -1/-1/-1->7->6 [18] -1/-1/-1->7->6 [19] -1/-1/-1->7->6 [20] -1/-1/-1->7->6 [21] -1/-1/-1->7->6 [22] -1/-1/-1->7->6 [23] -1/-1/-1->7->6 +ywang29-vrdb-test2-worker-0:732985:740020 [3] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:732987:740026 [5] NCCL INFO Trees [0] 6/-1/-1->5->4 [1] 6/-1/-1->5->4 [2] 6/-1/-1->5->4 [3] 6/-1/-1->5->4 [4] 6/-1/-1->5->4 [5] 6/-1/-1->5->4 [6] 6/-1/-1->5->4 [7] 6/-1/-1->5->4 [8] 6/-1/-1->5->4 [9] 6/-1/-1->5->4 [10] 6/-1/-1->5->4 [11] 6/-1/-1->5->4 [12] 6/-1/-1->5->4 [13] 6/-1/-1->5->4 [14] 6/-1/-1->5->4 [15] 6/-1/-1->5->4 [16] 6/-1/-1->5->4 [17] 6/-1/-1->5->4 [18] 6/-1/-1->5->4 [19] 6/-1/-1->5->4 [20] 6/-1/-1->5->4 [21] 6/-1/-1->5->4 [22] 6/-1/-1->5->4 [23] 6/-1/-1->5->4 +ywang29-vrdb-test2-worker-0:732982:740019 [0] NCCL INFO Channel 08/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:732989:740023 [7] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:732987:740026 [5] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:732982:740019 [0] NCCL INFO Channel 09/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:732982:740019 [0] NCCL INFO Channel 10/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:732982:740019 [0] NCCL INFO Channel 11/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:732982:740019 [0] NCCL INFO Channel 12/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:732982:740019 [0] NCCL INFO Channel 13/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:732982:740019 [0] NCCL INFO Channel 14/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:732983:740025 [1] NCCL INFO Trees [0] 2/-1/-1->1->0 [1] 2/-1/-1->1->0 [2] 2/-1/-1->1->0 [3] 2/-1/-1->1->0 [4] 2/-1/-1->1->0 [5] 2/-1/-1->1->0 [6] 2/-1/-1->1->0 [7] 2/-1/-1->1->0 [8] 2/-1/-1->1->0 [9] 2/-1/-1->1->0 [10] 2/-1/-1->1->0 [11] 2/-1/-1->1->0 [12] 2/-1/-1->1->0 [13] 2/-1/-1->1->0 [14] 2/-1/-1->1->0 [15] 2/-1/-1->1->0 [16] 2/-1/-1->1->0 [17] 2/-1/-1->1->0 [18] 2/-1/-1->1->0 [19] 2/-1/-1->1->0 [20] 2/-1/-1->1->0 [21] 2/-1/-1->1->0 [22] 2/-1/-1->1->0 [23] 2/-1/-1->1->0 +ywang29-vrdb-test2-worker-0:732984:740022 [2] NCCL INFO Trees [0] 3/-1/-1->2->1 [1] 3/-1/-1->2->1 [2] 3/-1/-1->2->1 [3] 3/-1/-1->2->1 [4] 3/-1/-1->2->1 [5] 3/-1/-1->2->1 [6] 3/-1/-1->2->1 [7] 3/-1/-1->2->1 [8] 3/-1/-1->2->1 [9] 3/-1/-1->2->1 [10] 3/-1/-1->2->1 [11] 3/-1/-1->2->1 [12] 3/-1/-1->2->1 [13] 3/-1/-1->2->1 [14] 3/-1/-1->2->1 [15] 3/-1/-1->2->1 [16] 3/-1/-1->2->1 [17] 3/-1/-1->2->1 [18] 3/-1/-1->2->1 [19] 3/-1/-1->2->1 [20] 3/-1/-1->2->1 [21] 3/-1/-1->2->1 [22] 3/-1/-1->2->1 [23] 3/-1/-1->2->1 +ywang29-vrdb-test2-worker-0:732982:740019 [0] NCCL INFO Channel 15/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:732983:740025 [1] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:732982:740019 [0] NCCL INFO Channel 16/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:732984:740022 [2] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:732982:740019 [0] NCCL INFO Channel 17/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:732988:740021 [6] NCCL INFO Trees [0] 7/-1/-1->6->5 [1] 7/-1/-1->6->5 [2] 7/-1/-1->6->5 [3] 7/-1/-1->6->5 [4] 7/-1/-1->6->5 [5] 7/-1/-1->6->5 [6] 7/-1/-1->6->5 [7] 7/-1/-1->6->5 [8] 7/-1/-1->6->5 [9] 7/-1/-1->6->5 [10] 7/-1/-1->6->5 [11] 7/-1/-1->6->5 [12] 7/-1/-1->6->5 [13] 7/-1/-1->6->5 [14] 7/-1/-1->6->5 [15] 7/-1/-1->6->5 [16] 7/-1/-1->6->5 [17] 7/-1/-1->6->5 [18] 7/-1/-1->6->5 [19] 7/-1/-1->6->5 [20] 7/-1/-1->6->5 [21] 7/-1/-1->6->5 [22] 7/-1/-1->6->5 [23] 7/-1/-1->6->5 +ywang29-vrdb-test2-worker-0:732982:740019 [0] NCCL INFO Channel 18/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:732988:740021 [6] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:732982:740019 [0] NCCL INFO Channel 19/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:732982:740019 [0] NCCL INFO Channel 20/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:732986:740024 [4] NCCL INFO Trees [0] 5/-1/-1->4->3 [1] 5/-1/-1->4->3 [2] 5/-1/-1->4->3 [3] 5/-1/-1->4->3 [4] 5/-1/-1->4->3 [5] 5/-1/-1->4->3 [6] 5/-1/-1->4->3 [7] 5/-1/-1->4->3 [8] 5/-1/-1->4->3 [9] 5/-1/-1->4->3 [10] 5/-1/-1->4->3 [11] 5/-1/-1->4->3 [12] 5/-1/-1->4->3 [13] 5/-1/-1->4->3 [14] 5/-1/-1->4->3 [15] 5/-1/-1->4->3 [16] 5/-1/-1->4->3 [17] 5/-1/-1->4->3 [18] 5/-1/-1->4->3 [19] 5/-1/-1->4->3 [20] 5/-1/-1->4->3 [21] 5/-1/-1->4->3 [22] 5/-1/-1->4->3 [23] 5/-1/-1->4->3 +ywang29-vrdb-test2-worker-0:732982:740019 [0] NCCL INFO Channel 21/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:732982:740019 [0] NCCL INFO Channel 22/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:732986:740024 [4] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:732982:740019 [0] NCCL INFO Channel 23/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:732982:740019 [0] NCCL INFO Trees [0] 1/-1/-1->0->-1 [1] 1/-1/-1->0->-1 [2] 1/-1/-1->0->-1 [3] 1/-1/-1->0->-1 [4] 1/-1/-1->0->-1 [5] 1/-1/-1->0->-1 [6] 1/-1/-1->0->-1 [7] 1/-1/-1->0->-1 [8] 1/-1/-1->0->-1 [9] 1/-1/-1->0->-1 [10] 1/-1/-1->0->-1 [11] 1/-1/-1->0->-1 [12] 1/-1/-1->0->-1 [13] 1/-1/-1->0->-1 [14] 1/-1/-1->0->-1 [15] 1/-1/-1->0->-1 [16] 1/-1/-1->0->-1 [17] 1/-1/-1->0->-1 [18] 1/-1/-1->0->-1 [19] 1/-1/-1->0->-1 [20] 1/-1/-1->0->-1 [21] 1/-1/-1->0->-1 [22] 1/-1/-1->0->-1 [23] 1/-1/-1->0->-1 +ywang29-vrdb-test2-worker-0:732982:740019 [0] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:732987:740026 [5] NCCL INFO Channel 00/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:740023 [7] NCCL INFO Channel 00/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:740020 [3] NCCL INFO Channel 00/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:740026 [5] NCCL INFO Channel 01/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732982:740019 [0] NCCL INFO Channel 00/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:740023 [7] NCCL INFO Channel 01/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:740020 [3] NCCL INFO Channel 01/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:740026 [5] NCCL INFO Channel 02/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732982:740019 [0] NCCL INFO Channel 01/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:740023 [7] NCCL INFO Channel 02/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:740020 [3] NCCL INFO Channel 02/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:740026 [5] NCCL INFO Channel 03/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732982:740019 [0] NCCL INFO Channel 02/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:740023 [7] NCCL INFO Channel 03/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:740022 [2] NCCL INFO Channel 00/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:740020 [3] NCCL INFO Channel 03/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:740026 [5] NCCL INFO Channel 04/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732982:740019 [0] NCCL INFO Channel 03/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:740025 [1] NCCL INFO Channel 00/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:740021 [6] NCCL INFO Channel 00/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:740023 [7] NCCL INFO Channel 04/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:740022 [2] NCCL INFO Channel 01/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:740020 [3] NCCL INFO Channel 04/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:740024 [4] NCCL INFO Channel 00/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:740026 [5] NCCL INFO Channel 05/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732982:740019 [0] NCCL INFO Channel 04/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:740025 [1] NCCL INFO Channel 01/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:740023 [7] NCCL INFO Channel 05/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:740020 [3] NCCL INFO Channel 05/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:740021 [6] NCCL INFO Channel 01/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:740022 [2] NCCL INFO Channel 02/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:740024 [4] NCCL INFO Channel 01/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:740026 [5] NCCL INFO Channel 06/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732982:740019 [0] NCCL INFO Channel 05/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:740025 [1] NCCL INFO Channel 02/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:740023 [7] NCCL INFO Channel 06/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:740020 [3] NCCL INFO Channel 06/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:740021 [6] NCCL INFO Channel 02/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:740022 [2] NCCL INFO Channel 03/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:740024 [4] NCCL INFO Channel 02/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:740026 [5] NCCL INFO Channel 07/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732982:740019 [0] NCCL INFO Channel 06/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:740025 [1] NCCL INFO Channel 03/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:740023 [7] NCCL INFO Channel 07/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:740020 [3] NCCL INFO Channel 07/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:740021 [6] NCCL INFO Channel 03/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:740022 [2] NCCL INFO Channel 04/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:740024 [4] NCCL INFO Channel 03/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:740026 [5] NCCL INFO Channel 08/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732982:740019 [0] NCCL INFO Channel 07/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:740025 [1] NCCL INFO Channel 04/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:740023 [7] NCCL INFO Channel 08/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:740020 [3] NCCL INFO Channel 08/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:740021 [6] NCCL INFO Channel 04/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:740022 [2] NCCL INFO Channel 05/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:740026 [5] NCCL INFO Channel 09/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:740024 [4] NCCL INFO Channel 04/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732982:740019 [0] NCCL INFO Channel 08/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:740025 [1] NCCL INFO Channel 05/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:740023 [7] NCCL INFO Channel 09/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:740020 [3] NCCL INFO Channel 09/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:740021 [6] NCCL INFO Channel 05/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:740026 [5] NCCL INFO Channel 10/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:740022 [2] NCCL INFO Channel 06/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:740024 [4] NCCL INFO Channel 05/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732982:740019 [0] NCCL INFO Channel 09/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:740025 [1] NCCL INFO Channel 06/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:740023 [7] NCCL INFO Channel 10/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:740020 [3] NCCL INFO Channel 10/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:740021 [6] NCCL INFO Channel 06/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:740026 [5] NCCL INFO Channel 11/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:740022 [2] NCCL INFO Channel 07/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:740024 [4] NCCL INFO Channel 06/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732982:740019 [0] NCCL INFO Channel 10/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:740023 [7] NCCL INFO Channel 11/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:740025 [1] NCCL INFO Channel 07/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:740020 [3] NCCL INFO Channel 11/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:740021 [6] NCCL INFO Channel 07/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:740026 [5] NCCL INFO Channel 12/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:740022 [2] NCCL INFO Channel 08/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732982:740019 [0] NCCL INFO Channel 11/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:740024 [4] NCCL INFO Channel 07/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:740023 [7] NCCL INFO Channel 12/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:740020 [3] NCCL INFO Channel 12/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:740025 [1] NCCL INFO Channel 08/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:740021 [6] NCCL INFO Channel 08/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:740026 [5] NCCL INFO Channel 13/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:740024 [4] NCCL INFO Channel 08/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732982:740019 [0] NCCL INFO Channel 12/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:740022 [2] NCCL INFO Channel 09/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:740023 [7] NCCL INFO Channel 13/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:740020 [3] NCCL INFO Channel 13/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:740025 [1] NCCL INFO Channel 09/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:740021 [6] NCCL INFO Channel 09/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:740026 [5] NCCL INFO Channel 14/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:740024 [4] NCCL INFO Channel 09/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732982:740019 [0] NCCL INFO Channel 13/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:740022 [2] NCCL INFO Channel 10/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:740023 [7] NCCL INFO Channel 14/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:740020 [3] NCCL INFO Channel 14/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:740025 [1] NCCL INFO Channel 10/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:740021 [6] NCCL INFO Channel 10/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:740026 [5] NCCL INFO Channel 15/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:740024 [4] NCCL INFO Channel 10/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732982:740019 [0] NCCL INFO Channel 14/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:740022 [2] NCCL INFO Channel 11/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:740023 [7] NCCL INFO Channel 15/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:740020 [3] NCCL INFO Channel 15/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:740025 [1] NCCL INFO Channel 11/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:740021 [6] NCCL INFO Channel 11/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:740026 [5] NCCL INFO Channel 16/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:740024 [4] NCCL INFO Channel 11/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732982:740019 [0] NCCL INFO Channel 15/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:740022 [2] NCCL INFO Channel 12/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:740023 [7] NCCL INFO Channel 16/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:740020 [3] NCCL INFO Channel 16/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:740025 [1] NCCL INFO Channel 12/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:740021 [6] NCCL INFO Channel 12/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:740026 [5] NCCL INFO Channel 17/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732982:740019 [0] NCCL INFO Channel 16/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:740024 [4] NCCL INFO Channel 12/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:740022 [2] NCCL INFO Channel 13/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:740023 [7] NCCL INFO Channel 17/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:740020 [3] NCCL INFO Channel 17/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:740025 [1] NCCL INFO Channel 13/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:740021 [6] NCCL INFO Channel 13/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:740026 [5] NCCL INFO Channel 18/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732982:740019 [0] NCCL INFO Channel 17/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:740024 [4] NCCL INFO Channel 13/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:740023 [7] NCCL INFO Channel 18/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:740022 [2] NCCL INFO Channel 14/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:740020 [3] NCCL INFO Channel 18/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:740025 [1] NCCL INFO Channel 14/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:740021 [6] NCCL INFO Channel 14/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:740026 [5] NCCL INFO Channel 19/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732982:740019 [0] NCCL INFO Channel 18/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:740024 [4] NCCL INFO Channel 14/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:740020 [3] NCCL INFO Channel 19/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:740023 [7] NCCL INFO Channel 19/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:740022 [2] NCCL INFO Channel 15/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:740025 [1] NCCL INFO Channel 15/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:740021 [6] NCCL INFO Channel 15/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:740026 [5] NCCL INFO Channel 20/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732982:740019 [0] NCCL INFO Channel 19/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:740024 [4] NCCL INFO Channel 15/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:740020 [3] NCCL INFO Channel 20/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:740023 [7] NCCL INFO Channel 20/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:740022 [2] NCCL INFO Channel 16/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:740025 [1] NCCL INFO Channel 16/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:740026 [5] NCCL INFO Channel 21/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:740021 [6] NCCL INFO Channel 16/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732982:740019 [0] NCCL INFO Channel 20/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:740020 [3] NCCL INFO Channel 21/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:740024 [4] NCCL INFO Channel 16/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:740023 [7] NCCL INFO Channel 21/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:740022 [2] NCCL INFO Channel 17/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:740025 [1] NCCL INFO Channel 17/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:740021 [6] NCCL INFO Channel 17/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:740026 [5] NCCL INFO Channel 22/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732982:740019 [0] NCCL INFO Channel 21/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:740020 [3] NCCL INFO Channel 22/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:740024 [4] NCCL INFO Channel 17/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:740023 [7] NCCL INFO Channel 22/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:740022 [2] NCCL INFO Channel 18/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:740025 [1] NCCL INFO Channel 18/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:740026 [5] NCCL INFO Channel 23/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732982:740019 [0] NCCL INFO Channel 22/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:740021 [6] NCCL INFO Channel 18/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:740020 [3] NCCL INFO Channel 23/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:740023 [7] NCCL INFO Channel 23/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:740024 [4] NCCL INFO Channel 18/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:740022 [2] NCCL INFO Channel 19/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:740025 [1] NCCL INFO Channel 19/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732982:740019 [0] NCCL INFO Channel 23/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:740021 [6] NCCL INFO Channel 19/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:740024 [4] NCCL INFO Channel 19/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:740022 [2] NCCL INFO Channel 20/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:740025 [1] NCCL INFO Channel 20/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:740021 [6] NCCL INFO Channel 20/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:740024 [4] NCCL INFO Channel 20/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:740022 [2] NCCL INFO Channel 21/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:740021 [6] NCCL INFO Channel 21/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:740025 [1] NCCL INFO Channel 21/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:740024 [4] NCCL INFO Channel 21/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:740022 [2] NCCL INFO Channel 22/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:740025 [1] NCCL INFO Channel 22/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:740021 [6] NCCL INFO Channel 22/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:740024 [4] NCCL INFO Channel 22/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:740022 [2] NCCL INFO Channel 23/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:740025 [1] NCCL INFO Channel 23/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:740021 [6] NCCL INFO Channel 23/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:740024 [4] NCCL INFO Channel 23/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:740022 [2] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:732983:740025 [1] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:732985:740020 [3] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:732982:740019 [0] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:732989:740023 [7] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:732989:740023 [7] NCCL INFO Channel 00/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:740023 [7] NCCL INFO Channel 01/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:740023 [7] NCCL INFO Channel 02/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:740023 [7] NCCL INFO Channel 03/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:740023 [7] NCCL INFO Channel 04/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:740022 [2] NCCL INFO Channel 00/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:740022 [2] NCCL INFO Channel 01/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:740023 [7] NCCL INFO Channel 05/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:740023 [7] NCCL INFO Channel 06/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:740022 [2] NCCL INFO Channel 02/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:740021 [6] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:732986:740024 [4] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:732987:740026 [5] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:732989:740023 [7] NCCL INFO Channel 07/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:740022 [2] NCCL INFO Channel 03/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:740023 [7] NCCL INFO Channel 08/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:740022 [2] NCCL INFO Channel 04/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:740023 [7] NCCL INFO Channel 09/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:740022 [2] NCCL INFO Channel 05/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:740023 [7] NCCL INFO Channel 10/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:740022 [2] NCCL INFO Channel 06/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:740023 [7] NCCL INFO Channel 11/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:740022 [2] NCCL INFO Channel 07/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:740023 [7] NCCL INFO Channel 12/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:740022 [2] NCCL INFO Channel 08/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:740023 [7] NCCL INFO Channel 13/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:740022 [2] NCCL INFO Channel 09/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:740023 [7] NCCL INFO Channel 14/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:740022 [2] NCCL INFO Channel 10/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:740023 [7] NCCL INFO Channel 15/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:740022 [2] NCCL INFO Channel 11/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:740023 [7] NCCL INFO Channel 16/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:740022 [2] NCCL INFO Channel 12/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:740023 [7] NCCL INFO Channel 17/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:740022 [2] NCCL INFO Channel 13/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:740023 [7] NCCL INFO Channel 18/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:740022 [2] NCCL INFO Channel 14/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:740023 [7] NCCL INFO Channel 19/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:740022 [2] NCCL INFO Channel 15/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:740023 [7] NCCL INFO Channel 20/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:740022 [2] NCCL INFO Channel 16/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:740023 [7] NCCL INFO Channel 21/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:740022 [2] NCCL INFO Channel 17/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:740023 [7] NCCL INFO Channel 22/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:740020 [3] NCCL INFO Channel 00/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:740022 [2] NCCL INFO Channel 18/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732989:740023 [7] NCCL INFO Channel 23/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:740020 [3] NCCL INFO Channel 01/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:740022 [2] NCCL INFO Channel 19/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:740020 [3] NCCL INFO Channel 02/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:740022 [2] NCCL INFO Channel 20/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:740020 [3] NCCL INFO Channel 03/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:740022 [2] NCCL INFO Channel 21/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:740020 [3] NCCL INFO Channel 04/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:740022 [2] NCCL INFO Channel 22/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:740020 [3] NCCL INFO Channel 05/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732984:740022 [2] NCCL INFO Channel 23/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:740020 [3] NCCL INFO Channel 06/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:740020 [3] NCCL INFO Channel 07/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:740020 [3] NCCL INFO Channel 08/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:740026 [5] NCCL INFO Channel 00/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:740020 [3] NCCL INFO Channel 09/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:740025 [1] NCCL INFO Channel 00/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:740026 [5] NCCL INFO Channel 01/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:740020 [3] NCCL INFO Channel 10/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:740021 [6] NCCL INFO Channel 00/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:740025 [1] NCCL INFO Channel 01/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:740026 [5] NCCL INFO Channel 02/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:740020 [3] NCCL INFO Channel 11/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:740021 [6] NCCL INFO Channel 01/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:740025 [1] NCCL INFO Channel 02/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:740026 [5] NCCL INFO Channel 03/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:740020 [3] NCCL INFO Channel 12/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:740025 [1] NCCL INFO Channel 03/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:740021 [6] NCCL INFO Channel 02/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:740026 [5] NCCL INFO Channel 04/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:740020 [3] NCCL INFO Channel 13/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:740024 [4] NCCL INFO Channel 00/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:740025 [1] NCCL INFO Channel 04/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:740021 [6] NCCL INFO Channel 03/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:740026 [5] NCCL INFO Channel 05/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:740020 [3] NCCL INFO Channel 14/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:740024 [4] NCCL INFO Channel 01/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:740025 [1] NCCL INFO Channel 05/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:740021 [6] NCCL INFO Channel 04/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:740026 [5] NCCL INFO Channel 06/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:740020 [3] NCCL INFO Channel 15/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:740024 [4] NCCL INFO Channel 02/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:740025 [1] NCCL INFO Channel 06/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:740026 [5] NCCL INFO Channel 07/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:740020 [3] NCCL INFO Channel 16/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:740025 [1] NCCL INFO Channel 07/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:740026 [5] NCCL INFO Channel 08/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:740020 [3] NCCL INFO Channel 17/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:740025 [1] NCCL INFO Channel 08/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:740026 [5] NCCL INFO Channel 09/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:740020 [3] NCCL INFO Channel 18/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:740025 [1] NCCL INFO Channel 09/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:740026 [5] NCCL INFO Channel 10/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:740020 [3] NCCL INFO Channel 19/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:740025 [1] NCCL INFO Channel 10/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:740026 [5] NCCL INFO Channel 11/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:740020 [3] NCCL INFO Channel 20/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:740025 [1] NCCL INFO Channel 11/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:740026 [5] NCCL INFO Channel 12/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:740020 [3] NCCL INFO Channel 21/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:740025 [1] NCCL INFO Channel 12/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:740026 [5] NCCL INFO Channel 13/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:740020 [3] NCCL INFO Channel 22/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:740025 [1] NCCL INFO Channel 13/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:740026 [5] NCCL INFO Channel 14/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732985:740020 [3] NCCL INFO Channel 23/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:740025 [1] NCCL INFO Channel 14/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:740026 [5] NCCL INFO Channel 15/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:740025 [1] NCCL INFO Channel 15/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:740026 [5] NCCL INFO Channel 16/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:740025 [1] NCCL INFO Channel 16/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:740026 [5] NCCL INFO Channel 17/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:740025 [1] NCCL INFO Channel 17/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:740026 [5] NCCL INFO Channel 18/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:740025 [1] NCCL INFO Channel 18/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:740026 [5] NCCL INFO Channel 19/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:740025 [1] NCCL INFO Channel 19/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:740026 [5] NCCL INFO Channel 20/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:740025 [1] NCCL INFO Channel 20/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:740026 [5] NCCL INFO Channel 21/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:740025 [1] NCCL INFO Channel 21/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:740026 [5] NCCL INFO Channel 22/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:740025 [1] NCCL INFO Channel 22/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732987:740026 [5] NCCL INFO Channel 23/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732983:740025 [1] NCCL INFO Channel 23/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:740024 [4] NCCL INFO Channel 03/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:740024 [4] NCCL INFO Channel 04/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:740021 [6] NCCL INFO Channel 05/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:740024 [4] NCCL INFO Channel 05/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:740021 [6] NCCL INFO Channel 06/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:740024 [4] NCCL INFO Channel 06/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:740021 [6] NCCL INFO Channel 07/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:740024 [4] NCCL INFO Channel 07/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:740024 [4] NCCL INFO Channel 08/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:740021 [6] NCCL INFO Channel 08/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:740024 [4] NCCL INFO Channel 09/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:740021 [6] NCCL INFO Channel 09/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:740024 [4] NCCL INFO Channel 10/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:740024 [4] NCCL INFO Channel 11/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:740021 [6] NCCL INFO Channel 10/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:740024 [4] NCCL INFO Channel 12/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:740021 [6] NCCL INFO Channel 11/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:740024 [4] NCCL INFO Channel 13/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:740021 [6] NCCL INFO Channel 12/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:740024 [4] NCCL INFO Channel 14/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:740021 [6] NCCL INFO Channel 13/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:740024 [4] NCCL INFO Channel 15/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:740021 [6] NCCL INFO Channel 14/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:740024 [4] NCCL INFO Channel 16/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:740021 [6] NCCL INFO Channel 15/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:740024 [4] NCCL INFO Channel 17/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:740021 [6] NCCL INFO Channel 16/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:740024 [4] NCCL INFO Channel 18/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:740021 [6] NCCL INFO Channel 17/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:740024 [4] NCCL INFO Channel 19/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:740021 [6] NCCL INFO Channel 18/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:740024 [4] NCCL INFO Channel 20/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:740021 [6] NCCL INFO Channel 19/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:740024 [4] NCCL INFO Channel 21/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:740021 [6] NCCL INFO Channel 20/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:740024 [4] NCCL INFO Channel 22/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:740021 [6] NCCL INFO Channel 21/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732986:740024 [4] NCCL INFO Channel 23/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:740021 [6] NCCL INFO Channel 22/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732988:740021 [6] NCCL INFO Channel 23/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:732982:740019 [0] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:732982:740019 [0] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:732982:740019 [0] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:732983:740025 [1] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:732983:740025 [1] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:732983:740025 [1] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:732984:740022 [2] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:732984:740022 [2] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:732984:740022 [2] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:732985:740020 [3] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:732985:740020 [3] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:732985:740020 [3] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:732986:740024 [4] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:732986:740024 [4] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:732986:740024 [4] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:732987:740026 [5] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:732987:740026 [5] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:732987:740026 [5] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:732989:740023 [7] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:732989:740023 [7] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:732989:740023 [7] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:732988:740021 [6] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:732988:740021 [6] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:732988:740021 [6] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:732982:740019 [0] NCCL INFO ncclCommInitRank comm 0x7fc52006b960 rank 0 nranks 8 cudaDev 0 nvmlDev 0 busId 101c0 commId 0x71ccc5801c8491c1 - Init COMPLETE +ywang29-vrdb-test2-worker-0:732986:740024 [4] NCCL INFO ncclCommInitRank comm 0x7fa3d006ac30 rank 4 nranks 8 cudaDev 4 nvmlDev 4 busId 901c0 commId 0x71ccc5801c8491c1 - Init COMPLETE +ywang29-vrdb-test2-worker-0:732988:740021 [6] NCCL INFO ncclCommInitRank comm 0x7f6bac06af80 rank 6 nranks 8 cudaDev 6 nvmlDev 6 busId a01c0 commId 0x71ccc5801c8491c1 - Init COMPLETE +ywang29-vrdb-test2-worker-0:732984:740022 [2] NCCL INFO ncclCommInitRank comm 0x7f422c06ac70 rank 2 nranks 8 cudaDev 2 nvmlDev 2 busId 201c0 commId 0x71ccc5801c8491c1 - Init COMPLETE +ywang29-vrdb-test2-worker-0:732989:740023 [7] NCCL INFO ncclCommInitRank comm 0x7f399c06aeb0 rank 7 nranks 8 cudaDev 7 nvmlDev 7 busId a01d0 commId 0x71ccc5801c8491c1 - Init COMPLETE +ywang29-vrdb-test2-worker-0:732987:740026 [5] NCCL INFO ncclCommInitRank comm 0x7fc0b006ab20 rank 5 nranks 8 cudaDev 5 nvmlDev 5 busId 901d0 commId 0x71ccc5801c8491c1 - Init COMPLETE +ywang29-vrdb-test2-worker-0:732985:740020 [3] NCCL INFO ncclCommInitRank comm 0x7f67b006b1a0 rank 3 nranks 8 cudaDev 3 nvmlDev 3 busId 201d0 commId 0x71ccc5801c8491c1 - Init COMPLETE +ywang29-vrdb-test2-worker-0:732983:740025 [1] NCCL INFO ncclCommInitRank comm 0x7f196406ac80 rank 1 nranks 8 cudaDev 1 nvmlDev 1 busId 101d0 commId 0x71ccc5801c8491c1 - Init COMPLETE + 0%| | 1/520 [00:13<1:59:56, 13.87s/it] {'loss': 2.0453, 'grad_norm': 0.004834356896382741, 'learning_rate': 0.004375, 'epoch': 0.0} + 0%| | 1/520 [00:13<1:59:56, 13.87s/it] 0%| | 2/520 [00:17<1:07:50, 7.86s/it] {'loss': 2.0549, 'grad_norm': 0.005249049061747471, 'learning_rate': 0.00875, 'epoch': 0.0} + 0%| | 2/520 [00:17<1:07:50, 7.86s/it] 1%| | 3/520 [00:21<51:07, 5.93s/it] {'loss': 2.1899, 'grad_norm': 0.006006508918899832, 'learning_rate': 0.013125000000000001, 'epoch': 0.01} + 1%| | 3/520 [00:21<51:07, 5.93s/it] 1%| | 4/520 [00:24<43:53, 5.10s/it] {'loss': 2.0656, 'grad_norm': 0.004962891911939091, 'learning_rate': 0.0175, 'epoch': 0.01} + 1%| | 4/520 [00:24<43:53, 5.10s/it] 1%| | 5/520 [00:28<39:52, 4.65s/it] {'loss': 2.2333, 'grad_norm': 0.005481623737507595, 'learning_rate': 0.021875000000000002, 'epoch': 0.01} + 1%| | 5/520 [00:28<39:52, 4.65s/it] 1%| | 6/520 [00:32<37:16, 4.35s/it] {'loss': 1.6754, 'grad_norm': 0.0028027712695852954, 'learning_rate': 0.026250000000000002, 'epoch': 0.01} + 1%| | 6/520 [00:32<37:16, 4.35s/it] 1%|▏ | 7/520 [00:36<35:04, 4.10s/it] {'loss': 2.0776, 'grad_norm': 0.005415654228103603, 'learning_rate': 0.030625000000000003, 'epoch': 0.01} + 1%|▏ | 7/520 [00:36<35:04, 4.10s/it] 2%|▏ | 8/520 [00:40<35:23, 4.15s/it] {'loss': 1.7605, 'grad_norm': 0.002180944237581593, 'learning_rate': 0.035, 'epoch': 0.02} + 2%|▏ | 8/520 [00:40<35:23, 4.15s/it] 2%|▏ | 9/520 [00:44<35:24, 4.16s/it] {'loss': 1.7354, 'grad_norm': 0.0018893325780273997, 'learning_rate': 0.03937500000000001, 'epoch': 0.02} + 2%|▏ | 9/520 [00:44<35:24, 4.16s/it] 2%|▏ | 10/520 [00:48<33:57, 4.00s/it] {'loss': 1.5628, 'grad_norm': 0.0012196611497503325, 'learning_rate': 0.043750000000000004, 'epoch': 0.02} + 2%|▏ | 10/520 [00:48<33:57, 4.00s/it] 2%|▏ | 11/520 [00:52<33:19, 3.93s/it] {'loss': 1.5441, 'grad_norm': 0.000696312943998993, 'learning_rate': 0.048125, 'epoch': 0.02} + 2%|▏ | 11/520 [00:52<33:19, 3.93s/it] 2%|▏ | 12/520 [00:55<32:27, 3.83s/it] {'loss': 1.4076, 'grad_norm': 0.0005106515413708146, 'learning_rate': 0.052500000000000005, 'epoch': 0.02} + 2%|▏ | 12/520 [00:55<32:27, 3.83s/it][2025-10-12 13:37:54,539] [WARNING] [stage3.py:2069:step] 1 pytorch allocator cache flushes since last step. this happens when there is high memory pressure and is detrimental to performance. if this is happening frequently consider adjusting settings to reduce memory consumption. If you are unable to make the cache flushes go away consider adding get_accelerator().empty_cache() calls in your training loop to ensure that all ranks flush their caches at the same time + 2%|▎ | 13/520 [01:00<33:43, 3.99s/it] {'loss': 1.5096, 'grad_norm': 0.0005569688397321211, 'learning_rate': 0.05687500000000001, 'epoch': 0.03} + 2%|▎ | 13/520 [01:00<33:43, 3.99s/it] 3%|▎ | 14/520 [01:03<32:50, 3.89s/it] {'loss': 1.531, 'grad_norm': 0.00047046215297355044, 'learning_rate': 0.061250000000000006, 'epoch': 0.03} + 3%|▎ | 14/520 [01:03<32:50, 3.89s/it] 3%|▎ | 15/520 [01:07<32:37, 3.88s/it] {'loss': 1.4437, 'grad_norm': 0.0003981479167569913, 'learning_rate': 0.065625, 'epoch': 0.03} + 3%|▎ | 15/520 [01:07<32:37, 3.88s/it] 3%|▎ | 16/520 [01:11<32:26, 3.86s/it] {'loss': 1.4086, 'grad_norm': 0.00039377689467717705, 'learning_rate': 0.07, 'epoch': 0.03} + 3%|▎ | 16/520 [01:11<32:26, 3.86s/it] 3%|▎ | 17/520 [01:15<32:24, 3.87s/it] {'loss': 1.5574, 'grad_norm': 0.0004061421117402763, 'learning_rate': 0.06999932005260384, 'epoch': 0.03} + 3%|▎ | 17/520 [01:15<32:24, 3.87s/it] 3%|▎ | 18/520 [01:18<32:00, 3.83s/it] {'loss': 1.4186, 'grad_norm': 0.0004195048327887794, 'learning_rate': 0.06999728023683412, 'epoch': 0.03} + 3%|▎ | 18/520 [01:18<32:00, 3.83s/it] 4%|▎ | 19/520 [01:22<31:41, 3.80s/it] {'loss': 1.3935, 'grad_norm': 0.0003532360334581685, 'learning_rate': 0.06999388063194613, 'epoch': 0.04} + 4%|▎ | 19/520 [01:22<31:41, 3.80s/it] 4%|▍ | 20/520 [01:26<31:10, 3.74s/it] {'loss': 1.3788, 'grad_norm': 0.00041849389714556745, 'learning_rate': 0.06998912137002858, 'epoch': 0.04} + 4%|▍ | 20/520 [01:26<31:10, 3.74s/it] 4%|▍ | 21/520 [01:29<30:58, 3.72s/it] {'loss': 1.3828, 'grad_norm': 0.000426806821921547, 'learning_rate': 0.06998300263599846, 'epoch': 0.04} + 4%|▍ | 21/520 [01:29<30:58, 3.72s/it] 4%|▍ | 22/520 [01:33<30:41, 3.70s/it] {'loss': 1.5069, 'grad_norm': 0.00043992207074016235, 'learning_rate': 0.06997552466759396, 'epoch': 0.04} + 4%|▍ | 22/520 [01:33<30:41, 3.70s/it] 4%|▍ | 23/520 [01:37<30:30, 3.68s/it] {'loss': 1.4446, 'grad_norm': 0.00042073019512395255, 'learning_rate': 0.06996668775536503, 'epoch': 0.04} + 4%|▍ | 23/520 [01:37<30:30, 3.68s/it] 5%|▍ | 24/520 [01:40<30:29, 3.69s/it] {'loss': 1.359, 'grad_norm': 0.0004174658973996765, 'learning_rate': 0.06995649224266229, 'epoch': 0.05} + 5%|▍ | 24/520 [01:40<30:29, 3.69s/it] 5%|▍ | 25/520 [01:44<30:20, 3.68s/it] {'loss': 1.4716, 'grad_norm': 0.0005358599205597377, 'learning_rate': 0.06994493852562358, 'epoch': 0.05} + 5%|▍ | 25/520 [01:44<30:20, 3.68s/it] 5%|▌ | 26/520 [01:48<30:17, 3.68s/it] {'loss': 1.3877, 'grad_norm': 0.000455602395206332, 'learning_rate': 0.06993202705315862, 'epoch': 0.05} + 5%|▌ | 26/520 [01:48<30:17, 3.68s/it] 5%|▌ | 27/520 [01:51<30:05, 3.66s/it] {'loss': 1.3208, 'grad_norm': 0.0004439372958797529, 'learning_rate': 0.06991775832693152, 'epoch': 0.05} + 5%|▌ | 27/520 [01:51<30:05, 3.66s/it] 5%|▌ | 28/520 [01:55<30:10, 3.68s/it] {'loss': 1.3547, 'grad_norm': 0.00047985368616899595, 'learning_rate': 0.06990213290134131, 'epoch': 0.05} + 5%|▌ | 28/520 [01:55<30:10, 3.68s/it] 6%|▌ | 29/520 [01:59<30:28, 3.72s/it] {'loss': 1.3642, 'grad_norm': 0.00048176106038920024, 'learning_rate': 0.06988515138350043, 'epoch': 0.06} + 6%|▌ | 29/520 [01:59<30:28, 3.72s/it] 6%|▌ | 30/520 [02:03<30:42, 3.76s/it] {'loss': 1.4212, 'grad_norm': 0.00043178559362331295, 'learning_rate': 0.0698668144332111, 'epoch': 0.06} + 6%|▌ | 30/520 [02:03<30:42, 3.76s/it] 6%|▌ | 31/520 [02:07<30:46, 3.78s/it] {'loss': 1.3234, 'grad_norm': 0.00042983643025928177, 'learning_rate': 0.06984712276293968, 'epoch': 0.06} + 6%|▌ | 31/520 [02:07<30:46, 3.78s/it] 6%|▌ | 32/520 [02:10<30:50, 3.79s/it] {'loss': 1.229, 'grad_norm': 0.00045616262412171625, 'learning_rate': 0.06982607713778906, 'epoch': 0.06} + 6%|▌ | 32/520 [02:10<30:50, 3.79s/it] 6%|▋ | 33/520 [02:14<30:45, 3.79s/it] {'loss': 1.3205, 'grad_norm': 0.0005203430349258656, 'learning_rate': 0.0698036783754688, 'epoch': 0.06} + 6%|▋ | 33/520 [02:14<30:45, 3.79s/it] 7%|▋ | 34/520 [02:18<30:24, 3.75s/it] {'loss': 1.3178, 'grad_norm': 0.0005399549396249944, 'learning_rate': 0.0697799273462635, 'epoch': 0.07} + 7%|▋ | 34/520 [02:18<30:24, 3.75s/it] 7%|▋ | 35/520 [02:22<30:12, 3.74s/it] {'loss': 1.3282, 'grad_norm': 0.0005875616549042669, 'learning_rate': 0.06975482497299888, 'epoch': 0.07} + 7%|▋ | 35/520 [02:22<30:12, 3.74s/it] 7%|▋ | 36/520 [02:25<29:55, 3.71s/it] {'loss': 1.4203, 'grad_norm': 0.0004944188557143451, 'learning_rate': 0.06972837223100603, 'epoch': 0.07} + 7%|▋ | 36/520 [02:25<29:55, 3.71s/it] 7%|▋ | 37/520 [02:29<29:44, 3.69s/it] {'loss': 1.3804, 'grad_norm': 0.0004584859012879818, 'learning_rate': 0.06970057014808337, 'epoch': 0.07} + 7%|▋ | 37/520 [02:29<29:44, 3.69s/it] 7%|▋ | 38/520 [02:33<29:33, 3.68s/it] {'loss': 1.4796, 'grad_norm': 0.0004832946154499517, 'learning_rate': 0.06967141980445685, 'epoch': 0.07} + 7%|▋ | 38/520 [02:33<29:33, 3.68s/it] 8%|▊ | 39/520 [02:36<29:20, 3.66s/it] {'loss': 1.357, 'grad_norm': 0.0006239712940270828, 'learning_rate': 0.06964092233273791, 'epoch': 0.07} + 8%|▊ | 39/520 [02:36<29:20, 3.66s/it] 8%|▊ | 40/520 [02:40<29:13, 3.65s/it] {'loss': 1.3765, 'grad_norm': 0.0005051039886447539, 'learning_rate': 0.0696090789178795, 'epoch': 0.08} + 8%|▊ | 40/520 [02:40<29:13, 3.65s/it] 8%|▊ | 41/520 [02:43<29:11, 3.66s/it] {'loss': 1.3525, 'grad_norm': 0.0005366388134435708, 'learning_rate': 0.06957589079713002, 'epoch': 0.08} + 8%|▊ | 41/520 [02:43<29:11, 3.66s/it] 8%|▊ | 42/520 [02:47<29:14, 3.67s/it] {'loss': 1.3452, 'grad_norm': 0.0007054269661167032, 'learning_rate': 0.06954135925998524, 'epoch': 0.08} + 8%|▊ | 42/520 [02:47<29:14, 3.67s/it] 8%|▊ | 43/520 [02:51<29:37, 3.73s/it] {'loss': 1.2702, 'grad_norm': 0.000544685996215645, 'learning_rate': 0.06950548564813826, 'epoch': 0.08} + 8%|▊ | 43/520 [02:51<29:37, 3.73s/it] 8%|▊ | 44/520 [02:55<29:29, 3.72s/it] {'loss': 1.3674, 'grad_norm': 0.0005922807689630181, 'learning_rate': 0.0694682713554273, 'epoch': 0.08} + 8%|▊ | 44/520 [02:55<29:29, 3.72s/it] 9%|▊ | 45/520 [02:58<29:18, 3.70s/it] {'loss': 1.3705, 'grad_norm': 0.0007117812313558786, 'learning_rate': 0.06942971782778155, 'epoch': 0.09} + 9%|▊ | 45/520 [02:58<29:18, 3.70s/it] 9%|▉ | 46/520 [03:02<29:12, 3.70s/it] {'loss': 1.4043, 'grad_norm': 0.0006088396209007887, 'learning_rate': 0.0693898265631651, 'epoch': 0.09} + 9%|▉ | 46/520 [03:02<29:12, 3.70s/it] 9%|▉ | 47/520 [03:06<28:53, 3.67s/it] {'loss': 1.3281, 'grad_norm': 0.0006815757453356814, 'learning_rate': 0.06934859911151858, 'epoch': 0.09} + 9%|▉ | 47/520 [03:06<28:53, 3.67s/it] 9%|▉ | 48/520 [03:09<28:41, 3.65s/it] {'loss': 1.3508, 'grad_norm': 0.0009331937558406166, 'learning_rate': 0.06930603707469904, 'epoch': 0.09} + 9%|▉ | 48/520 [03:09<28:41, 3.65s/it] 9%|▉ | 49/520 [03:13<28:37, 3.65s/it] {'loss': 1.3667, 'grad_norm': 0.0007246329196436865, 'learning_rate': 0.0692621421064177, 'epoch': 0.09} + 9%|▉ | 49/520 [03:13<28:37, 3.65s/it] 10%|▉ | 50/520 [03:17<28:27, 3.63s/it] {'loss': 1.3624, 'grad_norm': 0.0006960810861336613, 'learning_rate': 0.06921691591217567, 'epoch': 0.1} + 10%|▉ | 50/520 [03:17<28:27, 3.63s/it] 10%|▉ | 51/520 [03:20<28:23, 3.63s/it] {'loss': 1.2928, 'grad_norm': 0.0007854266506720693, 'learning_rate': 0.06917036024919768, 'epoch': 0.1} + 10%|▉ | 51/520 [03:20<28:23, 3.63s/it] 10%|█ | 52/520 [03:24<28:18, 3.63s/it] {'loss': 1.4296, 'grad_norm': 0.0008218193274407449, 'learning_rate': 0.06912247692636384, 'epoch': 0.1} + 10%|█ | 52/520 [03:24<28:18, 3.63s/it] 10%|█ | 53/520 [03:27<28:16, 3.63s/it] {'loss': 1.3972, 'grad_norm': 0.0007462363886377517, 'learning_rate': 0.06907326780413932, 'epoch': 0.1} + 10%|█ | 53/520 [03:27<28:16, 3.63s/it] 10%|█ | 54/520 [03:31<28:10, 3.63s/it] {'loss': 1.3382, 'grad_norm': 0.0008018530358332771, 'learning_rate': 0.06902273479450212, 'epoch': 0.1} + 10%|█ | 54/520 [03:31<28:10, 3.63s/it] 11%|█ | 55/520 [03:35<28:11, 3.64s/it] {'loss': 1.2778, 'grad_norm': 0.0007596252235741617, 'learning_rate': 0.0689708798608687, 'epoch': 0.11} + 11%|█ | 55/520 [03:35<28:11, 3.64s/it] 11%|█ | 56/520 [03:38<28:10, 3.64s/it] {'loss': 1.4104, 'grad_norm': 0.0007573478965629791, 'learning_rate': 0.06891770501801774, 'epoch': 0.11} + 11%|█ | 56/520 [03:38<28:10, 3.64s/it] 11%|█ | 57/520 [03:42<28:05, 3.64s/it] {'loss': 1.2879, 'grad_norm': 0.0008232042396662973, 'learning_rate': 0.06886321233201188, 'epoch': 0.11} + 11%|█ | 57/520 [03:42<28:05, 3.64s/it] 11%|█ | 58/520 [03:46<27:59, 3.64s/it] {'loss': 1.4193, 'grad_norm': 0.0006778650627922748, 'learning_rate': 0.06880740392011739, 'epoch': 0.11} + 11%|█ | 58/520 [03:46<27:59, 3.64s/it] 11%|█▏ | 59/520 [03:49<28:04, 3.65s/it] {'loss': 1.2313, 'grad_norm': 0.0007639023713777867, 'learning_rate': 0.06875028195072198, 'epoch': 0.11} + 11%|█▏ | 59/520 [03:49<28:04, 3.65s/it] 12%|█▏ | 60/520 [03:53<27:56, 3.65s/it] {'loss': 1.3367, 'grad_norm': 0.0007782198364592023, 'learning_rate': 0.06869184864325042, 'epoch': 0.12} + 12%|█▏ | 60/520 [03:53<27:56, 3.65s/it] 12%|█▏ | 61/520 [03:57<27:52, 3.64s/it] {'loss': 1.3055, 'grad_norm': 0.0008342266590567714, 'learning_rate': 0.0686321062680785, 'epoch': 0.12} + 12%|█▏ | 61/520 [03:57<27:52, 3.64s/it] 12%|█▏ | 62/520 [04:00<27:49, 3.65s/it] {'loss': 1.3178, 'grad_norm': 0.0008892620125847733, 'learning_rate': 0.06857105714644458, 'epoch': 0.12} + 12%|█▏ | 62/520 [04:00<27:49, 3.65s/it] 12%|█▏ | 63/520 [04:04<27:44, 3.64s/it] {'loss': 1.3181, 'grad_norm': 0.0008777222150280926, 'learning_rate': 0.06850870365035965, 'epoch': 0.12} + 12%|█▏ | 63/520 [04:04<27:44, 3.64s/it] 12%|█▏ | 64/520 [04:08<27:49, 3.66s/it] {'loss': 1.3376, 'grad_norm': 0.0008677211101743564, 'learning_rate': 0.06844504820251493, 'epoch': 0.12} + 12%|█▏ | 64/520 [04:08<27:49, 3.66s/it] 12%|█▎ | 65/520 [04:11<28:14, 3.72s/it] {'loss': 1.344, 'grad_norm': 0.0009413017685872237, 'learning_rate': 0.06838009327618795, 'epoch': 0.12} + 12%|█▎ | 65/520 [04:11<28:14, 3.72s/it] 13%|█▎ | 66/520 [04:15<28:05, 3.71s/it] {'loss': 1.2815, 'grad_norm': 0.000801938679099666, 'learning_rate': 0.06831384139514629, 'epoch': 0.13} + 13%|█▎ | 66/520 [04:15<28:05, 3.71s/it] 13%|█▎ | 67/520 [04:19<27:50, 3.69s/it] {'loss': 1.2093, 'grad_norm': 0.0009084495883622171, 'learning_rate': 0.06824629513354961, 'epoch': 0.13} + 13%|█▎ | 67/520 [04:19<27:50, 3.69s/it] 13%|█▎ | 68/520 [04:22<27:48, 3.69s/it] {'loss': 1.2728, 'grad_norm': 0.0008484524462980759, 'learning_rate': 0.06817745711584962, 'epoch': 0.13} + 13%|█▎ | 68/520 [04:22<27:48, 3.69s/it] 13%|█▎ | 69/520 [04:26<27:35, 3.67s/it] {'loss': 1.2504, 'grad_norm': 0.0009371746350629662, 'learning_rate': 0.06810733001668806, 'epoch': 0.13} + 13%|█▎ | 69/520 [04:26<27:35, 3.67s/it] 13%|█▎ | 70/520 [04:30<27:35, 3.68s/it] {'loss': 1.2854, 'grad_norm': 0.0010365787853363635, 'learning_rate': 0.06803591656079287, 'epoch': 0.13} + 13%|█▎ | 70/520 [04:30<27:35, 3.68s/it] 14%|█▎ | 71/520 [04:33<27:33, 3.68s/it] {'loss': 1.2247, 'grad_norm': 0.0008907851573921511, 'learning_rate': 0.06796321952287222, 'epoch': 0.14} + 14%|█▎ | 71/520 [04:33<27:33, 3.68s/it] 14%|█▍ | 72/520 [04:37<27:54, 3.74s/it] {'loss': 1.3668, 'grad_norm': 0.0008948726671267764, 'learning_rate': 0.06788924172750681, 'epoch': 0.14} + 14%|█▍ | 72/520 [04:37<27:54, 3.74s/it] 14%|█▍ | 73/520 [04:41<27:51, 3.74s/it] {'loss': 1.196, 'grad_norm': 0.0009765736044038474, 'learning_rate': 0.06781398604903999, 'epoch': 0.14} + 14%|█▍ | 73/520 [04:41<27:51, 3.74s/it] 14%|█▍ | 74/520 [04:45<27:33, 3.71s/it] {'loss': 1.2961, 'grad_norm': 0.0009915304938356527, 'learning_rate': 0.0677374554114662, 'epoch': 0.14} + 14%|█▍ | 74/520 [04:45<27:33, 3.71s/it] 14%|█▍ | 75/520 [04:48<27:25, 3.70s/it] {'loss': 1.2197, 'grad_norm': 0.0008743241094896198, 'learning_rate': 0.06765965278831733, 'epoch': 0.14} + 14%|█▍ | 75/520 [04:48<27:25, 3.70s/it] 15%|█▍ | 76/520 [04:52<27:19, 3.69s/it] {'loss': 1.335, 'grad_norm': 0.0008004269290104737, 'learning_rate': 0.06758058120254716, 'epoch': 0.15} + 15%|█▍ | 76/520 [04:52<27:19, 3.69s/it] 15%|█▍ | 77/520 [04:56<27:24, 3.71s/it] {'loss': 1.1531, 'grad_norm': 0.0010274371394073457, 'learning_rate': 0.06750024372641387, 'epoch': 0.15} + 15%|█▍ | 77/520 [04:56<27:24, 3.71s/it] 15%|█▌ | 78/520 [04:59<27:16, 3.70s/it] {'loss': 1.2519, 'grad_norm': 0.0009213397124293374, 'learning_rate': 0.06741864348136081, 'epoch': 0.15} + 15%|█▌ | 78/520 [04:59<27:16, 3.70s/it] 15%|█▌ | 79/520 [05:03<27:36, 3.76s/it] {'loss': 1.2358, 'grad_norm': 0.0009030210332278625, 'learning_rate': 0.06733578363789504, 'epoch': 0.15} + 15%|█▌ | 79/520 [05:03<27:36, 3.76s/it] 15%|█▌ | 80/520 [05:07<27:48, 3.79s/it] {'loss': 1.3119, 'grad_norm': 0.0009420161711685643, 'learning_rate': 0.06725166741546428, 'epoch': 0.15} + 15%|█▌ | 80/520 [05:07<27:48, 3.79s/it] 16%|█▌ | 81/520 [05:11<27:48, 3.80s/it] {'loss': 1.3778, 'grad_norm': 0.001223360794364062, 'learning_rate': 0.06716629808233172, 'epoch': 0.16} + 16%|█▌ | 81/520 [05:11<27:48, 3.80s/it] 16%|█▌ | 82/520 [05:15<27:30, 3.77s/it] {'loss': 1.3038, 'grad_norm': 0.0009561695722518595, 'learning_rate': 0.06707967895544917, 'epoch': 0.16} + 16%|█▌ | 82/520 [05:15<27:30, 3.77s/it] 16%|█▌ | 83/520 [05:18<27:11, 3.73s/it] {'loss': 1.3127, 'grad_norm': 0.001086212378757713, 'learning_rate': 0.06699181340032802, 'epoch': 0.16} + 16%|█▌ | 83/520 [05:18<27:11, 3.73s/it] 16%|█▌ | 84/520 [05:22<26:56, 3.71s/it] {'loss': 1.3308, 'grad_norm': 0.0011377989373741204, 'learning_rate': 0.06690270483090857, 'epoch': 0.16} + 16%|█▌ | 84/520 [05:22<26:56, 3.71s/it] 16%|█▋ | 85/520 [05:26<26:49, 3.70s/it] {'loss': 1.3686, 'grad_norm': 0.0010115163296713521, 'learning_rate': 0.0668123567094274, 'epoch': 0.16} + 16%|█▋ | 85/520 [05:26<26:49, 3.70s/it] 17%|█▋ | 86/520 [05:29<26:36, 3.68s/it] {'loss': 1.3532, 'grad_norm': 0.0009788853501928232, 'learning_rate': 0.06672077254628275, 'epoch': 0.17} + 17%|█▋ | 86/520 [05:29<26:36, 3.68s/it] 17%|█▋ | 87/520 [05:33<26:34, 3.68s/it] {'loss': 1.2687, 'grad_norm': 0.0009573372309124434, 'learning_rate': 0.0666279558998983, 'epoch': 0.17} + 17%|█▋ | 87/520 [05:33<26:34, 3.68s/it] 17%|█▋ | 88/520 [05:37<26:27, 3.67s/it] {'loss': 1.2004, 'grad_norm': 0.0008004271361894321, 'learning_rate': 0.06653391037658467, 'epoch': 0.17} + 17%|█▋ | 88/520 [05:37<26:27, 3.67s/it] 17%|█▋ | 89/520 [05:40<26:20, 3.67s/it] {'loss': 1.3115, 'grad_norm': 0.0010687733125116872, 'learning_rate': 0.06643863963039955, 'epoch': 0.17} + 17%|█▋ | 89/520 [05:40<26:20, 3.67s/it] 17%|█▋ | 90/520 [05:44<26:19, 3.67s/it] {'loss': 1.2506, 'grad_norm': 0.001071644680940967, 'learning_rate': 0.06634214736300553, 'epoch': 0.17} + 17%|█▋ | 90/520 [05:44<26:19, 3.67s/it] 18%|█▊ | 91/520 [05:48<26:09, 3.66s/it] {'loss': 1.314, 'grad_norm': 0.0009601565128910398, 'learning_rate': 0.0662444373235264, 'epoch': 0.17} + 18%|█▊ | 91/520 [05:48<26:09, 3.66s/it] 18%|█▊ | 92/520 [05:51<25:58, 3.64s/it] {'loss': 1.2513, 'grad_norm': 0.0010662111356706987, 'learning_rate': 0.06614551330840142, 'epoch': 0.18} + 18%|█▊ | 92/520 [05:51<25:58, 3.64s/it] 18%|█▊ | 93/520 [05:55<26:02, 3.66s/it] {'loss': 1.2655, 'grad_norm': 0.0011015224142611945, 'learning_rate': 0.06604537916123776, 'epoch': 0.18} + 18%|█▊ | 93/520 [05:55<26:02, 3.66s/it] 18%|█▊ | 94/520 [05:59<26:18, 3.70s/it] {'loss': 1.3411, 'grad_norm': 0.0010828732946658313, 'learning_rate': 0.06594403877266135, 'epoch': 0.18} + 18%|█▊ | 94/520 [05:59<26:18, 3.70s/it] 18%|█▊ | 95/520 [06:03<26:21, 3.72s/it] {'loss': 1.2511, 'grad_norm': 0.0012512408333559973, 'learning_rate': 0.06584149608016548, 'epoch': 0.18} + 18%|█▊ | 95/520 [06:03<26:21, 3.72s/it] 18%|█▊ | 96/520 [06:06<26:22, 3.73s/it] {'loss': 1.2695, 'grad_norm': 0.0009659740368205414, 'learning_rate': 0.06573775506795801, 'epoch': 0.18} + 18%|█▊ | 96/520 [06:06<26:22, 3.73s/it] 19%|█▊ | 97/520 [06:10<26:15, 3.73s/it] {'loss': 1.2433, 'grad_norm': 0.0012238577480780256, 'learning_rate': 0.06563281976680639, 'epoch': 0.19} + 19%|█▊ | 97/520 [06:10<26:15, 3.73s/it] 19%|█▉ | 98/520 [06:14<26:01, 3.70s/it] {'loss': 1.2419, 'grad_norm': 0.0009185350065593367, 'learning_rate': 0.06552669425388119, 'epoch': 0.19} + 19%|█▉ | 98/520 [06:14<26:01, 3.70s/it] 19%|█▉ | 99/520 [06:17<25:56, 3.70s/it] {'loss': 1.2345, 'grad_norm': 0.0011116045404477085, 'learning_rate': 0.06541938265259763, 'epoch': 0.19} + 19%|█▉ | 99/520 [06:17<25:56, 3.70s/it] 19%|█▉ | 100/520 [06:21<25:45, 3.68s/it] {'loss': 1.2213, 'grad_norm': 0.001019380950934114, 'learning_rate': 0.06531088913245536, 'epoch': 0.19} + 19%|█▉ | 100/520 [06:21<25:45, 3.68s/it] 19%|█▉ | 101/520 [06:25<25:36, 3.67s/it] {'loss': 1.2669, 'grad_norm': 0.0011300100926507704, 'learning_rate': 0.06520121790887647, 'epoch': 0.19} + 19%|█▉ | 101/520 [06:25<25:36, 3.67s/it] 20%|█▉ | 102/520 [06:28<25:28, 3.66s/it] {'loss': 1.2572, 'grad_norm': 0.001145426110789352, 'learning_rate': 0.06509037324304166, 'epoch': 0.2} + 20%|█▉ | 102/520 [06:28<25:28, 3.66s/it] 20%|█▉ | 103/520 [06:32<25:17, 3.64s/it] {'loss': 1.2075, 'grad_norm': 0.0010298617693041052, 'learning_rate': 0.06497835944172481, 'epoch': 0.2} + 20%|█▉ | 103/520 [06:32<25:17, 3.64s/it] 20%|██ | 104/520 [06:35<25:13, 3.64s/it] {'loss': 1.2725, 'grad_norm': 0.001146346625466391, 'learning_rate': 0.06486518085712545, 'epoch': 0.2} + 20%|██ | 104/520 [06:35<25:13, 3.64s/it] 20%|██ | 105/520 [06:39<25:08, 3.64s/it] {'loss': 1.2622, 'grad_norm': 0.0010403794885597095, 'learning_rate': 0.06475084188669983, 'epoch': 0.2} + 20%|██ | 105/520 [06:39<25:08, 3.64s/it] 20%|██ | 106/520 [06:43<25:05, 3.64s/it] {'loss': 1.2404, 'grad_norm': 0.0009688089039834874, 'learning_rate': 0.06463534697298995, 'epoch': 0.2} + 20%|██ | 106/520 [06:43<25:05, 3.64s/it] 21%|██ | 107/520 [06:46<25:02, 3.64s/it] {'loss': 1.2162, 'grad_norm': 0.0010271530692141663, 'learning_rate': 0.06451870060345101, 'epoch': 0.21} + 21%|██ | 107/520 [06:46<25:02, 3.64s/it] 21%|██ | 108/520 [06:50<25:07, 3.66s/it] {'loss': 1.2184, 'grad_norm': 0.001180934795646103, 'learning_rate': 0.06440090731027702, 'epoch': 0.21} + 21%|██ | 108/520 [06:50<25:07, 3.66s/it] 21%|██ | 109/520 [06:54<25:08, 3.67s/it] {'loss': 1.1904, 'grad_norm': 0.0009537696544875849, 'learning_rate': 0.0642819716702247, 'epoch': 0.21} + 21%|██ | 109/520 [06:54<25:08, 3.67s/it] 21%|██ | 110/520 [06:57<25:07, 3.68s/it] {'loss': 1.3873, 'grad_norm': 0.0011325895011174527, 'learning_rate': 0.06416189830443572, 'epoch': 0.21} + 21%|██ | 110/520 [06:57<25:07, 3.68s/it] 21%|██▏ | 111/520 [07:01<25:27, 3.73s/it] {'loss': 1.3843, 'grad_norm': 0.0011717757503984108, 'learning_rate': 0.06404069187825707, 'epoch': 0.21} + 21%|██▏ | 111/520 [07:01<25:27, 3.73s/it] 22%|██▏ | 112/520 [07:05<25:39, 3.77s/it] {'loss': 1.2777, 'grad_norm': 0.0010954499260405824, 'learning_rate': 0.06391835710105982, 'epoch': 0.22} + 22%|██▏ | 112/520 [07:05<25:39, 3.77s/it] 22%|██▏ | 113/520 [07:09<25:45, 3.80s/it] {'loss': 1.1789, 'grad_norm': 0.0010696143062359777, 'learning_rate': 0.06379489872605618, 'epoch': 0.22} + 22%|██▏ | 113/520 [07:09<25:45, 3.80s/it] 22%|██▏ | 114/520 [07:13<25:49, 3.82s/it] {'loss': 1.2747, 'grad_norm': 0.0011188732182217053, 'learning_rate': 0.06367032155011472, 'epoch': 0.22} + 22%|██▏ | 114/520 [07:13<25:49, 3.82s/it] 22%|██▏ | 115/520 [07:17<25:41, 3.81s/it] {'loss': 1.3468, 'grad_norm': 0.0010908421444297016, 'learning_rate': 0.06354463041357411, 'epoch': 0.22} + 22%|██▏ | 115/520 [07:17<25:41, 3.81s/it] 22%|██▏ | 116/520 [07:20<25:24, 3.77s/it] {'loss': 1.3688, 'grad_norm': 0.001050602421465488, 'learning_rate': 0.063417830200055, 'epoch': 0.22} + 22%|██▏ | 116/520 [07:20<25:24, 3.77s/it] 22%|██▎ | 117/520 [07:24<25:12, 3.75s/it] {'loss': 1.3291, 'grad_norm': 0.0011494168444789014, 'learning_rate': 0.06328992583627019, 'epoch': 0.23} + 22%|██▎ | 117/520 [07:24<25:12, 3.75s/it] 23%|██▎ | 118/520 [07:28<24:56, 3.72s/it] {'loss': 1.267, 'grad_norm': 0.0010459887954407515, 'learning_rate': 0.0631609222918334, 'epoch': 0.23} + 23%|██▎ | 118/520 [07:28<24:56, 3.72s/it] 23%|██▎ | 119/520 [07:31<24:46, 3.71s/it] {'loss': 1.2175, 'grad_norm': 0.0011465730398795364, 'learning_rate': 0.06303082457906599, 'epoch': 0.23} + 23%|██▎ | 119/520 [07:31<24:46, 3.71s/it] 23%|██▎ | 120/520 [07:35<24:38, 3.70s/it] {'loss': 1.2294, 'grad_norm': 0.001342006178296207, 'learning_rate': 0.06289963775280229, 'epoch': 0.23} + 23%|██▎ | 120/520 [07:35<24:38, 3.70s/it] 23%|██▎ | 121/520 [07:39<24:28, 3.68s/it] {'loss': 1.273, 'grad_norm': 0.0011629490939945126, 'learning_rate': 0.06276736691019323, 'epoch': 0.23} + 23%|██▎ | 121/520 [07:39<24:28, 3.68s/it] 23%|██▎ | 122/520 [07:42<24:21, 3.67s/it] {'loss': 1.197, 'grad_norm': 0.0010898778170924655, 'learning_rate': 0.06263401719050825, 'epoch': 0.23} + 23%|██▎ | 122/520 [07:42<24:21, 3.67s/it] 24%|██▎ | 123/520 [07:46<24:17, 3.67s/it] {'loss': 1.2721, 'grad_norm': 0.001053013990898796, 'learning_rate': 0.06249959377493559, 'epoch': 0.24} + 24%|██▎ | 123/520 [07:46<24:17, 3.67s/it] 24%|██▍ | 124/520 [07:50<24:10, 3.66s/it] {'loss': 1.2405, 'grad_norm': 0.0012564982804835656, 'learning_rate': 0.06236410188638105, 'epoch': 0.24} + 24%|██▍ | 124/520 [07:50<24:10, 3.66s/it] 24%|██▍ | 125/520 [07:53<24:14, 3.68s/it] {'loss': 1.2401, 'grad_norm': 0.00114884298545511, 'learning_rate': 0.06222754678926502, 'epoch': 0.24} + 24%|██▍ | 125/520 [07:53<24:14, 3.68s/it] 24%|██▍ | 126/520 [07:58<25:29, 3.88s/it] {'loss': 1.204, 'grad_norm': 0.000922537808275188, 'learning_rate': 0.062089933789317965, 'epoch': 0.24} + 24%|██▍ | 126/520 [07:58<25:29, 3.88s/it] 24%|██▍ | 127/520 [08:01<25:01, 3.82s/it] {'loss': 1.2323, 'grad_norm': 0.0013482239199936728, 'learning_rate': 0.061951268233374206, 'epoch': 0.24} + 24%|██▍ | 127/520 [08:01<25:01, 3.82s/it] 25%|██▍ | 128/520 [08:05<24:39, 3.77s/it] {'loss': 1.2631, 'grad_norm': 0.0011350697944410924, 'learning_rate': 0.061811555509164236, 'epoch': 0.25} + 25%|██▍ | 128/520 [08:05<24:39, 3.77s/it] 25%|██▍ | 129/520 [08:09<24:18, 3.73s/it] {'loss': 1.2392, 'grad_norm': 0.0010594607206743266, 'learning_rate': 0.06167080104510538, 'epoch': 0.25} + 25%|██▍ | 129/520 [08:09<24:18, 3.73s/it] 25%|██▌ | 130/520 [08:12<24:03, 3.70s/it] {'loss': 1.2592, 'grad_norm': 0.0010838614564077718, 'learning_rate': 0.061529010310090865, 'epoch': 0.25} + 25%|██▌ | 130/520 [08:12<24:03, 3.70s/it] 25%|██▌ | 131/520 [08:16<23:53, 3.68s/it] {'loss': 1.1743, 'grad_norm': 0.0009858605392951626, 'learning_rate': 0.0613861888132773, 'epoch': 0.25} + 25%|██▌ | 131/520 [08:16<23:53, 3.68s/it] 25%|██▌ | 132/520 [08:20<23:45, 3.67s/it] {'loss': 1.3108, 'grad_norm': 0.0013008304351199936, 'learning_rate': 0.0612423421038707, 'epoch': 0.25} + 25%|██▌ | 132/520 [08:20<23:45, 3.67s/it] 26%|██▌ | 133/520 [08:23<23:39, 3.67s/it] {'loss': 1.2217, 'grad_norm': 0.0012964059970402167, 'learning_rate': 0.0610974757709108, 'epoch': 0.26} + 26%|██▌ | 133/520 [08:23<23:39, 3.67s/it] 26%|██▌ | 134/520 [08:27<23:29, 3.65s/it] {'loss': 1.2916, 'grad_norm': 0.001178755323095066, 'learning_rate': 0.06095159544305394, 'epoch': 0.26} + 26%|██▌ | 134/520 [08:27<23:29, 3.65s/it] 26%|██▌ | 135/520 [08:31<23:25, 3.65s/it] {'loss': 1.3435, 'grad_norm': 0.0011764911025929435, 'learning_rate': 0.06080470678835435, 'epoch': 0.26} + 26%|██▌ | 135/520 [08:31<23:25, 3.65s/it] 26%|██▌ | 136/520 [08:34<23:20, 3.65s/it] {'loss': 1.3095, 'grad_norm': 0.0012549785872826198, 'learning_rate': 0.06065681551404392, 'epoch': 0.26} + 26%|██▌ | 136/520 [08:34<23:20, 3.65s/it] 26%|██▋ | 137/520 [08:38<23:14, 3.64s/it] {'loss': 1.2212, 'grad_norm': 0.0013157229006059406, 'learning_rate': 0.0605079273663105, 'epoch': 0.26} + 26%|██▋ | 137/520 [08:38<23:14, 3.64s/it] 27%|██▋ | 138/520 [08:42<23:27, 3.69s/it] {'loss': 1.229, 'grad_norm': 0.0010779869256269403, 'learning_rate': 0.060358048130074546, 'epoch': 0.27} + 27%|██▋ | 138/520 [08:42<23:27, 3.69s/it] 27%|██▋ | 139/520 [08:46<23:44, 3.74s/it] {'loss': 1.088, 'grad_norm': 0.001056947202614913, 'learning_rate': 0.060207183628764434, 'epoch': 0.27} + 27%|██▋ | 139/520 [08:46<23:44, 3.74s/it] 27%|██▋ | 140/520 [08:49<23:47, 3.76s/it] {'loss': 1.2294, 'grad_norm': 0.0010347441629184812, 'learning_rate': 0.06005533972409015, 'epoch': 0.27} + 27%|██▋ | 140/520 [08:49<23:47, 3.76s/it] 27%|██▋ | 141/520 [08:53<23:31, 3.73s/it] {'loss': 1.319, 'grad_norm': 0.0010719792363632448, 'learning_rate': 0.05990252231581556, 'epoch': 0.27} + 27%|██▋ | 141/520 [08:53<23:31, 3.73s/it] 27%|██▋ | 142/520 [08:57<23:23, 3.71s/it] {'loss': 1.2349, 'grad_norm': 0.0010236953818169778, 'learning_rate': 0.05974873734152917, 'epoch': 0.27} + 27%|██▋ | 142/520 [08:57<23:23, 3.71s/it] 28%|██▊ | 143/520 [09:00<23:21, 3.72s/it] {'loss': 1.2492, 'grad_norm': 0.0012452790964310682, 'learning_rate': 0.059593990776413426, 'epoch': 0.28} + 28%|██▊ | 143/520 [09:00<23:21, 3.72s/it] 28%|██▊ | 144/520 [09:04<23:11, 3.70s/it] {'loss': 1.2402, 'grad_norm': 0.001259518318415582, 'learning_rate': 0.05943828863301255, 'epoch': 0.28} + 28%|██▊ | 144/520 [09:04<23:11, 3.70s/it] 28%|██▊ | 145/520 [09:08<23:03, 3.69s/it] {'loss': 1.1582, 'grad_norm': 0.0011043503560172851, 'learning_rate': 0.05928163696099897, 'epoch': 0.28} + 28%|██▊ | 145/520 [09:08<23:03, 3.69s/it] 28%|██▊ | 146/520 [09:11<22:59, 3.69s/it] {'loss': 1.282, 'grad_norm': 0.0011639800310102029, 'learning_rate': 0.059124041846938154, 'epoch': 0.28} + 28%|██▊ | 146/520 [09:11<22:59, 3.69s/it] 28%|██▊ | 147/520 [09:15<22:51, 3.68s/it] {'loss': 1.2052, 'grad_norm': 0.0011936527781822026, 'learning_rate': 0.05896550941405228, 'epoch': 0.28} + 28%|██▊ | 147/520 [09:15<22:51, 3.68s/it] 28%|██▊ | 148/520 [09:19<22:43, 3.66s/it] {'loss': 1.2319, 'grad_norm': 0.0011471658746596504, 'learning_rate': 0.05880604582198218, 'epoch': 0.28} + 28%|██▊ | 148/520 [09:19<22:43, 3.66s/it] 29%|██▊ | 149/520 [09:22<22:37, 3.66s/it] {'loss': 1.1698, 'grad_norm': 0.0011627950938010985, 'learning_rate': 0.05864565726654812, 'epoch': 0.29} + 29%|██▊ | 149/520 [09:22<22:37, 3.66s/it] 29%|██▉ | 150/520 [09:26<22:31, 3.65s/it] {'loss': 1.3787, 'grad_norm': 0.001180835403648768, 'learning_rate': 0.05848434997950896, 'epoch': 0.29} + 29%|██▉ | 150/520 [09:26<22:31, 3.65s/it] 29%|██▉ | 151/520 [09:30<22:27, 3.65s/it] {'loss': 1.2217, 'grad_norm': 0.001225517744148505, 'learning_rate': 0.05832213022832014, 'epoch': 0.29} + 29%|██▉ | 151/520 [09:30<22:27, 3.65s/it] 29%|██▉ | 152/520 [09:33<22:21, 3.65s/it] {'loss': 1.1877, 'grad_norm': 0.001276531582717886, 'learning_rate': 0.058159004315890085, 'epoch': 0.29} + 29%|██▉ | 152/520 [09:33<22:21, 3.65s/it] 29%|██▉ | 153/520 [09:37<22:24, 3.66s/it] {'loss': 1.217, 'grad_norm': 0.0011737225759289563, 'learning_rate': 0.057994978580335324, 'epoch': 0.29} + 29%|██▉ | 153/520 [09:37<22:24, 3.66s/it] 30%|██▉ | 154/520 [09:41<22:22, 3.67s/it] {'loss': 1.2978, 'grad_norm': 0.0011072904695351789, 'learning_rate': 0.05783005939473425, 'epoch': 0.3} + 30%|██▉ | 154/520 [09:41<22:22, 3.67s/it] 30%|██▉ | 155/520 [09:44<22:17, 3.66s/it] {'loss': 1.2148, 'grad_norm': 0.0012632221275415433, 'learning_rate': 0.05766425316687947, 'epoch': 0.3} + 30%|██▉ | 155/520 [09:44<22:17, 3.66s/it] 30%|███ | 156/520 [09:48<22:17, 3.67s/it] {'loss': 1.2411, 'grad_norm': 0.0013244936850634501, 'learning_rate': 0.05749756633902888, 'epoch': 0.3} + 30%|███ | 156/520 [09:48<22:17, 3.67s/it] 30%|███ | 157/520 [09:52<22:29, 3.72s/it] {'loss': 1.2789, 'grad_norm': 0.0011359232391795445, 'learning_rate': 0.05733000538765531, 'epoch': 0.3} + 30%|███ | 157/520 [09:52<22:29, 3.72s/it] 30%|███ | 158/520 [09:56<22:39, 3.75s/it] {'loss': 1.225, 'grad_norm': 0.0012343114853006305, 'learning_rate': 0.0571615768231949, 'epoch': 0.3} + 30%|███ | 158/520 [09:56<22:39, 3.75s/it] 31%|███ | 159/520 [09:59<22:43, 3.78s/it] {'loss': 1.2706, 'grad_norm': 0.0012065327386661725, 'learning_rate': 0.05699228718979416, 'epoch': 0.31} + 31%|███ | 159/520 [09:59<22:43, 3.78s/it] 31%|███ | 160/520 [10:03<22:30, 3.75s/it] {'loss': 1.2756, 'grad_norm': 0.001253628355615387, 'learning_rate': 0.05682214306505568, 'epoch': 0.31} + 31%|███ | 160/520 [10:03<22:30, 3.75s/it] 31%|███ | 161/520 [10:07<22:18, 3.73s/it] {'loss': 1.2477, 'grad_norm': 0.0011908425044719704, 'learning_rate': 0.05665115105978259, 'epoch': 0.31} + 31%|███ | 161/520 [10:07<22:18, 3.73s/it] 31%|███ | 162/520 [10:11<22:10, 3.72s/it] {'loss': 1.2195, 'grad_norm': 0.0011348925690148402, 'learning_rate': 0.05647931781772167, 'epoch': 0.31} + 31%|███ | 162/520 [10:11<22:10, 3.72s/it] 31%|███▏ | 163/520 [10:14<21:59, 3.70s/it] {'loss': 1.1586, 'grad_norm': 0.0013671144848579606, 'learning_rate': 0.05630665001530523, 'epoch': 0.31} + 31%|███▏ | 163/520 [10:14<21:59, 3.70s/it] 32%|███▏ | 164/520 [10:18<21:51, 3.68s/it] {'loss': 1.1183, 'grad_norm': 0.0011714948598145277, 'learning_rate': 0.05613315436139171, 'epoch': 0.32} + 32%|███▏ | 164/520 [10:18<21:51, 3.68s/it] 32%|███▏ | 165/520 [10:22<21:45, 3.68s/it] {'loss': 1.2782, 'grad_norm': 0.0011860802233111868, 'learning_rate': 0.05595883759700502, 'epoch': 0.32} + 32%|███▏ | 165/520 [10:22<21:45, 3.68s/it] 32%|███▏ | 166/520 [10:25<21:38, 3.67s/it] {'loss': 1.2334, 'grad_norm': 0.0013107774111360043, 'learning_rate': 0.055783706495072555, 'epoch': 0.32} + 32%|███▏ | 166/520 [10:25<21:38, 3.67s/it] 32%|███▏ | 167/520 [10:29<21:36, 3.67s/it] {'loss': 1.2303, 'grad_norm': 0.0011995279489064304, 'learning_rate': 0.055607767860162166, 'epoch': 0.32} + 32%|███▏ | 167/520 [10:29<21:36, 3.67s/it] 32%|███▏ | 168/520 [10:32<21:28, 3.66s/it] {'loss': 1.1708, 'grad_norm': 0.0011790061593050995, 'learning_rate': 0.05543102852821765, 'epoch': 0.32} + 32%|███▏ | 168/520 [10:32<21:28, 3.66s/it] 32%|███▎ | 169/520 [10:36<21:24, 3.66s/it] {'loss': 1.2422, 'grad_norm': 0.0011433860269197504, 'learning_rate': 0.05525349536629321, 'epoch': 0.33} + 32%|███▎ | 169/520 [10:36<21:24, 3.66s/it] 33%|███▎ | 170/520 [10:40<21:22, 3.66s/it] {'loss': 1.1786, 'grad_norm': 0.0010099519493884264, 'learning_rate': 0.055075175272286614, 'epoch': 0.33} + 33%|███▎ | 170/520 [10:40<21:22, 3.66s/it] 33%|███▎ | 171/520 [10:43<21:17, 3.66s/it] {'loss': 1.1914, 'grad_norm': 0.0013019153279719222, 'learning_rate': 0.05489607517467124, 'epoch': 0.33} + 33%|███▎ | 171/520 [10:43<21:17, 3.66s/it] 33%|███▎ | 172/520 [10:47<21:11, 3.65s/it] {'loss': 1.2614, 'grad_norm': 0.001192786467516857, 'learning_rate': 0.054716202032226775, 'epoch': 0.33} + 33%|███▎ | 172/520 [10:47<21:11, 3.65s/it] 33%|███▎ | 173/520 [10:51<21:12, 3.67s/it] {'loss': 1.1907, 'grad_norm': 0.0012012640099000665, 'learning_rate': 0.05453556283376895, 'epoch': 0.33} + 33%|███▎ | 173/520 [10:51<21:12, 3.67s/it] 33%|███▎ | 174/520 [10:55<21:20, 3.70s/it] {'loss': 1.2446, 'grad_norm': 0.00125987020045839, 'learning_rate': 0.05435416459787788, 'epoch': 0.33} + 33%|███▎ | 174/520 [10:55<21:20, 3.70s/it] 34%|███▎ | 175/520 [10:58<21:19, 3.71s/it] {'loss': 1.1606, 'grad_norm': 0.0011520561503531059, 'learning_rate': 0.054172014372625504, 'epoch': 0.34} + 34%|███▎ | 175/520 [10:58<21:19, 3.71s/it] 34%|███▍ | 176/520 [11:02<21:06, 3.68s/it] {'loss': 1.2424, 'grad_norm': 0.0011844299818735617, 'learning_rate': 0.053989119235301586, 'epoch': 0.34} + 34%|███▍ | 176/520 [11:02<21:06, 3.68s/it] 34%|███▍ | 177/520 [11:06<20:59, 3.67s/it] {'loss': 1.132, 'grad_norm': 0.0012470132277429095, 'learning_rate': 0.05380548629213885, 'epoch': 0.34} + 34%|███▍ | 177/520 [11:06<20:59, 3.67s/it] 34%|███▍ | 178/520 [11:09<20:50, 3.66s/it] {'loss': 1.2303, 'grad_norm': 0.0013078253428381952, 'learning_rate': 0.05362112267803679, 'epoch': 0.34} + 34%|███▍ | 178/520 [11:09<20:50, 3.66s/it] 34%|███▍ | 179/520 [11:13<20:44, 3.65s/it] {'loss': 1.3022, 'grad_norm': 0.0011721606829448125, 'learning_rate': 0.053436035556284525, 'epoch': 0.34} + 34%|███▍ | 179/520 [11:13<20:44, 3.65s/it] 35%|███▍ | 180/520 [11:16<20:37, 3.64s/it] {'loss': 1.2193, 'grad_norm': 0.0012252337597409163, 'learning_rate': 0.053250232118282434, 'epoch': 0.35} + 35%|███▍ | 180/520 [11:16<20:37, 3.64s/it] 35%|███▍ | 181/520 [11:20<20:31, 3.63s/it] {'loss': 1.2029, 'grad_norm': 0.0011115748670379475, 'learning_rate': 0.05306371958326274, 'epoch': 0.35} + 35%|███▍ | 181/520 [11:20<20:31, 3.63s/it] 35%|███▌ | 182/520 [11:24<20:26, 3.63s/it] {'loss': 1.2134, 'grad_norm': 0.0012521344069158444, 'learning_rate': 0.05287650519800901, 'epoch': 0.35} + 35%|███▌ | 182/520 [11:24<20:26, 3.63s/it] 35%|███▌ | 183/520 [11:27<20:42, 3.69s/it] {'loss': 1.2286, 'grad_norm': 0.0011776719152099624, 'learning_rate': 0.052688596236574584, 'epoch': 0.35} + 35%|███▌ | 183/520 [11:27<20:42, 3.69s/it] 35%|███▌ | 184/520 [11:31<20:49, 3.72s/it] {'loss': 1.1799, 'grad_norm': 0.0013063868635144371, 'learning_rate': 0.052500000000000005, 'epoch': 0.35} + 35%|███▌ | 184/520 [11:31<20:49, 3.72s/it] 36%|███▌ | 185/520 [11:35<20:43, 3.71s/it] {'loss': 1.2933, 'grad_norm': 0.0012002384930155942, 'learning_rate': 0.05231072381602926, 'epoch': 0.36} + 36%|███▌ | 185/520 [11:35<20:43, 3.71s/it] 36%|███▌ | 186/520 [11:39<20:41, 3.72s/it] {'loss': 1.1992, 'grad_norm': 0.0012759828832949371, 'learning_rate': 0.05212077503882513, 'epoch': 0.36} + 36%|███▌ | 186/520 [11:39<20:41, 3.72s/it] 36%|███▌ | 187/520 [11:42<20:37, 3.72s/it] {'loss': 1.1906, 'grad_norm': 0.0014032157594525562, 'learning_rate': 0.0519301610486834, 'epoch': 0.36} + 36%|███▌ | 187/520 [11:42<20:37, 3.72s/it] 36%|███▌ | 188/520 [11:46<20:23, 3.69s/it] {'loss': 1.2864, 'grad_norm': 0.0012768883924064328, 'learning_rate': 0.05173888925174614, 'epoch': 0.36} + 36%|███▌ | 188/520 [11:46<20:23, 3.69s/it] 36%|███▋ | 189/520 [11:50<20:18, 3.68s/it] {'loss': 1.282, 'grad_norm': 0.0011482166447054235, 'learning_rate': 0.05154696707971395, 'epoch': 0.36} + 36%|███▋ | 189/520 [11:50<20:18, 3.68s/it] 37%|███▋ | 190/520 [11:53<20:19, 3.69s/it] {'loss': 1.205, 'grad_norm': 0.0013358083782011117, 'learning_rate': 0.05135440198955717, 'epoch': 0.37} + 37%|███▋ | 190/520 [11:53<20:19, 3.69s/it] 37%|███▋ | 191/520 [11:57<20:24, 3.72s/it] {'loss': 1.1678, 'grad_norm': 0.0012169679520394093, 'learning_rate': 0.0511612014632262, 'epoch': 0.37} + 37%|███▋ | 191/520 [11:57<20:24, 3.72s/it] 37%|███▋ | 192/520 [12:01<20:23, 3.73s/it] {'loss': 1.2448, 'grad_norm': 0.0011762793797160513, 'learning_rate': 0.05096737300736071, 'epoch': 0.37} + 37%|███▋ | 192/520 [12:01<20:23, 3.73s/it] 37%|███▋ | 193/520 [12:05<20:22, 3.74s/it] {'loss': 1.1894, 'grad_norm': 0.0013056284361939968, 'learning_rate': 0.05077292415299809, 'epoch': 0.37} + 37%|███▋ | 193/520 [12:05<20:22, 3.74s/it] 37%|███▋ | 194/520 [12:08<20:11, 3.72s/it] {'loss': 1.0791, 'grad_norm': 0.0010894032521724416, 'learning_rate': 0.05057786245528073, 'epoch': 0.37} + 37%|███▋ | 194/520 [12:08<20:11, 3.72s/it] 38%|███▊ | 195/520 [12:12<20:04, 3.71s/it] {'loss': 1.2571, 'grad_norm': 0.0012048621746560482, 'learning_rate': 0.05038219549316258, 'epoch': 0.38} + 38%|███▊ | 195/520 [12:12<20:04, 3.71s/it] 38%|███▊ | 196/520 [12:16<19:54, 3.69s/it] {'loss': 1.2456, 'grad_norm': 0.0014502378091603928, 'learning_rate': 0.050185930869114534, 'epoch': 0.38} + 38%|███▊ | 196/520 [12:16<19:54, 3.69s/it] 38%|███▊ | 197/520 [12:19<19:54, 3.70s/it] {'loss': 1.1938, 'grad_norm': 0.0013095536861016234, 'learning_rate': 0.0499890762088292, 'epoch': 0.38} + 38%|███▊ | 197/520 [12:19<19:54, 3.70s/it] 38%|███▊ | 198/520 [12:23<19:46, 3.68s/it] {'loss': 1.2614, 'grad_norm': 0.0014135220372501467, 'learning_rate': 0.04979163916092449, 'epoch': 0.38} + 38%|███▊ | 198/520 [12:23<19:46, 3.68s/it] 38%|███▊ | 199/520 [12:27<19:44, 3.69s/it] {'loss': 1.1752, 'grad_norm': 0.0012855730562306496, 'learning_rate': 0.049593627396646495, 'epoch': 0.38} + 38%|███▊ | 199/520 [12:27<19:44, 3.69s/it] 38%|███▊ | 200/520 [12:31<19:48, 3.72s/it] {'loss': 1.1459, 'grad_norm': 0.0012414227346798992, 'learning_rate': 0.04939504860957142, 'epoch': 0.38} + 38%|███▊ | 200/520 [12:31<19:48, 3.72s/it] 39%|███▊ | 201/520 [12:34<19:51, 3.74s/it] {'loss': 1.159, 'grad_norm': 0.0010927418496246846, 'learning_rate': 0.04919591051530663, 'epoch': 0.39} + 39%|███▊ | 201/520 [12:34<19:51, 3.74s/it] 39%|███▉ | 202/520 [12:38<19:50, 3.74s/it] {'loss': 1.1927, 'grad_norm': 0.0012981391926833276, 'learning_rate': 0.04899622085119093, 'epoch': 0.39} + 39%|███▉ | 202/520 [12:38<19:50, 3.74s/it] 39%|███▉ | 203/520 [12:42<19:50, 3.75s/it] {'loss': 1.231, 'grad_norm': 0.0013067852553232337, 'learning_rate': 0.04879598737599389, 'epoch': 0.39} + 39%|███▉ | 203/520 [12:42<19:50, 3.75s/it] 39%|███▉ | 204/520 [12:46<19:48, 3.76s/it] {'loss': 1.2427, 'grad_norm': 0.001292607652165953, 'learning_rate': 0.04859521786961432, 'epoch': 0.39} + 39%|███▉ | 204/520 [12:46<19:48, 3.76s/it] 39%|███▉ | 205/520 [12:49<19:41, 3.75s/it] {'loss': 1.1593, 'grad_norm': 0.0011390688380992107, 'learning_rate': 0.04839392013277815, 'epoch': 0.39} + 39%|███▉ | 205/520 [12:49<19:41, 3.75s/it] 40%|███▉ | 206/520 [12:53<19:27, 3.72s/it] {'loss': 1.2611, 'grad_norm': 0.00120607735736459, 'learning_rate': 0.04819210198673519, 'epoch': 0.4} + 40%|███▉ | 206/520 [12:53<19:27, 3.72s/it] 40%|███▉ | 207/520 [12:57<19:17, 3.70s/it] {'loss': 1.1216, 'grad_norm': 0.0010551784039824661, 'learning_rate': 0.04798977127295533, 'epoch': 0.4} + 40%|███▉ | 207/520 [12:57<19:17, 3.70s/it] 40%|████ | 208/520 [13:00<19:05, 3.67s/it] {'loss': 1.2662, 'grad_norm': 0.0013740753384215954, 'learning_rate': 0.04778693585282384, 'epoch': 0.4} + 40%|████ | 208/520 [13:00<19:05, 3.67s/it] 40%|████ | 209/520 [13:04<19:14, 3.71s/it] {'loss': 1.1827, 'grad_norm': 0.0011997646863038639, 'learning_rate': 0.04758360360733587, 'epoch': 0.4} + 40%|████ | 209/520 [13:04<19:14, 3.71s/it] 40%|████ | 210/520 [13:08<19:25, 3.76s/it] {'loss': 1.2478, 'grad_norm': 0.0012639645233521837, 'learning_rate': 0.047379782436790356, 'epoch': 0.4} + 40%|████ | 210/520 [13:08<19:25, 3.76s/it] 41%|████ | 211/520 [13:12<19:38, 3.81s/it] {'loss': 1.2456, 'grad_norm': 0.001142193849164317, 'learning_rate': 0.04717548026048295, 'epoch': 0.41} + 41%|████ | 211/520 [13:12<19:38, 3.81s/it] 41%|████ | 212/520 [13:16<19:42, 3.84s/it] {'loss': 1.2497, 'grad_norm': 0.0011969146470948452, 'learning_rate': 0.04697070501639841, 'epoch': 0.41} + 41%|████ | 212/520 [13:16<19:42, 3.84s/it] 41%|████ | 213/520 [13:20<19:45, 3.86s/it] {'loss': 1.202, 'grad_norm': 0.0014231833400619718, 'learning_rate': 0.04676546466090209, 'epoch': 0.41} + 41%|████ | 213/520 [13:20<19:45, 3.86s/it] 41%|████ | 214/520 [13:24<19:44, 3.87s/it] {'loss': 1.1969, 'grad_norm': 0.0012719108790985995, 'learning_rate': 0.046559767168430855, 'epoch': 0.41} + 41%|████ | 214/520 [13:24<19:44, 3.87s/it] 41%|████▏ | 215/520 [13:28<19:44, 3.88s/it] {'loss': 1.0993, 'grad_norm': 0.0011892530917979563, 'learning_rate': 0.046353620531183255, 'epoch': 0.41} + 41%|████▏ | 215/520 [13:28<19:44, 3.88s/it] 42%|████▏ | 216/520 [13:31<19:46, 3.90s/it] {'loss': 1.1138, 'grad_norm': 0.0012265282887414582, 'learning_rate': 0.046147032758808965, 'epoch': 0.42} + 42%|████▏ | 216/520 [13:31<19:46, 3.90s/it] 42%|████▏ | 217/520 [13:35<19:50, 3.93s/it] {'loss': 1.2442, 'grad_norm': 0.0012911165224588992, 'learning_rate': 0.04594001187809756, 'epoch': 0.42} + 42%|████▏ | 217/520 [13:35<19:50, 3.93s/it] 42%|████▏ | 218/520 [13:39<19:46, 3.93s/it] {'loss': 1.2183, 'grad_norm': 0.0013737636352914463, 'learning_rate': 0.04573256593266671, 'epoch': 0.42} + 42%|████▏ | 218/520 [13:39<19:46, 3.93s/it] 42%|████▏ | 219/520 [13:43<19:41, 3.92s/it] {'loss': 1.2396, 'grad_norm': 0.0011978664094353649, 'learning_rate': 0.04552470298264956, 'epoch': 0.42} + 42%|████▏ | 219/520 [13:43<19:41, 3.92s/it] 42%|████▏ | 220/520 [13:47<19:25, 3.89s/it] {'loss': 1.1422, 'grad_norm': 0.0011568721922732333, 'learning_rate': 0.04531643110438165, 'epoch': 0.42} + 42%|████▏ | 220/520 [13:47<19:25, 3.89s/it] 42%|████▎ | 221/520 [13:51<19:04, 3.83s/it] {'loss': 1.2317, 'grad_norm': 0.0012906290360530045, 'learning_rate': 0.04510775839008705, 'epoch': 0.42} + 42%|████▎ | 221/520 [13:51<19:04, 3.83s/it] 43%|████▎ | 222/520 [13:54<18:46, 3.78s/it] {'loss': 1.1728, 'grad_norm': 0.0012635643649546678, 'learning_rate': 0.04489869294756397, 'epoch': 0.43} + 43%|████▎ | 222/520 [13:54<18:46, 3.78s/it] 43%|████▎ | 223/520 [13:58<18:32, 3.75s/it] {'loss': 1.1747, 'grad_norm': 0.0012283702288669136, 'learning_rate': 0.04468924289986973, 'epoch': 0.43} + 43%|████▎ | 223/520 [13:58<18:32, 3.75s/it] 43%|████▎ | 224/520 [14:02<18:24, 3.73s/it] {'loss': 1.1833, 'grad_norm': 0.0011027394577030585, 'learning_rate': 0.04447941638500519, 'epoch': 0.43} + 43%|████▎ | 224/520 [14:02<18:24, 3.73s/it] 43%|████▎ | 225/520 [14:06<18:15, 3.71s/it] {'loss': 1.1734, 'grad_norm': 0.0013182092140257384, 'learning_rate': 0.044269221555598454, 'epoch': 0.43} + 43%|████▎ | 225/520 [14:06<18:15, 3.71s/it] 43%|████▎ | 226/520 [14:09<18:05, 3.69s/it] {'loss': 1.273, 'grad_norm': 0.0012390349831090453, 'learning_rate': 0.04405866657858823, 'epoch': 0.43} + 43%|████▎ | 226/520 [14:09<18:05, 3.69s/it] 44%|████▎ | 227/520 [14:13<17:57, 3.68s/it] {'loss': 1.254, 'grad_norm': 0.0012359200548670674, 'learning_rate': 0.04384775963490641, 'epoch': 0.44} + 44%|████▎ | 227/520 [14:13<17:57, 3.68s/it] 44%|████▍ | 228/520 [14:16<17:53, 3.68s/it] {'loss': 1.2301, 'grad_norm': 0.0012163047537345915, 'learning_rate': 0.04363650891916027, 'epoch': 0.44} + 44%|████▍ | 228/520 [14:16<17:53, 3.68s/it] 44%|████▍ | 229/520 [14:20<17:46, 3.66s/it] {'loss': 1.231, 'grad_norm': 0.0011712332393287602, 'learning_rate': 0.043424922639314065, 'epoch': 0.44} + 44%|████▍ | 229/520 [14:20<17:46, 3.66s/it] 44%|████▍ | 230/520 [14:24<17:42, 3.66s/it] {'loss': 1.1247, 'grad_norm': 0.0012223674108524736, 'learning_rate': 0.04321300901637004, 'epoch': 0.44} + 44%|████▍ | 230/520 [14:24<17:42, 3.66s/it] 44%|████▍ | 231/520 [14:27<17:37, 3.66s/it] {'loss': 1.1938, 'grad_norm': 0.001181712209974143, 'learning_rate': 0.04300077628404914, 'epoch': 0.44} + 44%|████▍ | 231/520 [14:27<17:37, 3.66s/it] 45%|████▍ | 232/520 [14:31<17:35, 3.67s/it] {'loss': 1.2693, 'grad_norm': 0.001237501555775087, 'learning_rate': 0.042788232688471005, 'epoch': 0.45} + 45%|████▍ | 232/520 [14:31<17:35, 3.67s/it] 45%|████▍ | 233/520 [14:35<17:30, 3.66s/it] {'loss': 1.1649, 'grad_norm': 0.0013114220690592327, 'learning_rate': 0.04257538648783361, 'epoch': 0.45} + 45%|████▍ | 233/520 [14:35<17:30, 3.66s/it] 45%|████▌ | 234/520 [14:38<17:23, 3.65s/it] {'loss': 1.15, 'grad_norm': 0.0013351722235596041, 'learning_rate': 0.042362245952092356, 'epoch': 0.45} + 45%|████▌ | 234/520 [14:38<17:23, 3.65s/it] 45%|████▌ | 235/520 [14:42<17:23, 3.66s/it] {'loss': 1.2025, 'grad_norm': 0.0013173830028704232, 'learning_rate': 0.04214881936263882, 'epoch': 0.45} + 45%|████▌ | 235/520 [14:42<17:23, 3.66s/it] 45%|████▌ | 236/520 [14:46<17:17, 3.65s/it] {'loss': 1.2628, 'grad_norm': 0.0011943613259579539, 'learning_rate': 0.04193511501197892, 'epoch': 0.45} + 45%|████▌ | 236/520 [14:46<17:17, 3.65s/it] 46%|████▌ | 237/520 [14:49<17:13, 3.65s/it] {'loss': 1.2635, 'grad_norm': 0.0012453365251387934, 'learning_rate': 0.04172114120341078, 'epoch': 0.46} + 46%|████▌ | 237/520 [14:49<17:13, 3.65s/it] 46%|████▌ | 238/520 [14:53<17:11, 3.66s/it] {'loss': 1.1978, 'grad_norm': 0.00131494015128837, 'learning_rate': 0.04150690625070202, 'epoch': 0.46} + 46%|████▌ | 238/520 [14:53<17:11, 3.66s/it] 46%|████▌ | 239/520 [14:57<17:24, 3.72s/it] {'loss': 1.2612, 'grad_norm': 0.001263085443112371, 'learning_rate': 0.041292418477766844, 'epoch': 0.46} + 46%|████▌ | 239/520 [14:57<17:24, 3.72s/it] 46%|████▌ | 240/520 [15:01<17:19, 3.71s/it] {'loss': 1.0986, 'grad_norm': 0.00119099224958702, 'learning_rate': 0.041077686218342575, 'epoch': 0.46} + 46%|████▌ | 240/520 [15:01<17:19, 3.71s/it] 46%|████▋ | 241/520 [15:04<17:11, 3.70s/it] {'loss': 1.1791, 'grad_norm': 0.0013099948317818102, 'learning_rate': 0.04086271781566578, 'epoch': 0.46} + 46%|████▋ | 241/520 [15:04<17:11, 3.70s/it] 47%|████▋ | 242/520 [15:08<17:07, 3.70s/it] {'loss': 1.1812, 'grad_norm': 0.001210535472255199, 'learning_rate': 0.04064752162214824, 'epoch': 0.47} + 47%|████▋ | 242/520 [15:08<17:07, 3.70s/it] 47%|████▋ | 243/520 [15:12<17:01, 3.69s/it] {'loss': 1.1767, 'grad_norm': 0.0012517541376063314, 'learning_rate': 0.04043210599905231, 'epoch': 0.47} + 47%|████▋ | 243/520 [15:12<17:01, 3.69s/it] 47%|████▋ | 244/520 [15:15<16:53, 3.67s/it] {'loss': 1.2697, 'grad_norm': 0.0012338112042604113, 'learning_rate': 0.04021647931616611, 'epoch': 0.47} + 47%|████▋ | 244/520 [15:15<16:53, 3.67s/it] 47%|████▋ | 245/520 [15:19<16:49, 3.67s/it] {'loss': 1.1606, 'grad_norm': 0.0013517006209517836, 'learning_rate': 0.040000649951478305, 'epoch': 0.47} + 47%|████▋ | 245/520 [15:19<16:49, 3.67s/it] 47%|████▋ | 246/520 [15:23<16:47, 3.68s/it] {'loss': 1.2635, 'grad_norm': 0.0012596578255944719, 'learning_rate': 0.039784626290852575, 'epoch': 0.47} + 47%|████▋ | 246/520 [15:23<16:47, 3.68s/it] 48%|████▊ | 247/520 [15:26<16:43, 3.67s/it] {'loss': 1.3292, 'grad_norm': 0.0013028474521635251, 'learning_rate': 0.03956841672770182, 'epoch': 0.47} + 48%|████▊ | 247/520 [15:26<16:43, 3.67s/it] 48%|████▊ | 248/520 [15:30<16:35, 3.66s/it] {'loss': 1.1512, 'grad_norm': 0.0012924519585120785, 'learning_rate': 0.03935202966266199, 'epoch': 0.48} + 48%|████▊ | 248/520 [15:30<16:35, 3.66s/it] 48%|████▊ | 249/520 [15:34<16:31, 3.66s/it] {'loss': 1.2443, 'grad_norm': 0.0013025711975898063, 'learning_rate': 0.03913547350326575, 'epoch': 0.48} + 48%|████▊ | 249/520 [15:34<16:31, 3.66s/it] 48%|████▊ | 250/520 [15:37<16:26, 3.66s/it] {'loss': 1.1995, 'grad_norm': 0.001377383857386791, 'learning_rate': 0.03891875666361578, 'epoch': 0.48} + 48%|████▊ | 250/520 [15:37<16:26, 3.66s/it] 48%|████▊ | 251/520 [15:41<16:27, 3.67s/it] {'loss': 1.2576, 'grad_norm': 0.001232872177107492, 'learning_rate': 0.03870188756405783, 'epoch': 0.48} + 48%|████▊ | 251/520 [15:41<16:27, 3.67s/it] 48%|████▊ | 252/520 [15:45<16:40, 3.73s/it] {'loss': 1.1717, 'grad_norm': 0.001178757816007794, 'learning_rate': 0.03848487463085359, 'epoch': 0.48} + 48%|████▊ | 252/520 [15:45<16:40, 3.73s/it] 49%|████▊ | 253/520 [15:49<16:49, 3.78s/it] {'loss': 1.2341, 'grad_norm': 0.0013623764451369698, 'learning_rate': 0.038267726295853274, 'epoch': 0.49} + 49%|████▊ | 253/520 [15:49<16:49, 3.78s/it] 49%|████▉ | 254/520 [15:53<16:51, 3.80s/it] {'loss': 1.1844, 'grad_norm': 0.0012347758548948524, 'learning_rate': 0.03805045099616804, 'epoch': 0.49} + 49%|████▉ | 254/520 [15:53<16:51, 3.80s/it] 49%|████▉ | 255/520 [15:56<16:40, 3.78s/it] {'loss': 1.1872, 'grad_norm': 0.0013436491169386603, 'learning_rate': 0.03783305717384212, 'epoch': 0.49} + 49%|████▉ | 255/520 [15:56<16:40, 3.78s/it] 49%|████▉ | 256/520 [16:00<16:22, 3.72s/it] {'loss': 1.2448, 'grad_norm': 0.0013460536486991747, 'learning_rate': 0.037615553275524856, 'epoch': 0.49} + 49%|████▉ | 256/520 [16:00<16:22, 3.72s/it] 49%|████▉ | 257/520 [16:03<16:12, 3.70s/it] {'loss': 1.2243, 'grad_norm': 0.0012949031993773875, 'learning_rate': 0.03739794775214249, 'epoch': 0.49} + 49%|████▉ | 257/520 [16:03<16:12, 3.70s/it] 50%|████▉ | 258/520 [16:07<16:05, 3.69s/it] {'loss': 1.2287, 'grad_norm': 0.0011083129173038101, 'learning_rate': 0.03718024905856983, 'epoch': 0.5} + 50%|████▉ | 258/520 [16:07<16:05, 3.69s/it] 50%|████▉ | 259/520 [16:11<16:01, 3.68s/it] {'loss': 1.3013, 'grad_norm': 0.001419577830058232, 'learning_rate': 0.03696246565330172, 'epoch': 0.5} + 50%|████▉ | 259/520 [16:11<16:01, 3.68s/it] 50%|█████ | 260/520 [16:14<15:54, 3.67s/it] {'loss': 1.2136, 'grad_norm': 0.0010108262569871276, 'learning_rate': 0.0367446059981244, 'epoch': 0.5} + 50%|█████ | 260/520 [16:14<15:54, 3.67s/it] 50%|█████ | 261/520 [16:18<15:50, 3.67s/it] {'loss': 1.1628, 'grad_norm': 0.0011994506405235937, 'learning_rate': 0.03652667855778677, 'epoch': 0.5} + 50%|█████ | 261/520 [16:18<15:50, 3.67s/it] 50%|█████ | 262/520 [16:22<15:47, 3.67s/it] {'loss': 1.1737, 'grad_norm': 0.001302728518186142, 'learning_rate': 0.03630869179967141, 'epoch': 0.5} + 50%|█████ | 262/520 [16:22<15:47, 3.67s/it] 51%|█████ | 263/520 [16:25<15:43, 3.67s/it] {'loss': 1.1882, 'grad_norm': 0.0012415583998281606, 'learning_rate': 0.03609065419346567, 'epoch': 0.51} + 51%|█████ | 263/520 [16:25<15:43, 3.67s/it] 51%|█████ | 264/520 [16:29<15:42, 3.68s/it] {'loss': 1.2534, 'grad_norm': 0.0012095790670797865, 'learning_rate': 0.03587257421083256, 'epoch': 0.51} + 51%|█████ | 264/520 [16:29<15:42, 3.68s/it] 51%|█████ | 265/520 [16:33<15:48, 3.72s/it] {'loss': 1.1844, 'grad_norm': 0.0014529707273225232, 'learning_rate': 0.03565446032508158, 'epoch': 0.51} + 51%|█████ | 265/520 [16:33<15:48, 3.72s/it] 51%|█████ | 266/520 [16:37<15:51, 3.75s/it] {'loss': 1.0647, 'grad_norm': 0.0011345984731638836, 'learning_rate': 0.03543632101083953, 'epoch': 0.51} + 51%|█████ | 266/520 [16:37<15:51, 3.75s/it] 51%|█████▏ | 267/520 [16:41<15:53, 3.77s/it] {'loss': 1.1797, 'grad_norm': 0.0012489278470318335, 'learning_rate': 0.03521816474372118, 'epoch': 0.51} + 51%|█████▏ | 267/520 [16:41<15:53, 3.77s/it] 52%|█████▏ | 268/520 [16:44<15:41, 3.73s/it] {'loss': 1.2837, 'grad_norm': 0.0012425284298360866, 'learning_rate': 0.035, 'epoch': 0.52} + 52%|█████▏ | 268/520 [16:44<15:41, 3.73s/it] 52%|█████▏ | 269/520 [16:48<15:31, 3.71s/it] {'loss': 1.283, 'grad_norm': 0.0013904452932550692, 'learning_rate': 0.03478183525627883, 'epoch': 0.52} + 52%|█████▏ | 269/520 [16:48<15:31, 3.71s/it] 52%|█████▏ | 270/520 [16:52<15:28, 3.71s/it] {'loss': 1.128, 'grad_norm': 0.0012302890177972015, 'learning_rate': 0.034563678989160476, 'epoch': 0.52} + 52%|█████▏ | 270/520 [16:52<15:28, 3.71s/it] 52%|█████▏ | 271/520 [16:55<15:21, 3.70s/it] {'loss': 1.2555, 'grad_norm': 0.0013223908617885265, 'learning_rate': 0.034345539674918435, 'epoch': 0.52} + 52%|█████▏ | 271/520 [16:55<15:21, 3.70s/it] 52%|█████▏ | 272/520 [16:59<15:14, 3.69s/it] {'loss': 1.1466, 'grad_norm': 0.0012311002601041957, 'learning_rate': 0.034127425789167457, 'epoch': 0.52} + 52%|█████▏ | 272/520 [16:59<15:14, 3.69s/it] 52%|█████▎ | 273/520 [17:03<15:11, 3.69s/it] {'loss': 1.2404, 'grad_norm': 0.0011988976105721488, 'learning_rate': 0.03390934580653435, 'epoch': 0.53} + 52%|█████▎ | 273/520 [17:03<15:11, 3.69s/it] 53%|█████▎ | 274/520 [17:06<15:05, 3.68s/it] {'loss': 1.2502, 'grad_norm': 0.0013840267346829512, 'learning_rate': 0.03369130820032861, 'epoch': 0.53} + 53%|█████▎ | 274/520 [17:06<15:05, 3.68s/it] 53%|█████▎ | 275/520 [17:10<15:00, 3.68s/it] {'loss': 1.1861, 'grad_norm': 0.0012981179315243766, 'learning_rate': 0.033473321442213246, 'epoch': 0.53} + 53%|█████▎ | 275/520 [17:10<15:00, 3.68s/it] 53%|█████▎ | 276/520 [17:14<14:59, 3.69s/it] {'loss': 1.2439, 'grad_norm': 0.0013965739355191434, 'learning_rate': 0.0332553940018756, 'epoch': 0.53} + 53%|█████▎ | 276/520 [17:14<14:59, 3.69s/it] 53%|█████▎ | 277/520 [17:17<14:58, 3.70s/it] {'loss': 1.254, 'grad_norm': 0.001161108327274977, 'learning_rate': 0.03303753434669829, 'epoch': 0.53} + 53%|█████▎ | 277/520 [17:17<14:58, 3.70s/it] 53%|█████▎ | 278/520 [17:21<14:54, 3.70s/it] {'loss': 1.1539, 'grad_norm': 0.0011990877040747461, 'learning_rate': 0.03281975094143018, 'epoch': 0.53} + 53%|█████▎ | 278/520 [17:21<14:54, 3.70s/it] 54%|█████▎ | 279/520 [17:25<14:48, 3.69s/it] {'loss': 1.1354, 'grad_norm': 0.0013171671303535933, 'learning_rate': 0.03260205224785752, 'epoch': 0.54} + 54%|█████▎ | 279/520 [17:25<14:48, 3.69s/it] 54%|█████▍ | 280/520 [17:28<14:44, 3.69s/it] {'loss': 1.1814, 'grad_norm': 0.0013645135648759602, 'learning_rate': 0.03238444672447516, 'epoch': 0.54} + 54%|█████▍ | 280/520 [17:28<14:44, 3.69s/it] 54%|█████▍ | 281/520 [17:32<14:40, 3.69s/it] {'loss': 1.2801, 'grad_norm': 0.0013447884340788603, 'learning_rate': 0.03216694282615789, 'epoch': 0.54} + 54%|█████▍ | 281/520 [17:32<14:40, 3.69s/it] 54%|█████▍ | 282/520 [17:36<14:35, 3.68s/it] {'loss': 1.1619, 'grad_norm': 0.0012381092252654303, 'learning_rate': 0.03194954900383196, 'epoch': 0.54} + 54%|█████▍ | 282/520 [17:36<14:35, 3.68s/it] 54%|█████▍ | 283/520 [17:39<14:30, 3.67s/it] {'loss': 1.2872, 'grad_norm': 0.0013954292177387883, 'learning_rate': 0.03173227370414673, 'epoch': 0.54} + 54%|█████▍ | 283/520 [17:39<14:30, 3.67s/it] 55%|█████▍ | 284/520 [17:43<14:26, 3.67s/it] {'loss': 1.1462, 'grad_norm': 0.001334969089696765, 'learning_rate': 0.03151512536914643, 'epoch': 0.55} + 55%|█████▍ | 284/520 [17:43<14:26, 3.67s/it] 55%|█████▍ | 285/520 [17:47<14:22, 3.67s/it] {'loss': 1.1842, 'grad_norm': 0.0012871058898653721, 'learning_rate': 0.03129811243594218, 'epoch': 0.55} + 55%|█████▍ | 285/520 [17:47<14:22, 3.67s/it] 55%|█████▌ | 286/520 [17:50<14:17, 3.66s/it] {'loss': 1.0673, 'grad_norm': 0.0012968463062362527, 'learning_rate': 0.031081243336384227, 'epoch': 0.55} + 55%|█████▌ | 286/520 [17:50<14:17, 3.66s/it] 55%|█████▌ | 287/520 [17:54<14:12, 3.66s/it] {'loss': 1.2942, 'grad_norm': 0.001257998632987496, 'learning_rate': 0.030864526496734254, 'epoch': 0.55} + 55%|█████▌ | 287/520 [17:54<14:12, 3.66s/it] 55%|█████▌ | 288/520 [17:58<14:09, 3.66s/it] {'loss': 1.309, 'grad_norm': 0.0011828319480231355, 'learning_rate': 0.030647970337338033, 'epoch': 0.55} + 55%|█████▌ | 288/520 [17:58<14:09, 3.66s/it] 56%|█████▌ | 289/520 [18:01<14:09, 3.68s/it] {'loss': 1.2083, 'grad_norm': 0.0012284368077593248, 'learning_rate': 0.030431583272298206, 'epoch': 0.56} + 56%|█████▌ | 289/520 [18:01<14:09, 3.68s/it] 56%|█████▌ | 290/520 [18:05<14:03, 3.67s/it] {'loss': 1.1276, 'grad_norm': 0.0011727523890474538, 'learning_rate': 0.03021537370914744, 'epoch': 0.56} + 56%|█████▌ | 290/520 [18:05<14:03, 3.67s/it] 56%|█████▌ | 291/520 [18:09<13:59, 3.67s/it] {'loss': 1.1592, 'grad_norm': 0.0012570753464564777, 'learning_rate': 0.02999935004852171, 'epoch': 0.56} + 56%|█████▌ | 291/520 [18:09<13:59, 3.67s/it] 56%|█████▌ | 292/520 [18:12<13:56, 3.67s/it] {'loss': 1.219, 'grad_norm': 0.001266149814251958, 'learning_rate': 0.029783520683833893, 'epoch': 0.56} + 56%|█████▌ | 292/520 [18:12<13:56, 3.67s/it] 56%|█████▋ | 293/520 [18:16<13:51, 3.66s/it] {'loss': 1.1702, 'grad_norm': 0.0013442620727082716, 'learning_rate': 0.029567894000947694, 'epoch': 0.56} + 56%|█████▋ | 293/520 [18:16<13:51, 3.66s/it] 57%|█████▋ | 294/520 [18:20<13:47, 3.66s/it] {'loss': 1.1929, 'grad_norm': 0.0014081723737820679, 'learning_rate': 0.02935247837785177, 'epoch': 0.57} + 57%|█████▋ | 294/520 [18:20<13:47, 3.66s/it] 57%|█████▋ | 295/520 [18:23<13:45, 3.67s/it] {'loss': 1.167, 'grad_norm': 0.001126863023566673, 'learning_rate': 0.029137282184334232, 'epoch': 0.57} + 57%|█████▋ | 295/520 [18:23<13:45, 3.67s/it] 57%|█████▋ | 296/520 [18:27<13:39, 3.66s/it] {'loss': 1.142, 'grad_norm': 0.0013151680377990077, 'learning_rate': 0.028922313781657442, 'epoch': 0.57} + 57%|█████▋ | 296/520 [18:27<13:39, 3.66s/it] 57%|█████▋ | 297/520 [18:31<13:40, 3.68s/it] {'loss': 1.271, 'grad_norm': 0.001395256851875004, 'learning_rate': 0.02870758152223316, 'epoch': 0.57} + 57%|█████▋ | 297/520 [18:31<13:40, 3.68s/it] 57%|█████▋ | 298/520 [18:34<13:34, 3.67s/it] {'loss': 1.2288, 'grad_norm': 0.0012169141514151718, 'learning_rate': 0.028493093749297992, 'epoch': 0.57} + 57%|█████▋ | 298/520 [18:34<13:34, 3.67s/it] 57%|█████▊ | 299/520 [18:38<13:31, 3.67s/it] {'loss': 1.2258, 'grad_norm': 0.0012246099894316629, 'learning_rate': 0.02827885879658924, 'epoch': 0.57} + 57%|█████▊ | 299/520 [18:38<13:31, 3.67s/it] 58%|█████▊ | 300/520 [18:42<13:26, 3.67s/it] {'loss': 1.2772, 'grad_norm': 0.0012795656573802614, 'learning_rate': 0.028064884988021094, 'epoch': 0.58} + 58%|█████▊ | 300/520 [18:42<13:26, 3.67s/it] 58%|█████▊ | 301/520 [18:45<13:21, 3.66s/it] {'loss': 1.2711, 'grad_norm': 0.0012990229202636038, 'learning_rate': 0.027851180637361195, 'epoch': 0.58} + 58%|█████▊ | 301/520 [18:45<13:21, 3.66s/it] 58%|█████▊ | 302/520 [18:49<13:18, 3.66s/it] {'loss': 1.2195, 'grad_norm': 0.0011941308422766157, 'learning_rate': 0.027637754047907657, 'epoch': 0.58} + 58%|█████▊ | 302/520 [18:49<13:18, 3.66s/it] 58%|█████▊ | 303/520 [18:53<13:16, 3.67s/it] {'loss': 1.1844, 'grad_norm': 0.0014171761252145196, 'learning_rate': 0.027424613512166404, 'epoch': 0.58} + 58%|█████▊ | 303/520 [18:53<13:16, 3.67s/it] 58%|█████▊ | 304/520 [18:57<13:18, 3.69s/it] {'loss': 1.1381, 'grad_norm': 0.001274074151953295, 'learning_rate': 0.027211767311529002, 'epoch': 0.58} + 58%|█████▊ | 304/520 [18:57<13:18, 3.69s/it] 59%|█████▊ | 305/520 [19:00<13:16, 3.71s/it] {'loss': 1.2838, 'grad_norm': 0.0014272346805225744, 'learning_rate': 0.026999223715950874, 'epoch': 0.59} + 59%|█████▊ | 305/520 [19:00<13:16, 3.71s/it] 59%|█████▉ | 306/520 [19:04<13:12, 3.70s/it] {'loss': 1.2339, 'grad_norm': 0.001309341718839588, 'learning_rate': 0.026786990983629976, 'epoch': 0.59} + 59%|█████▉ | 306/520 [19:04<13:12, 3.70s/it] 59%|█████▉ | 307/520 [19:08<13:08, 3.70s/it] {'loss': 1.1765, 'grad_norm': 0.0012436846851193821, 'learning_rate': 0.026575077360685956, 'epoch': 0.59} + 59%|█████▉ | 307/520 [19:08<13:08, 3.70s/it] 59%|█████▉ | 308/520 [19:11<13:12, 3.74s/it] {'loss': 1.2932, 'grad_norm': 0.0012267636771061304, 'learning_rate': 0.026363491080839723, 'epoch': 0.59} + 59%|█████▉ | 308/520 [19:12<13:12, 3.74s/it] 59%|█████▉ | 309/520 [19:16<13:25, 3.82s/it] {'loss': 1.1878, 'grad_norm': 0.0012521901637287869, 'learning_rate': 0.026152240365093584, 'epoch': 0.59} + 59%|█████▉ | 309/520 [19:16<13:25, 3.82s/it] 60%|█████▉ | 310/520 [19:19<13:09, 3.76s/it] {'loss': 1.1644, 'grad_norm': 0.001284304558521113, 'learning_rate': 0.025941333421411774, 'epoch': 0.6} + 60%|█████▉ | 310/520 [19:19<13:09, 3.76s/it] 60%|█████▉ | 311/520 [19:23<12:58, 3.72s/it] {'loss': 1.1552, 'grad_norm': 0.0012669680517853148, 'learning_rate': 0.025730778444401546, 'epoch': 0.6} + 60%|█████▉ | 311/520 [19:23<12:58, 3.72s/it] 60%|██████ | 312/520 [19:26<12:54, 3.72s/it] {'loss': 1.1356, 'grad_norm': 0.0013135664696211078, 'learning_rate': 0.02552058361499483, 'epoch': 0.6} + 60%|██████ | 312/520 [19:26<12:54, 3.72s/it] 60%|██████ | 313/520 [19:30<12:45, 3.70s/it] {'loss': 1.1101, 'grad_norm': 0.0011719022243807718, 'learning_rate': 0.025310757100130275, 'epoch': 0.6} + 60%|██████ | 313/520 [19:30<12:45, 3.70s/it] 60%|██████ | 314/520 [19:34<13:02, 3.80s/it] {'loss': 1.1552, 'grad_norm': 0.0012328092090707636, 'learning_rate': 0.02510130705243604, 'epoch': 0.6} + 60%|██████ | 314/520 [19:34<13:02, 3.80s/it] 61%|██████ | 315/520 [19:38<12:50, 3.76s/it] {'loss': 1.1866, 'grad_norm': 0.0013700493022709862, 'learning_rate': 0.024892241609912963, 'epoch': 0.61} + 61%|██████ | 315/520 [19:38<12:50, 3.76s/it] 61%|██████ | 316/520 [19:42<13:10, 3.87s/it] {'loss': 1.1457, 'grad_norm': 0.0013433081368902713, 'learning_rate': 0.024683568895618353, 'epoch': 0.61} + 61%|██████ | 316/520 [19:42<13:10, 3.87s/it] 61%|██████ | 317/520 [19:46<12:52, 3.81s/it] {'loss': 1.14, 'grad_norm': 0.0011539600921451766, 'learning_rate': 0.024475297017350447, 'epoch': 0.61} + 61%|██████ | 317/520 [19:46<12:52, 3.81s/it] 61%|██████ | 318/520 [19:49<12:40, 3.77s/it] {'loss': 1.2543, 'grad_norm': 0.0013478336903533963, 'learning_rate': 0.024267434067333307, 'epoch': 0.61} + 61%|██████ | 318/520 [19:49<12:40, 3.77s/it] 61%|██████▏ | 319/520 [19:53<12:51, 3.84s/it] {'loss': 1.1375, 'grad_norm': 0.0011571363284162512, 'learning_rate': 0.02405998812190245, 'epoch': 0.61} + 61%|██████▏ | 319/520 [19:53<12:51, 3.84s/it] 62%|██████▏ | 320/520 [19:57<12:35, 3.78s/it] {'loss': 1.076, 'grad_norm': 0.0012853927821300857, 'learning_rate': 0.023852967241191052, 'epoch': 0.62} + 62%|██████▏ | 320/520 [19:57<12:35, 3.78s/it] 62%|██████▏ | 321/520 [20:01<12:26, 3.75s/it] {'loss': 1.2756, 'grad_norm': 0.0012582762275082354, 'learning_rate': 0.02364637946881676, 'epoch': 0.62} + 62%|██████▏ | 321/520 [20:01<12:26, 3.75s/it] 62%|██████▏ | 322/520 [20:04<12:16, 3.72s/it] {'loss': 1.0868, 'grad_norm': 0.0012139307388543425, 'learning_rate': 0.023440232831569166, 'epoch': 0.62} + 62%|██████▏ | 322/520 [20:04<12:16, 3.72s/it] 62%|██████▏ | 323/520 [20:08<12:08, 3.70s/it] {'loss': 1.1634, 'grad_norm': 0.0012560377379184878, 'learning_rate': 0.02323453533909793, 'epoch': 0.62} + 62%|██████▏ | 323/520 [20:08<12:08, 3.70s/it] 62%|██████▏ | 324/520 [20:12<12:04, 3.70s/it] {'loss': 1.222, 'grad_norm': 0.001311379326082302, 'learning_rate': 0.023029294983601598, 'epoch': 0.62} + 62%|██████▏ | 324/520 [20:12<12:04, 3.70s/it] 62%|██████▎ | 325/520 [20:15<11:58, 3.69s/it] {'loss': 1.212, 'grad_norm': 0.001341664912706582, 'learning_rate': 0.022824519739517046, 'epoch': 0.62} + 62%|██████▎ | 325/520 [20:15<11:58, 3.69s/it] 63%|██████▎ | 326/520 [20:19<11:53, 3.68s/it] {'loss': 1.2154, 'grad_norm': 0.001370412192008163, 'learning_rate': 0.022620217563209654, 'epoch': 0.63} + 63%|██████▎ | 326/520 [20:19<11:53, 3.68s/it] 63%|██████▎ | 327/520 [20:23<11:48, 3.67s/it] {'loss': 1.191, 'grad_norm': 0.0013029597998068278, 'learning_rate': 0.02241639639266414, 'epoch': 0.63} + 63%|██████▎ | 327/520 [20:23<11:48, 3.67s/it] 63%|██████▎ | 328/520 [20:26<11:43, 3.66s/it] {'loss': 1.2549, 'grad_norm': 0.001347023513413778, 'learning_rate': 0.022213064147176175, 'epoch': 0.63} + 63%|██████▎ | 328/520 [20:26<11:43, 3.66s/it] 63%|██████▎ | 329/520 [20:30<11:37, 3.65s/it] {'loss': 1.1391, 'grad_norm': 0.0011582370836060308, 'learning_rate': 0.022010228727044674, 'epoch': 0.63} + 63%|██████▎ | 329/520 [20:30<11:37, 3.65s/it] 63%|██████▎ | 330/520 [20:34<11:33, 3.65s/it] {'loss': 1.2169, 'grad_norm': 0.0012143592452230412, 'learning_rate': 0.02180789801326482, 'epoch': 0.63} + 63%|██████▎ | 330/520 [20:34<11:33, 3.65s/it] 64%|██████▎ | 331/520 [20:37<11:29, 3.65s/it] {'loss': 1.173, 'grad_norm': 0.0013531147415962535, 'learning_rate': 0.02160607986722186, 'epoch': 0.64} + 64%|██████▎ | 331/520 [20:37<11:29, 3.65s/it] 64%|██████▍ | 332/520 [20:41<11:26, 3.65s/it] {'loss': 1.2186, 'grad_norm': 0.0011609406467837182, 'learning_rate': 0.02140478213038569, 'epoch': 0.64} + 64%|██████▍ | 332/520 [20:41<11:26, 3.65s/it] 64%|██████▍ | 333/520 [20:44<11:22, 3.65s/it] {'loss': 1.3065, 'grad_norm': 0.0013705453140908641, 'learning_rate': 0.02120401262400613, 'epoch': 0.64} + 64%|██████▍ | 333/520 [20:44<11:22, 3.65s/it] 64%|██████▍ | 334/520 [20:48<11:18, 3.65s/it] {'loss': 1.22, 'grad_norm': 0.0013638374545905056, 'learning_rate': 0.021003779148809073, 'epoch': 0.64} + 64%|██████▍ | 334/520 [20:48<11:18, 3.65s/it] 64%|██████▍ | 335/520 [20:52<11:14, 3.65s/it] {'loss': 1.2197, 'grad_norm': 0.0012425535619398406, 'learning_rate': 0.02080408948469338, 'epoch': 0.64} + 64%|██████▍ | 335/520 [20:52<11:14, 3.65s/it] 65%|██████▍ | 336/520 [20:55<11:10, 3.65s/it] {'loss': 1.1339, 'grad_norm': 0.0014303136042996374, 'learning_rate': 0.020604951390428606, 'epoch': 0.65} + 65%|██████▍ | 336/520 [20:55<11:10, 3.65s/it] 65%|██████▍ | 337/520 [20:59<11:09, 3.66s/it] {'loss': 1.1233, 'grad_norm': 0.0013362403762127444, 'learning_rate': 0.02040637260335353, 'epoch': 0.65} + 65%|██████▍ | 337/520 [20:59<11:09, 3.66s/it] 65%|██████▌ | 338/520 [21:03<11:06, 3.66s/it] {'loss': 1.2289, 'grad_norm': 0.0013007234089351474, 'learning_rate': 0.020208360839075526, 'epoch': 0.65} + 65%|██████▌ | 338/520 [21:03<11:06, 3.66s/it] 65%|██████▌ | 339/520 [21:06<11:02, 3.66s/it] {'loss': 1.1682, 'grad_norm': 0.0013667767427492498, 'learning_rate': 0.020010923791170802, 'epoch': 0.65} + 65%|██████▌ | 339/520 [21:06<11:02, 3.66s/it] 65%|██████▌ | 340/520 [21:10<10:59, 3.67s/it] {'loss': 1.1571, 'grad_norm': 0.0012575274356006255, 'learning_rate': 0.01981406913088547, 'epoch': 0.65} + 65%|██████▌ | 340/520 [21:10<10:59, 3.67s/it] 66%|██████▌ | 341/520 [21:14<10:59, 3.69s/it] {'loss': 1.1907, 'grad_norm': 0.0013574467297950228, 'learning_rate': 0.019617804506837425, 'epoch': 0.66} + 66%|██████▌ | 341/520 [21:14<10:59, 3.69s/it] 66%|██████▌ | 342/520 [21:17<10:54, 3.68s/it] {'loss': 1.2, 'grad_norm': 0.001463386112421856, 'learning_rate': 0.01942213754471927, 'epoch': 0.66} + 66%|██████▌ | 342/520 [21:17<10:54, 3.68s/it] 66%|██████▌ | 343/520 [21:21<10:51, 3.68s/it] {'loss': 1.1425, 'grad_norm': 0.0010383215620322652, 'learning_rate': 0.019227075847001913, 'epoch': 0.66} + 66%|██████▌ | 343/520 [21:21<10:51, 3.68s/it] 66%|██████▌ | 344/520 [21:25<10:46, 3.67s/it] {'loss': 1.147, 'grad_norm': 0.0011987570896659428, 'learning_rate': 0.019032626992639298, 'epoch': 0.66} + 66%|██████▌ | 344/520 [21:25<10:46, 3.67s/it] 66%|██████▋ | 345/520 [21:28<10:41, 3.67s/it] {'loss': 1.2446, 'grad_norm': 0.0013358609909062667, 'learning_rate': 0.018838798536773824, 'epoch': 0.66} + 66%|██████▋ | 345/520 [21:28<10:41, 3.67s/it] 67%|██████▋ | 346/520 [21:32<10:37, 3.66s/it] {'loss': 1.1656, 'grad_norm': 0.0012198660261445525, 'learning_rate': 0.01864559801044283, 'epoch': 0.67} + 67%|██████▋ | 346/520 [21:32<10:37, 3.66s/it] 67%|██████▋ | 347/520 [21:36<10:30, 3.65s/it] {'loss': 1.1639, 'grad_norm': 0.0011845604141496626, 'learning_rate': 0.01845303292028606, 'epoch': 0.67} + 67%|██████▋ | 347/520 [21:36<10:30, 3.65s/it]Token indices sequence length is longer than the specified maximum sequence length for this model (2778 > 2048). Running this sequence through the model will result in indexing errors + 67%|██████▋ | 348/520 [21:39<10:27, 3.65s/it] {'loss': 1.12, 'grad_norm': 0.001525911452776613, 'learning_rate': 0.018261110748253876, 'epoch': 0.67} + 67%|██████▋ | 348/520 [21:39<10:27, 3.65s/it] 67%|██████▋ | 349/520 [21:43<10:25, 3.66s/it] {'loss': 1.1541, 'grad_norm': 0.001320390303433302, 'learning_rate': 0.01806983895131661, 'epoch': 0.67} + 67%|██████▋ | 349/520 [21:43<10:25, 3.66s/it] 67%|██████▋ | 350/520 [21:47<10:25, 3.68s/it] {'loss': 1.2004, 'grad_norm': 0.0013486623588722139, 'learning_rate': 0.01787922496117489, 'epoch': 0.67} + 67%|██████▋ | 350/520 [21:47<10:25, 3.68s/it] 68%|██████▊ | 351/520 [21:50<10:22, 3.68s/it] {'loss': 1.1144, 'grad_norm': 0.0012406312510197155, 'learning_rate': 0.01768927618397074, 'epoch': 0.68} + 68%|██████▊ | 351/520 [21:50<10:22, 3.68s/it] 68%|██████▊ | 352/520 [21:54<10:16, 3.67s/it] {'loss': 1.2238, 'grad_norm': 0.0012051640730361053, 'learning_rate': 0.01750000000000001, 'epoch': 0.68} + 68%|██████▊ | 352/520 [21:54<10:16, 3.67s/it] 68%|██████▊ | 353/520 [21:58<10:14, 3.68s/it] {'loss': 1.1432, 'grad_norm': 0.0010577372347082882, 'learning_rate': 0.017311403763425437, 'epoch': 0.68} + 68%|██████▊ | 353/520 [21:58<10:14, 3.68s/it] 68%|██████▊ | 354/520 [22:01<10:10, 3.68s/it] {'loss': 1.2308, 'grad_norm': 0.001160315805581365, 'learning_rate': 0.017123494801991015, 'epoch': 0.68} + 68%|██████▊ | 354/520 [22:01<10:10, 3.68s/it] 68%|██████▊ | 355/520 [22:05<10:04, 3.66s/it] {'loss': 1.1764, 'grad_norm': 0.0013092298675680698, 'learning_rate': 0.016936280416737268, 'epoch': 0.68} + 68%|██████▊ | 355/520 [22:05<10:04, 3.66s/it] 68%|██████▊ | 356/520 [22:09<10:00, 3.66s/it] {'loss': 1.1773, 'grad_norm': 0.0013298183599922487, 'learning_rate': 0.016749767881717573, 'epoch': 0.68} + 68%|██████▊ | 356/520 [22:09<10:00, 3.66s/it] 69%|██████▊ | 357/520 [22:12<09:56, 3.66s/it] {'loss': 1.2127, 'grad_norm': 0.0012403765256565314, 'learning_rate': 0.016563964443715475, 'epoch': 0.69} + 69%|██████▊ | 357/520 [22:12<09:56, 3.66s/it] 69%|██████▉ | 358/520 [22:16<09:59, 3.70s/it] {'loss': 1.1394, 'grad_norm': 0.0013143561300793679, 'learning_rate': 0.016378877321963227, 'epoch': 0.69} + 69%|██████▉ | 358/520 [22:16<09:59, 3.70s/it] 69%|██████▉ | 359/520 [22:20<10:01, 3.73s/it] {'loss': 1.1768, 'grad_norm': 0.0012672029435439557, 'learning_rate': 0.016194513707861163, 'epoch': 0.69} + 69%|██████▉ | 359/520 [22:20<10:01, 3.73s/it] 69%|██████▉ | 360/520 [22:24<10:01, 3.76s/it] {'loss': 1.1795, 'grad_norm': 0.0012278845161732125, 'learning_rate': 0.016010880764698424, 'epoch': 0.69} + 69%|██████▉ | 360/520 [22:24<10:01, 3.76s/it] 69%|██████▉ | 361/520 [22:28<09:57, 3.76s/it] {'loss': 1.2011, 'grad_norm': 0.00112201892707463, 'learning_rate': 0.01582798562737451, 'epoch': 0.69} + 69%|██████▉ | 361/520 [22:28<09:57, 3.76s/it] 70%|██████▉ | 362/520 [22:31<09:55, 3.77s/it] {'loss': 1.183, 'grad_norm': 0.0014161396026761614, 'learning_rate': 0.01564583540212212, 'epoch': 0.7} + 70%|██████▉ | 362/520 [22:31<09:55, 3.77s/it] 70%|██████▉ | 363/520 [22:35<09:52, 3.78s/it] {'loss': 1.218, 'grad_norm': 0.0012827249196735286, 'learning_rate': 0.015464437166231068, 'epoch': 0.7} + 70%|██████▉ | 363/520 [22:35<09:52, 3.78s/it] 70%|███████ | 364/520 [22:39<09:51, 3.79s/it] {'loss': 1.2103, 'grad_norm': 0.001265055716525186, 'learning_rate': 0.01528379796777323, 'epoch': 0.7} + 70%|███████ | 364/520 [22:39<09:51, 3.79s/it] 70%|███████ | 365/520 [22:43<09:44, 3.77s/it] {'loss': 1.2669, 'grad_norm': 0.0013197982702874821, 'learning_rate': 0.015103924825328774, 'epoch': 0.7} + 70%|███████ | 365/520 [22:43<09:44, 3.77s/it] 70%|███████ | 366/520 [22:46<09:35, 3.74s/it] {'loss': 1.2375, 'grad_norm': 0.0012534546080369316, 'learning_rate': 0.014924824727713398, 'epoch': 0.7} + 70%|███████ | 366/520 [22:46<09:35, 3.74s/it] 71%|███████ | 367/520 [22:50<09:28, 3.71s/it] {'loss': 1.2316, 'grad_norm': 0.0013352218401193438, 'learning_rate': 0.014746504633706801, 'epoch': 0.71} + 71%|███████ | 367/520 [22:50<09:28, 3.71s/it] 71%|███████ | 368/520 [22:54<09:22, 3.70s/it] {'loss': 1.0842, 'grad_norm': 0.0013128179525754675, 'learning_rate': 0.014568971471782365, 'epoch': 0.71} + 71%|███████ | 368/520 [22:54<09:22, 3.70s/it] 71%|███████ | 369/520 [22:57<09:17, 3.69s/it] {'loss': 1.1779, 'grad_norm': 0.0011517692622274225, 'learning_rate': 0.014392232139837839, 'epoch': 0.71} + 71%|███████ | 369/520 [22:57<09:17, 3.69s/it] 71%|███████ | 370/520 [23:01<09:11, 3.68s/it] {'loss': 1.1476, 'grad_norm': 0.0012379064545881989, 'learning_rate': 0.01421629350492745, 'epoch': 0.71} + 71%|███████ | 370/520 [23:01<09:11, 3.68s/it] 71%|███████▏ | 371/520 [23:05<09:06, 3.67s/it] {'loss': 1.131, 'grad_norm': 0.0013693486937965205, 'learning_rate': 0.014041162402994993, 'epoch': 0.71} + 71%|███████▏ | 371/520 [23:05<09:06, 3.67s/it] 72%|███████▏ | 372/520 [23:08<09:02, 3.66s/it] {'loss': 1.2421, 'grad_norm': 0.0011157417223105798, 'learning_rate': 0.013866845638608285, 'epoch': 0.72} + 72%|███████▏ | 372/520 [23:08<09:02, 3.66s/it] 72%|███████▏ | 373/520 [23:12<08:57, 3.66s/it] {'loss': 1.1327, 'grad_norm': 0.0013561057608889391, 'learning_rate': 0.013693349984694778, 'epoch': 0.72} + 72%|███████▏ | 373/520 [23:12<08:57, 3.66s/it] 72%|███████▏ | 374/520 [23:16<08:53, 3.65s/it] {'loss': 1.2306, 'grad_norm': 0.0013733015882458016, 'learning_rate': 0.013520682182278348, 'epoch': 0.72} + 72%|███████▏ | 374/520 [23:16<08:53, 3.65s/it] 72%|███████▏ | 375/520 [23:19<08:49, 3.65s/it] {'loss': 1.1473, 'grad_norm': 0.0013196140254702299, 'learning_rate': 0.013348848940217414, 'epoch': 0.72} + 72%|███████▏ | 375/520 [23:19<08:49, 3.65s/it] 72%|███████▏ | 376/520 [23:23<08:46, 3.66s/it] {'loss': 1.2534, 'grad_norm': 0.0012456184518684081, 'learning_rate': 0.013177856934944328, 'epoch': 0.72} + 72%|███████▏ | 376/520 [23:23<08:46, 3.66s/it] 72%|███████▎ | 377/520 [23:27<08:43, 3.66s/it] {'loss': 1.1842, 'grad_norm': 0.001402536305267108, 'learning_rate': 0.013007712810205846, 'epoch': 0.72} + 72%|███████▎ | 377/520 [23:27<08:43, 3.66s/it] 73%|███████▎ | 378/520 [23:30<08:39, 3.66s/it] {'loss': 1.2482, 'grad_norm': 0.0012246715337501201, 'learning_rate': 0.012838423176805112, 'epoch': 0.73} + 73%|███████▎ | 378/520 [23:30<08:39, 3.66s/it] 73%|███████▎ | 379/520 [23:34<08:35, 3.66s/it] {'loss': 1.2098, 'grad_norm': 0.0011992844011412696, 'learning_rate': 0.012669994612344705, 'epoch': 0.73} + 73%|███████▎ | 379/520 [23:34<08:35, 3.66s/it] 73%|███████▎ | 380/520 [23:38<08:31, 3.66s/it] {'loss': 1.2228, 'grad_norm': 0.0012794706429285795, 'learning_rate': 0.012502433660971124, 'epoch': 0.73} + 73%|███████▎ | 380/520 [23:38<08:31, 3.66s/it] 73%|███████▎ | 381/520 [23:41<08:35, 3.71s/it] {'loss': 1.2228, 'grad_norm': 0.0012322404781645727, 'learning_rate': 0.012335746833120541, 'epoch': 0.73} + 73%|███████▎ | 381/520 [23:41<08:35, 3.71s/it] 73%|███████▎ | 382/520 [23:45<08:37, 3.75s/it] {'loss': 1.1961, 'grad_norm': 0.0011598061670215199, 'learning_rate': 0.012169940605265772, 'epoch': 0.73} + 73%|███████▎ | 382/520 [23:45<08:37, 3.75s/it] 74%|███████▎ | 383/520 [23:49<08:30, 3.73s/it] {'loss': 1.0693, 'grad_norm': 0.001409126983282906, 'learning_rate': 0.01200502141966469, 'epoch': 0.74} + 74%|███████▎ | 383/520 [23:49<08:30, 3.73s/it] 74%|███████▍ | 384/520 [23:53<08:26, 3.73s/it] {'loss': 1.2104, 'grad_norm': 0.0010805196300306779, 'learning_rate': 0.01184099568410993, 'epoch': 0.74} + 74%|███████▍ | 384/520 [23:53<08:26, 3.73s/it] 74%|███████▍ | 385/520 [23:56<08:19, 3.70s/it] {'loss': 1.2138, 'grad_norm': 0.0012273185910940563, 'learning_rate': 0.011677869771679865, 'epoch': 0.74} + 74%|███████▍ | 385/520 [23:56<08:19, 3.70s/it] 74%|███████▍ | 386/520 [24:00<08:14, 3.69s/it] {'loss': 1.1604, 'grad_norm': 0.0011129481637346793, 'learning_rate': 0.011515650020491054, 'epoch': 0.74} + 74%|███████▍ | 386/520 [24:00<08:14, 3.69s/it] 74%|███████▍ | 387/520 [24:04<08:09, 3.68s/it] {'loss': 1.2397, 'grad_norm': 0.0012267881311715672, 'learning_rate': 0.011354342733451893, 'epoch': 0.74} + 74%|███████▍ | 387/520 [24:04<08:09, 3.68s/it] 75%|███████▍ | 388/520 [24:07<08:05, 3.68s/it] {'loss': 1.1224, 'grad_norm': 0.0012373591022383783, 'learning_rate': 0.011193954178017817, 'epoch': 0.75} + 75%|███████▍ | 388/520 [24:07<08:05, 3.68s/it] 75%|███████▍ | 389/520 [24:11<08:02, 3.68s/it] {'loss': 1.1707, 'grad_norm': 0.0016410433007983682, 'learning_rate': 0.011034490585947729, 'epoch': 0.75} + 75%|███████▍ | 389/520 [24:11<08:02, 3.68s/it] 75%|███████▌ | 390/520 [24:15<07:57, 3.68s/it] {'loss': 1.2382, 'grad_norm': 0.001223730995124632, 'learning_rate': 0.010875958153061855, 'epoch': 0.75} + 75%|███████▌ | 390/520 [24:15<07:57, 3.68s/it] 75%|███████▌ | 391/520 [24:18<07:55, 3.68s/it] {'loss': 1.2943, 'grad_norm': 0.0012921295370741561, 'learning_rate': 0.010718363039001044, 'epoch': 0.75} + 75%|███████▌ | 391/520 [24:18<07:55, 3.68s/it] 75%|███████▌ | 392/520 [24:22<07:50, 3.68s/it] {'loss': 1.123, 'grad_norm': 0.001259159656393515, 'learning_rate': 0.010561711366987456, 'epoch': 0.75} + 75%|███████▌ | 392/520 [24:22<07:50, 3.68s/it] 76%|███████▌ | 393/520 [24:26<07:46, 3.67s/it] {'loss': 1.1066, 'grad_norm': 0.0010753703732635878, 'learning_rate': 0.01040600922358658, 'epoch': 0.76} + 76%|███████▌ | 393/520 [24:26<07:46, 3.67s/it] 76%|███████▌ | 394/520 [24:29<07:44, 3.69s/it] {'loss': 1.1955, 'grad_norm': 0.0013394676979479258, 'learning_rate': 0.01025126265847084, 'epoch': 0.76} + 76%|███████▌ | 394/520 [24:29<07:44, 3.69s/it] 76%|███████▌ | 395/520 [24:33<07:40, 3.69s/it] {'loss': 1.1634, 'grad_norm': 0.0013868238964430926, 'learning_rate': 0.010097477684184454, 'epoch': 0.76} + 76%|███████▌ | 395/520 [24:33<07:40, 3.69s/it] 76%|███████▌ | 396/520 [24:37<07:35, 3.67s/it] {'loss': 1.2345, 'grad_norm': 0.0013583638616027104, 'learning_rate': 0.009944660275909855, 'epoch': 0.76} + 76%|███████▌ | 396/520 [24:37<07:35, 3.67s/it] 76%|███████▋ | 397/520 [24:40<07:33, 3.69s/it] {'loss': 1.2089, 'grad_norm': 0.0012620213818264657, 'learning_rate': 0.009792816371235578, 'epoch': 0.76} + 76%|███████▋ | 397/520 [24:40<07:33, 3.69s/it] 77%|███████▋ | 398/520 [24:44<07:29, 3.68s/it] {'loss': 1.1995, 'grad_norm': 0.0013576661436786026, 'learning_rate': 0.009641951869925457, 'epoch': 0.77} + 77%|███████▋ | 398/520 [24:44<07:29, 3.68s/it] 77%|███████▋ | 399/520 [24:48<07:26, 3.69s/it] {'loss': 1.1337, 'grad_norm': 0.0011969540985520327, 'learning_rate': 0.00949207263368951, 'epoch': 0.77} + 77%|███████▋ | 399/520 [24:48<07:26, 3.69s/it] 77%|███████▋ | 400/520 [24:52<07:22, 3.69s/it] {'loss': 1.1685, 'grad_norm': 0.0011546294158398798, 'learning_rate': 0.009343184485956087, 'epoch': 0.77} + 77%|███████▋ | 400/520 [44:33<07:22, 3.69s/it] 77%|███████▋ | 401/520 [44:37<11:50:29, 358.23s/it] {'loss': 1.0479, 'grad_norm': 0.0013979945564417129, 'learning_rate': 0.009195293211645661, 'epoch': 0.77} + 77%|███████▋ | 401/520 [44:37<11:50:29, 358.23s/it] 77%|███████▋ | 402/520 [44:41<8:15:19, 251.86s/it] {'loss': 1.178, 'grad_norm': 0.001331502123991238, 'learning_rate': 0.009048404556946065, 'epoch': 0.77} + 77%|███████▋ | 402/520 [44:41<8:15:19, 251.86s/it] 78%|███████▊ | 403/520 [44:44<5:45:57, 177.42s/it] {'loss': 1.1959, 'grad_norm': 0.001399658354618957, 'learning_rate': 0.008902524229089204, 'epoch': 0.78} + 78%|███████▊ | 403/520 [44:44<5:45:57, 177.42s/it] 78%|███████▊ | 404/520 [44:48<4:02:13, 125.29s/it] {'loss': 1.1118, 'grad_norm': 0.0015027002673988763, 'learning_rate': 0.0087576578961293, 'epoch': 0.78} + 78%|███████▊ | 404/520 [44:48<4:02:13, 125.29s/it] 78%|███████▊ | 405/520 [44:52<2:50:13, 88.82s/it] {'loss': 1.15, 'grad_norm': 0.0012079038199843356, 'learning_rate': 0.008613811186722707, 'epoch': 0.78} + 78%|███████▊ | 405/520 [44:52<2:50:13, 88.82s/it] 78%|███████▊ | 406/520 [44:55<2:00:14, 63.29s/it] {'loss': 1.078, 'grad_norm': 0.001496380021360577, 'learning_rate': 0.008470989689909142, 'epoch': 0.78} + 78%|███████▊ | 406/520 [44:55<2:00:14, 63.29s/it] 78%|███████▊ | 407/520 [44:59<1:25:31, 45.41s/it] {'loss': 1.2729, 'grad_norm': 0.0013029492513362251, 'learning_rate': 0.008329198954894623, 'epoch': 0.78} + 78%|███████▊ | 407/520 [44:59<1:25:31, 45.41s/it] 78%|███████▊ | 408/520 [45:03<1:01:23, 32.89s/it] {'loss': 1.1927, 'grad_norm': 0.0014418417222284681, 'learning_rate': 0.008188444490835774, 'epoch': 0.78} + 78%|███████▊ | 408/520 [45:03<1:01:23, 32.89s/it] 79%|███████▊ | 409/520 [45:07<44:42, 24.17s/it] {'loss': 1.3072, 'grad_norm': 0.0013623585785449638, 'learning_rate': 0.008048731766625803, 'epoch': 0.79} + 79%|███████▊ | 409/520 [45:07<44:42, 24.17s/it] 79%|███████▉ | 410/520 [45:10<33:05, 18.05s/it] {'loss': 1.0474, 'grad_norm': 0.0013247808376012603, 'learning_rate': 0.007910066210682042, 'epoch': 0.79} + 79%|███████▉ | 410/520 [45:10<33:05, 18.05s/it] 79%|███████▉ | 411/520 [45:14<25:02, 13.78s/it] {'loss': 1.2873, 'grad_norm': 0.0013784747534306306, 'learning_rate': 0.0077724532107349845, 'epoch': 0.79} + 79%|███████▉ | 411/520 [45:14<25:02, 13.78s/it] 79%|███████▉ | 412/520 [45:18<19:23, 10.78s/it] {'loss': 1.1977, 'grad_norm': 0.001271829609772669, 'learning_rate': 0.0076358981136189575, 'epoch': 0.79} + 79%|███████▉ | 412/520 [45:18<19:23, 10.78s/it] 79%|███████▉ | 413/520 [45:22<15:25, 8.65s/it] {'loss': 1.1703, 'grad_norm': 0.0012077660174441198, 'learning_rate': 0.007500406225064428, 'epoch': 0.79} + 79%|███████▉ | 413/520 [45:22<15:25, 8.65s/it] 80%|███████▉ | 414/520 [45:25<12:39, 7.16s/it] {'loss': 0.9792, 'grad_norm': 0.0010559102132620772, 'learning_rate': 0.0073659828094917645, 'epoch': 0.8} + 80%|███████▉ | 414/520 [45:25<12:39, 7.16s/it] 80%|███████▉ | 415/520 [45:29<10:42, 6.12s/it] {'loss': 1.1796, 'grad_norm': 0.0012296803304891258, 'learning_rate': 0.007232633089806774, 'epoch': 0.8} + 80%|███████▉ | 415/520 [45:29<10:42, 6.12s/it] 80%|████████ | 416/520 [45:33<09:22, 5.41s/it] {'loss': 1.0819, 'grad_norm': 0.0014162053839968012, 'learning_rate': 0.007100362247197725, 'epoch': 0.8} + 80%|████████ | 416/520 [45:33<09:22, 5.41s/it] 80%|████████ | 417/520 [45:36<08:22, 4.88s/it] {'loss': 1.2433, 'grad_norm': 0.0012756006616889014, 'learning_rate': 0.006969175420934026, 'epoch': 0.8} + 80%|████████ | 417/520 [45:36<08:22, 4.88s/it] 80%|████████ | 418/520 [45:40<07:40, 4.52s/it] {'loss': 1.2383, 'grad_norm': 0.001196194175782273, 'learning_rate': 0.0068390777081666085, 'epoch': 0.8} + 80%|████████ | 418/520 [45:40<07:40, 4.52s/it] 81%|████████ | 419/520 [45:44<07:11, 4.27s/it] {'loss': 1.2339, 'grad_norm': 0.0014099070182814948, 'learning_rate': 0.006710074163729818, 'epoch': 0.81} + 81%|████████ | 419/520 [45:44<07:11, 4.27s/it] 81%|████████ | 420/520 [45:48<06:48, 4.09s/it] {'loss': 1.1241, 'grad_norm': 0.001346350420164084, 'learning_rate': 0.006582169799945023, 'epoch': 0.81} + 81%|████████ | 420/520 [45:48<06:48, 4.09s/it] 81%|████████ | 421/520 [45:51<06:31, 3.95s/it] {'loss': 1.0626, 'grad_norm': 0.0013760116220815086, 'learning_rate': 0.006455369586425895, 'epoch': 0.81} + 81%|████████ | 421/520 [45:51<06:31, 3.95s/it] 81%|████████ | 422/520 [45:55<06:19, 3.87s/it] {'loss': 1.1863, 'grad_norm': 0.0013651550354527315, 'learning_rate': 0.006329678449885284, 'epoch': 0.81} + 81%|████████ | 422/520 [45:55<06:19, 3.87s/it] 81%|████████▏ | 423/520 [45:58<06:08, 3.80s/it] {'loss': 1.1526, 'grad_norm': 0.0013982401721233753, 'learning_rate': 0.006205101273943833, 'epoch': 0.81} + 81%|████████▏ | 423/520 [45:58<06:08, 3.80s/it] 82%|████████▏ | 424/520 [46:02<06:02, 3.77s/it] {'loss': 1.2479, 'grad_norm': 0.0011550599876903244, 'learning_rate': 0.006081642898940187, 'epoch': 0.82} + 82%|████████▏ | 424/520 [46:02<06:02, 3.77s/it] 82%|████████▏ | 425/520 [46:06<05:54, 3.73s/it] {'loss': 1.1651, 'grad_norm': 0.0012757213577086232, 'learning_rate': 0.005959308121742939, 'epoch': 0.82} + 82%|████████▏ | 425/520 [46:06<05:54, 3.73s/it] 82%|████████▏ | 426/520 [46:09<05:47, 3.70s/it] {'loss': 1.2056, 'grad_norm': 0.0016928272523417103, 'learning_rate': 0.005838101695564292, 'epoch': 0.82} + 82%|████████▏ | 426/520 [46:09<05:47, 3.70s/it] 82%|████████▏ | 427/520 [46:13<05:42, 3.68s/it] {'loss': 1.0973, 'grad_norm': 0.0012369965522982564, 'learning_rate': 0.0057180283297753095, 'epoch': 0.82} + 82%|████████▏ | 427/520 [46:13<05:42, 3.68s/it] 82%|████████▏ | 428/520 [46:17<05:37, 3.66s/it] {'loss': 1.0954, 'grad_norm': 0.0014056735849371166, 'learning_rate': 0.005599092689723002, 'epoch': 0.82} + 82%|████████▏ | 428/520 [46:17<05:37, 3.66s/it] 82%|████████▎ | 429/520 [46:20<05:33, 3.66s/it] {'loss': 1.1974, 'grad_norm': 0.0013117301938224632, 'learning_rate': 0.005481299396549008, 'epoch': 0.82} + 82%|████████▎ | 429/520 [46:20<05:33, 3.66s/it]Token indices sequence length is longer than the specified maximum sequence length for this model (2076 > 2048). Running this sequence through the model will result in indexing errors + 83%|████████▎ | 430/520 [46:24<05:29, 3.66s/it] {'loss': 1.195, 'grad_norm': 0.0012447332829931932, 'learning_rate': 0.005364653027010056, 'epoch': 0.83} + 83%|████████▎ | 430/520 [46:24<05:29, 3.66s/it] 83%|████████▎ | 431/520 [46:28<05:25, 3.66s/it] {'loss': 1.1387, 'grad_norm': 0.0012222814702945628, 'learning_rate': 0.005249158113300181, 'epoch': 0.83} + 83%|████████▎ | 431/520 [46:28<05:25, 3.66s/it] 83%|████████▎ | 432/520 [46:31<05:22, 3.67s/it] {'loss': 1.0997, 'grad_norm': 0.0013509695448575521, 'learning_rate': 0.005134819142874554, 'epoch': 0.83} + 83%|████████▎ | 432/520 [46:31<05:22, 3.67s/it] 83%|████████▎ | 433/520 [46:35<05:18, 3.66s/it] {'loss': 1.234, 'grad_norm': 0.001288271595769924, 'learning_rate': 0.005021640558275204, 'epoch': 0.83} + 83%|████████▎ | 433/520 [46:35<05:18, 3.66s/it] 83%|████████▎ | 434/520 [46:39<05:15, 3.66s/it] {'loss': 0.9899, 'grad_norm': 0.001337554869466896, 'learning_rate': 0.00490962675695834, 'epoch': 0.83} + 83%|████████▎ | 434/520 [46:39<05:15, 3.66s/it] 84%|████████▎ | 435/520 [46:42<05:11, 3.66s/it] {'loss': 1.2648, 'grad_norm': 0.0013870549423171793, 'learning_rate': 0.004798782091123544, 'epoch': 0.84} + 84%|████████▎ | 435/520 [46:42<05:11, 3.66s/it] 84%|████████▍ | 436/520 [46:46<05:07, 3.66s/it] {'loss': 1.077, 'grad_norm': 0.0013757920338344002, 'learning_rate': 0.004689110867544645, 'epoch': 0.84} + 84%|████████▍ | 436/520 [46:46<05:07, 3.66s/it] 84%|████████▍ | 437/520 [46:50<05:03, 3.66s/it] {'loss': 1.2871, 'grad_norm': 0.0013187191552207409, 'learning_rate': 0.004580617347402376, 'epoch': 0.84} + 84%|████████▍ | 437/520 [46:50<05:03, 3.66s/it] 84%|████████▍ | 438/520 [46:53<05:00, 3.67s/it] {'loss': 1.1058, 'grad_norm': 0.0013108469227964673, 'learning_rate': 0.004473305746118814, 'epoch': 0.84} + 84%|████████▍ | 438/520 [46:53<05:00, 3.67s/it] 84%|████████▍ | 439/520 [46:57<04:57, 3.67s/it] {'loss': 1.1253, 'grad_norm': 0.0010399341899998916, 'learning_rate': 0.0043671802331936216, 'epoch': 0.84} + 84%|████████▍ | 439/520 [46:57<04:57, 3.67s/it] 85%|████████▍ | 440/520 [47:01<04:53, 3.67s/it] {'loss': 1.1442, 'grad_norm': 0.0013582165403399405, 'learning_rate': 0.0042622449320419975, 'epoch': 0.85} + 85%|████████▍ | 440/520 [47:01<04:53, 3.67s/it] 85%|████████▍ | 441/520 [47:05<04:54, 3.73s/it] {'loss': 1.1343, 'grad_norm': 0.0012265082151035563, 'learning_rate': 0.004158503919834516, 'epoch': 0.85} + 85%|████████▍ | 441/520 [47:05<04:54, 3.73s/it] 85%|████████▌ | 442/520 [47:08<04:53, 3.76s/it] {'loss': 1.2065, 'grad_norm': 0.0014611423038326525, 'learning_rate': 0.004055961227338662, 'epoch': 0.85} + 85%|████████▌ | 442/520 [47:08<04:53, 3.76s/it] 85%|████████▌ | 443/520 [47:12<04:51, 3.78s/it] {'loss': 1.2121, 'grad_norm': 0.0012596887143696246, 'learning_rate': 0.00395462083876224, 'epoch': 0.85} + 85%|████████▌ | 443/520 [47:12<04:51, 3.78s/it] 85%|████████▌ | 444/520 [47:16<04:46, 3.77s/it] {'loss': 1.1822, 'grad_norm': 0.0011658903097740093, 'learning_rate': 0.003854486691598601, 'epoch': 0.85} + 85%|████████▌ | 444/520 [47:16<04:46, 3.77s/it] 86%|████████▌ | 445/520 [47:20<04:39, 3.73s/it] {'loss': 1.1077, 'grad_norm': 0.0012338321518858186, 'learning_rate': 0.003755562676473604, 'epoch': 0.86} + 86%|████████▌ | 445/520 [47:20<04:39, 3.73s/it] 86%|████████▌ | 446/520 [47:23<04:35, 3.72s/it] {'loss': 1.213, 'grad_norm': 0.0011311744462722021, 'learning_rate': 0.0036578526369944677, 'epoch': 0.86} + 86%|████████▌ | 446/520 [47:23<04:35, 3.72s/it] 86%|████████▌ | 447/520 [47:27<04:30, 3.70s/it] {'loss': 1.1754, 'grad_norm': 0.0012799124560174664, 'learning_rate': 0.003561360369600459, 'epoch': 0.86} + 86%|████████▌ | 447/520 [47:27<04:30, 3.70s/it] 86%|████████▌ | 448/520 [47:31<04:25, 3.69s/it] {'loss': 1.1779, 'grad_norm': 0.0014429927294357158, 'learning_rate': 0.003466089623415334, 'epoch': 0.86} + 86%|████████▌ | 448/520 [47:31<04:25, 3.69s/it] 86%|████████▋ | 449/520 [47:34<04:21, 3.68s/it] {'loss': 1.1753, 'grad_norm': 0.0012506980779415072, 'learning_rate': 0.0033720441001017236, 'epoch': 0.86} + 86%|████████▋ | 449/520 [47:34<04:21, 3.68s/it] 87%|████████▋ | 450/520 [47:38<04:16, 3.67s/it] {'loss': 1.2028, 'grad_norm': 0.0013354085333177098, 'learning_rate': 0.0032792274537172526, 'epoch': 0.87} + 87%|████████▋ | 450/520 [47:38<04:16, 3.67s/it] 87%|████████▋ | 451/520 [47:42<04:13, 3.67s/it] {'loss': 1.2053, 'grad_norm': 0.0013447344843189536, 'learning_rate': 0.003187643290572617, 'epoch': 0.87} + 87%|████████▋ | 451/520 [47:42<04:13, 3.67s/it] 87%|████████▋ | 452/520 [47:45<04:09, 3.67s/it] {'loss': 1.2228, 'grad_norm': 0.0011952034956579389, 'learning_rate': 0.0030972951690914404, 'epoch': 0.87} + 87%|████████▋ | 452/520 [47:45<04:09, 3.67s/it] 87%|████████▋ | 453/520 [47:49<04:04, 3.66s/it] {'loss': 1.1985, 'grad_norm': 0.0011990289409755698, 'learning_rate': 0.003008186599671995, 'epoch': 0.87} + 87%|████████▋ | 453/520 [47:49<04:04, 3.66s/it] 87%|████████▋ | 454/520 [47:52<04:00, 3.65s/it] {'loss': 1.1154, 'grad_norm': 0.0012926915506730162, 'learning_rate': 0.0029203210445508335, 'epoch': 0.87} + 87%|████████▋ | 454/520 [47:52<04:00, 3.65s/it] 88%|████████▊ | 455/520 [47:56<03:55, 3.63s/it] {'loss': 1.2517, 'grad_norm': 0.0013044561101496935, 'learning_rate': 0.0028337019176682774, 'epoch': 0.88} + 88%|████████▊ | 455/520 [47:56<03:55, 3.63s/it] 88%|████████▊ | 456/520 [48:00<03:51, 3.62s/it] {'loss': 1.1916, 'grad_norm': 0.0013447826700337024, 'learning_rate': 0.0027483325845357293, 'epoch': 0.88} + 88%|████████▊ | 456/520 [48:00<03:51, 3.62s/it] 88%|████████▊ | 457/520 [48:03<03:48, 3.63s/it] {'loss': 1.0745, 'grad_norm': 0.0011122130417397264, 'learning_rate': 0.0026642163621049646, 'epoch': 0.88} + 88%|████████▊ | 457/520 [48:03<03:48, 3.63s/it] 88%|████████▊ | 458/520 [48:07<03:45, 3.64s/it] {'loss': 1.3032, 'grad_norm': 0.0013928411175924567, 'learning_rate': 0.0025813565186391975, 'epoch': 0.88} + 88%|████████▊ | 458/520 [48:07<03:45, 3.64s/it] 88%|████████▊ | 459/520 [48:11<03:41, 3.64s/it] {'loss': 1.2358, 'grad_norm': 0.0012707723363443101, 'learning_rate': 0.0024997562735861257, 'epoch': 0.88} + 88%|████████▊ | 459/520 [48:11<03:41, 3.64s/it] 88%|████████▊ | 460/520 [48:14<03:37, 3.63s/it] {'loss': 1.1274, 'grad_norm': 0.001286504701605234, 'learning_rate': 0.0024194187974528555, 'epoch': 0.88} + 88%|████████▊ | 460/520 [48:14<03:37, 3.63s/it] 89%|████████▊ | 461/520 [48:18<03:34, 3.63s/it] {'loss': 1.1587, 'grad_norm': 0.0009149015738430847, 'learning_rate': 0.0023403472116826727, 'epoch': 0.89} + 89%|████████▊ | 461/520 [48:18<03:34, 3.63s/it] 89%|████████▉ | 462/520 [48:21<03:30, 3.63s/it] {'loss': 1.2604, 'grad_norm': 0.00121230218682169, 'learning_rate': 0.0022625445885338102, 'epoch': 0.89} + 89%|████████▉ | 462/520 [48:21<03:30, 3.63s/it] 89%|████████▉ | 463/520 [48:25<03:27, 3.63s/it] {'loss': 1.1052, 'grad_norm': 0.0013671757239736972, 'learning_rate': 0.002186013950960032, 'epoch': 0.89} + 89%|████████▉ | 463/520 [48:25<03:27, 3.63s/it] 89%|████████▉ | 464/520 [48:29<03:23, 3.64s/it] {'loss': 1.2195, 'grad_norm': 0.0013128830067696627, 'learning_rate': 0.0021107582724932093, 'epoch': 0.89} + 89%|████████▉ | 464/520 [48:29<03:23, 3.64s/it] 89%|████████▉ | 465/520 [48:32<03:20, 3.64s/it] {'loss': 1.3213, 'grad_norm': 0.001336264998904904, 'learning_rate': 0.002036780477127779, 'epoch': 0.89} + 89%|████████▉ | 465/520 [48:32<03:20, 3.64s/it] 90%|████████▉ | 466/520 [48:36<03:16, 3.64s/it] {'loss': 1.2214, 'grad_norm': 0.0011745461675090989, 'learning_rate': 0.0019640834392071352, 'epoch': 0.9} + 90%|████████▉ | 466/520 [48:36<03:16, 3.64s/it] 90%|████████▉ | 467/520 [48:40<03:13, 3.65s/it] {'loss': 1.1499, 'grad_norm': 0.0011826418701880604, 'learning_rate': 0.0018926699833119395, 'epoch': 0.9} + 90%|████████▉ | 467/520 [48:40<03:13, 3.65s/it] 90%|█████████ | 468/520 [48:43<03:09, 3.65s/it] {'loss': 1.1806, 'grad_norm': 0.0014739357565698495, 'learning_rate': 0.0018225428841503906, 'epoch': 0.9} + 90%|█████████ | 468/520 [48:43<03:09, 3.65s/it] 90%|█████████ | 469/520 [48:47<03:05, 3.64s/it] {'loss': 1.2544, 'grad_norm': 0.0013538487953721468, 'learning_rate': 0.0017537048664503904, 'epoch': 0.9} + 90%|█████████ | 469/520 [48:47<03:05, 3.64s/it] 90%|█████████ | 470/520 [48:51<03:06, 3.72s/it] {'loss': 1.1233, 'grad_norm': 0.001190606423656073, 'learning_rate': 0.0016861586048537176, 'epoch': 0.9} + 90%|█████████ | 470/520 [48:51<03:06, 3.72s/it] 91%|█████████ | 471/520 [48:55<03:01, 3.70s/it] {'loss': 1.158, 'grad_norm': 0.0014042147332900422, 'learning_rate': 0.0016199067238120615, 'epoch': 0.91} + 91%|█████████ | 471/520 [48:55<03:01, 3.70s/it] 91%|█████████ | 472/520 [48:58<02:57, 3.69s/it] {'loss': 1.1232, 'grad_norm': 0.0012395378991416125, 'learning_rate': 0.0015549517974850725, 'epoch': 0.91} + 91%|█████████ | 472/520 [48:58<02:57, 3.69s/it] 91%|█████████ | 473/520 [49:02<02:53, 3.69s/it] {'loss': 1.1989, 'grad_norm': 0.001377149775076386, 'learning_rate': 0.0014912963496403677, 'epoch': 0.91} + 91%|█████████ | 473/520 [49:02<02:53, 3.69s/it] 91%|█████████ | 474/520 [49:06<02:49, 3.68s/it] {'loss': 1.1833, 'grad_norm': 0.0011682481406036904, 'learning_rate': 0.0014289428535554285, 'epoch': 0.91} + 91%|█████████ | 474/520 [49:06<02:49, 3.68s/it] 91%|█████████▏| 475/520 [49:09<02:45, 3.68s/it] {'loss': 1.1012, 'grad_norm': 0.0011808895152998792, 'learning_rate': 0.001367893731921518, 'epoch': 0.91} + 91%|█████████▏| 475/520 [49:09<02:45, 3.68s/it] 92%|█████████▏| 476/520 [49:13<02:41, 3.67s/it] {'loss': 1.1774, 'grad_norm': 0.0013743815257492998, 'learning_rate': 0.0013081513567495791, 'epoch': 0.92} + 92%|█████████▏| 476/520 [49:13<02:41, 3.67s/it] 92%|█████████▏| 477/520 [49:17<02:37, 3.65s/it] {'loss': 1.1742, 'grad_norm': 0.0014232079508802904, 'learning_rate': 0.001249718049278032, 'epoch': 0.92} + 92%|█████████▏| 477/520 [49:17<02:37, 3.65s/it] 92%|█████████▏| 478/520 [49:20<02:33, 3.65s/it] {'loss': 1.1158, 'grad_norm': 0.001245204133898875, 'learning_rate': 0.001192596079882613, 'epoch': 0.92} + 92%|█████████▏| 478/520 [49:20<02:33, 3.65s/it] 92%|█████████▏| 479/520 [49:24<02:29, 3.64s/it] {'loss': 1.1626, 'grad_norm': 0.001353961368879092, 'learning_rate': 0.0011367876679881362, 'epoch': 0.92} + 92%|█████████▏| 479/520 [49:24<02:29, 3.64s/it] 92%|█████████▏| 480/520 [49:27<02:25, 3.64s/it] {'loss': 1.1701, 'grad_norm': 0.0011734705071345909, 'learning_rate': 0.0010822949819822754, 'epoch': 0.92} + 92%|█████████▏| 480/520 [49:27<02:25, 3.64s/it] 92%|█████████▎| 481/520 [49:31<02:22, 3.66s/it] {'loss': 1.1621, 'grad_norm': 0.0011228719987949126, 'learning_rate': 0.0010291201391313166, 'epoch': 0.93} + 92%|█████████▎| 481/520 [49:31<02:22, 3.66s/it] 93%|█████████▎| 482/520 [49:35<02:19, 3.66s/it] {'loss': 1.1834, 'grad_norm': 0.001148369083598399, 'learning_rate': 0.0009772652054978925, 'epoch': 0.93} + 93%|█████████▎| 482/520 [49:35<02:19, 3.66s/it] 93%|█████████▎| 483/520 [49:38<02:15, 3.66s/it] {'loss': 1.1844, 'grad_norm': 0.001286904953028503, 'learning_rate': 0.0009267321958606829, 'epoch': 0.93} + 93%|█████████▎| 483/520 [49:38<02:15, 3.66s/it] 93%|█████████▎| 484/520 [49:42<02:11, 3.66s/it] {'loss': 1.1972, 'grad_norm': 0.001339202584702413, 'learning_rate': 0.0008775230736361734, 'epoch': 0.93} + 93%|█████████▎| 484/520 [49:42<02:11, 3.66s/it] 93%|█████████▎| 485/520 [49:46<02:08, 3.66s/it] {'loss': 1.1456, 'grad_norm': 0.001263712597591797, 'learning_rate': 0.0008296397508023324, 'epoch': 0.93} + 93%|█████████▎| 485/520 [49:46<02:08, 3.66s/it] 93%|█████████▎| 486/520 [49:49<02:03, 3.64s/it] {'loss': 1.2669, 'grad_norm': 0.0013353470255456142, 'learning_rate': 0.0007830840878243412, 'epoch': 0.93} + 93%|█████████▎| 486/520 [49:49<02:03, 3.64s/it] 94%|█████████▎| 487/520 [49:53<02:00, 3.66s/it] {'loss': 1.1328, 'grad_norm': 0.0012453313531103057, 'learning_rate': 0.0007378578935823072, 'epoch': 0.94} + 94%|█████████▎| 487/520 [49:53<02:00, 3.66s/it] 94%|█████████▍| 488/520 [49:57<01:56, 3.64s/it] {'loss': 1.0764, 'grad_norm': 0.001318003476399678, 'learning_rate': 0.0006939629253009661, 'epoch': 0.94} + 94%|█████████▍| 488/520 [49:57<01:56, 3.64s/it] 94%|█████████▍| 489/520 [50:00<01:53, 3.65s/it] {'loss': 1.1821, 'grad_norm': 0.0010828187302708053, 'learning_rate': 0.0006514008884814321, 'epoch': 0.94} + 94%|█████████▍| 489/520 [50:00<01:53, 3.65s/it] 94%|█████████▍| 490/520 [50:04<01:48, 3.63s/it] {'loss': 1.1891, 'grad_norm': 0.0013174969270532708, 'learning_rate': 0.0006101734368349105, 'epoch': 0.94} + 94%|█████████▍| 490/520 [50:04<01:48, 3.63s/it] 94%|█████████▍| 491/520 [50:08<01:44, 3.62s/it] {'loss': 1.1552, 'grad_norm': 0.0013874951121286593, 'learning_rate': 0.0005702821722184538, 'epoch': 0.94} + 94%|█████████▍| 491/520 [50:08<01:44, 3.62s/it] 95%|█████████▍| 492/520 [50:11<01:41, 3.61s/it] {'loss': 1.2658, 'grad_norm': 0.0013201768747347453, 'learning_rate': 0.0005317286445727194, 'epoch': 0.95} + 95%|█████████▍| 492/520 [50:11<01:41, 3.61s/it] 95%|█████████▍| 493/520 [50:15<01:37, 3.61s/it] {'loss': 1.1797, 'grad_norm': 0.0013554783973331094, 'learning_rate': 0.000494514351861744, 'epoch': 0.95} + 95%|█████████▍| 493/520 [50:15<01:37, 3.61s/it] 95%|█████████▌| 494/520 [50:18<01:34, 3.62s/it] {'loss': 1.1917, 'grad_norm': 0.0011765053030060577, 'learning_rate': 0.0004586407400147619, 'epoch': 0.95} + 95%|█████████▌| 494/520 [50:18<01:34, 3.62s/it] 95%|█████████▌| 495/520 [50:22<01:30, 3.60s/it] {'loss': 1.1793, 'grad_norm': 0.0013320986347618617, 'learning_rate': 0.000424109202869985, 'epoch': 0.95} + 95%|█████████▌| 495/520 [50:22<01:30, 3.60s/it] 95%|█████████▌| 496/520 [50:26<01:26, 3.62s/it] {'loss': 1.1142, 'grad_norm': 0.0013898387670674593, 'learning_rate': 0.0003909210821205017, 'epoch': 0.95} + 95%|█████████▌| 496/520 [50:26<01:26, 3.62s/it] 96%|█████████▌| 497/520 [50:29<01:23, 3.62s/it] {'loss': 1.1144, 'grad_norm': 0.0011025699223335985, 'learning_rate': 0.0003590776672620905, 'epoch': 0.96} + 96%|█████████▌| 497/520 [50:29<01:23, 3.62s/it] 96%|█████████▌| 498/520 [50:33<01:20, 3.64s/it] {'loss': 1.1682, 'grad_norm': 0.001283792259187783, 'learning_rate': 0.0003285801955431517, 'epoch': 0.96} + 96%|█████████▌| 498/520 [50:33<01:20, 3.64s/it] 96%|█████████▌| 499/520 [50:37<01:16, 3.65s/it] {'loss': 1.251, 'grad_norm': 0.0011776167419112176, 'learning_rate': 0.00029942985191663666, 'epoch': 0.96} + 96%|█████████▌| 499/520 [50:37<01:16, 3.65s/it] 96%|█████████▌| 500/520 [50:40<01:12, 3.65s/it] {'loss': 1.2932, 'grad_norm': 0.0015022306517585638, 'learning_rate': 0.0002716277689939778, 'epoch': 0.96} + 96%|█████████▌| 500/520 [50:40<01:12, 3.65s/it] 96%|█████████▋| 501/520 [50:44<01:09, 3.66s/it] {'loss': 1.1551, 'grad_norm': 0.0013100977831646634, 'learning_rate': 0.0002451750270011133, 'epoch': 0.96} + 96%|█████████▋| 501/520 [50:44<01:09, 3.66s/it] 97%|█████████▋| 502/520 [50:48<01:06, 3.67s/it] {'loss': 1.2036, 'grad_norm': 0.001239820244280822, 'learning_rate': 0.00022007265373650889, 'epoch': 0.97} + 97%|█████████▋| 502/520 [50:48<01:06, 3.67s/it] 97%|█████████▋| 503/520 [50:51<01:02, 3.67s/it] {'loss': 1.1399, 'grad_norm': 0.0012158247292281888, 'learning_rate': 0.0001963216245312083, 'epoch': 0.97} + 97%|█████████▋| 503/520 [50:51<01:02, 3.67s/it] 97%|█████████▋| 504/520 [50:55<00:58, 3.67s/it] {'loss': 1.1938, 'grad_norm': 0.0015030528963919647, 'learning_rate': 0.0001739228622109507, 'epoch': 0.97} + 97%|█████████▋| 504/520 [50:55<00:58, 3.67s/it] 97%|█████████▋| 505/520 [50:59<00:54, 3.66s/it] {'loss': 1.2244, 'grad_norm': 0.0013006339954026724, 'learning_rate': 0.00015287723706031655, 'epoch': 0.97} + 97%|█████████▋| 505/520 [50:59<00:54, 3.66s/it] 97%|█████████▋| 506/520 [51:02<00:51, 3.65s/it] {'loss': 1.1577, 'grad_norm': 0.0013686984830410247, 'learning_rate': 0.00013318556678890592, 'epoch': 0.97} + 97%|█████████▋| 506/520 [51:02<00:51, 3.65s/it] 98%|█████████▊| 507/520 [51:06<00:47, 3.65s/it] {'loss': 1.2834, 'grad_norm': 0.0011269776243512338, 'learning_rate': 0.00011484861649957213, 'epoch': 0.97} + 98%|█████████▊| 507/520 [51:06<00:47, 3.65s/it] 98%|█████████▊| 508/520 [51:09<00:43, 3.64s/it] {'loss': 1.2808, 'grad_norm': 0.0013482577744701988, 'learning_rate': 9.786709865869549e-05, 'epoch': 0.98} + 98%|█████████▊| 508/520 [51:09<00:43, 3.64s/it] 98%|█████████▊| 509/520 [51:13<00:40, 3.65s/it] {'loss': 1.2468, 'grad_norm': 0.0012465622052021798, 'learning_rate': 8.22416730684894e-05, 'epoch': 0.98} + 98%|█████████▊| 509/520 [51:13<00:40, 3.65s/it] 98%|█████████▊| 510/520 [51:17<00:36, 3.65s/it] {'loss': 1.1968, 'grad_norm': 0.0012989908178302825, 'learning_rate': 6.797294684138533e-05, 'epoch': 0.98} + 98%|█████████▊| 510/520 [51:17<00:36, 3.65s/it] 98%|█████████▊| 511/520 [51:20<00:32, 3.65s/it] {'loss': 1.1558, 'grad_norm': 0.0012677292030156893, 'learning_rate': 5.506147437641884e-05, 'epoch': 0.98} + 98%|█████████▊| 511/520 [51:20<00:32, 3.65s/it] 98%|█████████▊| 512/520 [51:24<00:29, 3.64s/it] {'loss': 1.053, 'grad_norm': 0.0012658114956323396, 'learning_rate': 4.350775733771795e-05, 'epoch': 0.98} + 98%|█████████▊| 512/520 [51:24<00:29, 3.64s/it] 99%|█████████▊| 513/520 [51:28<00:25, 3.66s/it] {'loss': 1.2473, 'grad_norm': 0.0014494602502988065, 'learning_rate': 3.3312244634977066e-05, 'epoch': 0.99} + 99%|█████████▊| 513/520 [51:28<00:25, 3.66s/it] 99%|█████████▉| 514/520 [51:31<00:21, 3.66s/it] {'loss': 1.216, 'grad_norm': 0.0011977513507102223, 'learning_rate': 2.4475332406048713e-05, 'epoch': 0.99} + 99%|█████████▉| 514/520 [51:31<00:21, 3.66s/it] 99%|█████████▉| 515/520 [51:35<00:18, 3.65s/it] {'loss': 1.2732, 'grad_norm': 0.0015413444376266008, 'learning_rate': 1.6997364001532512e-05, 'epoch': 0.99} + 99%|█████████▉| 515/520 [51:35<00:18, 3.65s/it] 99%|█████████▉| 516/520 [51:39<00:14, 3.66s/it] {'loss': 1.1847, 'grad_norm': 0.0013115471461120846, 'learning_rate': 1.087862997143141e-05, 'epoch': 0.99} + 99%|█████████▉| 516/520 [51:39<00:14, 3.66s/it] 99%|█████████▉| 517/520 [51:42<00:10, 3.64s/it] {'loss': 1.1812, 'grad_norm': 0.001185100859151014, 'learning_rate': 6.119368053875141e-06, 'epoch': 0.99} + 99%|█████████▉| 517/520 [51:42<00:10, 3.64s/it] 100%|█████████▉| 518/520 [51:46<00:07, 3.61s/it] {'loss': 1.1857, 'grad_norm': 0.0013708625892964502, 'learning_rate': 2.719763165879852e-06, 'epoch': 1.0} + 100%|█████████▉| 518/520 [51:46<00:07, 3.61s/it] 100%|█████████▉| 519/520 [51:49<00:03, 3.61s/it] {'loss': 1.1606, 'grad_norm': 0.001246638956466991, 'learning_rate': 6.79947396163283e-07, 'epoch': 1.0} + 100%|█████████▉| 519/520 [51:49<00:03, 3.61s/it] 100%|██████████| 520/520 [51:54<00:00, 3.89s/it] {'loss': 1.144, 'grad_norm': 0.0010970428584545064, 'learning_rate': 0.0, 'epoch': 1.0} + 100%|██████████| 520/520 [51:54<00:00, 3.89s/it] {'train_runtime': 3114.5486, 'train_samples_per_second': 21.361, 'train_steps_per_second': 0.167, 'train_loss': 1.2381137208296702, 'epoch': 1.0} + 100%|██████████| 520/520 [51:54<00:00, 3.89s/it] 100%|██████████| 520/520 [51:54<00:00, 5.99s/it] +[2025-10-12 14:28:54,890] [INFO] [launch.py:348:main] Process 732987 exits successfully. +[2025-10-12 14:28:54,890] [INFO] [launch.py:348:main] Process 732988 exits successfully. +[2025-10-12 14:28:55,892] [INFO] [launch.py:348:main] Process 732984 exits successfully. +[2025-10-12 14:28:55,892] [INFO] [launch.py:348:main] Process 732986 exits successfully. +[2025-10-12 14:28:55,892] [INFO] [launch.py:348:main] Process 732985 exits successfully. +[2025-10-12 14:28:55,893] [INFO] [launch.py:348:main] Process 732983 exits successfully. +[2025-10-12 14:28:56,894] [INFO] [launch.py:348:main] Process 732989 exits successfully. +[2025-10-12 14:28:59,898] [INFO] [launch.py:348:main] Process 732982 exits successfully. +==== EXPERIMENT COMPLETED: qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_7e-2_connector-3.0_0.5_7e-2_ablation ==== +Log File: qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_7e-2_connector-3.0_0.5_7e-2_ablation_20251012_132440.log +Timestamp: 2025-10-12 14:29:02 +===================================== diff --git a/logs_oct11/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_9e-1_connector-3.0_0.5_9e-1_ablation_20251012_081416.log b/logs_oct11/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_9e-1_connector-3.0_0.5_9e-1_ablation_20251012_081416.log new file mode 100644 index 0000000000000000000000000000000000000000..e98f6a094d159a9917e0d1a2d7e14584c3fe8d73 --- /dev/null +++ b/logs_oct11/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_9e-1_connector-3.0_0.5_9e-1_ablation_20251012_081416.log @@ -0,0 +1,2312 @@ +==== STARTING EXPERIMENT: qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_9e-1_connector-3.0_0.5_9e-1_ablation ==== +Log File: qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_9e-1_connector-3.0_0.5_9e-1_ablation_20251012_081416.log +Timestamp: 2025-10-12 08:14:16 +===================================== +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 08:14:19,481] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 08:14:22,195] [WARNING] [runner.py:202:fetch_hostfile] Unable to find hostfile, will proceed with training with local resources only. +[2025-10-12 08:14:22,197] [INFO] [runner.py:568:main] cmd = /opt/conda/envs/tinyllava/bin/python3.10 -u -m deepspeed.launcher.launch --world_info=eyJsb2NhbGhvc3QiOiBbMCwgMSwgMiwgMywgNCwgNSwgNiwgN119 --master_addr=127.0.0.1 --master_port=29501 --enable_each_rank_log=None tinyllava/train/train.py --deepspeed ./scripts/zero3.json --data_path /s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json --image_folder /s3-code/ywang29/datasets/tinyllava --is_multimodal True --conv_version qwen2_base --model_name_or_path Qwen/Qwen2.5-0.5B --vision_tower google/siglip-so400m-patch14-384 --vision_tower2 --connector_type mlp2x_gelu --mm_vision_select_layer -2 --image_aspect_ratio square --attn_implementation flash_attention_2 --bf16 True --training_recipe common --tune_type_llm full --tune_type_vision_tower frozen --tune_vision_tower_from_layer 0 --tune_type_connector full --group_by_modality_length True --pretrained_model_path /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain --output_dir /nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_9e-1_connector-3.0_0.5_9e-1_ablation --num_train_epochs 1 --per_device_train_batch_size 4 --per_device_eval_batch_size 4 --gradient_accumulation_steps 4 --evaluation_strategy no --learning_rate 9e-1 --weight_decay 0. --warmup_ratio 0.03 --lr_scheduler_type cosine --logging_steps 1 --tf32 False --model_max_length 2048 --gradient_checkpointing True --dataloader_num_workers 8 --lazy_preprocess True --report_to tensorboard --tokenizer_use_fast False --run_name tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune --subnet_mode_text both --subnet_type_text None --mask_type_text soft --init_mean_text 3.0 --temperature_attn_text 0.5 --temperature_mlp_text 0.5 --backward_type_text normal --masked_layers_text all --subnet_mode_vision both --subnet_type_vision None --mask_type_vision soft --init_mean_vision 3.0 --temperature_attn_vision 0.5 --temperature_mlp_vision 0.5 --backward_type_vision normal --masked_layers_vision all --subnet_type_connector global --mask_type_connector soft --init_mean_connector 3.0 --temperature_connector 0.5 --backward_type_connector normal --mm_projector_lr 9e-1 --seed 42 --mask_model llm-connector --save_strategy steps --save_steps 50000 --save_total_limit 1 --train_data_ratio 0.1 +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 08:14:24,818] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 08:14:25,951] [INFO] [launch.py:138:main] 0 NCCL_VERSION=2.21.5 +[2025-10-12 08:14:25,951] [INFO] [launch.py:138:main] 0 NCCL_SOCKET_IFNAME=eth +[2025-10-12 08:14:25,951] [INFO] [launch.py:145:main] WORLD INFO DICT: {'localhost': [0, 1, 2, 3, 4, 5, 6, 7]} +[2025-10-12 08:14:25,951] [INFO] [launch.py:151:main] nnodes=1, num_local_procs=8, node_rank=0 +[2025-10-12 08:14:25,951] [INFO] [launch.py:162:main] global_rank_mapping=defaultdict(, {'localhost': [0, 1, 2, 3, 4, 5, 6, 7]}) +[2025-10-12 08:14:25,951] [INFO] [launch.py:163:main] dist_world_size=8 +[2025-10-12 08:14:25,951] [INFO] [launch.py:165:main] Setting CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 +[2025-10-12 08:14:25,953] [INFO] [launch.py:253:main] process 531651 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=0', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_9e-1_connector-3.0_0.5_9e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '9e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '9e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 08:14:25,955] [INFO] [launch.py:253:main] process 531652 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=1', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_9e-1_connector-3.0_0.5_9e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '9e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '9e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 08:14:25,957] [INFO] [launch.py:253:main] process 531653 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=2', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_9e-1_connector-3.0_0.5_9e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '9e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '9e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 08:14:25,959] [INFO] [launch.py:253:main] process 531654 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=3', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_9e-1_connector-3.0_0.5_9e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '9e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '9e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 08:14:25,961] [INFO] [launch.py:253:main] process 531655 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=4', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_9e-1_connector-3.0_0.5_9e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '9e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '9e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 08:14:25,963] [INFO] [launch.py:253:main] process 531656 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=5', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_9e-1_connector-3.0_0.5_9e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '9e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '9e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 08:14:25,964] [INFO] [launch.py:253:main] process 531657 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=6', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_9e-1_connector-3.0_0.5_9e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '9e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '9e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 08:14:25,966] [INFO] [launch.py:253:main] process 531658 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=7', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_9e-1_connector-3.0_0.5_9e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '9e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '9e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 08:14:32,652] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 08:14:32,872] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 08:14:32,965] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 08:14:33,055] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 08:14:33,081] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 08:14:33,081] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 08:14:33,083] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 08:14:33,084] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 08:14:33,084] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 08:14:33,268] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 08:14:33,370] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 08:14:33,495] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 08:14:33,495] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 08:14:33,495] [INFO] [comm.py:668:init_distributed] Initializing TorchBackend in DeepSpeed with backend nccl +[2025-10-12 08:14:33,497] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 08:14:33,497] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 08:14:33,502] [INFO] [comm.py:637:init_distributed] cdb=None +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +{'llm': {'model_name_or_path': 'Qwen/Qwen2.5-0.5B', 'cache_dir': None, 'attn_implementation': 'flash_attention_2', 'subnet_mode': 'both', 'subnet_type': 'None', 'sparsity_attn': None, 'sparsity_mlp': None, 'threshold_attn': None, 'threshold_mlp': None, 'temperature_attn': 0.5, 'temperature_mlp': 0.5, 'masked_layers': 'all', 'mask_type': 'soft', 'backward_type': 'normal'}, 'vision_tower': {'model_name_or_path': 'google/siglip-so400m-patch14-384'}, 'connector': {'connector_type': 'mlp2x_gelu', 'subnet_type': 'global', 'threshold': None, 'sparsity': None, 'temperature': 0.5, 'mask_type': 'soft', 'backward_type': 'normal'}} +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +TinyLlavaConfig { + "backward_type_connector": "normal", + "cache_dir": null, + "connector_type": "mlp2x_gelu", + "hidden_size": 896, + "ignore_index": -100, + "image_aspect_ratio": "square", + "image_token_index": -200, + "llm_model_name_or_path": "Qwen/Qwen2.5-0.5B", + "mask_model": [ + "llm", + "connector" + ], + "mask_type_connector": "soft", + "model_type": "tinyllava", + "num_queries": 128, + "num_resampler_layers": 3, + "pad_token": null, + "resampler_hidden_size": 768, + "sparsity_connector": null, + "subnet_type_connector": "global", + "temperature_connector": 0.5, + "text_config": { + "_name_or_path": "Qwen/Qwen2.5-0.5B", + "architectures": [ + "Qwen2ForCausalLM" + ], + "backward_type": "normal", + "bos_token_id": 151643, + "eos_token_id": 151643, + "hidden_size": 896, + "intermediate_size": 4864, + "mask_type": "soft", + "masked_layers": "all", + "max_position_embeddings": 32768, + "max_window_layers": 24, + "model_type": "qwen2", + "num_attention_heads": 14, + "num_hidden_layers": 24, + "num_key_value_heads": 2, + "rope_theta": 1000000.0, + "sliding_window": 32768, + "subnet_mode": "both", + "subnet_type": "None", + "temperature_attn": 0.5, + "temperature_mlp": 0.5, + "tie_word_embeddings": true, + "torch_dtype": "bfloat16", + "use_mrope": false, + "use_sliding_window": false, + "vocab_size": 151936 + }, + "threshold_connector": null, + "tokenizer_model_max_length": 2048, + "tokenizer_name_or_path": "Qwen/Qwen2.5-0.5B", + "tokenizer_padding_side": "right", + "tokenizer_use_fast": false, + "transformers_version": "4.40.1", + "tune_type_connector": "frozen", + "tune_type_llm": "frozen", + "tune_type_vision_tower": "frozen", + "tune_vision_tower_from_layer": -1, + "use_cache": false, + "vision_config": { + "hidden_act": "gelu_pytorch_tanh", + "hidden_size": 1152, + "image_size": 384, + "intermediate_size": 4304, + "layer_norm_eps": 1e-06, + "model_name_or_path": "google/siglip-so400m-patch14-384", + "model_name_or_path2": "", + "model_type": "siglip_vision_model", + "num_attention_heads": 16, + "num_hidden_layers": 27, + "patch_size": 14 + }, + "vision_feature_layer": -2, + "vision_feature_select_strategy": "patch", + "vision_hidden_size": 1152, + "vision_model_name_or_path": "google/siglip-so400m-patch14-384", + "vision_model_name_or_path2": "", + "vocab_size": 151936 +} + +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +ywang29-vrdb-test2-worker-0:531651:531651 [0] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:531651:531651 [0] NCCL INFO Bootstrap : Using eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:531651:531651 [0] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test2-worker-0:531651:531651 [0] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test2-worker-0:531651:531651 [0] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test2-worker-0:531651:531651 [0] NCCL INFO cudaDriverVersion 12040 +NCCL version 2.21.5+cuda12.1 +ywang29-vrdb-test2-worker-0:531653:531653 [2] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test2-worker-0:531653:531653 [2] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:531653:531653 [2] NCCL INFO Bootstrap : Using eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:531653:531653 [2] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test2-worker-0:531653:531653 [2] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test2-worker-0:531653:531653 [2] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test2-worker-0:531658:531658 [7] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test2-worker-0:531658:531658 [7] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:531658:531658 [7] NCCL INFO Bootstrap : Using eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:531656:531656 [5] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test2-worker-0:531658:531658 [7] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test2-worker-0:531658:531658 [7] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test2-worker-0:531658:531658 [7] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test2-worker-0:531656:531656 [5] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:531656:531656 [5] NCCL INFO Bootstrap : Using eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:531656:531656 [5] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test2-worker-0:531656:531656 [5] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test2-worker-0:531656:531656 [5] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test2-worker-0:531651:533252 [0] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:531651:533252 [0] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test2-worker-0:531651:533252 [0] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:531651:533252 [0] NCCL INFO NET/Socket : Using [0]eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:531651:533252 [0] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test2-worker-0:531651:533252 [0] NCCL INFO Using network Socket +ywang29-vrdb-test2-worker-0:531653:533253 [2] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:531653:533253 [2] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test2-worker-0:531653:533253 [2] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:531653:533253 [2] NCCL INFO NET/Socket : Using [0]eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:531653:533253 [2] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test2-worker-0:531653:533253 [2] NCCL INFO Using network Socket +ywang29-vrdb-test2-worker-0:531656:533255 [5] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:531658:533254 [7] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:531656:533255 [5] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test2-worker-0:531658:533254 [7] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test2-worker-0:531658:533254 [7] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:531656:533255 [5] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:531656:533255 [5] NCCL INFO NET/Socket : Using [0]eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:531658:533254 [7] NCCL INFO NET/Socket : Using [0]eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:531656:533255 [5] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test2-worker-0:531658:533254 [7] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test2-worker-0:531658:533254 [7] NCCL INFO Using network Socket +ywang29-vrdb-test2-worker-0:531656:533255 [5] NCCL INFO Using network Socket +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test2-worker-0:531655:531655 [4] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test2-worker-0:531655:531655 [4] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:531655:531655 [4] NCCL INFO Bootstrap : Using eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:531655:531655 [4] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test2-worker-0:531655:531655 [4] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test2-worker-0:531655:531655 [4] NCCL INFO NET/Plugin: Using internal network plugin. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test2-worker-0:531657:531657 [6] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test2-worker-0:531657:531657 [6] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:531657:531657 [6] NCCL INFO Bootstrap : Using eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:531657:531657 [6] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test2-worker-0:531657:531657 [6] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test2-worker-0:531657:531657 [6] NCCL INFO NET/Plugin: Using internal network plugin. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test2-worker-0:531655:533256 [4] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:531655:533256 [4] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test2-worker-0:531655:533256 [4] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:531655:533256 [4] NCCL INFO NET/Socket : Using [0]eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:531655:533256 [4] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test2-worker-0:531655:533256 [4] NCCL INFO Using network Socket +ywang29-vrdb-test2-worker-0:531654:531654 [3] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test2-worker-0:531654:531654 [3] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:531654:531654 [3] NCCL INFO Bootstrap : Using eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:531654:531654 [3] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test2-worker-0:531654:531654 [3] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test2-worker-0:531654:531654 [3] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test2-worker-0:531657:533257 [6] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:531657:533257 [6] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test2-worker-0:531657:533257 [6] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:531657:533257 [6] NCCL INFO NET/Socket : Using [0]eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:531657:533257 [6] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test2-worker-0:531657:533257 [6] NCCL INFO Using network Socket +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test2-worker-0:531652:531652 [1] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test2-worker-0:531652:531652 [1] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:531652:531652 [1] NCCL INFO Bootstrap : Using eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:531652:531652 [1] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test2-worker-0:531652:531652 [1] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test2-worker-0:531652:531652 [1] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test2-worker-0:531654:533258 [3] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:531654:533258 [3] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test2-worker-0:531654:533258 [3] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:531654:533258 [3] NCCL INFO NET/Socket : Using [0]eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:531654:533258 [3] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test2-worker-0:531654:533258 [3] NCCL INFO Using network Socket +ywang29-vrdb-test2-worker-0:531652:533259 [1] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:531652:533259 [1] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test2-worker-0:531652:533259 [1] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:531652:533259 [1] NCCL INFO NET/Socket : Using [0]eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:531652:533259 [1] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test2-worker-0:531652:533259 [1] NCCL INFO Using network Socket +ywang29-vrdb-test2-worker-0:531652:533259 [1] NCCL INFO ncclCommInitRank comm 0x563a4ecada20 rank 1 nranks 8 cudaDev 1 nvmlDev 1 busId 101d0 commId 0xd24d9de844a9958f - Init START +ywang29-vrdb-test2-worker-0:531651:533252 [0] NCCL INFO ncclCommInitRank comm 0x561bc96df090 rank 0 nranks 8 cudaDev 0 nvmlDev 0 busId 101c0 commId 0xd24d9de844a9958f - Init START +ywang29-vrdb-test2-worker-0:531655:533256 [4] NCCL INFO ncclCommInitRank comm 0x55d2f6e63520 rank 4 nranks 8 cudaDev 4 nvmlDev 4 busId 901c0 commId 0xd24d9de844a9958f - Init START +ywang29-vrdb-test2-worker-0:531654:533258 [3] NCCL INFO ncclCommInitRank comm 0x5630a211eaf0 rank 3 nranks 8 cudaDev 3 nvmlDev 3 busId 201d0 commId 0xd24d9de844a9958f - Init START +ywang29-vrdb-test2-worker-0:531653:533253 [2] NCCL INFO ncclCommInitRank comm 0x55e93dcd8f80 rank 2 nranks 8 cudaDev 2 nvmlDev 2 busId 201c0 commId 0xd24d9de844a9958f - Init START +ywang29-vrdb-test2-worker-0:531657:533257 [6] NCCL INFO ncclCommInitRank comm 0x5652e7ecec80 rank 6 nranks 8 cudaDev 6 nvmlDev 6 busId a01c0 commId 0xd24d9de844a9958f - Init START +ywang29-vrdb-test2-worker-0:531656:533255 [5] NCCL INFO ncclCommInitRank comm 0x557d844f03a0 rank 5 nranks 8 cudaDev 5 nvmlDev 5 busId 901d0 commId 0xd24d9de844a9958f - Init START +ywang29-vrdb-test2-worker-0:531658:533254 [7] NCCL INFO ncclCommInitRank comm 0x557e48e7ab20 rank 7 nranks 8 cudaDev 7 nvmlDev 7 busId a01d0 commId 0xd24d9de844a9958f - Init START +ywang29-vrdb-test2-worker-0:531655:533256 [4] NCCL INFO Setting affinity for GPU 4 to ffffff00,0000ffff,ff000000 +ywang29-vrdb-test2-worker-0:531655:533256 [4] NCCL INFO NVLS multicast support is not available on dev 4 +ywang29-vrdb-test2-worker-0:531653:533253 [2] NCCL INFO Setting affinity for GPU 2 to ff,ffff0000,00ffffff +ywang29-vrdb-test2-worker-0:531653:533253 [2] NCCL INFO NVLS multicast support is not available on dev 2 +ywang29-vrdb-test2-worker-0:531652:533259 [1] NCCL INFO Setting affinity for GPU 1 to ff,ffff0000,00ffffff +ywang29-vrdb-test2-worker-0:531652:533259 [1] NCCL INFO NVLS multicast support is not available on dev 1 +ywang29-vrdb-test2-worker-0:531651:533252 [0] NCCL INFO Setting affinity for GPU 0 to ff,ffff0000,00ffffff +ywang29-vrdb-test2-worker-0:531651:533252 [0] NCCL INFO NVLS multicast support is not available on dev 0 +ywang29-vrdb-test2-worker-0:531658:533254 [7] NCCL INFO Setting affinity for GPU 7 to ffffff00,0000ffff,ff000000 +ywang29-vrdb-test2-worker-0:531658:533254 [7] NCCL INFO NVLS multicast support is not available on dev 7 +ywang29-vrdb-test2-worker-0:531657:533257 [6] NCCL INFO Setting affinity for GPU 6 to ffffff00,0000ffff,ff000000 +ywang29-vrdb-test2-worker-0:531657:533257 [6] NCCL INFO NVLS multicast support is not available on dev 6 +ywang29-vrdb-test2-worker-0:531656:533255 [5] NCCL INFO Setting affinity for GPU 5 to ffffff00,0000ffff,ff000000 +ywang29-vrdb-test2-worker-0:531656:533255 [5] NCCL INFO NVLS multicast support is not available on dev 5 +ywang29-vrdb-test2-worker-0:531654:533258 [3] NCCL INFO Setting affinity for GPU 3 to ff,ffff0000,00ffffff +ywang29-vrdb-test2-worker-0:531654:533258 [3] NCCL INFO NVLS multicast support is not available on dev 3 +ywang29-vrdb-test2-worker-0:531653:533253 [2] NCCL INFO comm 0x55e93dcd8f80 rank 2 nRanks 8 nNodes 1 localRanks 8 localRank 2 MNNVL 0 +ywang29-vrdb-test2-worker-0:531657:533257 [6] NCCL INFO comm 0x5652e7ecec80 rank 6 nRanks 8 nNodes 1 localRanks 8 localRank 6 MNNVL 0 +ywang29-vrdb-test2-worker-0:531656:533255 [5] NCCL INFO comm 0x557d844f03a0 rank 5 nRanks 8 nNodes 1 localRanks 8 localRank 5 MNNVL 0 +ywang29-vrdb-test2-worker-0:531652:533259 [1] NCCL INFO comm 0x563a4ecada20 rank 1 nRanks 8 nNodes 1 localRanks 8 localRank 1 MNNVL 0 +ywang29-vrdb-test2-worker-0:531651:533252 [0] NCCL INFO comm 0x561bc96df090 rank 0 nRanks 8 nNodes 1 localRanks 8 localRank 0 MNNVL 0 +ywang29-vrdb-test2-worker-0:531654:533258 [3] NCCL INFO comm 0x5630a211eaf0 rank 3 nRanks 8 nNodes 1 localRanks 8 localRank 3 MNNVL 0 +ywang29-vrdb-test2-worker-0:531658:533254 [7] NCCL INFO comm 0x557e48e7ab20 rank 7 nRanks 8 nNodes 1 localRanks 8 localRank 7 MNNVL 0 +ywang29-vrdb-test2-worker-0:531655:533256 [4] NCCL INFO comm 0x55d2f6e63520 rank 4 nRanks 8 nNodes 1 localRanks 8 localRank 4 MNNVL 0 +ywang29-vrdb-test2-worker-0:531651:533252 [0] NCCL INFO Channel 00/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:531651:533252 [0] NCCL INFO Channel 01/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:531651:533252 [0] NCCL INFO Channel 02/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:531651:533252 [0] NCCL INFO Channel 03/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:531651:533252 [0] NCCL INFO Channel 04/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:531651:533252 [0] NCCL INFO Channel 05/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:531656:533255 [5] NCCL INFO Trees [0] 6/-1/-1->5->4 [1] 6/-1/-1->5->4 [2] 6/-1/-1->5->4 [3] 6/-1/-1->5->4 [4] 6/-1/-1->5->4 [5] 6/-1/-1->5->4 [6] 6/-1/-1->5->4 [7] 6/-1/-1->5->4 [8] 6/-1/-1->5->4 [9] 6/-1/-1->5->4 [10] 6/-1/-1->5->4 [11] 6/-1/-1->5->4 [12] 6/-1/-1->5->4 [13] 6/-1/-1->5->4 [14] 6/-1/-1->5->4 [15] 6/-1/-1->5->4 [16] 6/-1/-1->5->4 [17] 6/-1/-1->5->4 [18] 6/-1/-1->5->4 [19] 6/-1/-1->5->4 [20] 6/-1/-1->5->4 [21] 6/-1/-1->5->4 [22] 6/-1/-1->5->4 [23] 6/-1/-1->5->4 +ywang29-vrdb-test2-worker-0:531651:533252 [0] NCCL INFO Channel 06/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:531654:533258 [3] NCCL INFO Trees [0] 4/-1/-1->3->2 [1] 4/-1/-1->3->2 [2] 4/-1/-1->3->2 [3] 4/-1/-1->3->2 [4] 4/-1/-1->3->2 [5] 4/-1/-1->3->2 [6] 4/-1/-1->3->2 [7] 4/-1/-1->3->2 [8] 4/-1/-1->3->2 [9] 4/-1/-1->3->2 [10] 4/-1/-1->3->2 [11] 4/-1/-1->3->2 [12] 4/-1/-1->3->2 [13] 4/-1/-1->3->2 [14] 4/-1/-1->3->2 [15] 4/-1/-1->3->2 [16] 4/-1/-1->3->2 [17] 4/-1/-1->3->2 [18] 4/-1/-1->3->2 [19] 4/-1/-1->3->2 [20] 4/-1/-1->3->2 [21] 4/-1/-1->3->2 [22] 4/-1/-1->3->2 [23] 4/-1/-1->3->2 +ywang29-vrdb-test2-worker-0:531653:533253 [2] NCCL INFO Trees [0] 3/-1/-1->2->1 [1] 3/-1/-1->2->1 [2] 3/-1/-1->2->1 [3] 3/-1/-1->2->1 [4] 3/-1/-1->2->1 [5] 3/-1/-1->2->1 [6] 3/-1/-1->2->1 [7] 3/-1/-1->2->1 [8] 3/-1/-1->2->1 [9] 3/-1/-1->2->1 [10] 3/-1/-1->2->1 [11] 3/-1/-1->2->1 [12] 3/-1/-1->2->1 [13] 3/-1/-1->2->1 [14] 3/-1/-1->2->1 [15] 3/-1/-1->2->1 [16] 3/-1/-1->2->1 [17] 3/-1/-1->2->1 [18] 3/-1/-1->2->1 [19] 3/-1/-1->2->1 [20] 3/-1/-1->2->1 [21] 3/-1/-1->2->1 [22] 3/-1/-1->2->1 [23] 3/-1/-1->2->1 +ywang29-vrdb-test2-worker-0:531651:533252 [0] NCCL INFO Channel 07/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:531656:533255 [5] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:531652:533259 [1] NCCL INFO Trees [0] 2/-1/-1->1->0 [1] 2/-1/-1->1->0 [2] 2/-1/-1->1->0 [3] 2/-1/-1->1->0 [4] 2/-1/-1->1->0 [5] 2/-1/-1->1->0 [6] 2/-1/-1->1->0 [7] 2/-1/-1->1->0 [8] 2/-1/-1->1->0 [9] 2/-1/-1->1->0 [10] 2/-1/-1->1->0 [11] 2/-1/-1->1->0 [12] 2/-1/-1->1->0 [13] 2/-1/-1->1->0 [14] 2/-1/-1->1->0 [15] 2/-1/-1->1->0 [16] 2/-1/-1->1->0 [17] 2/-1/-1->1->0 [18] 2/-1/-1->1->0 [19] 2/-1/-1->1->0 [20] 2/-1/-1->1->0 [21] 2/-1/-1->1->0 [22] 2/-1/-1->1->0 [23] 2/-1/-1->1->0 +ywang29-vrdb-test2-worker-0:531654:533258 [3] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:531651:533252 [0] NCCL INFO Channel 08/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:531658:533254 [7] NCCL INFO Trees [0] -1/-1/-1->7->6 [1] -1/-1/-1->7->6 [2] -1/-1/-1->7->6 [3] -1/-1/-1->7->6 [4] -1/-1/-1->7->6 [5] -1/-1/-1->7->6 [6] -1/-1/-1->7->6 [7] -1/-1/-1->7->6 [8] -1/-1/-1->7->6 [9] -1/-1/-1->7->6 [10] -1/-1/-1->7->6 [11] -1/-1/-1->7->6 [12] -1/-1/-1->7->6 [13] -1/-1/-1->7->6 [14] -1/-1/-1->7->6 [15] -1/-1/-1->7->6 [16] -1/-1/-1->7->6 [17] -1/-1/-1->7->6 [18] -1/-1/-1->7->6 [19] -1/-1/-1->7->6 [20] -1/-1/-1->7->6 [21] -1/-1/-1->7->6 [22] -1/-1/-1->7->6 [23] -1/-1/-1->7->6 +ywang29-vrdb-test2-worker-0:531653:533253 [2] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:531651:533252 [0] NCCL INFO Channel 09/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:531655:533256 [4] NCCL INFO Trees [0] 5/-1/-1->4->3 [1] 5/-1/-1->4->3 [2] 5/-1/-1->4->3 [3] 5/-1/-1->4->3 [4] 5/-1/-1->4->3 [5] 5/-1/-1->4->3 [6] 5/-1/-1->4->3 [7] 5/-1/-1->4->3 [8] 5/-1/-1->4->3 [9] 5/-1/-1->4->3 [10] 5/-1/-1->4->3 [11] 5/-1/-1->4->3 [12] 5/-1/-1->4->3 [13] 5/-1/-1->4->3 [14] 5/-1/-1->4->3 [15] 5/-1/-1->4->3 [16] 5/-1/-1->4->3 [17] 5/-1/-1->4->3 [18] 5/-1/-1->4->3 [19] 5/-1/-1->4->3 [20] 5/-1/-1->4->3 [21] 5/-1/-1->4->3 [22] 5/-1/-1->4->3 [23] 5/-1/-1->4->3 +ywang29-vrdb-test2-worker-0:531651:533252 [0] NCCL INFO Channel 10/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:531652:533259 [1] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:531658:533254 [7] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:531651:533252 [0] NCCL INFO Channel 11/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:531655:533256 [4] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:531657:533257 [6] NCCL INFO Trees [0] 7/-1/-1->6->5 [1] 7/-1/-1->6->5 [2] 7/-1/-1->6->5 [3] 7/-1/-1->6->5 [4] 7/-1/-1->6->5 [5] 7/-1/-1->6->5 [6] 7/-1/-1->6->5 [7] 7/-1/-1->6->5 [8] 7/-1/-1->6->5 [9] 7/-1/-1->6->5 [10] 7/-1/-1->6->5 [11] 7/-1/-1->6->5 [12] 7/-1/-1->6->5 [13] 7/-1/-1->6->5 [14] 7/-1/-1->6->5 [15] 7/-1/-1->6->5 [16] 7/-1/-1->6->5 [17] 7/-1/-1->6->5 [18] 7/-1/-1->6->5 [19] 7/-1/-1->6->5 [20] 7/-1/-1->6->5 [21] 7/-1/-1->6->5 [22] 7/-1/-1->6->5 [23] 7/-1/-1->6->5 +ywang29-vrdb-test2-worker-0:531651:533252 [0] NCCL INFO Channel 12/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:531651:533252 [0] NCCL INFO Channel 13/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:531651:533252 [0] NCCL INFO Channel 14/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:531651:533252 [0] NCCL INFO Channel 15/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:531657:533257 [6] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:531651:533252 [0] NCCL INFO Channel 16/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:531651:533252 [0] NCCL INFO Channel 17/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:531651:533252 [0] NCCL INFO Channel 18/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:531651:533252 [0] NCCL INFO Channel 19/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:531651:533252 [0] NCCL INFO Channel 20/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:531651:533252 [0] NCCL INFO Channel 21/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:531651:533252 [0] NCCL INFO Channel 22/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:531651:533252 [0] NCCL INFO Channel 23/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:531651:533252 [0] NCCL INFO Trees [0] 1/-1/-1->0->-1 [1] 1/-1/-1->0->-1 [2] 1/-1/-1->0->-1 [3] 1/-1/-1->0->-1 [4] 1/-1/-1->0->-1 [5] 1/-1/-1->0->-1 [6] 1/-1/-1->0->-1 [7] 1/-1/-1->0->-1 [8] 1/-1/-1->0->-1 [9] 1/-1/-1->0->-1 [10] 1/-1/-1->0->-1 [11] 1/-1/-1->0->-1 [12] 1/-1/-1->0->-1 [13] 1/-1/-1->0->-1 [14] 1/-1/-1->0->-1 [15] 1/-1/-1->0->-1 [16] 1/-1/-1->0->-1 [17] 1/-1/-1->0->-1 [18] 1/-1/-1->0->-1 [19] 1/-1/-1->0->-1 [20] 1/-1/-1->0->-1 [21] 1/-1/-1->0->-1 [22] 1/-1/-1->0->-1 [23] 1/-1/-1->0->-1 +ywang29-vrdb-test2-worker-0:531651:533252 [0] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:531656:533255 [5] NCCL INFO Channel 00/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:533259 [1] NCCL INFO Channel 00/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:533253 [2] NCCL INFO Channel 00/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:533258 [3] NCCL INFO Channel 00/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:533257 [6] NCCL INFO Channel 00/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531651:533252 [0] NCCL INFO Channel 00/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:533254 [7] NCCL INFO Channel 00/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:533259 [1] NCCL INFO Channel 01/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:533255 [5] NCCL INFO Channel 01/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:533253 [2] NCCL INFO Channel 01/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:533257 [6] NCCL INFO Channel 01/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:533258 [3] NCCL INFO Channel 01/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531651:533252 [0] NCCL INFO Channel 01/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:533259 [1] NCCL INFO Channel 02/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:533254 [7] NCCL INFO Channel 01/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:533255 [5] NCCL INFO Channel 02/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:533253 [2] NCCL INFO Channel 02/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:533257 [6] NCCL INFO Channel 02/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:533258 [3] NCCL INFO Channel 02/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531651:533252 [0] NCCL INFO Channel 02/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:533259 [1] NCCL INFO Channel 03/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:533254 [7] NCCL INFO Channel 02/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:533255 [5] NCCL INFO Channel 03/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:533253 [2] NCCL INFO Channel 03/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:533257 [6] NCCL INFO Channel 03/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:533258 [3] NCCL INFO Channel 03/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531651:533252 [0] NCCL INFO Channel 03/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:533259 [1] NCCL INFO Channel 04/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:533254 [7] NCCL INFO Channel 03/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:533255 [5] NCCL INFO Channel 04/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:533253 [2] NCCL INFO Channel 04/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:533257 [6] NCCL INFO Channel 04/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:533258 [3] NCCL INFO Channel 04/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531651:533252 [0] NCCL INFO Channel 04/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:533259 [1] NCCL INFO Channel 05/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:533254 [7] NCCL INFO Channel 04/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:533255 [5] NCCL INFO Channel 05/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:533253 [2] NCCL INFO Channel 05/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:533257 [6] NCCL INFO Channel 05/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:533258 [3] NCCL INFO Channel 05/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531651:533252 [0] NCCL INFO Channel 05/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:533259 [1] NCCL INFO Channel 06/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:533254 [7] NCCL INFO Channel 05/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:533255 [5] NCCL INFO Channel 06/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:533253 [2] NCCL INFO Channel 06/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:533257 [6] NCCL INFO Channel 06/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:533258 [3] NCCL INFO Channel 06/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531651:533252 [0] NCCL INFO Channel 06/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:533259 [1] NCCL INFO Channel 07/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:533254 [7] NCCL INFO Channel 06/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:533255 [5] NCCL INFO Channel 07/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:533253 [2] NCCL INFO Channel 07/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:533257 [6] NCCL INFO Channel 07/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:533258 [3] NCCL INFO Channel 07/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531651:533252 [0] NCCL INFO Channel 07/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:533259 [1] NCCL INFO Channel 08/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:533254 [7] NCCL INFO Channel 07/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:533255 [5] NCCL INFO Channel 08/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:533253 [2] NCCL INFO Channel 08/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:533257 [6] NCCL INFO Channel 08/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:533258 [3] NCCL INFO Channel 08/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531651:533252 [0] NCCL INFO Channel 08/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:533259 [1] NCCL INFO Channel 09/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:533254 [7] NCCL INFO Channel 08/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:533255 [5] NCCL INFO Channel 09/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:533253 [2] NCCL INFO Channel 09/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:533257 [6] NCCL INFO Channel 09/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:533258 [3] NCCL INFO Channel 09/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531651:533252 [0] NCCL INFO Channel 09/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:533259 [1] NCCL INFO Channel 10/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:533254 [7] NCCL INFO Channel 09/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:533255 [5] NCCL INFO Channel 10/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:533253 [2] NCCL INFO Channel 10/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:533257 [6] NCCL INFO Channel 10/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:533258 [3] NCCL INFO Channel 10/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531651:533252 [0] NCCL INFO Channel 10/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:533259 [1] NCCL INFO Channel 11/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:533255 [5] NCCL INFO Channel 11/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:533254 [7] NCCL INFO Channel 10/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:533253 [2] NCCL INFO Channel 11/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:533257 [6] NCCL INFO Channel 11/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:533258 [3] NCCL INFO Channel 11/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531651:533252 [0] NCCL INFO Channel 11/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:533259 [1] NCCL INFO Channel 12/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:533255 [5] NCCL INFO Channel 12/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:533254 [7] NCCL INFO Channel 11/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:533256 [4] NCCL INFO Channel 00/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:533253 [2] NCCL INFO Channel 12/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:533257 [6] NCCL INFO Channel 12/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:533258 [3] NCCL INFO Channel 12/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531651:533252 [0] NCCL INFO Channel 12/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:533259 [1] NCCL INFO Channel 13/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:533255 [5] NCCL INFO Channel 13/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:533254 [7] NCCL INFO Channel 12/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:533256 [4] NCCL INFO Channel 01/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:533253 [2] NCCL INFO Channel 13/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:533257 [6] NCCL INFO Channel 13/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:533258 [3] NCCL INFO Channel 13/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531651:533252 [0] NCCL INFO Channel 13/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:533259 [1] NCCL INFO Channel 14/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:533255 [5] NCCL INFO Channel 14/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:533254 [7] NCCL INFO Channel 13/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:533256 [4] NCCL INFO Channel 02/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:533253 [2] NCCL INFO Channel 14/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:533257 [6] NCCL INFO Channel 14/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:533258 [3] NCCL INFO Channel 14/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531651:533252 [0] NCCL INFO Channel 14/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:533259 [1] NCCL INFO Channel 15/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:533255 [5] NCCL INFO Channel 15/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:533254 [7] NCCL INFO Channel 14/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:533253 [2] NCCL INFO Channel 15/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:533256 [4] NCCL INFO Channel 03/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:533257 [6] NCCL INFO Channel 15/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:533258 [3] NCCL INFO Channel 15/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531651:533252 [0] NCCL INFO Channel 15/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:533259 [1] NCCL INFO Channel 16/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:533255 [5] NCCL INFO Channel 16/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:533254 [7] NCCL INFO Channel 15/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:533253 [2] NCCL INFO Channel 16/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:533256 [4] NCCL INFO Channel 04/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:533257 [6] NCCL INFO Channel 16/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:533258 [3] NCCL INFO Channel 16/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:533259 [1] NCCL INFO Channel 17/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531651:533252 [0] NCCL INFO Channel 16/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:533255 [5] NCCL INFO Channel 17/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:533254 [7] NCCL INFO Channel 16/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:533253 [2] NCCL INFO Channel 17/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:533256 [4] NCCL INFO Channel 05/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:533257 [6] NCCL INFO Channel 17/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:533258 [3] NCCL INFO Channel 17/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:533259 [1] NCCL INFO Channel 18/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531651:533252 [0] NCCL INFO Channel 17/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:533255 [5] NCCL INFO Channel 18/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:533254 [7] NCCL INFO Channel 17/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:533253 [2] NCCL INFO Channel 18/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:533256 [4] NCCL INFO Channel 06/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:533258 [3] NCCL INFO Channel 18/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:533257 [6] NCCL INFO Channel 18/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:533259 [1] NCCL INFO Channel 19/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531651:533252 [0] NCCL INFO Channel 18/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:533255 [5] NCCL INFO Channel 19/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:533254 [7] NCCL INFO Channel 18/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:533253 [2] NCCL INFO Channel 19/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:533256 [4] NCCL INFO Channel 07/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:533258 [3] NCCL INFO Channel 19/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:533257 [6] NCCL INFO Channel 19/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:533259 [1] NCCL INFO Channel 20/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531651:533252 [0] NCCL INFO Channel 19/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:533255 [5] NCCL INFO Channel 20/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:533254 [7] NCCL INFO Channel 19/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:533253 [2] NCCL INFO Channel 20/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:533256 [4] NCCL INFO Channel 08/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:533258 [3] NCCL INFO Channel 20/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:533257 [6] NCCL INFO Channel 20/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:533259 [1] NCCL INFO Channel 21/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531651:533252 [0] NCCL INFO Channel 20/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:533255 [5] NCCL INFO Channel 21/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:533254 [7] NCCL INFO Channel 20/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:533253 [2] NCCL INFO Channel 21/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:533256 [4] NCCL INFO Channel 09/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:533258 [3] NCCL INFO Channel 21/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:533257 [6] NCCL INFO Channel 21/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:533259 [1] NCCL INFO Channel 22/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531651:533252 [0] NCCL INFO Channel 21/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:533255 [5] NCCL INFO Channel 22/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:533254 [7] NCCL INFO Channel 21/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:533253 [2] NCCL INFO Channel 22/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:533256 [4] NCCL INFO Channel 10/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:533258 [3] NCCL INFO Channel 22/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:533257 [6] NCCL INFO Channel 22/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:533259 [1] NCCL INFO Channel 23/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531651:533252 [0] NCCL INFO Channel 22/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:533255 [5] NCCL INFO Channel 23/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:533254 [7] NCCL INFO Channel 22/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:533253 [2] NCCL INFO Channel 23/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:533256 [4] NCCL INFO Channel 11/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:533258 [3] NCCL INFO Channel 23/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:533257 [6] NCCL INFO Channel 23/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531651:533252 [0] NCCL INFO Channel 23/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:533254 [7] NCCL INFO Channel 23/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:533256 [4] NCCL INFO Channel 12/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:533256 [4] NCCL INFO Channel 13/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:533256 [4] NCCL INFO Channel 14/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:533256 [4] NCCL INFO Channel 15/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:533256 [4] NCCL INFO Channel 16/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:533256 [4] NCCL INFO Channel 17/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:533256 [4] NCCL INFO Channel 18/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:533256 [4] NCCL INFO Channel 19/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:533256 [4] NCCL INFO Channel 20/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:533256 [4] NCCL INFO Channel 21/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:533256 [4] NCCL INFO Channel 22/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:533256 [4] NCCL INFO Channel 23/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:533259 [1] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:531651:533252 [0] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:531653:533253 [2] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:531654:533258 [3] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:531655:533256 [4] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:531658:533254 [7] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:531658:533254 [7] NCCL INFO Channel 00/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:533255 [5] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:531657:533257 [6] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:531658:533254 [7] NCCL INFO Channel 01/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:533254 [7] NCCL INFO Channel 02/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:533254 [7] NCCL INFO Channel 03/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:533254 [7] NCCL INFO Channel 04/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:533254 [7] NCCL INFO Channel 05/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:533259 [1] NCCL INFO Channel 00/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:533254 [7] NCCL INFO Channel 06/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:533259 [1] NCCL INFO Channel 01/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:533254 [7] NCCL INFO Channel 07/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:533259 [1] NCCL INFO Channel 02/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:533253 [2] NCCL INFO Channel 00/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:533254 [7] NCCL INFO Channel 08/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:533259 [1] NCCL INFO Channel 03/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:533253 [2] NCCL INFO Channel 01/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:533254 [7] NCCL INFO Channel 09/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:533259 [1] NCCL INFO Channel 04/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:533253 [2] NCCL INFO Channel 02/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:533254 [7] NCCL INFO Channel 10/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:533259 [1] NCCL INFO Channel 05/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:533253 [2] NCCL INFO Channel 03/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:533254 [7] NCCL INFO Channel 11/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:533259 [1] NCCL INFO Channel 06/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:533253 [2] NCCL INFO Channel 04/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:533254 [7] NCCL INFO Channel 12/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:533258 [3] NCCL INFO Channel 00/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:533259 [1] NCCL INFO Channel 07/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:533253 [2] NCCL INFO Channel 05/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:533254 [7] NCCL INFO Channel 13/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:533258 [3] NCCL INFO Channel 01/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:533259 [1] NCCL INFO Channel 08/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:533253 [2] NCCL INFO Channel 06/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:533254 [7] NCCL INFO Channel 14/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:533258 [3] NCCL INFO Channel 02/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:533259 [1] NCCL INFO Channel 09/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:533253 [2] NCCL INFO Channel 07/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:533254 [7] NCCL INFO Channel 15/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:533258 [3] NCCL INFO Channel 03/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:533259 [1] NCCL INFO Channel 10/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:533253 [2] NCCL INFO Channel 08/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:533254 [7] NCCL INFO Channel 16/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:533258 [3] NCCL INFO Channel 04/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:533259 [1] NCCL INFO Channel 11/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:533253 [2] NCCL INFO Channel 09/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:533254 [7] NCCL INFO Channel 17/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:533258 [3] NCCL INFO Channel 05/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:533259 [1] NCCL INFO Channel 12/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:533253 [2] NCCL INFO Channel 10/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:533254 [7] NCCL INFO Channel 18/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:533258 [3] NCCL INFO Channel 06/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:533259 [1] NCCL INFO Channel 13/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:533253 [2] NCCL INFO Channel 11/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:533254 [7] NCCL INFO Channel 19/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:533256 [4] NCCL INFO Channel 00/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:533258 [3] NCCL INFO Channel 07/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:533259 [1] NCCL INFO Channel 14/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:533253 [2] NCCL INFO Channel 12/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:533254 [7] NCCL INFO Channel 20/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:533256 [4] NCCL INFO Channel 01/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:533258 [3] NCCL INFO Channel 08/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:533259 [1] NCCL INFO Channel 15/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:533253 [2] NCCL INFO Channel 13/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:533254 [7] NCCL INFO Channel 21/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:533256 [4] NCCL INFO Channel 02/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:533258 [3] NCCL INFO Channel 09/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:533259 [1] NCCL INFO Channel 16/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:533253 [2] NCCL INFO Channel 14/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:533254 [7] NCCL INFO Channel 22/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:533256 [4] NCCL INFO Channel 03/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:533258 [3] NCCL INFO Channel 10/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:533259 [1] NCCL INFO Channel 17/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:533253 [2] NCCL INFO Channel 15/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:533254 [7] NCCL INFO Channel 23/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:533256 [4] NCCL INFO Channel 04/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:533258 [3] NCCL INFO Channel 11/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:533259 [1] NCCL INFO Channel 18/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:533253 [2] NCCL INFO Channel 16/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:533256 [4] NCCL INFO Channel 05/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:533255 [5] NCCL INFO Channel 00/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:533257 [6] NCCL INFO Channel 00/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:533258 [3] NCCL INFO Channel 12/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:533259 [1] NCCL INFO Channel 19/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:533253 [2] NCCL INFO Channel 17/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:533255 [5] NCCL INFO Channel 01/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:533256 [4] NCCL INFO Channel 06/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:533257 [6] NCCL INFO Channel 01/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:533258 [3] NCCL INFO Channel 13/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:533259 [1] NCCL INFO Channel 20/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:533253 [2] NCCL INFO Channel 18/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:533255 [5] NCCL INFO Channel 02/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:533256 [4] NCCL INFO Channel 07/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:533257 [6] NCCL INFO Channel 02/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:533258 [3] NCCL INFO Channel 14/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:533259 [1] NCCL INFO Channel 21/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:533253 [2] NCCL INFO Channel 19/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:533255 [5] NCCL INFO Channel 03/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:533256 [4] NCCL INFO Channel 08/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:533257 [6] NCCL INFO Channel 03/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:533258 [3] NCCL INFO Channel 15/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:533259 [1] NCCL INFO Channel 22/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:533253 [2] NCCL INFO Channel 20/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:533255 [5] NCCL INFO Channel 04/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:533256 [4] NCCL INFO Channel 09/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:533257 [6] NCCL INFO Channel 04/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:533258 [3] NCCL INFO Channel 16/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:533259 [1] NCCL INFO Channel 23/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:533253 [2] NCCL INFO Channel 21/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:533255 [5] NCCL INFO Channel 05/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:533256 [4] NCCL INFO Channel 10/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:533257 [6] NCCL INFO Channel 05/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:533258 [3] NCCL INFO Channel 17/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:533253 [2] NCCL INFO Channel 22/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:533255 [5] NCCL INFO Channel 06/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:533256 [4] NCCL INFO Channel 11/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:533258 [3] NCCL INFO Channel 18/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:533257 [6] NCCL INFO Channel 06/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:533253 [2] NCCL INFO Channel 23/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:533255 [5] NCCL INFO Channel 07/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:533256 [4] NCCL INFO Channel 12/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:533258 [3] NCCL INFO Channel 19/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:533257 [6] NCCL INFO Channel 07/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:533255 [5] NCCL INFO Channel 08/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:533256 [4] NCCL INFO Channel 13/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:533257 [6] NCCL INFO Channel 08/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:533258 [3] NCCL INFO Channel 20/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:533258 [3] NCCL INFO Channel 21/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:533257 [6] NCCL INFO Channel 09/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:533255 [5] NCCL INFO Channel 09/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:533258 [3] NCCL INFO Channel 22/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:533257 [6] NCCL INFO Channel 10/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:533255 [5] NCCL INFO Channel 10/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:533256 [4] NCCL INFO Channel 14/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:533257 [6] NCCL INFO Channel 11/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:533256 [4] NCCL INFO Channel 15/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:533257 [6] NCCL INFO Channel 12/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:533256 [4] NCCL INFO Channel 16/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:533258 [3] NCCL INFO Channel 23/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:533257 [6] NCCL INFO Channel 13/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:533256 [4] NCCL INFO Channel 17/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:533255 [5] NCCL INFO Channel 11/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:533257 [6] NCCL INFO Channel 14/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:533255 [5] NCCL INFO Channel 12/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:533256 [4] NCCL INFO Channel 18/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:533257 [6] NCCL INFO Channel 15/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:533255 [5] NCCL INFO Channel 13/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:533257 [6] NCCL INFO Channel 16/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:533255 [5] NCCL INFO Channel 14/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:533257 [6] NCCL INFO Channel 17/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:533255 [5] NCCL INFO Channel 15/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:533257 [6] NCCL INFO Channel 18/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:533255 [5] NCCL INFO Channel 16/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:533256 [4] NCCL INFO Channel 19/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:533255 [5] NCCL INFO Channel 17/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:533257 [6] NCCL INFO Channel 19/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:533256 [4] NCCL INFO Channel 20/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:533255 [5] NCCL INFO Channel 18/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:533255 [5] NCCL INFO Channel 19/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:533255 [5] NCCL INFO Channel 20/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:533255 [5] NCCL INFO Channel 21/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:533255 [5] NCCL INFO Channel 22/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:533255 [5] NCCL INFO Channel 23/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:533256 [4] NCCL INFO Channel 21/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:533257 [6] NCCL INFO Channel 20/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:533256 [4] NCCL INFO Channel 22/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:533257 [6] NCCL INFO Channel 21/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:533256 [4] NCCL INFO Channel 23/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:533257 [6] NCCL INFO Channel 22/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:533257 [6] NCCL INFO Channel 23/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531651:533252 [0] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:531651:533252 [0] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:531651:533252 [0] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:531652:533259 [1] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:531652:533259 [1] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:531652:533259 [1] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:531653:533253 [2] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:531653:533253 [2] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:531653:533253 [2] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:531654:533258 [3] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:531654:533258 [3] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:531654:533258 [3] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:531658:533254 [7] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:531658:533254 [7] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:531658:533254 [7] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:531655:533256 [4] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:531655:533256 [4] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:531655:533256 [4] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:531657:533257 [6] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:531657:533257 [6] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:531657:533257 [6] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:531656:533255 [5] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:531656:533255 [5] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:531656:533255 [5] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:531653:533253 [2] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test2-worker-0:531654:533258 [3] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test2-worker-0:531651:533252 [0] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test2-worker-0:531652:533259 [1] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test2-worker-0:531653:533253 [2] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test2-worker-0:531654:533258 [3] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test2-worker-0:531651:533252 [0] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test2-worker-0:531652:533259 [1] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test2-worker-0:531653:533253 [2] NCCL INFO ncclCommInitRank comm 0x55e93dcd8f80 rank 2 nranks 8 cudaDev 2 nvmlDev 2 busId 201c0 commId 0xd24d9de844a9958f - Init COMPLETE +ywang29-vrdb-test2-worker-0:531654:533258 [3] NCCL INFO ncclCommInitRank comm 0x5630a211eaf0 rank 3 nranks 8 cudaDev 3 nvmlDev 3 busId 201d0 commId 0xd24d9de844a9958f - Init COMPLETE +ywang29-vrdb-test2-worker-0:531651:533252 [0] NCCL INFO ncclCommInitRank comm 0x561bc96df090 rank 0 nranks 8 cudaDev 0 nvmlDev 0 busId 101c0 commId 0xd24d9de844a9958f - Init COMPLETE +ywang29-vrdb-test2-worker-0:531652:533259 [1] NCCL INFO ncclCommInitRank comm 0x563a4ecada20 rank 1 nranks 8 cudaDev 1 nvmlDev 1 busId 101d0 commId 0xd24d9de844a9958f - Init COMPLETE +ywang29-vrdb-test2-worker-0:531657:533257 [6] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test2-worker-0:531656:533255 [5] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test2-worker-0:531658:533254 [7] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test2-worker-0:531657:533257 [6] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test2-worker-0:531655:533256 [4] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test2-worker-0:531656:533255 [5] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test2-worker-0:531658:533254 [7] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test2-worker-0:531655:533256 [4] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test2-worker-0:531657:533257 [6] NCCL INFO ncclCommInitRank comm 0x5652e7ecec80 rank 6 nranks 8 cudaDev 6 nvmlDev 6 busId a01c0 commId 0xd24d9de844a9958f - Init COMPLETE +ywang29-vrdb-test2-worker-0:531656:533255 [5] NCCL INFO ncclCommInitRank comm 0x557d844f03a0 rank 5 nranks 8 cudaDev 5 nvmlDev 5 busId 901d0 commId 0xd24d9de844a9958f - Init COMPLETE +ywang29-vrdb-test2-worker-0:531658:533254 [7] NCCL INFO ncclCommInitRank comm 0x557e48e7ab20 rank 7 nranks 8 cudaDev 7 nvmlDev 7 busId a01d0 commId 0xd24d9de844a9958f - Init COMPLETE +ywang29-vrdb-test2-worker-0:531655:533256 [4] NCCL INFO ncclCommInitRank comm 0x55d2f6e63520 rank 4 nranks 8 cudaDev 4 nvmlDev 4 busId 901c0 commId 0xd24d9de844a9958f - Init COMPLETE +[2025-10-12 08:15:16,923] [INFO] [partition_parameters.py:343:__exit__] finished initializing model - num_params = 459, num_elems = 0.99B +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.laSome weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +loading language model from loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model + /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model + /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +yers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +[2025-10-12 08:31:45,183] [INFO] [partition_parameters.py:343:__exit__] finished initializing model - num_params = 907, num_elems = 1.42B +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +TinyLlavaForConditionalGeneration( + (language_model): Qwen2ForCausalLM( + (model): Qwen2Model( + (embed_tokens): Embedding(151936, 896) + (layers): ModuleList( + (0-23): 24 x Qwen2DecoderLayer( + (self_attn): Qwen2FlashAttention2( + (q_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=True) + (k_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=128, bias=True) + (v_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=128, bias=True) + (o_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=False) + (rotary_emb): Qwen2RotaryEmbedding() + ) + (mlp): Qwen2MLP( + (gate_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=4864, bias=False) + (up_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=4864, bias=False) + (down_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=4864, out_features=896, bias=False) + (act_fn): SiLU() + ) + (input_layernorm): Qwen2RMSNorm() + (post_attention_layernorm): Qwen2RMSNorm() + ) + ) + (norm): Qwen2RMSNorm() + ) + (lm_head): Linear(in_features=896, out_features=151936, bias=False) + ) + (vision_tower): SIGLIPVisionTower( + (_vision_tower): SiglipVisionModel( + (vision_model): SiglipVisionTransformer( + (embeddings): SiglipVisionEmbeddings( + (patch_embedding): Conv2d(3, 1152, kernel_size=(14, 14), stride=(14, 14), padding=valid) + (position_embedding): Embedding(729, 1152) + ) + (encoder): SiglipEncoder( + (layers): ModuleList( + (0-26): 27 x SiglipEncoderLayer( + (self_attn): SiglipAttention( + (k_proj): Linear(in_features=1152, out_features=1152, bias=True) + (v_proj): Linear(in_features=1152, out_features=1152, bias=True) + (q_proj): Linear(in_features=1152, out_features=1152, bias=True) + (out_proj): Linear(in_features=1152, out_features=1152, bias=True) + ) + (layer_norm1): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + (layer_norm2): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + ) + ) + ) + (post_layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (head): SiglipMultiheadAttentionPoolingHead( + (attention): MultiheadAttention( + (out_proj): NonDynamicallyQuantizableLinear(in_features=1152, out_features=1152, bias=True) + ) + (layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + ) + ) + ) + ) + (connector): MLPConnector( + (_connector): Sequential( + (0): SupermaskLinearSparsity_SoftForward_Normal(in_features=1152, out_features=896, bias=True) + (1): GELU(approximate='none') + (2): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=True) + ) + ) +) +Pre-training init language_model.model.layers.0.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.0.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.0.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.0.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.0.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.0.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.0.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.mlp.down_proj.scores: Mean=3.000000 +Pre-training init connector._connector.0.scores: Mean=3.000005 +Pre-training init connector._connector.2.scores: Mean=2.999970 +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +2025-10-12 08:32:03,397 | INFO: Total Parameters: 1283756736, Total Trainable Parameters: 359661568 +2025-10-12 08:32:03,402 | INFO: Trainable Parameters: +language_model.model.layers.0.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.0.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.0.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.0.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.0.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.0.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.0.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.1.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.1.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.1.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.1.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.1.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.1.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.1.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.2.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.2.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.2.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.2.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.2.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.2.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.2.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.3.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.3.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.3.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.3.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.3.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.3.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.3.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.4.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.4.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.4.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.4.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.4.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.4.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.4.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.5.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.5.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.5.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.5.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.5.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.5.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.5.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.6.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.6.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.6.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.6.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.6.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.6.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.6.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.7.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.7.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.7.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.7.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.7.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.7.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.7.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.8.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.8.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.8.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.8.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.8.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.8.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.8.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.9.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.9.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.9.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.9.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.9.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.9.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.9.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.10.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.10.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.10.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.10.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.10.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.10.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.10.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.11.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.11.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.11.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.11.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.11.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.11.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.11.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.12.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.12.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.12.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.12.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.12.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.12.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.12.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.13.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.13.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.13.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.13.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.13.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.13.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.13.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.14.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.14.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.14.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.14.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.14.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.14.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.14.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.15.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.15.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.15.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.15.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.15.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.15.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.15.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.16.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.16.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.16.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.16.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.16.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.16.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.16.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.17.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.17.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.17.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.17.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.17.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.17.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.17.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.18.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.18.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.18.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.18.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.18.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.18.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.18.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.19.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.19.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.19.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.19.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.19.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.19.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.19.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.20.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.20.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.20.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.20.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.20.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.20.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.20.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.21.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.21.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.21.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.21.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.21.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.21.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.21.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.22.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.22.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.22.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.22.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.22.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.22.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.22.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.23.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.23.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.23.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.23.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.23.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.23.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.23.mlp.down_proj.scores: 4358144 parameters +connector._connector.0.scores: 1032192 parameters +connector._connector.2.scores: 802816 parameters +Parameter Offload: Total persistent parameters: 486464 in 403 params + 0%| | 0/520 [00:006->5 [1] 7/-1/-1->6->5 [2] 7/-1/-1->6->5 [3] 7/-1/-1->6->5 [4] 7/-1/-1->6->5 [5] 7/-1/-1->6->5 [6] 7/-1/-1->6->5 [7] 7/-1/-1->6->5 [8] 7/-1/-1->6->5 [9] 7/-1/-1->6->5 [10] 7/-1/-1->6->5 [11] 7/-1/-1->6->5 [12] 7/-1/-1->6->5 [13] 7/-1/-1->6->5 [14] 7/-1/-1->6->5 [15] 7/-1/-1->6->5 [16] 7/-1/-1->6->5 [17] 7/-1/-1->6->5 [18] 7/-1/-1->6->5 [19] 7/-1/-1->6->5 [20] 7/-1/-1->6->5 [21] 7/-1/-1->6->5 [22] 7/-1/-1->6->5 [23] 7/-1/-1->6->5 +ywang29-vrdb-test2-worker-0:531658:538763 [7] NCCL INFO Trees [0] -1/-1/-1->7->6 [1] -1/-1/-1->7->6 [2] -1/-1/-1->7->6 [3] -1/-1/-1->7->6 [4] -1/-1/-1->7->6 [5] -1/-1/-1->7->6 [6] -1/-1/-1->7->6 [7] -1/-1/-1->7->6 [8] -1/-1/-1->7->6 [9] -1/-1/-1->7->6 [10] -1/-1/-1->7->6 [11] -1/-1/-1->7->6 [12] -1/-1/-1->7->6 [13] -1/-1/-1->7->6 [14] -1/-1/-1->7->6 [15] -1/-1/-1->7->6 [16] -1/-1/-1->7->6 [17] -1/-1/-1->7->6 [18] -1/-1/-1->7->6 [19] -1/-1/-1->7->6 [20] -1/-1/-1->7->6 [21] -1/-1/-1->7->6 [22] -1/-1/-1->7->6 [23] -1/-1/-1->7->6 +ywang29-vrdb-test2-worker-0:531657:538759 [6] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:531651:538758 [0] NCCL INFO Channel 00/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:531658:538763 [7] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:531656:538764 [5] NCCL INFO Trees [0] 6/-1/-1->5->4 [1] 6/-1/-1->5->4 [2] 6/-1/-1->5->4 [3] 6/-1/-1->5->4 [4] 6/-1/-1->5->4 [5] 6/-1/-1->5->4 [6] 6/-1/-1->5->4 [7] 6/-1/-1->5->4 [8] 6/-1/-1->5->4 [9] 6/-1/-1->5->4 [10] 6/-1/-1->5->4 [11] 6/-1/-1->5->4 [12] 6/-1/-1->5->4 [13] 6/-1/-1->5->4 [14] 6/-1/-1->5->4 [15] 6/-1/-1->5->4 [16] 6/-1/-1->5->4 [17] 6/-1/-1->5->4 [18] 6/-1/-1->5->4 [19] 6/-1/-1->5->4 [20] 6/-1/-1->5->4 [21] 6/-1/-1->5->4 [22] 6/-1/-1->5->4 [23] 6/-1/-1->5->4 +ywang29-vrdb-test2-worker-0:531656:538764 [5] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:531651:538758 [0] NCCL INFO Channel 01/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:531655:538760 [4] NCCL INFO Trees [0] 5/-1/-1->4->3 [1] 5/-1/-1->4->3 [2] 5/-1/-1->4->3 [3] 5/-1/-1->4->3 [4] 5/-1/-1->4->3 [5] 5/-1/-1->4->3 [6] 5/-1/-1->4->3 [7] 5/-1/-1->4->3 [8] 5/-1/-1->4->3 [9] 5/-1/-1->4->3 [10] 5/-1/-1->4->3 [11] 5/-1/-1->4->3 [12] 5/-1/-1->4->3 [13] 5/-1/-1->4->3 [14] 5/-1/-1->4->3 [15] 5/-1/-1->4->3 [16] 5/-1/-1->4->3 [17] 5/-1/-1->4->3 [18] 5/-1/-1->4->3 [19] 5/-1/-1->4->3 [20] 5/-1/-1->4->3 [21] 5/-1/-1->4->3 [22] 5/-1/-1->4->3 [23] 5/-1/-1->4->3 +ywang29-vrdb-test2-worker-0:531655:538760 [4] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:531654:538762 [3] NCCL INFO Trees [0] 4/-1/-1->3->2 [1] 4/-1/-1->3->2 [2] 4/-1/-1->3->2 [3] 4/-1/-1->3->2 [4] 4/-1/-1->3->2 [5] 4/-1/-1->3->2 [6] 4/-1/-1->3->2 [7] 4/-1/-1->3->2 [8] 4/-1/-1->3->2 [9] 4/-1/-1->3->2 [10] 4/-1/-1->3->2 [11] 4/-1/-1->3->2 [12] 4/-1/-1->3->2 [13] 4/-1/-1->3->2 [14] 4/-1/-1->3->2 [15] 4/-1/-1->3->2 [16] 4/-1/-1->3->2 [17] 4/-1/-1->3->2 [18] 4/-1/-1->3->2 [19] 4/-1/-1->3->2 [20] 4/-1/-1->3->2 [21] 4/-1/-1->3->2 [22] 4/-1/-1->3->2 [23] 4/-1/-1->3->2 +ywang29-vrdb-test2-worker-0:531651:538758 [0] NCCL INFO Channel 02/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:531653:538765 [2] NCCL INFO Trees [0] 3/-1/-1->2->1 [1] 3/-1/-1->2->1 [2] 3/-1/-1->2->1 [3] 3/-1/-1->2->1 [4] 3/-1/-1->2->1 [5] 3/-1/-1->2->1 [6] 3/-1/-1->2->1 [7] 3/-1/-1->2->1 [8] 3/-1/-1->2->1 [9] 3/-1/-1->2->1 [10] 3/-1/-1->2->1 [11] 3/-1/-1->2->1 [12] 3/-1/-1->2->1 [13] 3/-1/-1->2->1 [14] 3/-1/-1->2->1 [15] 3/-1/-1->2->1 [16] 3/-1/-1->2->1 [17] 3/-1/-1->2->1 [18] 3/-1/-1->2->1 [19] 3/-1/-1->2->1 [20] 3/-1/-1->2->1 [21] 3/-1/-1->2->1 [22] 3/-1/-1->2->1 [23] 3/-1/-1->2->1 +ywang29-vrdb-test2-worker-0:531654:538762 [3] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:531653:538765 [2] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:531651:538758 [0] NCCL INFO Channel 03/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:531651:538758 [0] NCCL INFO Channel 04/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:531651:538758 [0] NCCL INFO Channel 05/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:531651:538758 [0] NCCL INFO Channel 06/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:531651:538758 [0] NCCL INFO Channel 07/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:531651:538758 [0] NCCL INFO Channel 08/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:531652:538761 [1] NCCL INFO Trees [0] 2/-1/-1->1->0 [1] 2/-1/-1->1->0 [2] 2/-1/-1->1->0 [3] 2/-1/-1->1->0 [4] 2/-1/-1->1->0 [5] 2/-1/-1->1->0 [6] 2/-1/-1->1->0 [7] 2/-1/-1->1->0 [8] 2/-1/-1->1->0 [9] 2/-1/-1->1->0 [10] 2/-1/-1->1->0 [11] 2/-1/-1->1->0 [12] 2/-1/-1->1->0 [13] 2/-1/-1->1->0 [14] 2/-1/-1->1->0 [15] 2/-1/-1->1->0 [16] 2/-1/-1->1->0 [17] 2/-1/-1->1->0 [18] 2/-1/-1->1->0 [19] 2/-1/-1->1->0 [20] 2/-1/-1->1->0 [21] 2/-1/-1->1->0 [22] 2/-1/-1->1->0 [23] 2/-1/-1->1->0 +ywang29-vrdb-test2-worker-0:531652:538761 [1] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:531651:538758 [0] NCCL INFO Channel 09/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:531651:538758 [0] NCCL INFO Channel 10/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:531651:538758 [0] NCCL INFO Channel 11/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:531651:538758 [0] NCCL INFO Channel 12/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:531651:538758 [0] NCCL INFO Channel 13/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:531651:538758 [0] NCCL INFO Channel 14/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:531651:538758 [0] NCCL INFO Channel 15/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:531651:538758 [0] NCCL INFO Channel 16/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:531651:538758 [0] NCCL INFO Channel 17/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:531651:538758 [0] NCCL INFO Channel 18/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:531651:538758 [0] NCCL INFO Channel 19/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:531651:538758 [0] NCCL INFO Channel 20/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:531651:538758 [0] NCCL INFO Channel 21/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:531651:538758 [0] NCCL INFO Channel 22/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:531651:538758 [0] NCCL INFO Channel 23/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:531651:538758 [0] NCCL INFO Trees [0] 1/-1/-1->0->-1 [1] 1/-1/-1->0->-1 [2] 1/-1/-1->0->-1 [3] 1/-1/-1->0->-1 [4] 1/-1/-1->0->-1 [5] 1/-1/-1->0->-1 [6] 1/-1/-1->0->-1 [7] 1/-1/-1->0->-1 [8] 1/-1/-1->0->-1 [9] 1/-1/-1->0->-1 [10] 1/-1/-1->0->-1 [11] 1/-1/-1->0->-1 [12] 1/-1/-1->0->-1 [13] 1/-1/-1->0->-1 [14] 1/-1/-1->0->-1 [15] 1/-1/-1->0->-1 [16] 1/-1/-1->0->-1 [17] 1/-1/-1->0->-1 [18] 1/-1/-1->0->-1 [19] 1/-1/-1->0->-1 [20] 1/-1/-1->0->-1 [21] 1/-1/-1->0->-1 [22] 1/-1/-1->0->-1 [23] 1/-1/-1->0->-1 +ywang29-vrdb-test2-worker-0:531651:538758 [0] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:531658:538763 [7] NCCL INFO Channel 00/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:538759 [6] NCCL INFO Channel 00/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:538760 [4] NCCL INFO Channel 00/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:538764 [5] NCCL INFO Channel 00/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:538765 [2] NCCL INFO Channel 00/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:538763 [7] NCCL INFO Channel 01/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:538759 [6] NCCL INFO Channel 01/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:538762 [3] NCCL INFO Channel 00/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:538760 [4] NCCL INFO Channel 01/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:538764 [5] NCCL INFO Channel 01/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:538765 [2] NCCL INFO Channel 01/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:538759 [6] NCCL INFO Channel 02/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:538763 [7] NCCL INFO Channel 02/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:538762 [3] NCCL INFO Channel 01/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:538760 [4] NCCL INFO Channel 02/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:538764 [5] NCCL INFO Channel 02/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:538765 [2] NCCL INFO Channel 02/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:538759 [6] NCCL INFO Channel 03/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:538763 [7] NCCL INFO Channel 03/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:538762 [3] NCCL INFO Channel 02/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:538760 [4] NCCL INFO Channel 03/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:538761 [1] NCCL INFO Channel 00/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:538764 [5] NCCL INFO Channel 03/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:538765 [2] NCCL INFO Channel 03/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531651:538758 [0] NCCL INFO Channel 00/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:538759 [6] NCCL INFO Channel 04/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:538763 [7] NCCL INFO Channel 04/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:538762 [3] NCCL INFO Channel 03/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:538760 [4] NCCL INFO Channel 04/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:538761 [1] NCCL INFO Channel 01/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:538764 [5] NCCL INFO Channel 04/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:538765 [2] NCCL INFO Channel 04/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:538759 [6] NCCL INFO Channel 05/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531651:538758 [0] NCCL INFO Channel 01/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:538763 [7] NCCL INFO Channel 05/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:538762 [3] NCCL INFO Channel 04/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:538760 [4] NCCL INFO Channel 05/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:538765 [2] NCCL INFO Channel 05/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:538761 [1] NCCL INFO Channel 02/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:538764 [5] NCCL INFO Channel 05/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:538759 [6] NCCL INFO Channel 06/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:538762 [3] NCCL INFO Channel 05/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531651:538758 [0] NCCL INFO Channel 02/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:538763 [7] NCCL INFO Channel 06/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:538760 [4] NCCL INFO Channel 06/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:538765 [2] NCCL INFO Channel 06/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:538764 [5] NCCL INFO Channel 06/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:538761 [1] NCCL INFO Channel 03/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:538759 [6] NCCL INFO Channel 07/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531651:538758 [0] NCCL INFO Channel 03/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:538762 [3] NCCL INFO Channel 06/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:538763 [7] NCCL INFO Channel 07/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:538760 [4] NCCL INFO Channel 07/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:538765 [2] NCCL INFO Channel 07/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:538764 [5] NCCL INFO Channel 07/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:538761 [1] NCCL INFO Channel 04/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:538759 [6] NCCL INFO Channel 08/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:538763 [7] NCCL INFO Channel 08/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531651:538758 [0] NCCL INFO Channel 04/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:538762 [3] NCCL INFO Channel 07/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:538760 [4] NCCL INFO Channel 08/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:538765 [2] NCCL INFO Channel 08/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:538764 [5] NCCL INFO Channel 08/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:538761 [1] NCCL INFO Channel 05/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:538759 [6] NCCL INFO Channel 09/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:538763 [7] NCCL INFO Channel 09/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:538762 [3] NCCL INFO Channel 08/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531651:538758 [0] NCCL INFO Channel 05/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:538760 [4] NCCL INFO Channel 09/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:538765 [2] NCCL INFO Channel 09/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:538764 [5] NCCL INFO Channel 09/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:538761 [1] NCCL INFO Channel 06/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:538759 [6] NCCL INFO Channel 10/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:538763 [7] NCCL INFO Channel 10/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:538762 [3] NCCL INFO Channel 09/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531651:538758 [0] NCCL INFO Channel 06/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:538760 [4] NCCL INFO Channel 10/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:538765 [2] NCCL INFO Channel 10/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:538761 [1] NCCL INFO Channel 07/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:538764 [5] NCCL INFO Channel 10/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:538759 [6] NCCL INFO Channel 11/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:538763 [7] NCCL INFO Channel 11/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:538762 [3] NCCL INFO Channel 10/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531651:538758 [0] NCCL INFO Channel 07/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:538760 [4] NCCL INFO Channel 11/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:538765 [2] NCCL INFO Channel 11/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:538761 [1] NCCL INFO Channel 08/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:538764 [5] NCCL INFO Channel 11/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:538759 [6] NCCL INFO Channel 12/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:538763 [7] NCCL INFO Channel 12/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:538762 [3] NCCL INFO Channel 11/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531651:538758 [0] NCCL INFO Channel 08/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:538760 [4] NCCL INFO Channel 12/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:538765 [2] NCCL INFO Channel 12/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:538761 [1] NCCL INFO Channel 09/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:538764 [5] NCCL INFO Channel 12/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:538759 [6] NCCL INFO Channel 13/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:538763 [7] NCCL INFO Channel 13/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:538762 [3] NCCL INFO Channel 12/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531651:538758 [0] NCCL INFO Channel 09/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:538760 [4] NCCL INFO Channel 13/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:538765 [2] NCCL INFO Channel 13/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:538761 [1] NCCL INFO Channel 10/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:538759 [6] NCCL INFO Channel 14/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:538764 [5] NCCL INFO Channel 13/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:538763 [7] NCCL INFO Channel 14/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:538762 [3] NCCL INFO Channel 13/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:538760 [4] NCCL INFO Channel 14/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531651:538758 [0] NCCL INFO Channel 10/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:538765 [2] NCCL INFO Channel 14/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:538761 [1] NCCL INFO Channel 11/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:538759 [6] NCCL INFO Channel 15/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:538764 [5] NCCL INFO Channel 14/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:538763 [7] NCCL INFO Channel 15/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:538762 [3] NCCL INFO Channel 14/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:538760 [4] NCCL INFO Channel 15/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531651:538758 [0] NCCL INFO Channel 11/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:538761 [1] NCCL INFO Channel 12/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:538759 [6] NCCL INFO Channel 16/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:538764 [5] NCCL INFO Channel 15/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:538765 [2] NCCL INFO Channel 15/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:538763 [7] NCCL INFO Channel 16/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:538762 [3] NCCL INFO Channel 15/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:538760 [4] NCCL INFO Channel 16/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531651:538758 [0] NCCL INFO Channel 12/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:538761 [1] NCCL INFO Channel 13/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:538764 [5] NCCL INFO Channel 16/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:538765 [2] NCCL INFO Channel 16/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:538759 [6] NCCL INFO Channel 17/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:538763 [7] NCCL INFO Channel 17/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:538760 [4] NCCL INFO Channel 17/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:538762 [3] NCCL INFO Channel 16/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531651:538758 [0] NCCL INFO Channel 13/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:538764 [5] NCCL INFO Channel 17/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:538761 [1] NCCL INFO Channel 14/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:538765 [2] NCCL INFO Channel 17/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:538759 [6] NCCL INFO Channel 18/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:538763 [7] NCCL INFO Channel 18/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:538760 [4] NCCL INFO Channel 18/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:538762 [3] NCCL INFO Channel 17/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531651:538758 [0] NCCL INFO Channel 14/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:538764 [5] NCCL INFO Channel 18/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:538761 [1] NCCL INFO Channel 15/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:538765 [2] NCCL INFO Channel 18/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:538759 [6] NCCL INFO Channel 19/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:538763 [7] NCCL INFO Channel 19/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:538760 [4] NCCL INFO Channel 19/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:538762 [3] NCCL INFO Channel 18/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531651:538758 [0] NCCL INFO Channel 15/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:538764 [5] NCCL INFO Channel 19/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:538765 [2] NCCL INFO Channel 19/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:538761 [1] NCCL INFO Channel 16/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:538759 [6] NCCL INFO Channel 20/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:538763 [7] NCCL INFO Channel 20/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:538760 [4] NCCL INFO Channel 20/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:538762 [3] NCCL INFO Channel 19/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531651:538758 [0] NCCL INFO Channel 16/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:538764 [5] NCCL INFO Channel 20/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:538759 [6] NCCL INFO Channel 21/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:538765 [2] NCCL INFO Channel 20/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:538763 [7] NCCL INFO Channel 21/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:538761 [1] NCCL INFO Channel 17/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:538760 [4] NCCL INFO Channel 21/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:538762 [3] NCCL INFO Channel 20/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531651:538758 [0] NCCL INFO Channel 17/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:538764 [5] NCCL INFO Channel 21/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:538759 [6] NCCL INFO Channel 22/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:538765 [2] NCCL INFO Channel 21/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:538763 [7] NCCL INFO Channel 22/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:538760 [4] NCCL INFO Channel 22/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:538761 [1] NCCL INFO Channel 18/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:538762 [3] NCCL INFO Channel 21/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531651:538758 [0] NCCL INFO Channel 18/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:538764 [5] NCCL INFO Channel 22/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:538759 [6] NCCL INFO Channel 23/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:538765 [2] NCCL INFO Channel 22/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:538763 [7] NCCL INFO Channel 23/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:538761 [1] NCCL INFO Channel 19/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:538762 [3] NCCL INFO Channel 22/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:538760 [4] NCCL INFO Channel 23/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531651:538758 [0] NCCL INFO Channel 19/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:538764 [5] NCCL INFO Channel 23/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:538765 [2] NCCL INFO Channel 23/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:538761 [1] NCCL INFO Channel 20/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:538762 [3] NCCL INFO Channel 23/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531651:538758 [0] NCCL INFO Channel 20/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:538761 [1] NCCL INFO Channel 21/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531651:538758 [0] NCCL INFO Channel 21/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531651:538758 [0] NCCL INFO Channel 22/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:538761 [1] NCCL INFO Channel 22/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531651:538758 [0] NCCL INFO Channel 23/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:538761 [1] NCCL INFO Channel 23/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:538765 [2] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:531654:538762 [3] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:531651:538758 [0] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:531652:538761 [1] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:531653:538765 [2] NCCL INFO Channel 00/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:538760 [4] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:531653:538765 [2] NCCL INFO Channel 01/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:538765 [2] NCCL INFO Channel 02/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:538763 [7] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:531656:538764 [5] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:531658:538763 [7] NCCL INFO Channel 00/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:538759 [6] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:531653:538765 [2] NCCL INFO Channel 03/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:538763 [7] NCCL INFO Channel 01/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:538765 [2] NCCL INFO Channel 04/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:538763 [7] NCCL INFO Channel 02/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:538765 [2] NCCL INFO Channel 05/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:538763 [7] NCCL INFO Channel 03/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:538765 [2] NCCL INFO Channel 06/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:538763 [7] NCCL INFO Channel 04/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:538765 [2] NCCL INFO Channel 07/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:538763 [7] NCCL INFO Channel 05/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:538765 [2] NCCL INFO Channel 08/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:538763 [7] NCCL INFO Channel 06/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:538765 [2] NCCL INFO Channel 09/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:538763 [7] NCCL INFO Channel 07/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:538765 [2] NCCL INFO Channel 10/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:538763 [7] NCCL INFO Channel 08/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:538765 [2] NCCL INFO Channel 11/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:538763 [7] NCCL INFO Channel 09/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:538765 [2] NCCL INFO Channel 12/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:538765 [2] NCCL INFO Channel 13/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:538763 [7] NCCL INFO Channel 10/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:538765 [2] NCCL INFO Channel 14/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:538763 [7] NCCL INFO Channel 11/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:538765 [2] NCCL INFO Channel 15/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:538763 [7] NCCL INFO Channel 12/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:538765 [2] NCCL INFO Channel 16/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:538763 [7] NCCL INFO Channel 13/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:538762 [3] NCCL INFO Channel 00/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:538765 [2] NCCL INFO Channel 17/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:538763 [7] NCCL INFO Channel 14/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:538762 [3] NCCL INFO Channel 01/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:538765 [2] NCCL INFO Channel 18/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:538761 [1] NCCL INFO Channel 00/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:538763 [7] NCCL INFO Channel 15/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:538762 [3] NCCL INFO Channel 02/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:538765 [2] NCCL INFO Channel 19/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:538761 [1] NCCL INFO Channel 01/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:538763 [7] NCCL INFO Channel 16/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:538762 [3] NCCL INFO Channel 03/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:538765 [2] NCCL INFO Channel 20/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:538761 [1] NCCL INFO Channel 02/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:538762 [3] NCCL INFO Channel 04/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:538765 [2] NCCL INFO Channel 21/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:538761 [1] NCCL INFO Channel 03/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:538762 [3] NCCL INFO Channel 05/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:538765 [2] NCCL INFO Channel 22/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:538761 [1] NCCL INFO Channel 04/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:538762 [3] NCCL INFO Channel 06/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531653:538765 [2] NCCL INFO Channel 23/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:538761 [1] NCCL INFO Channel 05/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:538762 [3] NCCL INFO Channel 07/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:538761 [1] NCCL INFO Channel 06/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:538762 [3] NCCL INFO Channel 08/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:538761 [1] NCCL INFO Channel 07/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:538762 [3] NCCL INFO Channel 09/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:538761 [1] NCCL INFO Channel 08/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:538764 [5] NCCL INFO Channel 00/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:538759 [6] NCCL INFO Channel 00/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:538762 [3] NCCL INFO Channel 10/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:538761 [1] NCCL INFO Channel 09/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:538764 [5] NCCL INFO Channel 01/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:538759 [6] NCCL INFO Channel 01/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:538762 [3] NCCL INFO Channel 11/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:538761 [1] NCCL INFO Channel 10/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:538764 [5] NCCL INFO Channel 02/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:538759 [6] NCCL INFO Channel 02/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:538762 [3] NCCL INFO Channel 12/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:538761 [1] NCCL INFO Channel 11/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:538764 [5] NCCL INFO Channel 03/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:538759 [6] NCCL INFO Channel 03/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:538762 [3] NCCL INFO Channel 13/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:538761 [1] NCCL INFO Channel 12/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:538764 [5] NCCL INFO Channel 04/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:538759 [6] NCCL INFO Channel 04/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:538762 [3] NCCL INFO Channel 14/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:538761 [1] NCCL INFO Channel 13/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:538764 [5] NCCL INFO Channel 05/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:538759 [6] NCCL INFO Channel 05/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:538762 [3] NCCL INFO Channel 15/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:538761 [1] NCCL INFO Channel 14/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:538764 [5] NCCL INFO Channel 06/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:538759 [6] NCCL INFO Channel 06/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:538762 [3] NCCL INFO Channel 16/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:538761 [1] NCCL INFO Channel 15/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:538764 [5] NCCL INFO Channel 07/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:538759 [6] NCCL INFO Channel 07/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:538762 [3] NCCL INFO Channel 17/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:538761 [1] NCCL INFO Channel 16/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:538764 [5] NCCL INFO Channel 08/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:538759 [6] NCCL INFO Channel 08/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:538762 [3] NCCL INFO Channel 18/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:538761 [1] NCCL INFO Channel 17/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:538764 [5] NCCL INFO Channel 09/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:538759 [6] NCCL INFO Channel 09/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:538762 [3] NCCL INFO Channel 19/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:538761 [1] NCCL INFO Channel 18/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:538764 [5] NCCL INFO Channel 10/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:538759 [6] NCCL INFO Channel 10/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:538762 [3] NCCL INFO Channel 20/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:538761 [1] NCCL INFO Channel 19/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:538764 [5] NCCL INFO Channel 11/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:538759 [6] NCCL INFO Channel 11/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:538762 [3] NCCL INFO Channel 21/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:538761 [1] NCCL INFO Channel 20/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:538763 [7] NCCL INFO Channel 17/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:538764 [5] NCCL INFO Channel 12/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:538759 [6] NCCL INFO Channel 12/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:538762 [3] NCCL INFO Channel 22/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:538761 [1] NCCL INFO Channel 21/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:538763 [7] NCCL INFO Channel 18/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:538764 [5] NCCL INFO Channel 13/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:538759 [6] NCCL INFO Channel 13/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531654:538762 [3] NCCL INFO Channel 23/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:538761 [1] NCCL INFO Channel 22/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:538763 [7] NCCL INFO Channel 19/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:538764 [5] NCCL INFO Channel 14/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:538759 [6] NCCL INFO Channel 14/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531652:538761 [1] NCCL INFO Channel 23/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:538763 [7] NCCL INFO Channel 20/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:538764 [5] NCCL INFO Channel 15/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:538759 [6] NCCL INFO Channel 15/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:538763 [7] NCCL INFO Channel 21/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:538760 [4] NCCL INFO Channel 00/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:538764 [5] NCCL INFO Channel 16/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:538759 [6] NCCL INFO Channel 16/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:538763 [7] NCCL INFO Channel 22/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:538760 [4] NCCL INFO Channel 01/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:538759 [6] NCCL INFO Channel 17/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:538764 [5] NCCL INFO Channel 17/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531658:538763 [7] NCCL INFO Channel 23/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:538760 [4] NCCL INFO Channel 02/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:538759 [6] NCCL INFO Channel 18/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:538759 [6] NCCL INFO Channel 19/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:538764 [5] NCCL INFO Channel 18/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:538760 [4] NCCL INFO Channel 03/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:538764 [5] NCCL INFO Channel 19/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:538759 [6] NCCL INFO Channel 20/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:538760 [4] NCCL INFO Channel 04/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:538764 [5] NCCL INFO Channel 20/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:538759 [6] NCCL INFO Channel 21/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:538764 [5] NCCL INFO Channel 21/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:538759 [6] NCCL INFO Channel 22/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:538760 [4] NCCL INFO Channel 05/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:538764 [5] NCCL INFO Channel 22/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:538760 [4] NCCL INFO Channel 06/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531657:538759 [6] NCCL INFO Channel 23/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531656:538764 [5] NCCL INFO Channel 23/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:538760 [4] NCCL INFO Channel 07/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:538760 [4] NCCL INFO Channel 08/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:538760 [4] NCCL INFO Channel 09/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:538760 [4] NCCL INFO Channel 10/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:538760 [4] NCCL INFO Channel 11/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:538760 [4] NCCL INFO Channel 12/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:538760 [4] NCCL INFO Channel 13/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:538760 [4] NCCL INFO Channel 14/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:538760 [4] NCCL INFO Channel 15/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:538760 [4] NCCL INFO Channel 16/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:538760 [4] NCCL INFO Channel 17/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:538760 [4] NCCL INFO Channel 18/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:538760 [4] NCCL INFO Channel 19/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:538760 [4] NCCL INFO Channel 20/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:538760 [4] NCCL INFO Channel 21/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:538760 [4] NCCL INFO Channel 22/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531655:538760 [4] NCCL INFO Channel 23/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:531651:538758 [0] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:531651:538758 [0] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:531651:538758 [0] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:531652:538761 [1] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:531652:538761 [1] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:531652:538761 [1] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:531658:538763 [7] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:531658:538763 [7] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:531658:538763 [7] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:531653:538765 [2] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:531653:538765 [2] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:531653:538765 [2] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:531654:538762 [3] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:531654:538762 [3] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:531654:538762 [3] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:531655:538760 [4] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:531655:538760 [4] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:531655:538760 [4] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:531657:538759 [6] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:531657:538759 [6] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:531656:538764 [5] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:531657:538759 [6] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:531656:538764 [5] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:531656:538764 [5] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:531654:538762 [3] NCCL INFO ncclCommInitRank comm 0x7f82d006a7d0 rank 3 nranks 8 cudaDev 3 nvmlDev 3 busId 201d0 commId 0x263709ab87944962 - Init COMPLETE +ywang29-vrdb-test2-worker-0:531655:538760 [4] NCCL INFO ncclCommInitRank comm 0x7f7a4406bba0 rank 4 nranks 8 cudaDev 4 nvmlDev 4 busId 901c0 commId 0x263709ab87944962 - Init COMPLETE +ywang29-vrdb-test2-worker-0:531657:538759 [6] NCCL INFO ncclCommInitRank comm 0x7fcfdc06b490 rank 6 nranks 8 cudaDev 6 nvmlDev 6 busId a01c0 commId 0x263709ab87944962 - Init COMPLETE +ywang29-vrdb-test2-worker-0:531658:538763 [7] NCCL INFO ncclCommInitRank comm 0x7f7f7006a680 rank 7 nranks 8 cudaDev 7 nvmlDev 7 busId a01d0 commId 0x263709ab87944962 - Init COMPLETE +ywang29-vrdb-test2-worker-0:531653:538765 [2] NCCL INFO ncclCommInitRank comm 0x7f2bc0069ce0 rank 2 nranks 8 cudaDev 2 nvmlDev 2 busId 201c0 commId 0x263709ab87944962 - Init COMPLETE +ywang29-vrdb-test2-worker-0:531656:538764 [5] NCCL INFO ncclCommInitRank comm 0x7fc04806af40 rank 5 nranks 8 cudaDev 5 nvmlDev 5 busId 901d0 commId 0x263709ab87944962 - Init COMPLETE +ywang29-vrdb-test2-worker-0:531652:538761 [1] NCCL INFO ncclCommInitRank comm 0x7f546c06b610 rank 1 nranks 8 cudaDev 1 nvmlDev 1 busId 101d0 commId 0x263709ab87944962 - Init COMPLETE +ywang29-vrdb-test2-worker-0:531651:538758 [0] NCCL INFO ncclCommInitRank comm 0x7f6ca006b3b0 rank 0 nranks 8 cudaDev 0 nvmlDev 0 busId 101c0 commId 0x263709ab87944962 - Init COMPLETE + 0%| | 1/520 [00:14<2:02:08, 14.12s/it] {'loss': 2.0453, 'grad_norm': 0.004834342655741857, 'learning_rate': 0.05625, 'epoch': 0.0} + 0%| | 1/520 [00:14<2:02:08, 14.12s/it] 0%| | 2/520 [00:17<1:08:43, 7.96s/it] {'loss': 2.0549, 'grad_norm': 0.005249094463000703, 'learning_rate': 0.1125, 'epoch': 0.0} + 0%| | 2/520 [00:17<1:08:43, 7.96s/it] 1%| | 3/520 [00:21<51:34, 5.99s/it] {'loss': 2.1899, 'grad_norm': 0.006004669552221527, 'learning_rate': 0.16875, 'epoch': 0.01} + 1%| | 3/520 [00:21<51:34, 5.99s/it] 1%| | 4/520 [00:24<43:21, 5.04s/it] {'loss': 1.6662, 'grad_norm': 0.0014854292319139368, 'learning_rate': 0.225, 'epoch': 0.01} + 1%| | 4/520 [00:25<43:21, 5.04s/it] 1%| | 5/520 [00:28<39:04, 4.55s/it] {'loss': 1.6543, 'grad_norm': 0.0007715164628665445, 'learning_rate': 0.28125, 'epoch': 0.01} + 1%| | 5/520 [00:28<39:04, 4.55s/it] 1%| | 6/520 [00:32<36:24, 4.25s/it] {'loss': 1.3717, 'grad_norm': 0.0005455655559058728, 'learning_rate': 0.3375, 'epoch': 0.01} + 1%| | 6/520 [00:32<36:24, 4.25s/it] 1%|▏ | 7/520 [00:35<34:36, 4.05s/it] {'loss': 1.4196, 'grad_norm': 0.0007208371149205127, 'learning_rate': 0.39375, 'epoch': 0.01} + 1%|▏ | 7/520 [00:35<34:36, 4.05s/it] 2%|▏ | 8/520 [00:40<35:12, 4.13s/it] {'loss': 1.4553, 'grad_norm': 0.0009351432944771337, 'learning_rate': 0.45, 'epoch': 0.02} + 2%|▏ | 8/520 [00:40<35:12, 4.13s/it] 2%|▏ | 9/520 [00:44<35:22, 4.15s/it] {'loss': 1.5258, 'grad_norm': 0.0014097436621606084, 'learning_rate': 0.50625, 'epoch': 0.02} + 2%|▏ | 9/520 [00:44<35:22, 4.15s/it] 2%|▏ | 10/520 [00:48<34:03, 4.01s/it] {'loss': 1.36, 'grad_norm': 0.0021721445767895033, 'learning_rate': 0.5625, 'epoch': 0.02} + 2%|▏ | 10/520 [00:48<34:03, 4.01s/it] 2%|▏ | 11/520 [00:51<33:25, 3.94s/it] {'loss': 1.5236, 'grad_norm': 0.003930048480545171, 'learning_rate': 0.61875, 'epoch': 0.02} + 2%|▏ | 11/520 [00:51<33:25, 3.94s/it] 2%|▏ | 12/520 [00:55<32:38, 3.85s/it] {'loss': 1.485, 'grad_norm': 0.005233456418958682, 'learning_rate': 0.675, 'epoch': 0.02} + 2%|▏ | 12/520 [00:55<32:38, 3.85s/it][2025-10-12 08:33:08,100] [WARNING] [stage3.py:2069:step] 1 pytorch allocator cache flushes since last step. this happens when there is high memory pressure and is detrimental to performance. if this is happening frequently consider adjusting settings to reduce memory consumption. If you are unable to make the cache flushes go away consider adding get_accelerator().empty_cache() calls in your training loop to ensure that all ranks flush their caches at the same time + 2%|▎ | 13/520 [00:59<33:50, 4.01s/it] {'loss': 1.824, 'grad_norm': 0.013339066614449033, 'learning_rate': 0.7312500000000001, 'epoch': 0.03} + 2%|▎ | 13/520 [00:59<33:50, 4.01s/it] 3%|▎ | 14/520 [01:03<32:58, 3.91s/it] {'loss': 2.0279, 'grad_norm': 0.023440117967150065, 'learning_rate': 0.7875, 'epoch': 0.03} + 3%|▎ | 14/520 [01:03<32:58, 3.91s/it] 3%|▎ | 15/520 [01:07<32:19, 3.84s/it] {'loss': 2.02, 'grad_norm': 0.020811526355123355, 'learning_rate': 0.84375, 'epoch': 0.03} + 3%|▎ | 15/520 [01:07<32:19, 3.84s/it] 3%|▎ | 16/520 [01:10<31:46, 3.78s/it] {'loss': 2.3644, 'grad_norm': 0.02339189032332367, 'learning_rate': 0.9, 'epoch': 0.03} + 3%|▎ | 16/520 [01:10<31:46, 3.78s/it] 3%|▎ | 17/520 [01:14<31:20, 3.74s/it] {'loss': 3.8357, 'grad_norm': 0.11489304537784638, 'learning_rate': 0.8999912578191922, 'epoch': 0.03} + 3%|▎ | 17/520 [01:14<31:20, 3.74s/it] 3%|▎ | 18/520 [01:18<31:07, 3.72s/it] {'loss': 7.0193, 'grad_norm': 0.2169400548698785, 'learning_rate': 0.8999650316164387, 'epoch': 0.03} + 3%|▎ | 18/520 [01:18<31:07, 3.72s/it] 4%|▎ | 19/520 [01:21<30:56, 3.70s/it] {'loss': 4.0828, 'grad_norm': 0.08146021159463933, 'learning_rate': 0.8999213224107359, 'epoch': 0.04} + 4%|▎ | 19/520 [01:21<30:56, 3.70s/it] 4%|▍ | 20/520 [01:25<30:49, 3.70s/it] {'loss': 3.3418, 'grad_norm': 0.07228251103077857, 'learning_rate': 0.8998601319003673, 'epoch': 0.04} + 4%|▍ | 20/520 [01:25<30:49, 3.70s/it] 4%|▍ | 21/520 [01:29<30:45, 3.70s/it] {'loss': 3.6403, 'grad_norm': 0.055643323055046115, 'learning_rate': 0.8997814624628374, 'epoch': 0.04} + 4%|▍ | 21/520 [01:29<30:45, 3.70s/it] 4%|▍ | 22/520 [01:32<30:33, 3.68s/it] {'loss': 3.6152, 'grad_norm': 0.054386438634032246, 'learning_rate': 0.8996853171547794, 'epoch': 0.04} + 4%|▍ | 22/520 [01:32<30:33, 3.68s/it] 4%|▍ | 23/520 [01:36<30:26, 3.68s/it] {'loss': 4.1822, 'grad_norm': 0.08271478399773145, 'learning_rate': 0.899571699711836, 'epoch': 0.04} + 4%|▍ | 23/520 [01:36<30:26, 3.68s/it] 5%|▍ | 24/520 [01:40<30:18, 3.67s/it] {'loss': 4.344, 'grad_norm': 0.06400956875079636, 'learning_rate': 0.899440614548515, 'epoch': 0.05} + 5%|▍ | 24/520 [01:40<30:18, 3.67s/it] 5%|▍ | 25/520 [01:43<30:18, 3.67s/it] {'loss': 3.3713, 'grad_norm': 0.02075828967554016, 'learning_rate': 0.8992920667580175, 'epoch': 0.05} + 5%|▍ | 25/520 [01:43<30:18, 3.67s/it] 5%|▌ | 26/520 [01:47<30:16, 3.68s/it] {'loss': 2.9867, 'grad_norm': 0.01859605232256852, 'learning_rate': 0.8991260621120394, 'epoch': 0.05} + 5%|▌ | 26/520 [01:47<30:16, 3.68s/it] 5%|▌ | 27/520 [01:51<30:11, 3.67s/it] {'loss': 2.4951, 'grad_norm': 0.005974212041367981, 'learning_rate': 0.898942607060548, 'epoch': 0.05} + 5%|▌ | 27/520 [01:51<30:11, 3.67s/it] 5%|▌ | 28/520 [01:54<30:04, 3.67s/it] {'loss': 2.3944, 'grad_norm': 0.006364630445574721, 'learning_rate': 0.898741708731531, 'epoch': 0.05} + 5%|▌ | 28/520 [01:54<30:04, 3.67s/it] 6%|▌ | 29/520 [01:58<30:08, 3.68s/it] {'loss': 2.2742, 'grad_norm': 0.004509660583951691, 'learning_rate': 0.8985233749307198, 'epoch': 0.06} + 6%|▌ | 29/520 [01:58<30:08, 3.68s/it] 6%|▌ | 30/520 [02:02<30:05, 3.68s/it] {'loss': 2.8261, 'grad_norm': 0.004492835388974667, 'learning_rate': 0.8982876141412856, 'epoch': 0.06} + 6%|▌ | 30/520 [02:02<30:05, 3.68s/it] 6%|▌ | 31/520 [02:06<29:56, 3.67s/it] {'loss': 2.1874, 'grad_norm': 0.004201116625428159, 'learning_rate': 0.8980344355235101, 'epoch': 0.06} + 6%|▌ | 31/520 [02:06<29:56, 3.67s/it] 6%|▌ | 32/520 [02:09<29:49, 3.67s/it] {'loss': 3.0295, 'grad_norm': 0.0069792022555074905, 'learning_rate': 0.8977638489144307, 'epoch': 0.06} + 6%|▌ | 32/520 [02:09<29:49, 3.67s/it] 6%|▋ | 33/520 [02:13<29:41, 3.66s/it] {'loss': 2.1066, 'grad_norm': 0.0029009502324039587, 'learning_rate': 0.8974758648274559, 'epoch': 0.06} + 6%|▋ | 33/520 [02:13<29:41, 3.66s/it] 7%|▋ | 34/520 [02:16<29:36, 3.66s/it] {'loss': 1.9923, 'grad_norm': 0.002750243422687705, 'learning_rate': 0.8971704944519593, 'epoch': 0.07} + 7%|▋ | 34/520 [02:16<29:36, 3.66s/it] 7%|▋ | 35/520 [02:20<29:28, 3.65s/it] {'loss': 2.0, 'grad_norm': 0.002249950415621659, 'learning_rate': 0.8968477496528428, 'epoch': 0.07} + 7%|▋ | 35/520 [02:20<29:28, 3.65s/it] 7%|▋ | 36/520 [02:24<29:21, 3.64s/it] {'loss': 2.1321, 'grad_norm': 0.0024446267651994378, 'learning_rate': 0.8965076429700775, 'epoch': 0.07} + 7%|▋ | 36/520 [02:24<29:21, 3.64s/it] 7%|▋ | 37/520 [02:27<29:23, 3.65s/it] {'loss': 2.3851, 'grad_norm': 0.0033720318927310177, 'learning_rate': 0.8961501876182147, 'epoch': 0.07} + 7%|▋ | 37/520 [02:27<29:23, 3.65s/it] 7%|▋ | 38/520 [02:31<29:19, 3.65s/it] {'loss': 2.1478, 'grad_norm': 0.0020492462787480618, 'learning_rate': 0.8957753974858738, 'epoch': 0.07} + 7%|▋ | 38/520 [02:31<29:19, 3.65s/it] 8%|▊ | 39/520 [02:35<29:11, 3.64s/it] {'loss': 1.9511, 'grad_norm': 0.001931134155959717, 'learning_rate': 0.8953832871352018, 'epoch': 0.07} + 8%|▊ | 39/520 [02:35<29:11, 3.64s/it] 8%|▊ | 40/520 [02:38<29:10, 3.65s/it] {'loss': 1.9357, 'grad_norm': 0.0017837044242085332, 'learning_rate': 0.8949738718013078, 'epoch': 0.08} + 8%|▊ | 40/520 [02:38<29:10, 3.65s/it] 8%|▊ | 41/520 [02:42<29:11, 3.66s/it] {'loss': 1.9342, 'grad_norm': 0.0020556244822901608, 'learning_rate': 0.8945471673916716, 'epoch': 0.08} + 8%|▊ | 41/520 [02:42<29:11, 3.66s/it] 8%|▊ | 42/520 [02:46<29:02, 3.65s/it] {'loss': 1.9945, 'grad_norm': 0.0025347498365796625, 'learning_rate': 0.8941031904855246, 'epoch': 0.08} + 8%|▊ | 42/520 [02:46<29:02, 3.65s/it] 8%|▊ | 43/520 [02:49<28:59, 3.65s/it] {'loss': 2.1615, 'grad_norm': 0.0021576961840093197, 'learning_rate': 0.8936419583332061, 'epoch': 0.08} + 8%|▊ | 43/520 [02:49<28:59, 3.65s/it] 8%|▊ | 44/520 [02:53<28:54, 3.64s/it] {'loss': 2.3012, 'grad_norm': 0.002626679532480975, 'learning_rate': 0.8931634888554937, 'epoch': 0.08} + 8%|▊ | 44/520 [02:53<28:54, 3.64s/it] 9%|▊ | 45/520 [02:57<28:50, 3.64s/it] {'loss': 1.8606, 'grad_norm': 0.0016633947012036408, 'learning_rate': 0.8926678006429056, 'epoch': 0.09} + 9%|▊ | 45/520 [02:57<28:50, 3.64s/it] 9%|▉ | 46/520 [03:00<28:47, 3.64s/it] {'loss': 2.2954, 'grad_norm': 0.0020059361639954676, 'learning_rate': 0.8921549129549798, 'epoch': 0.09} + 9%|▉ | 46/520 [03:00<28:47, 3.64s/it] 9%|▉ | 47/520 [03:04<28:37, 3.63s/it] {'loss': 1.8725, 'grad_norm': 0.0015252595859714016, 'learning_rate': 0.8916248457195245, 'epoch': 0.09} + 9%|▉ | 47/520 [03:04<28:37, 3.63s/it] 9%|▉ | 48/520 [03:07<28:38, 3.64s/it] {'loss': 1.8357, 'grad_norm': 0.001731216044336632, 'learning_rate': 0.8910776195318447, 'epoch': 0.09} + 9%|▉ | 48/520 [03:07<28:38, 3.64s/it] 9%|▉ | 49/520 [03:11<28:35, 3.64s/it] {'loss': 1.8187, 'grad_norm': 0.0016542608227869179, 'learning_rate': 0.8905132556539418, 'epoch': 0.09} + 9%|▉ | 49/520 [03:11<28:35, 3.64s/it] 10%|▉ | 50/520 [03:15<28:28, 3.64s/it] {'loss': 1.8181, 'grad_norm': 0.0012193108789326085, 'learning_rate': 0.8899317760136871, 'epoch': 0.1} + 10%|▉ | 50/520 [03:15<28:28, 3.64s/it] 10%|▉ | 51/520 [03:18<28:25, 3.64s/it] {'loss': 1.7089, 'grad_norm': 0.0014488467863337377, 'learning_rate': 0.88933320320397, 'epoch': 0.1} + 10%|▉ | 51/520 [03:18<28:25, 3.64s/it] 10%|█ | 52/520 [03:22<28:20, 3.63s/it] {'loss': 1.875, 'grad_norm': 0.001639744368334642, 'learning_rate': 0.8887175604818206, 'epoch': 0.1} + 10%|█ | 52/520 [03:22<28:20, 3.63s/it] 10%|█ | 53/520 [03:26<28:13, 3.63s/it] {'loss': 1.8742, 'grad_norm': 0.0012832110186795046, 'learning_rate': 0.8880848717675055, 'epoch': 0.1} + 10%|█ | 53/520 [03:26<28:13, 3.63s/it] 10%|█ | 54/520 [03:29<28:09, 3.63s/it] {'loss': 1.7217, 'grad_norm': 0.0012114423046959142, 'learning_rate': 0.8874351616435986, 'epoch': 0.1} + 10%|█ | 54/520 [03:29<28:09, 3.63s/it] 11%|█ | 55/520 [03:33<28:09, 3.63s/it] {'loss': 1.7252, 'grad_norm': 0.0011625044518257387, 'learning_rate': 0.8867684553540259, 'epoch': 0.11} + 11%|█ | 55/520 [03:33<28:09, 3.63s/it] 11%|█ | 56/520 [03:37<28:06, 3.64s/it] {'loss': 1.8568, 'grad_norm': 0.0011962665211794918, 'learning_rate': 0.8860847788030851, 'epoch': 0.11} + 11%|█ | 56/520 [03:37<28:06, 3.64s/it] 11%|█ | 57/520 [03:40<28:00, 3.63s/it] {'loss': 1.7021, 'grad_norm': 0.0013973197237610998, 'learning_rate': 0.8853841585544384, 'epoch': 0.11} + 11%|█ | 57/520 [03:40<28:00, 3.63s/it] 11%|█ | 58/520 [03:44<27:56, 3.63s/it] {'loss': 1.8759, 'grad_norm': 0.001108036762828575, 'learning_rate': 0.8846666218300807, 'epoch': 0.11} + 11%|█ | 58/520 [03:44<27:56, 3.63s/it] 11%|█▏ | 59/520 [03:47<27:57, 3.64s/it] {'loss': 1.9094, 'grad_norm': 0.0013055153387259689, 'learning_rate': 0.8839321965092826, 'epoch': 0.11} + 11%|█▏ | 59/520 [03:47<27:57, 3.64s/it] 12%|█▏ | 60/520 [03:51<27:55, 3.64s/it] {'loss': 1.7939, 'grad_norm': 0.0010871467454737834, 'learning_rate': 0.8831809111275054, 'epoch': 0.12} + 12%|█▏ | 60/520 [03:51<27:55, 3.64s/it] 12%|█▏ | 61/520 [03:55<27:53, 3.65s/it] {'loss': 2.0784, 'grad_norm': 0.0013428491001244445, 'learning_rate': 0.8824127948752949, 'epoch': 0.12} + 12%|█▏ | 61/520 [03:55<27:53, 3.65s/it] 12%|█▏ | 62/520 [03:58<27:53, 3.65s/it] {'loss': 1.7474, 'grad_norm': 0.0010726989960033921, 'learning_rate': 0.8816278775971446, 'epoch': 0.12} + 12%|█▏ | 62/520 [03:58<27:53, 3.65s/it] 12%|█▏ | 63/520 [04:02<27:49, 3.65s/it] {'loss': 1.7586, 'grad_norm': 0.000981689461658485, 'learning_rate': 0.8808261897903381, 'epoch': 0.12} + 12%|█▏ | 63/520 [04:02<27:49, 3.65s/it] 12%|█▏ | 64/520 [04:06<27:49, 3.66s/it] {'loss': 1.7565, 'grad_norm': 0.0011021159030391917, 'learning_rate': 0.8800077626037633, 'epoch': 0.12} + 12%|█▏ | 64/520 [04:06<27:49, 3.66s/it] 12%|█▎ | 65/520 [04:09<27:42, 3.65s/it] {'loss': 1.793, 'grad_norm': 0.001273139180149171, 'learning_rate': 0.8791726278367021, 'epoch': 0.12} + 12%|█▎ | 65/520 [04:09<27:42, 3.65s/it] 13%|█▎ | 66/520 [04:13<27:35, 3.65s/it] {'loss': 1.745, 'grad_norm': 0.0011151463112299097, 'learning_rate': 0.8783208179375951, 'epoch': 0.13} + 13%|█▎ | 66/520 [04:13<27:35, 3.65s/it] 13%|█▎ | 67/520 [04:17<27:36, 3.66s/it] {'loss': 1.5914, 'grad_norm': 0.0010901349798467122, 'learning_rate': 0.8774523660027806, 'epoch': 0.13} + 13%|█▎ | 67/520 [04:17<27:36, 3.66s/it] 13%|█▎ | 68/520 [04:20<27:30, 3.65s/it] {'loss': 1.6302, 'grad_norm': 0.0011586095363339468, 'learning_rate': 0.8765673057752094, 'epoch': 0.13} + 13%|█▎ | 68/520 [04:20<27:30, 3.65s/it] 13%|█▎ | 69/520 [04:24<27:26, 3.65s/it] {'loss': 1.6112, 'grad_norm': 0.0013241519381821274, 'learning_rate': 0.8756656716431321, 'epoch': 0.13} + 13%|█▎ | 69/520 [04:24<27:26, 3.65s/it] 13%|█▎ | 70/520 [04:28<27:23, 3.65s/it] {'loss': 1.6881, 'grad_norm': 0.0011207315164303469, 'learning_rate': 0.8747474986387654, 'epoch': 0.13} + 13%|█▎ | 70/520 [04:28<27:23, 3.65s/it] 14%|█▎ | 71/520 [04:31<27:19, 3.65s/it] {'loss': 1.5773, 'grad_norm': 0.001217708908778682, 'learning_rate': 0.8738128224369286, 'epoch': 0.14} + 14%|█▎ | 71/520 [04:31<27:19, 3.65s/it] 14%|█▍ | 72/520 [04:35<27:13, 3.65s/it] {'loss': 1.7376, 'grad_norm': 0.0011087327619000598, 'learning_rate': 0.8728616793536588, 'epoch': 0.14} + 14%|█▍ | 72/520 [04:35<27:13, 3.65s/it] 14%|█▍ | 73/520 [04:39<27:12, 3.65s/it] {'loss': 1.538, 'grad_norm': 0.0009239166606567668, 'learning_rate': 0.8718941063447997, 'epoch': 0.14} + 14%|█▍ | 73/520 [04:39<27:12, 3.65s/it] 14%|█▍ | 74/520 [04:42<27:07, 3.65s/it] {'loss': 1.6904, 'grad_norm': 0.0011289850629814934, 'learning_rate': 0.8709101410045653, 'epoch': 0.14} + 14%|█▍ | 74/520 [04:42<27:07, 3.65s/it] 14%|█▍ | 75/520 [04:46<27:01, 3.64s/it] {'loss': 1.5811, 'grad_norm': 0.001169858364186847, 'learning_rate': 0.8699098215640799, 'epoch': 0.14} + 14%|█▍ | 75/520 [04:46<27:01, 3.64s/it] 15%|█▍ | 76/520 [04:50<27:07, 3.67s/it] {'loss': 2.024, 'grad_norm': 0.001008543379026055, 'learning_rate': 0.8688931868898919, 'epoch': 0.15} + 15%|█▍ | 76/520 [04:50<27:07, 3.67s/it] 15%|█▍ | 77/520 [04:53<27:25, 3.71s/it] {'loss': 1.5132, 'grad_norm': 0.00119958115887291, 'learning_rate': 0.8678602764824641, 'epoch': 0.15} + 15%|█▍ | 77/520 [04:53<27:25, 3.71s/it] 15%|█▌ | 78/520 [04:57<27:44, 3.77s/it] {'loss': 1.6307, 'grad_norm': 0.0010270654880059566, 'learning_rate': 0.8668111304746389, 'epoch': 0.15} + 15%|█▌ | 78/520 [04:57<27:44, 3.77s/it] 15%|█▌ | 79/520 [05:01<27:47, 3.78s/it] {'loss': 1.6035, 'grad_norm': 0.0009633497012778588, 'learning_rate': 0.8657457896300791, 'epoch': 0.15} + 15%|█▌ | 79/520 [05:01<27:47, 3.78s/it] 15%|█▌ | 80/520 [05:05<27:59, 3.82s/it] {'loss': 2.0501, 'grad_norm': 0.0013283716065271179, 'learning_rate': 0.8646642953416834, 'epoch': 0.15} + 15%|█▌ | 80/520 [05:05<27:59, 3.82s/it] 16%|█▌ | 81/520 [05:09<28:00, 3.83s/it] {'loss': 1.819, 'grad_norm': 0.0011474662125611418, 'learning_rate': 0.8635666896299793, 'epoch': 0.16} + 16%|█▌ | 81/520 [05:09<28:00, 3.83s/it] 16%|█▌ | 82/520 [05:13<27:59, 3.83s/it] {'loss': 1.7012, 'grad_norm': 0.0008247772717886263, 'learning_rate': 0.8624530151414893, 'epoch': 0.16} + 16%|█▌ | 82/520 [05:13<27:59, 3.83s/it] 16%|█▌ | 83/520 [05:17<28:05, 3.86s/it] {'loss': 1.7178, 'grad_norm': 0.0011262138492133468, 'learning_rate': 0.8613233151470745, 'epoch': 0.16} + 16%|█▌ | 83/520 [05:17<28:05, 3.86s/it] 16%|█▌ | 84/520 [05:20<27:47, 3.82s/it] {'loss': 1.726, 'grad_norm': 0.0013053526924338649, 'learning_rate': 0.860177633540253, 'epoch': 0.16} + 16%|█▌ | 84/520 [05:20<27:47, 3.82s/it] 16%|█▋ | 85/520 [05:24<27:28, 3.79s/it] {'loss': 1.7173, 'grad_norm': 0.001308930606345037, 'learning_rate': 0.859016014835495, 'epoch': 0.16} + 16%|█▋ | 85/520 [05:24<27:28, 3.79s/it] 17%|█▋ | 86/520 [05:28<27:07, 3.75s/it] {'loss': 1.8025, 'grad_norm': 0.0008838140073237333, 'learning_rate': 0.8578385041664925, 'epoch': 0.17} + 17%|█▋ | 86/520 [05:28<27:07, 3.75s/it] 17%|█▋ | 87/520 [05:31<26:58, 3.74s/it] {'loss': 2.005, 'grad_norm': 0.0015374798672575661, 'learning_rate': 0.8566451472844065, 'epoch': 0.17} + 17%|█▋ | 87/520 [05:31<26:58, 3.74s/it] 17%|█▋ | 88/520 [05:35<26:46, 3.72s/it] {'loss': 2.1336, 'grad_norm': 0.0010382735506870146, 'learning_rate': 0.8554359905560885, 'epoch': 0.17} + 17%|█▋ | 88/520 [05:35<26:46, 3.72s/it] 17%|█▋ | 89/520 [05:39<26:34, 3.70s/it] {'loss': 1.6819, 'grad_norm': 0.0008852626415659759, 'learning_rate': 0.8542110809622799, 'epoch': 0.17} + 17%|█▋ | 89/520 [05:39<26:34, 3.70s/it] 17%|█▋ | 90/520 [05:42<26:27, 3.69s/it] {'loss': 1.6076, 'grad_norm': 0.0007661042704893362, 'learning_rate': 0.8529704660957854, 'epoch': 0.17} + 17%|█▋ | 90/520 [05:42<26:27, 3.69s/it] 18%|█▊ | 91/520 [05:46<26:16, 3.68s/it] {'loss': 1.7041, 'grad_norm': 0.0008909202415165524, 'learning_rate': 0.8517141941596252, 'epoch': 0.17} + 18%|█▊ | 91/520 [05:46<26:16, 3.68s/it] 18%|█▊ | 92/520 [05:50<26:05, 3.66s/it] {'loss': 1.6282, 'grad_norm': 0.001040100399044249, 'learning_rate': 0.850442313965161, 'epoch': 0.18} + 18%|█▊ | 92/520 [05:50<26:05, 3.66s/it] 18%|█▊ | 93/520 [05:53<26:01, 3.66s/it] {'loss': 1.6023, 'grad_norm': 0.001070866353309159, 'learning_rate': 0.8491548749301997, 'epoch': 0.18} + 18%|█▊ | 93/520 [05:53<26:01, 3.66s/it] 18%|█▊ | 94/520 [05:57<25:56, 3.65s/it] {'loss': 1.7381, 'grad_norm': 0.0010980089263019847, 'learning_rate': 0.8478519270770745, 'epoch': 0.18} + 18%|█▊ | 94/520 [05:57<25:56, 3.65s/it] 18%|█▊ | 95/520 [06:01<25:46, 3.64s/it] {'loss': 1.6185, 'grad_norm': 0.0010563483159591654, 'learning_rate': 0.846533521030699, 'epoch': 0.18} + 18%|█▊ | 95/520 [06:01<25:46, 3.64s/it] 18%|█▊ | 96/520 [06:04<25:42, 3.64s/it] {'loss': 1.6117, 'grad_norm': 0.0006690253758658411, 'learning_rate': 0.8451997080166028, 'epoch': 0.18} + 18%|█▊ | 96/520 [06:04<25:42, 3.64s/it] 19%|█▊ | 97/520 [06:08<25:37, 3.64s/it] {'loss': 1.6013, 'grad_norm': 0.0009310490865431162, 'learning_rate': 0.8438505398589392, 'epoch': 0.19} + 19%|█▊ | 97/520 [06:08<25:37, 3.64s/it] 19%|█▉ | 98/520 [06:11<25:31, 3.63s/it] {'loss': 1.5951, 'grad_norm': 0.0009031052147119113, 'learning_rate': 0.8424860689784724, 'epoch': 0.19} + 19%|█▉ | 98/520 [06:11<25:31, 3.63s/it] 19%|█▉ | 99/520 [06:15<25:24, 3.62s/it] {'loss': 1.5999, 'grad_norm': 0.0007415268245646513, 'learning_rate': 0.8411063483905409, 'epoch': 0.19} + 19%|█▉ | 99/520 [06:15<25:24, 3.62s/it] 19%|█▉ | 100/520 [06:19<25:21, 3.62s/it] {'loss': 1.8417, 'grad_norm': 0.0010214472019761218, 'learning_rate': 0.8397114317029974, 'epoch': 0.19} + 19%|█▉ | 100/520 [06:19<25:21, 3.62s/it] 19%|█▉ | 101/520 [06:22<25:18, 3.62s/it] {'loss': 1.6075, 'grad_norm': 0.0009218120547283883, 'learning_rate': 0.8383013731141259, 'epoch': 0.19} + 19%|█▉ | 101/520 [06:22<25:18, 3.62s/it] 20%|█▉ | 102/520 [06:26<25:15, 3.63s/it] {'loss': 1.6088, 'grad_norm': 0.001140256221307287, 'learning_rate': 0.8368762274105357, 'epoch': 0.2} + 20%|█▉ | 102/520 [06:26<25:15, 3.63s/it] 20%|█▉ | 103/520 [06:30<25:11, 3.62s/it] {'loss': 1.5521, 'grad_norm': 0.0007675423679799907, 'learning_rate': 0.8354360499650332, 'epoch': 0.2} + 20%|█▉ | 103/520 [06:30<25:11, 3.62s/it] 20%|██ | 104/520 [06:33<25:10, 3.63s/it] {'loss': 1.6333, 'grad_norm': 0.0008793227100664041, 'learning_rate': 0.8339808967344701, 'epoch': 0.2} + 20%|██ | 104/520 [06:33<25:10, 3.63s/it] 20%|██ | 105/520 [06:37<25:10, 3.64s/it] {'loss': 1.6213, 'grad_norm': 0.0007131948395534324, 'learning_rate': 0.8325108242575692, 'epoch': 0.2} + 20%|██ | 105/520 [06:37<25:10, 3.64s/it] 20%|██ | 106/520 [06:41<25:07, 3.64s/it] {'loss': 1.8364, 'grad_norm': 0.0008293254760945663, 'learning_rate': 0.8310258896527278, 'epoch': 0.2} + 20%|██ | 106/520 [06:41<25:07, 3.64s/it] 21%|██ | 107/520 [06:44<25:08, 3.65s/it] {'loss': 1.8242, 'grad_norm': 0.0009274595983288102, 'learning_rate': 0.8295261506157986, 'epoch': 0.21} + 21%|██ | 107/520 [06:44<25:08, 3.65s/it] 21%|██ | 108/520 [06:48<25:05, 3.65s/it] {'loss': 1.562, 'grad_norm': 0.0007771494869275998, 'learning_rate': 0.8280116654178472, 'epoch': 0.21} + 21%|██ | 108/520 [06:48<25:05, 3.65s/it] 21%|██ | 109/520 [06:52<24:57, 3.64s/it] {'loss': 1.8241, 'grad_norm': 0.00074628424133812, 'learning_rate': 0.8264824929028889, 'epoch': 0.21} + 21%|██ | 109/520 [06:52<24:57, 3.64s/it] 21%|██ | 110/520 [06:55<24:50, 3.64s/it] {'loss': 1.7523, 'grad_norm': 0.0008341475387250509, 'learning_rate': 0.824938692485602, 'epoch': 0.21} + 21%|██ | 110/520 [06:55<24:50, 3.64s/it] 21%|██▏ | 111/520 [06:59<24:47, 3.64s/it] {'loss': 1.7761, 'grad_norm': 0.0011008792869427817, 'learning_rate': 0.8233803241490194, 'epoch': 0.21} + 21%|██▏ | 111/520 [06:59<24:47, 3.64s/it] 22%|██▏ | 112/520 [07:02<24:43, 3.63s/it] {'loss': 1.6678, 'grad_norm': 0.0008917972757539291, 'learning_rate': 0.8218074484421978, 'epoch': 0.22} + 22%|██▏ | 112/520 [07:02<24:43, 3.63s/it] 22%|██▏ | 113/520 [07:06<24:36, 3.63s/it] {'loss': 1.4987, 'grad_norm': 0.0008807408529995577, 'learning_rate': 0.820220126477865, 'epoch': 0.22} + 22%|██▏ | 113/520 [07:06<24:36, 3.63s/it] 22%|██▏ | 114/520 [07:10<24:36, 3.64s/it] {'loss': 1.6275, 'grad_norm': 0.0008745938132972718, 'learning_rate': 0.8186184199300464, 'epoch': 0.22} + 22%|██▏ | 114/520 [07:10<24:36, 3.64s/it] 22%|██▏ | 115/520 [07:13<24:33, 3.64s/it] {'loss': 1.7582, 'grad_norm': 0.0008033437642061044, 'learning_rate': 0.817002391031667, 'epoch': 0.22} + 22%|██▏ | 115/520 [07:13<24:33, 3.64s/it] 22%|██▏ | 116/520 [07:17<24:30, 3.64s/it] {'loss': 1.7268, 'grad_norm': 0.0007283224778531482, 'learning_rate': 0.8153721025721355, 'epoch': 0.22} + 22%|██▏ | 116/520 [07:17<24:30, 3.64s/it] 22%|██▎ | 117/520 [07:21<24:25, 3.64s/it] {'loss': 1.7146, 'grad_norm': 0.0008556202306291427, 'learning_rate': 0.8137276178949024, 'epoch': 0.23} + 22%|██▎ | 117/520 [07:21<24:25, 3.64s/it] 23%|██▎ | 118/520 [07:24<24:23, 3.64s/it] {'loss': 1.5807, 'grad_norm': 0.0006479043898626938, 'learning_rate': 0.8120690008950008, 'epoch': 0.23} + 23%|██▎ | 118/520 [07:24<24:23, 3.64s/it] 23%|██▎ | 119/520 [07:28<24:22, 3.65s/it] {'loss': 1.5189, 'grad_norm': 0.0007841772547126423, 'learning_rate': 0.8103963160165627, 'epoch': 0.23} + 23%|██▎ | 119/520 [07:28<24:22, 3.65s/it] 23%|██▎ | 120/520 [07:32<24:21, 3.65s/it] {'loss': 1.5668, 'grad_norm': 0.0009080578356813871, 'learning_rate': 0.8087096282503151, 'epoch': 0.23} + 23%|██▎ | 120/520 [07:32<24:21, 3.65s/it] 23%|██▎ | 121/520 [07:35<24:15, 3.65s/it] {'loss': 1.6096, 'grad_norm': 0.0008093463034572253, 'learning_rate': 0.8070090031310558, 'epoch': 0.23} + 23%|██▎ | 121/520 [07:35<24:15, 3.65s/it] 23%|██▎ | 122/520 [07:39<24:11, 3.65s/it] {'loss': 1.4807, 'grad_norm': 0.000699057835626335, 'learning_rate': 0.8052945067351059, 'epoch': 0.23} + 23%|██▎ | 122/520 [07:39<24:11, 3.65s/it] 24%|██▎ | 123/520 [07:43<24:15, 3.67s/it] {'loss': 1.8926, 'grad_norm': 0.0008373524204046949, 'learning_rate': 0.8035662056777432, 'epoch': 0.24} + 24%|██▎ | 123/520 [07:43<24:15, 3.67s/it] 24%|██▍ | 124/520 [07:46<24:22, 3.69s/it] {'loss': 1.6004, 'grad_norm': 0.0006924049772028669, 'learning_rate': 0.8018241671106134, 'epoch': 0.24} + 24%|██▍ | 124/520 [07:46<24:22, 3.69s/it] 24%|██▍ | 125/520 [07:50<24:14, 3.68s/it] {'loss': 1.5624, 'grad_norm': 0.0008484726505127982, 'learning_rate': 0.8000684587191217, 'epoch': 0.24} + 24%|██▍ | 125/520 [07:50<24:14, 3.68s/it] 24%|██▍ | 126/520 [07:54<25:26, 3.88s/it] {'loss': 1.7477, 'grad_norm': 0.0008121281595216517, 'learning_rate': 0.7982991487198023, 'epoch': 0.24} + 24%|██▍ | 126/520 [07:54<25:26, 3.88s/it] 24%|██▍ | 127/520 [07:58<25:05, 3.83s/it] {'loss': 1.5523, 'grad_norm': 0.0006723027824231054, 'learning_rate': 0.7965163058576683, 'epoch': 0.24} + 24%|██▍ | 127/520 [07:58<25:05, 3.83s/it] 25%|██▍ | 128/520 [08:02<24:54, 3.81s/it] {'loss': 1.6166, 'grad_norm': 0.0006691425110854235, 'learning_rate': 0.7947199994035401, 'epoch': 0.25} + 25%|██▍ | 128/520 [08:02<24:54, 3.81s/it] 25%|██▍ | 129/520 [08:05<24:35, 3.77s/it] {'loss': 1.5226, 'grad_norm': 0.000549847051647941, 'learning_rate': 0.7929102991513549, 'epoch': 0.25} + 25%|██▍ | 129/520 [08:05<24:35, 3.77s/it] 25%|██▌ | 130/520 [08:09<24:32, 3.78s/it] {'loss': 1.5984, 'grad_norm': 0.0006215857522558675, 'learning_rate': 0.7910872754154539, 'epoch': 0.25} + 25%|██▌ | 130/520 [08:09<24:32, 3.78s/it] 25%|██▌ | 131/520 [08:13<24:20, 3.76s/it] {'loss': 1.7489, 'grad_norm': 0.000776407809067261, 'learning_rate': 0.7892509990278509, 'epoch': 0.25} + 25%|██▌ | 131/520 [08:13<24:20, 3.76s/it] 25%|██▌ | 132/520 [08:17<24:20, 3.77s/it] {'loss': 1.6546, 'grad_norm': 0.0007410813109296242, 'learning_rate': 0.7874015413354805, 'epoch': 0.25} + 25%|██▌ | 132/520 [08:17<24:20, 3.77s/it] 26%|██▌ | 133/520 [08:20<24:00, 3.72s/it] {'loss': 1.5619, 'grad_norm': 0.0008033547897737006, 'learning_rate': 0.7855389741974245, 'epoch': 0.26} + 26%|██▌ | 133/520 [08:20<24:00, 3.72s/it] 26%|██▌ | 134/520 [08:24<23:52, 3.71s/it] {'loss': 1.6389, 'grad_norm': 0.0007162286209621575, 'learning_rate': 0.783663369982122, 'epoch': 0.26} + 26%|██▌ | 134/520 [08:24<23:52, 3.71s/it] 26%|██▌ | 135/520 [08:28<23:52, 3.72s/it] {'loss': 1.7113, 'grad_norm': 0.0010615554741252948, 'learning_rate': 0.7817748015645558, 'epoch': 0.26} + 26%|██▌ | 135/520 [08:28<23:52, 3.72s/it] 26%|██▌ | 136/520 [08:32<23:52, 3.73s/it] {'loss': 1.6349, 'grad_norm': 0.0008513737168117877, 'learning_rate': 0.7798733423234219, 'epoch': 0.26} + 26%|██▌ | 136/520 [08:32<23:52, 3.73s/it] 26%|██▋ | 137/520 [08:35<23:39, 3.71s/it] {'loss': 1.5485, 'grad_norm': 0.0008575557795754851, 'learning_rate': 0.7779590661382778, 'epoch': 0.26} + 26%|██▋ | 137/520 [08:35<23:39, 3.71s/it] 27%|██▋ | 138/520 [08:39<23:30, 3.69s/it] {'loss': 1.5486, 'grad_norm': 0.0008909293764369443, 'learning_rate': 0.7760320473866728, 'epoch': 0.27} + 27%|██▋ | 138/520 [08:39<23:30, 3.69s/it] 27%|██▋ | 139/520 [08:43<23:25, 3.69s/it] {'loss': 1.6394, 'grad_norm': 0.0008738552917529709, 'learning_rate': 0.774092360941257, 'epoch': 0.27} + 27%|██▋ | 139/520 [08:43<23:25, 3.69s/it] 27%|██▋ | 140/520 [08:46<23:17, 3.68s/it] {'loss': 1.7833, 'grad_norm': 0.0006363820585917117, 'learning_rate': 0.7721400821668734, 'epoch': 0.27} + 27%|██▋ | 140/520 [08:46<23:17, 3.68s/it] 27%|██▋ | 141/520 [08:50<23:16, 3.68s/it] {'loss': 1.6648, 'grad_norm': 0.0006385892202520176, 'learning_rate': 0.7701752869176286, 'epoch': 0.27} + 27%|██▋ | 141/520 [08:50<23:16, 3.68s/it] 27%|██▋ | 142/520 [08:54<23:22, 3.71s/it] {'loss': 1.8158, 'grad_norm': 0.0007188996148125342, 'learning_rate': 0.7681980515339464, 'epoch': 0.27} + 27%|██▋ | 142/520 [08:54<23:22, 3.71s/it] 28%|██▊ | 143/520 [08:57<23:11, 3.69s/it] {'loss': 1.606, 'grad_norm': 0.0007957449592263231, 'learning_rate': 0.7662084528396012, 'epoch': 0.28} + 28%|██▊ | 143/520 [08:57<23:11, 3.69s/it] 28%|██▊ | 144/520 [09:01<23:01, 3.67s/it] {'loss': 1.5346, 'grad_norm': 0.0008055934011340957, 'learning_rate': 0.7642065681387328, 'epoch': 0.28} + 28%|██▊ | 144/520 [09:01<23:01, 3.67s/it] 28%|██▊ | 145/520 [09:05<23:00, 3.68s/it] {'loss': 1.4539, 'grad_norm': 0.0006867991320579556, 'learning_rate': 0.7621924752128437, 'epoch': 0.28} + 28%|██▊ | 145/520 [09:05<23:00, 3.68s/it] 28%|██▊ | 146/520 [09:08<22:56, 3.68s/it] {'loss': 1.8765, 'grad_norm': 0.0007063393266862569, 'learning_rate': 0.7601662523177761, 'epoch': 0.28} + 28%|██▊ | 146/520 [09:08<22:56, 3.68s/it] 28%|██▊ | 147/520 [09:12<22:54, 3.68s/it] {'loss': 1.5016, 'grad_norm': 0.000815388053079538, 'learning_rate': 0.7581279781806721, 'epoch': 0.28} + 28%|██▊ | 147/520 [09:12<22:54, 3.68s/it] 28%|██▊ | 148/520 [09:16<22:46, 3.67s/it] {'loss': 1.554, 'grad_norm': 0.0007511802702056903, 'learning_rate': 0.7560777319969136, 'epoch': 0.28} + 28%|██▊ | 148/520 [09:16<22:46, 3.67s/it] 29%|██▊ | 149/520 [09:19<22:50, 3.69s/it] {'loss': 1.5051, 'grad_norm': 0.0006577039475622353, 'learning_rate': 0.7540155934270472, 'epoch': 0.29} + 29%|██▊ | 149/520 [09:19<22:50, 3.69s/it] 29%|██▉ | 150/520 [09:23<23:05, 3.74s/it] {'loss': 1.7283, 'grad_norm': 0.0006859864182405612, 'learning_rate': 0.7519416425936866, 'epoch': 0.29} + 29%|██▉ | 150/520 [09:23<23:05, 3.74s/it] 29%|██▉ | 151/520 [09:27<22:54, 3.72s/it] {'loss': 1.5274, 'grad_norm': 0.0007096317942789763, 'learning_rate': 0.7498559600784018, 'epoch': 0.29} + 29%|██▉ | 151/520 [09:27<22:54, 3.72s/it] 29%|██▉ | 152/520 [09:31<22:44, 3.71s/it] {'loss': 1.495, 'grad_norm': 0.0007145937728330766, 'learning_rate': 0.7477586269185867, 'epoch': 0.29} + 29%|██▉ | 152/520 [09:31<22:44, 3.71s/it] 29%|██▉ | 153/520 [09:34<22:37, 3.70s/it] {'loss': 1.5414, 'grad_norm': 0.0006549530815006231, 'learning_rate': 0.7456497246043112, 'epoch': 0.29} + 29%|██▉ | 153/520 [09:34<22:37, 3.70s/it] 30%|██▉ | 154/520 [09:38<22:34, 3.70s/it] {'loss': 1.6371, 'grad_norm': 0.0008619793168661297, 'learning_rate': 0.7435293350751545, 'epoch': 0.3} + 30%|██▉ | 154/520 [09:38<22:34, 3.70s/it] 30%|██▉ | 155/520 [09:42<22:24, 3.68s/it] {'loss': 1.522, 'grad_norm': 0.0008373535741311937, 'learning_rate': 0.7413975407170217, 'epoch': 0.3} + 30%|██▉ | 155/520 [09:42<22:24, 3.68s/it] 30%|███ | 156/520 [09:45<22:18, 3.68s/it] {'loss': 1.5909, 'grad_norm': 0.0008086778691720074, 'learning_rate': 0.7392544243589427, 'epoch': 0.3} + 30%|███ | 156/520 [09:45<22:18, 3.68s/it] 30%|███ | 157/520 [09:49<22:16, 3.68s/it] {'loss': 1.9012, 'grad_norm': 0.0008664116944304308, 'learning_rate': 0.7371000692698539, 'epoch': 0.3} + 30%|███ | 157/520 [09:49<22:16, 3.68s/it] 30%|███ | 158/520 [09:53<22:07, 3.67s/it] {'loss': 1.5261, 'grad_norm': 0.0006044668452445373, 'learning_rate': 0.734934559155363, 'epoch': 0.3} + 30%|███ | 158/520 [09:53<22:07, 3.67s/it] 31%|███ | 159/520 [09:56<21:59, 3.66s/it] {'loss': 1.5557, 'grad_norm': 0.0005483346977437766, 'learning_rate': 0.7327579781544963, 'epoch': 0.31} + 31%|███ | 159/520 [09:56<21:59, 3.66s/it] 31%|███ | 160/520 [10:00<22:14, 3.71s/it] {'loss': 1.598, 'grad_norm': 0.0008230451222006694, 'learning_rate': 0.7305704108364302, 'epoch': 0.31} + 31%|███ | 160/520 [10:00<22:14, 3.71s/it] 31%|███ | 161/520 [10:04<22:27, 3.75s/it] {'loss': 1.5868, 'grad_norm': 0.0006089171447497124, 'learning_rate': 0.7283719421972047, 'epoch': 0.31} + 31%|███ | 161/520 [10:04<22:27, 3.75s/it] 31%|███ | 162/520 [10:08<22:38, 3.79s/it] {'loss': 1.7734, 'grad_norm': 0.0008369865186139046, 'learning_rate': 0.7261626576564214, 'epoch': 0.31} + 31%|███ | 162/520 [10:08<22:38, 3.79s/it] 31%|███▏ | 163/520 [10:12<22:39, 3.81s/it] {'loss': 1.4763, 'grad_norm': 0.0007044482644304686, 'learning_rate': 0.7239426430539243, 'epoch': 0.31} + 31%|███▏ | 163/520 [10:12<22:39, 3.81s/it] 32%|███▏ | 164/520 [10:16<22:40, 3.82s/it] {'loss': 1.4124, 'grad_norm': 0.0006297687594235789, 'learning_rate': 0.7217119846464648, 'epoch': 0.32} + 32%|███▏ | 164/520 [10:16<22:40, 3.82s/it] 32%|███▏ | 165/520 [10:19<22:40, 3.83s/it] {'loss': 1.5533, 'grad_norm': 0.0007917000671387158, 'learning_rate': 0.7194707691043501, 'epoch': 0.32} + 32%|███▏ | 165/520 [10:19<22:40, 3.83s/it] 32%|███▏ | 166/520 [10:23<22:34, 3.83s/it] {'loss': 1.5485, 'grad_norm': 0.0010074072611269283, 'learning_rate': 0.7172190835080757, 'epoch': 0.32} + 32%|███▏ | 166/520 [10:23<22:34, 3.83s/it] 32%|███▏ | 167/520 [10:27<22:37, 3.84s/it] {'loss': 1.5411, 'grad_norm': 0.0008860505177161304, 'learning_rate': 0.7149570153449422, 'epoch': 0.32} + 32%|███▏ | 167/520 [10:27<22:37, 3.84s/it] 32%|███▏ | 168/520 [10:31<22:32, 3.84s/it] {'loss': 1.4937, 'grad_norm': 0.0005595012919088888, 'learning_rate': 0.7126846525056555, 'epoch': 0.32} + 32%|███▏ | 168/520 [10:31<22:32, 3.84s/it] 32%|███▎ | 169/520 [10:35<22:24, 3.83s/it] {'loss': 1.5556, 'grad_norm': 0.0008395403699179325, 'learning_rate': 0.7104020832809127, 'epoch': 0.33} + 32%|███▎ | 169/520 [10:35<22:24, 3.83s/it] 33%|███▎ | 170/520 [10:39<22:30, 3.86s/it] {'loss': 1.6789, 'grad_norm': 0.0007985778020966015, 'learning_rate': 0.7081093963579708, 'epoch': 0.33} + 33%|███▎ | 170/520 [10:39<22:30, 3.86s/it] 33%|███▎ | 171/520 [10:42<22:19, 3.84s/it] {'loss': 1.4844, 'grad_norm': 0.0006786773254929278, 'learning_rate': 0.7058066808172017, 'epoch': 0.33} + 33%|███▎ | 171/520 [10:42<22:19, 3.84s/it] 33%|███▎ | 172/520 [10:46<22:03, 3.80s/it] {'loss': 1.5759, 'grad_norm': 0.0007374646383319259, 'learning_rate': 0.7034940261286299, 'epoch': 0.33} + 33%|███▎ | 172/520 [10:46<22:03, 3.80s/it] 33%|███▎ | 173/520 [10:50<21:54, 3.79s/it] {'loss': 1.4859, 'grad_norm': 0.0006618814102536802, 'learning_rate': 0.7011715221484579, 'epoch': 0.33} + 33%|███▎ | 173/520 [10:50<21:54, 3.79s/it] 33%|███▎ | 174/520 [10:54<21:48, 3.78s/it] {'loss': 1.581, 'grad_norm': 0.000581926434140775, 'learning_rate': 0.6988392591155728, 'epoch': 0.33} + 33%|███▎ | 174/520 [10:54<21:48, 3.78s/it] 34%|███▎ | 175/520 [10:57<21:40, 3.77s/it] {'loss': 1.4735, 'grad_norm': 0.0005424819515362076, 'learning_rate': 0.6964973276480422, 'epoch': 0.34} + 34%|███▎ | 175/520 [10:57<21:40, 3.77s/it] 34%|███▍ | 176/520 [11:01<21:26, 3.74s/it] {'loss': 1.7919, 'grad_norm': 0.0008248530365110395, 'learning_rate': 0.6941458187395918, 'epoch': 0.34} + 34%|███▍ | 176/520 [11:01<21:26, 3.74s/it] 34%|███▍ | 177/520 [11:05<21:17, 3.72s/it] {'loss': 1.6693, 'grad_norm': 0.0007748275263126676, 'learning_rate': 0.6917848237560709, 'epoch': 0.34} + 34%|███▍ | 177/520 [11:05<21:17, 3.72s/it] 34%|███▍ | 178/520 [11:08<21:05, 3.70s/it] {'loss': 1.5385, 'grad_norm': 0.0009322477763705424, 'learning_rate': 0.6894144344319014, 'epoch': 0.34} + 34%|███▍ | 178/520 [11:08<21:05, 3.70s/it] 34%|███▍ | 179/520 [11:12<20:54, 3.68s/it] {'loss': 1.6464, 'grad_norm': 0.0008031387239381223, 'learning_rate': 0.6870347428665153, 'epoch': 0.34} + 34%|███▍ | 179/520 [11:12<20:54, 3.68s/it] 35%|███▍ | 180/520 [11:16<20:58, 3.70s/it] {'loss': 1.5438, 'grad_norm': 0.000764416044386618, 'learning_rate': 0.6846458415207741, 'epoch': 0.35} + 35%|███▍ | 180/520 [11:16<20:58, 3.70s/it] 35%|███▍ | 181/520 [11:20<21:11, 3.75s/it] {'loss': 1.5188, 'grad_norm': 0.0006381010697227001, 'learning_rate': 0.682247823213378, 'epoch': 0.35} + 35%|███▍ | 181/520 [11:20<21:11, 3.75s/it] 35%|███▌ | 182/520 [11:24<21:19, 3.79s/it] {'loss': 1.5623, 'grad_norm': 0.0008467282975867742, 'learning_rate': 0.6798407811172587, 'epoch': 0.35} + 35%|███▌ | 182/520 [11:24<21:19, 3.79s/it] 35%|███▌ | 183/520 [11:27<21:24, 3.81s/it] {'loss': 1.5617, 'grad_norm': 0.0007420790183556593, 'learning_rate': 0.6774248087559589, 'epoch': 0.35} + 35%|███▌ | 183/520 [11:27<21:24, 3.81s/it] 35%|███▌ | 184/520 [11:31<21:26, 3.83s/it] {'loss': 1.4764, 'grad_norm': 0.0007911229252822237, 'learning_rate': 0.675, 'epoch': 0.35} + 35%|███▌ | 184/520 [11:31<21:26, 3.83s/it] 36%|███▌ | 185/520 [11:35<21:23, 3.83s/it] {'loss': 1.6557, 'grad_norm': 0.0005958810541026588, 'learning_rate': 0.6725664490632334, 'epoch': 0.36} + 36%|███▌ | 185/520 [11:35<21:23, 3.83s/it] 36%|███▌ | 186/520 [11:39<21:22, 3.84s/it] {'loss': 1.4929, 'grad_norm': 0.000918962150673565, 'learning_rate': 0.6701242504991802, 'epoch': 0.36} + 36%|███▌ | 186/520 [11:39<21:22, 3.84s/it] 36%|███▌ | 187/520 [11:43<21:18, 3.84s/it] {'loss': 1.5035, 'grad_norm': 0.0009114137631223443, 'learning_rate': 0.667673499197358, 'epoch': 0.36} + 36%|███▌ | 187/520 [11:43<21:18, 3.84s/it] 36%|███▌ | 188/520 [11:47<21:07, 3.82s/it] {'loss': 1.5756, 'grad_norm': 0.0007271414908245055, 'learning_rate': 0.6652142903795932, 'epoch': 0.36} + 36%|███▌ | 188/520 [11:47<21:07, 3.82s/it] 36%|███▋ | 189/520 [11:50<21:04, 3.82s/it] {'loss': 1.6112, 'grad_norm': 0.000653753558792773, 'learning_rate': 0.6627467195963223, 'epoch': 0.36} + 36%|███▋ | 189/520 [11:50<21:04, 3.82s/it] 37%|███▋ | 190/520 [11:54<20:57, 3.81s/it] {'loss': 1.4872, 'grad_norm': 0.0008132832210742688, 'learning_rate': 0.6602708827228779, 'epoch': 0.37} + 37%|███▋ | 190/520 [11:54<20:57, 3.81s/it] 37%|███▋ | 191/520 [11:58<20:56, 3.82s/it] {'loss': 1.463, 'grad_norm': 0.0008496623113880274, 'learning_rate': 0.6577868759557653, 'epoch': 0.37} + 37%|███▋ | 191/520 [11:58<20:56, 3.82s/it] 37%|███▋ | 192/520 [12:02<20:52, 3.82s/it] {'loss': 1.5612, 'grad_norm': 0.0005840155005597547, 'learning_rate': 0.6552947958089234, 'epoch': 0.37} + 37%|███▋ | 192/520 [12:02<20:52, 3.82s/it] 37%|███▋ | 193/520 [12:06<20:50, 3.82s/it] {'loss': 1.7313, 'grad_norm': 0.0012113024731043895, 'learning_rate': 0.6527947391099754, 'epoch': 0.37} + 37%|███▋ | 193/520 [12:06<20:50, 3.82s/it] 37%|███▋ | 194/520 [12:10<20:46, 3.82s/it] {'loss': 1.6072, 'grad_norm': 0.0005775475603255291, 'learning_rate': 0.6502868029964665, 'epoch': 0.37} + 37%|███▋ | 194/520 [12:10<20:46, 3.82s/it] 38%|███▊ | 195/520 [12:13<20:43, 3.83s/it] {'loss': 1.5467, 'grad_norm': 0.0008389246478201119, 'learning_rate': 0.6477710849120902, 'epoch': 0.38} + 38%|███▊ | 195/520 [12:13<20:43, 3.83s/it] 38%|███▊ | 196/520 [12:17<20:34, 3.81s/it] {'loss': 1.5496, 'grad_norm': 0.0007221227017550442, 'learning_rate': 0.6452476826029011, 'epoch': 0.38} + 38%|███▊ | 196/520 [12:17<20:34, 3.81s/it] 38%|███▊ | 197/520 [12:21<20:18, 3.77s/it] {'loss': 1.4987, 'grad_norm': 0.0008933553098380504, 'learning_rate': 0.6427166941135182, 'epoch': 0.38} + 38%|███▊ | 197/520 [12:21<20:18, 3.77s/it] 38%|███▊ | 198/520 [12:24<20:05, 3.74s/it] {'loss': 1.6005, 'grad_norm': 0.0007201243699090968, 'learning_rate': 0.6401782177833147, 'epoch': 0.38} + 38%|███▊ | 198/520 [12:24<20:05, 3.74s/it] 38%|███▊ | 199/520 [12:28<20:11, 3.77s/it] {'loss': 1.4918, 'grad_norm': 0.0005790893633833629, 'learning_rate': 0.6376323522425977, 'epoch': 0.38} + 38%|███▊ | 199/520 [12:28<20:11, 3.77s/it] 38%|███▊ | 200/520 [12:32<19:56, 3.74s/it] {'loss': 1.6528, 'grad_norm': 0.0006255260339692199, 'learning_rate': 0.6350791964087753, 'epoch': 0.38} + 38%|███▊ | 200/520 [12:32<19:56, 3.74s/it] 39%|███▊ | 201/520 [12:36<19:48, 3.72s/it] {'loss': 1.6636, 'grad_norm': 0.0006312748016565646, 'learning_rate': 0.6325188494825138, 'epoch': 0.39} + 39%|███▊ | 201/520 [12:36<19:48, 3.72s/it] 39%|███▉ | 202/520 [12:39<19:39, 3.71s/it] {'loss': 1.4963, 'grad_norm': 0.0006100071476820832, 'learning_rate': 0.6299514109438834, 'epoch': 0.39} + 39%|███▉ | 202/520 [12:39<19:39, 3.71s/it] 39%|███▉ | 203/520 [12:43<19:28, 3.69s/it] {'loss': 1.5326, 'grad_norm': 0.0010793883288169736, 'learning_rate': 0.6273769805484928, 'epoch': 0.39} + 39%|███▉ | 203/520 [12:43<19:28, 3.69s/it] 39%|███▉ | 204/520 [12:47<19:39, 3.73s/it] {'loss': 1.5912, 'grad_norm': 0.0007542238730074619, 'learning_rate': 0.6247956583236127, 'epoch': 0.39} + 39%|███▉ | 204/520 [12:47<19:39, 3.73s/it] 39%|███▉ | 205/520 [12:50<19:31, 3.72s/it] {'loss': 1.6915, 'grad_norm': 0.0007223588612249585, 'learning_rate': 0.6222075445642904, 'epoch': 0.39} + 39%|███▉ | 205/520 [12:51<19:31, 3.72s/it] 40%|███▉ | 206/520 [12:54<19:23, 3.71s/it] {'loss': 1.6159, 'grad_norm': 0.0006829722284889455, 'learning_rate': 0.6196127398294524, 'epoch': 0.4} + 40%|███▉ | 206/520 [12:54<19:23, 3.71s/it] 40%|███▉ | 207/520 [12:58<19:17, 3.70s/it] {'loss': 1.6756, 'grad_norm': 0.000740297416659731, 'learning_rate': 0.617011344937997, 'epoch': 0.4} + 40%|███▉ | 207/520 [12:58<19:17, 3.70s/it] 40%|████ | 208/520 [13:02<19:10, 3.69s/it] {'loss': 1.559, 'grad_norm': 0.0006735843854135178, 'learning_rate': 0.6144034609648779, 'epoch': 0.4} + 40%|████ | 208/520 [13:02<19:10, 3.69s/it] 40%|████ | 209/520 [13:05<19:07, 3.69s/it] {'loss': 1.5068, 'grad_norm': 0.0008546755428660746, 'learning_rate': 0.6117891892371754, 'epoch': 0.4} + 40%|████ | 209/520 [13:05<19:07, 3.69s/it] 40%|████ | 210/520 [13:09<19:01, 3.68s/it] {'loss': 1.5591, 'grad_norm': 0.0010707369563025845, 'learning_rate': 0.6091686313301616, 'epoch': 0.4} + 40%|████ | 210/520 [13:09<19:01, 3.68s/it] 41%|████ | 211/520 [13:13<19:04, 3.70s/it] {'loss': 1.5688, 'grad_norm': 0.0004954528326903046, 'learning_rate': 0.6065418890633522, 'epoch': 0.41} + 41%|████ | 211/520 [13:13<19:04, 3.70s/it] 41%|████ | 212/520 [13:16<19:13, 3.75s/it] {'loss': 1.5658, 'grad_norm': 0.0006284675181244947, 'learning_rate': 0.603909064496551, 'epoch': 0.41} + 41%|████ | 212/520 [13:16<19:13, 3.75s/it] 41%|████ | 213/520 [13:20<19:25, 3.80s/it] {'loss': 1.5192, 'grad_norm': 0.0010801617768292336, 'learning_rate': 0.6012702599258839, 'epoch': 0.41} + 41%|████ | 213/520 [13:20<19:25, 3.80s/it] 41%|████ | 214/520 [13:24<19:26, 3.81s/it] {'loss': 1.5534, 'grad_norm': 0.0007524112444118947, 'learning_rate': 0.5986255778798253, 'epoch': 0.41} + 41%|████ | 214/520 [13:24<19:26, 3.81s/it] 41%|████▏ | 215/520 [13:28<19:30, 3.84s/it] {'loss': 1.5968, 'grad_norm': 0.0006230709328410912, 'learning_rate': 0.5959751211152132, 'epoch': 0.41} + 41%|████▏ | 215/520 [13:28<19:30, 3.84s/it] 42%|████▏ | 216/520 [13:32<19:33, 3.86s/it] {'loss': 1.4155, 'grad_norm': 0.00113820890703073, 'learning_rate': 0.593318992613258, 'epoch': 0.42} + 42%|████▏ | 216/520 [13:32<19:33, 3.86s/it] 42%|████▏ | 217/520 [13:36<19:32, 3.87s/it] {'loss': 1.5629, 'grad_norm': 0.0008887036587733264, 'learning_rate': 0.59065729557554, 'epoch': 0.42} + 42%|████▏ | 217/520 [13:36<19:32, 3.87s/it] 42%|████▏ | 218/520 [13:40<19:15, 3.83s/it] {'loss': 1.5595, 'grad_norm': 0.0005649765342398792, 'learning_rate': 0.5879901334200005, 'epoch': 0.42} + 42%|████▏ | 218/520 [13:40<19:15, 3.83s/it] 42%|████▏ | 219/520 [13:43<18:54, 3.77s/it] {'loss': 1.5122, 'grad_norm': 0.0007539084898295857, 'learning_rate': 0.5853176097769228, 'epoch': 0.42} + 42%|████▏ | 219/520 [13:43<18:54, 3.77s/it] 42%|████▏ | 220/520 [13:47<18:38, 3.73s/it] {'loss': 1.651, 'grad_norm': 0.0012402100398893826, 'learning_rate': 0.5826398284849069, 'epoch': 0.42} + 42%|████▏ | 220/520 [13:47<18:38, 3.73s/it] 42%|████▎ | 221/520 [13:51<18:28, 3.71s/it] {'loss': 1.5489, 'grad_norm': 0.0006697275444826561, 'learning_rate': 0.5799568935868334, 'epoch': 0.42} + 42%|████▎ | 221/520 [13:51<18:28, 3.71s/it] 43%|████▎ | 222/520 [13:54<18:20, 3.69s/it] {'loss': 1.4293, 'grad_norm': 0.0006087031322753994, 'learning_rate': 0.5772689093258224, 'epoch': 0.43} + 43%|████▎ | 222/520 [13:54<18:20, 3.69s/it] 43%|████▎ | 223/520 [13:58<18:17, 3.69s/it] {'loss': 1.4521, 'grad_norm': 0.0005450734138932322, 'learning_rate': 0.5745759801411822, 'epoch': 0.43} + 43%|████▎ | 223/520 [13:58<18:17, 3.69s/it] 43%|████▎ | 224/520 [14:02<18:13, 3.69s/it] {'loss': 1.9869, 'grad_norm': 0.0012637218524597662, 'learning_rate': 0.5718782106643524, 'epoch': 0.43} + 43%|████▎ | 224/520 [14:02<18:13, 3.69s/it] 43%|████▎ | 225/520 [14:05<18:10, 3.70s/it] {'loss': 1.456, 'grad_norm': 0.0006298442812839934, 'learning_rate': 0.5691757057148372, 'epoch': 0.43} + 43%|████▎ | 225/520 [14:05<18:10, 3.70s/it] 43%|████▎ | 226/520 [14:09<18:11, 3.71s/it] {'loss': 1.569, 'grad_norm': 0.0005994694234779792, 'learning_rate': 0.5664685702961344, 'epoch': 0.43} + 43%|████▎ | 226/520 [14:09<18:11, 3.71s/it] 44%|████▎ | 227/520 [14:13<18:06, 3.71s/it] {'loss': 1.5498, 'grad_norm': 0.0006221824824515306, 'learning_rate': 0.5637569095916538, 'epoch': 0.44} + 44%|████▎ | 227/520 [14:13<18:06, 3.71s/it] 44%|████▍ | 228/520 [14:17<18:10, 3.74s/it] {'loss': 1.8325, 'grad_norm': 0.0006575332488049972, 'learning_rate': 0.5610408289606321, 'epoch': 0.44} + 44%|████▍ | 228/520 [14:17<18:10, 3.74s/it] 44%|████▍ | 229/520 [14:20<18:08, 3.74s/it] {'loss': 1.5642, 'grad_norm': 0.0005087319490705094, 'learning_rate': 0.558320433934038, 'epoch': 0.44} + 44%|████▍ | 229/520 [14:20<18:08, 3.74s/it] 44%|████▍ | 230/520 [14:24<18:05, 3.74s/it] {'loss': 1.4011, 'grad_norm': 0.0005427527233183642, 'learning_rate': 0.5555958302104719, 'epoch': 0.44} + 44%|████▍ | 230/520 [14:24<18:05, 3.74s/it] 44%|████▍ | 231/520 [14:28<17:56, 3.72s/it] {'loss': 1.479, 'grad_norm': 0.000585087730261699, 'learning_rate': 0.5528671236520604, 'epoch': 0.44} + 44%|████▍ | 231/520 [14:28<17:56, 3.72s/it] 45%|████▍ | 232/520 [14:31<17:48, 3.71s/it] {'loss': 1.8608, 'grad_norm': 0.0007220104956152157, 'learning_rate': 0.5501344202803414, 'epoch': 0.45} + 45%|████▍ | 232/520 [14:31<17:48, 3.71s/it] 45%|████▍ | 233/520 [14:35<17:41, 3.70s/it] {'loss': 1.7281, 'grad_norm': 0.0007627642862337501, 'learning_rate': 0.5473978262721463, 'epoch': 0.45} + 45%|████▍ | 233/520 [14:35<17:41, 3.70s/it] 45%|████▌ | 234/520 [14:39<17:39, 3.70s/it] {'loss': 1.4506, 'grad_norm': 0.0009501625339783826, 'learning_rate': 0.5446574479554731, 'epoch': 0.45} + 45%|████▌ | 234/520 [14:39<17:39, 3.70s/it] 45%|████▌ | 235/520 [14:43<17:46, 3.74s/it] {'loss': 1.4833, 'grad_norm': 0.00052022020105066, 'learning_rate': 0.5419133918053562, 'epoch': 0.45} + 45%|████▌ | 235/520 [14:43<17:46, 3.74s/it] 45%|████▌ | 236/520 [14:46<17:43, 3.75s/it] {'loss': 1.6262, 'grad_norm': 0.0009358307988002196, 'learning_rate': 0.539165764439729, 'epoch': 0.45} + 45%|████▌ | 236/520 [14:46<17:43, 3.75s/it] 46%|████▌ | 237/520 [14:50<17:37, 3.74s/it] {'loss': 1.5563, 'grad_norm': 0.0009611458053729552, 'learning_rate': 0.5364146726152813, 'epoch': 0.46} + 46%|████▌ | 237/520 [14:50<17:37, 3.74s/it] 46%|████▌ | 238/520 [14:54<17:33, 3.73s/it] {'loss': 1.5105, 'grad_norm': 0.0006070653839387783, 'learning_rate': 0.5336602232233116, 'epoch': 0.46} + 46%|████▌ | 238/520 [14:54<17:33, 3.73s/it] 46%|████▌ | 239/520 [14:58<17:24, 3.72s/it] {'loss': 1.6401, 'grad_norm': 0.0006215097404557503, 'learning_rate': 0.5309025232855737, 'epoch': 0.46} + 46%|████▌ | 239/520 [14:58<17:24, 3.72s/it] 46%|████▌ | 240/520 [15:01<17:18, 3.71s/it] {'loss': 1.3517, 'grad_norm': 0.0007973488610947354, 'learning_rate': 0.5281416799501187, 'epoch': 0.46} + 46%|████▌ | 240/520 [15:01<17:18, 3.71s/it] 46%|████▋ | 241/520 [15:05<17:24, 3.74s/it] {'loss': 1.4378, 'grad_norm': 0.0006001060008757433, 'learning_rate': 0.5253778004871315, 'epoch': 0.46} + 46%|████▋ | 241/520 [15:05<17:24, 3.74s/it] 47%|████▋ | 242/520 [15:09<17:31, 3.78s/it] {'loss': 1.4853, 'grad_norm': 0.0005327875805651306, 'learning_rate': 0.522610992284763, 'epoch': 0.47} + 47%|████▋ | 242/520 [15:09<17:31, 3.78s/it] 47%|████▋ | 243/520 [15:13<17:33, 3.80s/it] {'loss': 1.4602, 'grad_norm': 0.0005452818087206564, 'learning_rate': 0.5198413628449582, 'epoch': 0.47} + 47%|████▋ | 243/520 [15:13<17:33, 3.80s/it] 47%|████▋ | 244/520 [15:17<17:34, 3.82s/it] {'loss': 1.6278, 'grad_norm': 0.0005380300090466224, 'learning_rate': 0.5170690197792784, 'epoch': 0.47} + 47%|████▋ | 244/520 [15:17<17:34, 3.82s/it] 47%|████▋ | 245/520 [15:20<17:20, 3.78s/it] {'loss': 1.4763, 'grad_norm': 0.000658739508066186, 'learning_rate': 0.514294070804721, 'epoch': 0.47} + 47%|████▋ | 245/520 [15:20<17:20, 3.78s/it] 47%|████▋ | 246/520 [15:24<17:08, 3.75s/it] {'loss': 1.8235, 'grad_norm': 0.0006408000623001473, 'learning_rate': 0.5115166237395331, 'epoch': 0.47} + 47%|████▋ | 246/520 [15:24<17:08, 3.75s/it] 48%|████▊ | 247/520 [15:28<17:02, 3.75s/it] {'loss': 1.6364, 'grad_norm': 0.0005282369515286827, 'learning_rate': 0.5087367864990233, 'epoch': 0.47} + 48%|████▊ | 247/520 [15:28<17:02, 3.75s/it] 48%|████▊ | 248/520 [15:32<17:09, 3.78s/it] {'loss': 1.4467, 'grad_norm': 0.0005463010786383606, 'learning_rate': 0.5059546670913684, 'epoch': 0.48} + 48%|████▊ | 248/520 [15:32<17:09, 3.78s/it] 48%|████▊ | 249/520 [15:36<17:13, 3.81s/it] {'loss': 1.5687, 'grad_norm': 0.0005145226863780186, 'learning_rate': 0.5031703736134169, 'epoch': 0.48} + 48%|████▊ | 249/520 [15:36<17:13, 3.81s/it] 48%|████▊ | 250/520 [15:39<17:14, 3.83s/it] {'loss': 1.522, 'grad_norm': 0.0006867335878060854, 'learning_rate': 0.5003840142464886, 'epoch': 0.48} + 48%|████▊ | 250/520 [15:39<17:14, 3.83s/it] 48%|████▊ | 251/520 [15:43<17:13, 3.84s/it] {'loss': 1.5782, 'grad_norm': 0.0005741723802480572, 'learning_rate': 0.49759569725217206, 'epoch': 0.48} + 48%|████▊ | 251/520 [15:43<17:13, 3.84s/it] 48%|████▊ | 252/520 [15:47<17:10, 3.85s/it] {'loss': 1.6812, 'grad_norm': 0.0006895647629046323, 'learning_rate': 0.4948055309681175, 'epoch': 0.48} + 48%|████▊ | 252/520 [15:47<17:10, 3.85s/it] 49%|████▊ | 253/520 [15:51<17:09, 3.85s/it] {'loss': 1.5836, 'grad_norm': 0.000806374997345504, 'learning_rate': 0.49201362380382774, 'epoch': 0.49} + 49%|████▊ | 253/520 [15:51<17:09, 3.85s/it] 49%|████▉ | 254/520 [15:55<17:02, 3.84s/it] {'loss': 1.4611, 'grad_norm': 0.0005117361029485381, 'learning_rate': 0.48922008423644625, 'epoch': 0.49} + 49%|████▉ | 254/520 [15:55<17:02, 3.84s/it] 49%|████▉ | 255/520 [15:59<17:01, 3.86s/it] {'loss': 1.504, 'grad_norm': 0.0008033018678374104, 'learning_rate': 0.48642502080654154, 'epoch': 0.49} + 49%|████▉ | 255/520 [15:59<17:01, 3.86s/it] 49%|████▉ | 256/520 [16:03<16:58, 3.86s/it] {'loss': 1.5498, 'grad_norm': 0.0009011069559848767, 'learning_rate': 0.48362854211389095, 'epoch': 0.49} + 49%|████▉ | 256/520 [16:03<16:58, 3.86s/it] 49%|████▉ | 257/520 [16:06<16:55, 3.86s/it] {'loss': 1.5692, 'grad_norm': 0.0005121568425118853, 'learning_rate': 0.4808307568132605, 'epoch': 0.49} + 49%|████▉ | 257/520 [16:06<16:55, 3.86s/it] 50%|████▉ | 258/520 [16:10<16:42, 3.83s/it] {'loss': 1.5799, 'grad_norm': 0.0005391120948217826, 'learning_rate': 0.4780317736101835, 'epoch': 0.5} + 50%|████▉ | 258/520 [16:10<16:42, 3.83s/it] 50%|████▉ | 259/520 [16:14<16:24, 3.77s/it] {'loss': 1.6256, 'grad_norm': 0.0005132040176088683, 'learning_rate': 0.4752317012567363, 'epoch': 0.5} + 50%|████▉ | 259/520 [16:14<16:24, 3.77s/it] 50%|█████ | 260/520 [16:17<16:09, 3.73s/it] {'loss': 1.7785, 'grad_norm': 0.0006850434861692008, 'learning_rate': 0.4724306485473137, 'epoch': 0.5} + 50%|█████ | 260/520 [16:17<16:09, 3.73s/it] 50%|█████ | 261/520 [16:21<15:59, 3.71s/it] {'loss': 1.6866, 'grad_norm': 0.000840569735444132, 'learning_rate': 0.4696287243144013, 'epoch': 0.5} + 50%|█████ | 261/520 [16:21<15:59, 3.71s/it] 50%|█████ | 262/520 [16:25<15:52, 3.69s/it] {'loss': 1.4754, 'grad_norm': 0.0005771597597112471, 'learning_rate': 0.46682603742434664, 'epoch': 0.5} + 50%|█████ | 262/520 [16:25<15:52, 3.69s/it] 51%|█████ | 263/520 [16:28<15:44, 3.68s/it] {'loss': 1.7339, 'grad_norm': 0.0007438276979890202, 'learning_rate': 0.46402269677312996, 'epoch': 0.51} + 51%|█████ | 263/520 [16:28<15:44, 3.68s/it] 51%|█████ | 264/520 [16:32<15:37, 3.66s/it] {'loss': 1.596, 'grad_norm': 0.0006657331960123404, 'learning_rate': 0.4612188112821328, 'epoch': 0.51} + 51%|█████ | 264/520 [16:32<15:37, 3.66s/it] 51%|█████ | 265/520 [16:36<15:33, 3.66s/it] {'loss': 1.4885, 'grad_norm': 0.0010745075672659089, 'learning_rate': 0.45841448989390604, 'epoch': 0.51} + 51%|█████ | 265/520 [16:36<15:33, 3.66s/it] 51%|█████ | 266/520 [16:39<15:29, 3.66s/it] {'loss': 1.3271, 'grad_norm': 0.0006021482788269592, 'learning_rate': 0.4556098415679368, 'epoch': 0.51} + 51%|█████ | 266/520 [16:39<15:29, 3.66s/it] 51%|█████▏ | 267/520 [16:43<15:25, 3.66s/it] {'loss': 1.4825, 'grad_norm': 0.0005205388775634902, 'learning_rate': 0.4528049752764151, 'epoch': 0.51} + 51%|█████▏ | 267/520 [16:43<15:25, 3.66s/it] 52%|█████▏ | 268/520 [16:47<15:21, 3.66s/it] {'loss': 1.8751, 'grad_norm': 0.0005571904556697638, 'learning_rate': 0.45, 'epoch': 0.52} + 52%|█████▏ | 268/520 [16:47<15:21, 3.66s/it] 52%|█████▏ | 269/520 [16:51<15:34, 3.72s/it] {'loss': 1.5847, 'grad_norm': 0.0008464649230339, 'learning_rate': 0.4471950247235849, 'epoch': 0.52} + 52%|█████▏ | 269/520 [16:51<15:34, 3.72s/it] 52%|█████▏ | 270/520 [16:54<15:38, 3.75s/it] {'loss': 1.6256, 'grad_norm': 0.0005776396385703337, 'learning_rate': 0.44439015843206325, 'epoch': 0.52} + 52%|█████▏ | 270/520 [16:54<15:38, 3.75s/it] 52%|█████▏ | 271/520 [16:58<15:40, 3.78s/it] {'loss': 1.5982, 'grad_norm': 0.0006900617687886516, 'learning_rate': 0.4415855101060941, 'epoch': 0.52} + 52%|█████▏ | 271/520 [16:58<15:40, 3.78s/it] 52%|█████▏ | 272/520 [17:02<15:43, 3.80s/it] {'loss': 1.6677, 'grad_norm': 0.0005650643386591377, 'learning_rate': 0.4387811887178673, 'epoch': 0.52} + 52%|█████▏ | 272/520 [17:02<15:43, 3.80s/it] 52%|█████▎ | 273/520 [17:06<15:38, 3.80s/it] {'loss': 1.771, 'grad_norm': 0.0005127133680678101, 'learning_rate': 0.43597730322687017, 'epoch': 0.53} + 52%|█████▎ | 273/520 [17:06<15:38, 3.80s/it] 53%|█████▎ | 274/520 [17:09<15:20, 3.74s/it] {'loss': 1.5291, 'grad_norm': 0.0009232482805423411, 'learning_rate': 0.4331739625756535, 'epoch': 0.53} + 53%|█████▎ | 274/520 [17:09<15:20, 3.74s/it] 53%|█████▎ | 275/520 [17:13<15:07, 3.70s/it] {'loss': 1.4663, 'grad_norm': 0.0004996129843846529, 'learning_rate': 0.43037127568559885, 'epoch': 0.53} + 53%|█████▎ | 275/520 [17:13<15:07, 3.70s/it] 53%|█████▎ | 276/520 [17:17<14:59, 3.69s/it] {'loss': 1.5661, 'grad_norm': 0.0005874859194159861, 'learning_rate': 0.42756935145268626, 'epoch': 0.53} + 53%|█████▎ | 276/520 [17:17<14:59, 3.69s/it] 53%|█████▎ | 277/520 [17:20<14:51, 3.67s/it] {'loss': 1.7678, 'grad_norm': 0.0006737837542466599, 'learning_rate': 0.42476829874326366, 'epoch': 0.53} + 53%|█████▎ | 277/520 [17:20<14:51, 3.67s/it] 53%|█████▎ | 278/520 [17:24<14:46, 3.66s/it] {'loss': 1.4345, 'grad_norm': 0.0005535365329072034, 'learning_rate': 0.42196822638981657, 'epoch': 0.53} + 53%|█████▎ | 278/520 [17:24<14:46, 3.66s/it] 54%|█████▎ | 279/520 [17:28<14:40, 3.65s/it] {'loss': 1.6842, 'grad_norm': 0.0012659684047237859, 'learning_rate': 0.41916924318673954, 'epoch': 0.54} + 54%|█████▎ | 279/520 [17:28<14:40, 3.65s/it] 54%|█████▍ | 280/520 [17:31<14:36, 3.65s/it] {'loss': 1.4737, 'grad_norm': 0.0006017634567640481, 'learning_rate': 0.4163714578861091, 'epoch': 0.54} + 54%|█████▍ | 280/520 [17:31<14:36, 3.65s/it] 54%|█████▍ | 281/520 [17:35<14:31, 3.65s/it] {'loss': 1.6012, 'grad_norm': 0.0008385092202609926, 'learning_rate': 0.4135749791934585, 'epoch': 0.54} + 54%|█████▍ | 281/520 [17:35<14:31, 3.65s/it] 54%|█████▍ | 282/520 [17:39<14:27, 3.65s/it] {'loss': 1.4301, 'grad_norm': 0.0006084826906977942, 'learning_rate': 0.41077991576355377, 'epoch': 0.54} + 54%|█████▍ | 282/520 [17:39<14:27, 3.65s/it] 54%|█████▍ | 283/520 [17:42<14:25, 3.65s/it] {'loss': 1.6531, 'grad_norm': 0.0005967436133985099, 'learning_rate': 0.4079863761961723, 'epoch': 0.54} + 54%|█████▍ | 283/520 [17:42<14:25, 3.65s/it] 55%|█████▍ | 284/520 [17:46<14:22, 3.65s/it] {'loss': 1.644, 'grad_norm': 0.0007529839641397584, 'learning_rate': 0.40519446903188255, 'epoch': 0.55} + 55%|█████▍ | 284/520 [17:46<14:22, 3.65s/it] 55%|█████▍ | 285/520 [17:50<14:32, 3.71s/it] {'loss': 1.4807, 'grad_norm': 0.0004951598174730033, 'learning_rate': 0.40240430274782807, 'epoch': 0.55} + 55%|█████▍ | 285/520 [17:50<14:32, 3.71s/it] 55%|█████▌ | 286/520 [17:54<14:39, 3.76s/it] {'loss': 1.3374, 'grad_norm': 0.0005714825223277545, 'learning_rate': 0.3996159857535115, 'epoch': 0.55} + 55%|█████▌ | 286/520 [17:54<14:39, 3.76s/it] 55%|█████▌ | 287/520 [17:57<14:43, 3.79s/it] {'loss': 1.6087, 'grad_norm': 0.0006211850834429666, 'learning_rate': 0.39682962638658326, 'epoch': 0.55} + 55%|█████▌ | 287/520 [17:57<14:43, 3.79s/it] 55%|█████▌ | 288/520 [18:01<14:37, 3.78s/it] {'loss': 1.6878, 'grad_norm': 0.0005567412018425939, 'learning_rate': 0.3940453329086318, 'epoch': 0.55} + 55%|█████▌ | 288/520 [18:01<14:37, 3.78s/it] 56%|█████▌ | 289/520 [18:05<14:23, 3.74s/it] {'loss': 1.4868, 'grad_norm': 0.0006023393398076616, 'learning_rate': 0.3912632135009769, 'epoch': 0.56} + 56%|█████▌ | 289/520 [18:05<14:23, 3.74s/it] 56%|█████▌ | 290/520 [18:08<14:13, 3.71s/it] {'loss': 1.43, 'grad_norm': 0.0005752419617525561, 'learning_rate': 0.3884833762604671, 'epoch': 0.56} + 56%|█████▌ | 290/520 [18:08<14:13, 3.71s/it] 56%|█████▌ | 291/520 [18:12<14:07, 3.70s/it] {'loss': 1.4595, 'grad_norm': 0.0006604394955571979, 'learning_rate': 0.3857059291952791, 'epoch': 0.56} + 56%|█████▌ | 291/520 [18:12<14:07, 3.70s/it] 56%|█████▌ | 292/520 [18:16<14:13, 3.74s/it] {'loss': 1.542, 'grad_norm': 0.0005430906951689499, 'learning_rate': 0.3829309802207215, 'epoch': 0.56} + 56%|█████▌ | 292/520 [18:16<14:13, 3.74s/it] 56%|█████▋ | 293/520 [18:20<14:17, 3.78s/it] {'loss': 1.451, 'grad_norm': 0.0007188179277693833, 'learning_rate': 0.3801586371550418, 'epoch': 0.56} + 56%|█████▋ | 293/520 [18:20<14:17, 3.78s/it] 57%|█████▋ | 294/520 [18:24<14:19, 3.80s/it] {'loss': 1.4936, 'grad_norm': 0.0008735787604839575, 'learning_rate': 0.377389007715237, 'epoch': 0.57} + 57%|█████▋ | 294/520 [18:24<14:19, 3.80s/it] 57%|█████▋ | 295/520 [18:28<14:19, 3.82s/it] {'loss': 1.743, 'grad_norm': 0.000586924852333245, 'learning_rate': 0.3746221995128687, 'epoch': 0.57} + 57%|█████▋ | 295/520 [18:28<14:19, 3.82s/it] 57%|█████▋ | 296/520 [18:31<14:19, 3.84s/it] {'loss': 1.4286, 'grad_norm': 0.0005138762404028277, 'learning_rate': 0.3718583200498814, 'epoch': 0.57} + 57%|█████▋ | 296/520 [18:31<14:19, 3.84s/it] 57%|█████▋ | 297/520 [18:35<14:17, 3.84s/it] {'loss': 1.6004, 'grad_norm': 0.0006448849596874113, 'learning_rate': 0.36909747671442633, 'epoch': 0.57} + 57%|█████▋ | 297/520 [18:35<14:17, 3.84s/it] 57%|█████▋ | 298/520 [18:39<14:13, 3.84s/it] {'loss': 1.5522, 'grad_norm': 0.0011854187542644891, 'learning_rate': 0.36633977677668844, 'epoch': 0.57} + 57%|█████▋ | 298/520 [18:39<14:13, 3.84s/it] 57%|█████▊ | 299/520 [18:43<14:09, 3.84s/it] {'loss': 1.7401, 'grad_norm': 0.0010566251941206196, 'learning_rate': 0.36358532738471877, 'epoch': 0.57} + 57%|█████▊ | 299/520 [18:43<14:09, 3.84s/it] 58%|█████▊ | 300/520 [18:47<14:05, 3.84s/it] {'loss': 1.6187, 'grad_norm': 0.0007489422386136172, 'learning_rate': 0.36083423556027117, 'epoch': 0.58} + 58%|█████▊ | 300/520 [18:47<14:05, 3.84s/it] 58%|█████▊ | 301/520 [18:51<14:01, 3.84s/it] {'loss': 1.5955, 'grad_norm': 0.0008214397499637073, 'learning_rate': 0.35808660819464394, 'epoch': 0.58} + 58%|█████▊ | 301/520 [18:51<14:01, 3.84s/it] 58%|█████▊ | 302/520 [18:54<13:55, 3.83s/it] {'loss': 1.7682, 'grad_norm': 0.0006666608860659562, 'learning_rate': 0.35534255204452697, 'epoch': 0.58} + 58%|█████▊ | 302/520 [18:55<13:55, 3.83s/it] 58%|█████▊ | 303/520 [18:58<13:36, 3.76s/it] {'loss': 1.4869, 'grad_norm': 0.0013269635730474257, 'learning_rate': 0.35260217372785374, 'epoch': 0.58} + 58%|█████▊ | 303/520 [18:58<13:36, 3.76s/it] 58%|█████▊ | 304/520 [19:02<13:23, 3.72s/it] {'loss': 1.6352, 'grad_norm': 0.0007913117897254674, 'learning_rate': 0.34986557971965854, 'epoch': 0.58} + 58%|█████▊ | 304/520 [19:02<13:23, 3.72s/it] 59%|█████▊ | 305/520 [19:05<13:13, 3.69s/it] {'loss': 1.6712, 'grad_norm': 0.0008659462335981605, 'learning_rate': 0.34713287634793977, 'epoch': 0.59} + 59%|█████▊ | 305/520 [19:05<13:13, 3.69s/it] 59%|█████▉ | 306/520 [19:09<13:05, 3.67s/it] {'loss': 1.5444, 'grad_norm': 0.0007150735633720339, 'learning_rate': 0.34440416978952826, 'epoch': 0.59} + 59%|█████▉ | 306/520 [19:09<13:05, 3.67s/it] 59%|█████▉ | 307/520 [19:13<13:20, 3.76s/it] {'loss': 1.493, 'grad_norm': 0.0006940494673743895, 'learning_rate': 0.3416795660659623, 'epoch': 0.59} + 59%|█████▉ | 307/520 [19:13<13:20, 3.76s/it] 59%|█████▉ | 308/520 [19:17<13:07, 3.71s/it] {'loss': 1.6192, 'grad_norm': 0.0005107716244409245, 'learning_rate': 0.33895917103936785, 'epoch': 0.59} + 59%|█████▉ | 308/520 [19:17<13:07, 3.71s/it] 59%|█████▉ | 309/520 [19:20<12:58, 3.69s/it] {'loss': 1.4933, 'grad_norm': 0.0004884729628317312, 'learning_rate': 0.3362430904083461, 'epoch': 0.59} + 59%|█████▉ | 309/520 [19:20<12:58, 3.69s/it] 60%|█████▉ | 310/520 [19:24<12:48, 3.66s/it] {'loss': 1.4569, 'grad_norm': 0.0006543278364767281, 'learning_rate': 0.3335314297038656, 'epoch': 0.6} + 60%|█████▉ | 310/520 [19:24<12:48, 3.66s/it] 60%|█████▉ | 311/520 [19:27<12:47, 3.67s/it] {'loss': 1.462, 'grad_norm': 0.0005632348316125208, 'learning_rate': 0.33082429428516275, 'epoch': 0.6} + 60%|█████▉ | 311/520 [19:27<12:47, 3.67s/it] 60%|██████ | 312/520 [19:31<12:42, 3.66s/it] {'loss': 1.4281, 'grad_norm': 0.0007772382042103036, 'learning_rate': 0.3281217893356478, 'epoch': 0.6} + 60%|██████ | 312/520 [19:31<12:42, 3.66s/it] 60%|██████ | 313/520 [19:35<12:38, 3.66s/it] {'loss': 1.383, 'grad_norm': 0.0007115326554277307, 'learning_rate': 0.3254240198588178, 'epoch': 0.6} + 60%|██████ | 313/520 [19:35<12:38, 3.66s/it] 60%|██████ | 314/520 [19:39<13:01, 3.80s/it] {'loss': 1.4323, 'grad_norm': 0.0006501082302601829, 'learning_rate': 0.32273109067417766, 'epoch': 0.6} + 60%|██████ | 314/520 [19:39<13:01, 3.80s/it] 61%|██████ | 315/520 [19:43<12:49, 3.75s/it] {'loss': 1.7626, 'grad_norm': 0.000808402166462667, 'learning_rate': 0.32004310641316663, 'epoch': 0.61} + 61%|██████ | 315/520 [19:43<12:49, 3.75s/it] 61%|██████ | 316/520 [19:47<13:10, 3.88s/it] {'loss': 1.4354, 'grad_norm': 0.0006896352459869312, 'learning_rate': 0.3173601715150931, 'epoch': 0.61} + 61%|██████ | 316/520 [19:47<13:10, 3.88s/it] 61%|██████ | 317/520 [19:50<12:51, 3.80s/it] {'loss': 1.4207, 'grad_norm': 0.000868820514276265, 'learning_rate': 0.31468239022307715, 'epoch': 0.61} + 61%|██████ | 317/520 [19:50<12:51, 3.80s/it] 61%|██████ | 318/520 [19:54<12:40, 3.76s/it] {'loss': 1.587, 'grad_norm': 0.0008670740969897607, 'learning_rate': 0.31200986657999963, 'epoch': 0.61} + 61%|██████ | 318/520 [30:47<12:40, 3.76s/it] 61%|██████▏ | 319/520 [30:51<11:08:54, 199.68s/it] {'loss': 1.4228, 'grad_norm': 0.0006163006486445086, 'learning_rate': 0.30934270442446005, 'epoch': 0.61} + 61%|██████▏ | 319/520 [30:51<11:08:54, 199.68s/it] 62%|██████▏ | 320/520 [30:54<7:49:32, 140.86s/it] {'loss': 1.357, 'grad_norm': 0.0005183965226942863, 'learning_rate': 0.3066810073867421, 'epoch': 0.62} + 62%|██████▏ | 320/520 [30:54<7:49:32, 140.86s/it] 62%|██████▏ | 321/520 [30:58<5:30:41, 99.71s/it] {'loss': 1.6114, 'grad_norm': 0.0006586900686046214, 'learning_rate': 0.30402487888478685, 'epoch': 0.62} + 62%|██████▏ | 321/520 [30:58<5:30:41, 99.71s/it] 62%|██████▏ | 322/520 [31:02<3:53:56, 70.89s/it] {'loss': 1.5643, 'grad_norm': 0.001270279333997281, 'learning_rate': 0.30137442212017496, 'epoch': 0.62} + 62%|██████▏ | 322/520 [31:02<3:53:56, 70.89s/it] 62%|██████▏ | 323/520 [31:05<2:46:31, 50.72s/it] {'loss': 1.6822, 'grad_norm': 0.0005225389295971243, 'learning_rate': 0.29872974007411623, 'epoch': 0.62} + 62%|██████▏ | 323/520 [31:05<2:46:31, 50.72s/it] 62%|██████▏ | 324/520 [31:09<1:59:33, 36.60s/it] {'loss': 1.5396, 'grad_norm': 0.0009513904884667106, 'learning_rate': 0.2960909355034491, 'epoch': 0.62} + 62%|██████▏ | 324/520 [31:09<1:59:33, 36.60s/it] 62%|██████▎ | 325/520 [31:13<1:26:48, 26.71s/it] {'loss': 1.5322, 'grad_norm': 0.000664450836296855, 'learning_rate': 0.2934581109366477, 'epoch': 0.62} + 62%|██████▎ | 325/520 [31:13<1:26:48, 26.71s/it] 63%|██████▎ | 326/520 [31:16<1:03:59, 19.79s/it] {'loss': 1.5107, 'grad_norm': 0.0005528575117569234, 'learning_rate': 0.2908313686698384, 'epoch': 0.63} + 63%|██████▎ | 326/520 [31:16<1:03:59, 19.79s/it] 63%|██████▎ | 327/520 [31:20<48:15, 15.00s/it] {'loss': 1.7465, 'grad_norm': 0.0009644508152634365, 'learning_rate': 0.2882108107628246, 'epoch': 0.63} + 63%|██████▎ | 327/520 [31:20<48:15, 15.00s/it] 63%|██████▎ | 328/520 [31:24<37:18, 11.66s/it] {'loss': 1.6046, 'grad_norm': 0.0005032980876631503, 'learning_rate': 0.2855965390351222, 'epoch': 0.63} + 63%|██████▎ | 328/520 [31:24<37:18, 11.66s/it] 63%|██████▎ | 329/520 [31:28<29:43, 9.34s/it] {'loss': 1.4095, 'grad_norm': 0.0005733497811015461, 'learning_rate': 0.28298865506200294, 'epoch': 0.63} + 63%|██████▎ | 329/520 [31:28<29:43, 9.34s/it] 63%|██████▎ | 330/520 [31:32<24:21, 7.69s/it] {'loss': 1.5134, 'grad_norm': 0.000720446205604842, 'learning_rate': 0.28038726017054766, 'epoch': 0.63} + 63%|██████▎ | 330/520 [31:32<24:21, 7.69s/it] 64%|██████▎ | 331/520 [31:36<20:36, 6.54s/it] {'loss': 1.4717, 'grad_norm': 0.0005839679851456226, 'learning_rate': 0.27779245543570963, 'epoch': 0.64} + 64%|██████▎ | 331/520 [31:36<20:36, 6.54s/it] 64%|██████▍ | 332/520 [31:40<17:58, 5.74s/it] {'loss': 1.7388, 'grad_norm': 0.0007988413163286246, 'learning_rate': 0.2752043416763874, 'epoch': 0.64} + 64%|██████▍ | 332/520 [31:40<17:58, 5.74s/it] 64%|██████▍ | 333/520 [31:43<15:54, 5.11s/it] {'loss': 1.6782, 'grad_norm': 0.0005445534598194279, 'learning_rate': 0.27262301945150735, 'epoch': 0.64} + 64%|██████▍ | 333/520 [31:43<15:54, 5.11s/it] 64%|██████▍ | 334/520 [31:47<14:27, 4.66s/it] {'loss': 1.5334, 'grad_norm': 0.000655145692744968, 'learning_rate': 0.2700485890561167, 'epoch': 0.64} + 64%|██████▍ | 334/520 [31:47<14:27, 4.66s/it] 64%|██████▍ | 335/520 [31:50<13:25, 4.35s/it] {'loss': 1.5205, 'grad_norm': 0.0004958076898034068, 'learning_rate': 0.26748115051748633, 'epoch': 0.64} + 64%|██████▍ | 335/520 [31:50<13:25, 4.35s/it] 65%|██████▍ | 336/520 [31:54<12:41, 4.14s/it] {'loss': 1.4195, 'grad_norm': 0.0007948039671677016, 'learning_rate': 0.2649208035912249, 'epoch': 0.65} + 65%|██████▍ | 336/520 [31:54<12:41, 4.14s/it] 65%|██████▍ | 337/520 [31:58<12:12, 4.00s/it] {'loss': 1.4103, 'grad_norm': 0.0005833827790231704, 'learning_rate': 0.2623676477574025, 'epoch': 0.65} + 65%|██████▍ | 337/520 [31:58<12:12, 4.00s/it] 65%|██████▌ | 338/520 [32:01<11:49, 3.90s/it] {'loss': 1.5616, 'grad_norm': 0.000648630535233091, 'learning_rate': 0.25982178221668534, 'epoch': 0.65} + 65%|██████▌ | 338/520 [32:01<11:49, 3.90s/it] 65%|██████▌ | 339/520 [32:05<11:33, 3.83s/it] {'loss': 1.4818, 'grad_norm': 0.0013919062995492466, 'learning_rate': 0.25728330588648174, 'epoch': 0.65} + 65%|██████▌ | 339/520 [32:05<11:33, 3.83s/it] 65%|██████▌ | 340/520 [32:09<11:21, 3.79s/it] {'loss': 1.4418, 'grad_norm': 0.0005936038792568147, 'learning_rate': 0.25475231739709886, 'epoch': 0.65} + 65%|██████▌ | 340/520 [32:09<11:21, 3.79s/it] 66%|██████▌ | 341/520 [32:12<11:11, 3.75s/it] {'loss': 1.4884, 'grad_norm': 0.0006417117836165739, 'learning_rate': 0.25222891508790973, 'epoch': 0.66} + 66%|██████▌ | 341/520 [32:12<11:11, 3.75s/it] 66%|██████▌ | 342/520 [32:16<11:01, 3.72s/it] {'loss': 1.7506, 'grad_norm': 0.0012985897869137598, 'learning_rate': 0.24971319700353342, 'epoch': 0.66} + 66%|██████▌ | 342/520 [32:16<11:01, 3.72s/it] 66%|██████▌ | 343/520 [32:20<10:58, 3.72s/it] {'loss': 1.6955, 'grad_norm': 0.0008963095586649477, 'learning_rate': 0.24720526089002456, 'epoch': 0.66} + 66%|██████▌ | 343/520 [32:20<10:58, 3.72s/it] 66%|██████▌ | 344/520 [32:23<10:53, 3.71s/it] {'loss': 1.4213, 'grad_norm': 0.0005000133669349044, 'learning_rate': 0.24470520419107664, 'epoch': 0.66} + 66%|██████▌ | 344/520 [32:23<10:53, 3.71s/it] 66%|██████▋ | 345/520 [32:27<10:46, 3.70s/it] {'loss': 1.5656, 'grad_norm': 0.0005843486890437183, 'learning_rate': 0.24221312404423484, 'epoch': 0.66} + 66%|██████▋ | 345/520 [32:27<10:46, 3.70s/it] 67%|██████▋ | 346/520 [32:31<10:42, 3.69s/it] {'loss': 1.6789, 'grad_norm': 0.0006123511378345746, 'learning_rate': 0.2397291172771221, 'epoch': 0.67} + 67%|██████▋ | 346/520 [32:31<10:42, 3.69s/it] 67%|██████▋ | 347/520 [32:34<10:37, 3.68s/it] {'loss': 1.441, 'grad_norm': 0.0004803529266263331, 'learning_rate': 0.2372532804036779, 'epoch': 0.67} + 67%|██████▋ | 347/520 [32:34<10:37, 3.68s/it]Token indices sequence length is longer than the specified maximum sequence length for this model (2778 > 2048). Running this sequence through the model will result in indexing errors + 67%|██████▋ | 348/520 [32:38<10:31, 3.67s/it] {'loss': 1.4505, 'grad_norm': 0.0007247597809724055, 'learning_rate': 0.23478570962040696, 'epoch': 0.67} + 67%|██████▋ | 348/520 [32:38<10:31, 3.67s/it] 67%|██████▋ | 349/520 [32:42<10:26, 3.66s/it] {'loss': 1.4722, 'grad_norm': 0.0005492207219215826, 'learning_rate': 0.23232650080264208, 'epoch': 0.67} + 67%|██████▋ | 349/520 [32:42<10:26, 3.66s/it] 67%|██████▋ | 350/520 [32:45<10:21, 3.65s/it] {'loss': 1.5034, 'grad_norm': 0.0005392640613507999, 'learning_rate': 0.22987574950082, 'epoch': 0.67} + 67%|██████▋ | 350/520 [32:45<10:21, 3.65s/it] 68%|██████▊ | 351/520 [32:49<10:21, 3.68s/it] {'loss': 1.3889, 'grad_norm': 0.0005491485159298918, 'learning_rate': 0.22743355093676668, 'epoch': 0.68} + 68%|██████▊ | 351/520 [32:49<10:21, 3.68s/it] 68%|██████▊ | 352/520 [32:53<10:16, 3.67s/it] {'loss': 1.5393, 'grad_norm': 0.0007256291667025259, 'learning_rate': 0.22500000000000012, 'epoch': 0.68} + 68%|██████▊ | 352/520 [32:53<10:16, 3.67s/it] 68%|██████▊ | 353/520 [32:56<10:12, 3.67s/it] {'loss': 1.6504, 'grad_norm': 0.000703941213633245, 'learning_rate': 0.22257519124404132, 'epoch': 0.68} + 68%|██████▊ | 353/520 [32:56<10:12, 3.67s/it] 68%|██████▊ | 354/520 [33:00<10:08, 3.67s/it] {'loss': 1.76, 'grad_norm': 0.00108473188655664, 'learning_rate': 0.2201592188827416, 'epoch': 0.68} + 68%|██████▊ | 354/520 [33:00<10:08, 3.67s/it] 68%|██████▊ | 355/520 [33:04<10:04, 3.66s/it] {'loss': 1.4734, 'grad_norm': 0.0007535086749237208, 'learning_rate': 0.21775217678662198, 'epoch': 0.68} + 68%|██████▊ | 355/520 [33:04<10:04, 3.66s/it] 68%|██████▊ | 356/520 [33:07<10:01, 3.67s/it] {'loss': 1.4653, 'grad_norm': 0.0007566994429889117, 'learning_rate': 0.2153541584792259, 'epoch': 0.68} + 68%|██████▊ | 356/520 [33:07<10:01, 3.67s/it] 69%|██████▊ | 357/520 [33:11<09:56, 3.66s/it] {'loss': 1.4704, 'grad_norm': 0.0006589748724199945, 'learning_rate': 0.21296525713348466, 'epoch': 0.69} + 69%|██████▊ | 357/520 [33:11<09:56, 3.66s/it] 69%|██████▉ | 358/520 [33:15<09:56, 3.68s/it] {'loss': 1.429, 'grad_norm': 0.000745482431718232, 'learning_rate': 0.2105855655680986, 'epoch': 0.69} + 69%|██████▉ | 358/520 [33:15<09:56, 3.68s/it] 69%|██████▉ | 359/520 [33:18<09:52, 3.68s/it] {'loss': 1.6938, 'grad_norm': 0.0013868715032804937, 'learning_rate': 0.2082151762439292, 'epoch': 0.69} + 69%|██████▉ | 359/520 [33:18<09:52, 3.68s/it] 69%|██████▉ | 360/520 [33:22<09:48, 3.68s/it] {'loss': 1.7081, 'grad_norm': 0.0006818293554379696, 'learning_rate': 0.2058541812604083, 'epoch': 0.69} + 69%|██████▉ | 360/520 [33:22<09:48, 3.68s/it] 69%|██████▉ | 361/520 [33:26<09:43, 3.67s/it] {'loss': 1.7076, 'grad_norm': 0.0008196387154059465, 'learning_rate': 0.20350267235195796, 'epoch': 0.69} + 69%|██████▉ | 361/520 [33:26<09:43, 3.67s/it] 70%|██████▉ | 362/520 [33:30<09:40, 3.68s/it] {'loss': 1.4584, 'grad_norm': 0.0008712221917868503, 'learning_rate': 0.20116074088442726, 'epoch': 0.7} + 70%|██████▉ | 362/520 [33:30<09:40, 3.68s/it] 70%|██████▉ | 363/520 [33:33<09:37, 3.68s/it] {'loss': 1.5615, 'grad_norm': 0.0005390138150445843, 'learning_rate': 0.19882847785154228, 'epoch': 0.7} + 70%|██████▉ | 363/520 [33:33<09:37, 3.68s/it] 70%|███████ | 364/520 [33:37<09:35, 3.69s/it] {'loss': 1.7424, 'grad_norm': 0.0006679408628525975, 'learning_rate': 0.1965059738713701, 'epoch': 0.7} + 70%|███████ | 364/520 [33:37<09:35, 3.69s/it] 70%|███████ | 365/520 [33:41<09:31, 3.69s/it] {'loss': 1.6064, 'grad_norm': 0.0007279645269717567, 'learning_rate': 0.1941933191827985, 'epoch': 0.7} + 70%|███████ | 365/520 [33:41<09:31, 3.69s/it] 70%|███████ | 366/520 [33:44<09:27, 3.69s/it] {'loss': 1.5518, 'grad_norm': 0.0005189728103956627, 'learning_rate': 0.1918906036420294, 'epoch': 0.7} + 70%|███████ | 366/520 [33:44<09:27, 3.69s/it] 71%|███████ | 367/520 [33:48<09:25, 3.70s/it] {'loss': 1.5627, 'grad_norm': 0.0005956079492461593, 'learning_rate': 0.18959791671908743, 'epoch': 0.71} + 71%|███████ | 367/520 [33:48<09:25, 3.70s/it] 71%|███████ | 368/520 [33:52<09:20, 3.69s/it] {'loss': 1.3786, 'grad_norm': 0.0005657732054834602, 'learning_rate': 0.18731534749434467, 'epoch': 0.71} + 71%|███████ | 368/520 [33:52<09:20, 3.69s/it] 71%|███████ | 369/520 [33:55<09:15, 3.68s/it] {'loss': 1.667, 'grad_norm': 0.000536314541527967, 'learning_rate': 0.18504298465505792, 'epoch': 0.71} + 71%|███████ | 369/520 [33:55<09:15, 3.68s/it] 71%|███████ | 370/520 [33:59<09:12, 3.68s/it] {'loss': 1.4441, 'grad_norm': 0.0007483115745755295, 'learning_rate': 0.18278091649192435, 'epoch': 0.71} + 71%|███████ | 370/520 [33:59<09:12, 3.68s/it] 71%|███████▏ | 371/520 [34:03<09:07, 3.67s/it] {'loss': 1.4312, 'grad_norm': 0.0006172200311435221, 'learning_rate': 0.18052923089564987, 'epoch': 0.71} + 71%|███████▏ | 371/520 [34:03<09:07, 3.67s/it] 72%|███████▏ | 372/520 [34:06<09:02, 3.67s/it] {'loss': 1.7811, 'grad_norm': 0.0008321272980215056, 'learning_rate': 0.17828801535353508, 'epoch': 0.72} + 72%|███████▏ | 372/520 [34:06<09:02, 3.67s/it] 72%|███████▏ | 373/520 [34:10<08:59, 3.67s/it] {'loss': 1.6604, 'grad_norm': 0.0010027553909830584, 'learning_rate': 0.17605735694607572, 'epoch': 0.72} + 72%|███████▏ | 373/520 [34:10<08:59, 3.67s/it] 72%|███████▏ | 374/520 [34:14<08:54, 3.66s/it] {'loss': 1.5384, 'grad_norm': 0.000598644129257854, 'learning_rate': 0.17383734234357875, 'epoch': 0.72} + 72%|███████▏ | 374/520 [34:14<08:54, 3.66s/it] 72%|███████▏ | 375/520 [34:17<08:56, 3.70s/it] {'loss': 1.4183, 'grad_norm': 0.0007011767919799764, 'learning_rate': 0.17162805780279533, 'epoch': 0.72} + 72%|███████▏ | 375/520 [34:17<08:56, 3.70s/it] 72%|███████▏ | 376/520 [34:21<08:58, 3.74s/it] {'loss': 1.5636, 'grad_norm': 0.0007140125215565691, 'learning_rate': 0.16942958916356993, 'epoch': 0.72} + 72%|███████▏ | 376/520 [34:21<08:58, 3.74s/it] 72%|███████▎ | 377/520 [34:25<08:57, 3.76s/it] {'loss': 1.5093, 'grad_norm': 0.0005699317785502749, 'learning_rate': 0.1672420218455037, 'epoch': 0.72} + 72%|███████▎ | 377/520 [34:25<08:57, 3.76s/it] 73%|███████▎ | 378/520 [34:29<08:57, 3.78s/it] {'loss': 1.5538, 'grad_norm': 0.0005611402741734944, 'learning_rate': 0.16506544084463715, 'epoch': 0.73} + 73%|███████▎ | 378/520 [34:29<08:57, 3.78s/it] 73%|███████▎ | 379/520 [34:33<08:55, 3.80s/it] {'loss': 1.5485, 'grad_norm': 0.0007228741743970552, 'learning_rate': 0.1628999307301462, 'epoch': 0.73} + 73%|███████▎ | 379/520 [34:33<08:55, 3.80s/it] 73%|███████▎ | 380/520 [34:37<08:51, 3.80s/it] {'loss': 1.7655, 'grad_norm': 0.0006112878217388256, 'learning_rate': 0.1607455756410573, 'epoch': 0.73} + 73%|███████▎ | 380/520 [34:37<08:51, 3.80s/it] 73%|███████▎ | 381/520 [34:40<08:42, 3.76s/it] {'loss': 1.543, 'grad_norm': 0.0004852615026631934, 'learning_rate': 0.15860245928297836, 'epoch': 0.73} + 73%|███████▎ | 381/520 [34:40<08:42, 3.76s/it] 73%|███████▎ | 382/520 [34:44<08:34, 3.73s/it] {'loss': 1.6886, 'grad_norm': 0.0006896794064199246, 'learning_rate': 0.15647066492484563, 'epoch': 0.73} + 73%|███████▎ | 382/520 [34:44<08:34, 3.73s/it] 74%|███████▎ | 383/520 [34:47<08:26, 3.70s/it] {'loss': 1.3735, 'grad_norm': 0.0008949439472157607, 'learning_rate': 0.15435027539568885, 'epoch': 0.74} + 74%|███████▎ | 383/520 [34:47<08:26, 3.70s/it] 74%|███████▍ | 384/520 [34:51<08:20, 3.68s/it] {'loss': 1.9346, 'grad_norm': 0.0008251987562215476, 'learning_rate': 0.1522413730814134, 'epoch': 0.74} + 74%|███████▍ | 384/520 [34:51<08:20, 3.68s/it] 74%|███████▍ | 385/520 [34:55<08:14, 3.66s/it] {'loss': 1.5179, 'grad_norm': 0.0006418911358522523, 'learning_rate': 0.15014403992159825, 'epoch': 0.74} + 74%|███████▍ | 385/520 [34:55<08:14, 3.66s/it] 74%|███████▍ | 386/520 [34:58<08:10, 3.66s/it] {'loss': 1.4329, 'grad_norm': 0.0005881559836029574, 'learning_rate': 0.14805835740631354, 'epoch': 0.74} + 74%|███████▍ | 386/520 [34:58<08:10, 3.66s/it] 74%|███████▍ | 387/520 [35:02<08:04, 3.65s/it] {'loss': 1.8003, 'grad_norm': 0.0006562843674360492, 'learning_rate': 0.1459844065729529, 'epoch': 0.74} + 74%|███████▍ | 387/520 [35:02<08:04, 3.65s/it] 75%|███████▍ | 388/520 [35:06<08:01, 3.64s/it] {'loss': 1.4007, 'grad_norm': 0.0004919339375941402, 'learning_rate': 0.1439222680030862, 'epoch': 0.75} + 75%|███████▍ | 388/520 [35:06<08:01, 3.64s/it] 75%|███████▍ | 389/520 [35:09<07:57, 3.65s/it] {'loss': 1.4848, 'grad_norm': 0.0009953007866048082, 'learning_rate': 0.14187202181932793, 'epoch': 0.75} + 75%|███████▍ | 389/520 [35:09<07:57, 3.65s/it] 75%|███████▌ | 390/520 [35:13<07:53, 3.65s/it] {'loss': 1.5416, 'grad_norm': 0.0005489084594865114, 'learning_rate': 0.13983374768222384, 'epoch': 0.75} + 75%|███████▌ | 390/520 [35:13<07:53, 3.65s/it] 75%|███████▌ | 391/520 [35:17<07:50, 3.65s/it] {'loss': 1.6343, 'grad_norm': 0.0005466074422118387, 'learning_rate': 0.13780752478715627, 'epoch': 0.75} + 75%|███████▌ | 391/520 [35:17<07:50, 3.65s/it] 75%|███████▌ | 392/520 [35:20<07:48, 3.66s/it] {'loss': 1.4379, 'grad_norm': 0.0005110146138191095, 'learning_rate': 0.13579343186126727, 'epoch': 0.75} + 75%|███████▌ | 392/520 [35:20<07:48, 3.66s/it] 76%|███████▌ | 393/520 [35:24<07:44, 3.66s/it] {'loss': 1.5728, 'grad_norm': 0.000824187380780973, 'learning_rate': 0.1337915471603989, 'epoch': 0.76} + 76%|███████▌ | 393/520 [35:24<07:44, 3.66s/it] 76%|███████▌ | 394/520 [35:28<07:39, 3.65s/it] {'loss': 1.5103, 'grad_norm': 0.0006718163086329292, 'learning_rate': 0.13180194846605364, 'epoch': 0.76} + 76%|███████▌ | 394/520 [35:28<07:39, 3.65s/it] 76%|███████▌ | 395/520 [35:31<07:36, 3.65s/it] {'loss': 1.4563, 'grad_norm': 0.000490044671493051, 'learning_rate': 0.12982471308237153, 'epoch': 0.76} + 76%|███████▌ | 395/520 [35:31<07:36, 3.65s/it] 76%|███████▌ | 396/520 [35:35<07:34, 3.66s/it] {'loss': 1.5486, 'grad_norm': 0.0008285717680781892, 'learning_rate': 0.1278599178331267, 'epoch': 0.76} + 76%|███████▌ | 396/520 [35:35<07:34, 3.66s/it] 76%|███████▋ | 397/520 [35:39<07:30, 3.66s/it] {'loss': 1.5305, 'grad_norm': 0.000529135568994151, 'learning_rate': 0.12590763905874314, 'epoch': 0.76} + 76%|███████▋ | 397/520 [35:39<07:30, 3.66s/it] 77%|███████▋ | 398/520 [35:42<07:26, 3.66s/it] {'loss': 1.4966, 'grad_norm': 0.0007889662572302601, 'learning_rate': 0.12396795261332731, 'epoch': 0.77} + 77%|███████▋ | 398/520 [35:42<07:26, 3.66s/it] 77%|███████▋ | 399/520 [35:46<07:23, 3.66s/it] {'loss': 1.6375, 'grad_norm': 0.0007496535605394598, 'learning_rate': 0.12204093386172225, 'epoch': 0.77} + 77%|███████▋ | 399/520 [35:46<07:23, 3.66s/it] 77%|███████▋ | 400/520 [35:50<07:19, 3.67s/it] {'loss': 1.6807, 'grad_norm': 0.0006526643069025286, 'learning_rate': 0.12012665767657825, 'epoch': 0.77} + 77%|███████▋ | 400/520 [35:50<07:19, 3.67s/it] 77%|███████▋ | 401/520 [35:53<07:15, 3.66s/it] {'loss': 1.2914, 'grad_norm': 0.0006367245743663037, 'learning_rate': 0.11822519843544421, 'epoch': 0.77} + 77%|███████▋ | 401/520 [35:53<07:15, 3.66s/it] 77%|███████▋ | 402/520 [35:57<07:11, 3.65s/it] {'loss': 1.4457, 'grad_norm': 0.0006072745667023345, 'learning_rate': 0.11633663001787797, 'epoch': 0.77} + 77%|███████▋ | 402/520 [35:57<07:11, 3.65s/it] 78%|███████▊ | 403/520 [36:01<07:08, 3.66s/it] {'loss': 1.4994, 'grad_norm': 0.0005199287332947598, 'learning_rate': 0.11446102580257549, 'epoch': 0.78} + 78%|███████▊ | 403/520 [36:01<07:08, 3.66s/it] 78%|███████▊ | 404/520 [36:04<07:05, 3.67s/it] {'loss': 1.4031, 'grad_norm': 0.0008355710339375408, 'learning_rate': 0.11259845866451956, 'epoch': 0.78} + 78%|███████▊ | 404/520 [36:04<07:05, 3.67s/it] 78%|███████▊ | 405/520 [36:08<07:01, 3.67s/it] {'loss': 1.6271, 'grad_norm': 0.0006514167473753808, 'learning_rate': 0.11074900097214908, 'epoch': 0.78} + 78%|███████▊ | 405/520 [36:08<07:01, 3.67s/it] 78%|███████▊ | 406/520 [36:12<06:57, 3.66s/it] {'loss': 1.5869, 'grad_norm': 0.0008845157260628537, 'learning_rate': 0.1089127245845461, 'epoch': 0.78} + 78%|███████▊ | 406/520 [36:12<06:57, 3.66s/it] 78%|███████▊ | 407/520 [36:15<06:53, 3.66s/it] {'loss': 1.6139, 'grad_norm': 0.0005997518097210534, 'learning_rate': 0.10708970084864515, 'epoch': 0.78} + 78%|███████▊ | 407/520 [36:15<06:53, 3.66s/it] 78%|███████▊ | 408/520 [36:19<06:49, 3.65s/it] {'loss': 1.4694, 'grad_norm': 0.0009132219833795238, 'learning_rate': 0.10528000059645995, 'epoch': 0.78} + 78%|███████▊ | 408/520 [36:19<06:49, 3.65s/it] 79%|███████▊ | 409/520 [36:22<06:44, 3.65s/it] {'loss': 1.6386, 'grad_norm': 0.0005999061441169911, 'learning_rate': 0.10348369414233174, 'epoch': 0.79} + 79%|███████▊ | 409/520 [36:22<06:44, 3.65s/it] 79%|███████▉ | 410/520 [36:26<06:40, 3.64s/it] {'loss': 1.3077, 'grad_norm': 0.0005886854014889135, 'learning_rate': 0.10170085128019768, 'epoch': 0.79} + 79%|███████▉ | 410/520 [36:26<06:40, 3.64s/it] 79%|███████▉ | 411/520 [36:30<06:37, 3.65s/it] {'loss': 1.5952, 'grad_norm': 0.0009081229274329448, 'learning_rate': 0.09993154128087836, 'epoch': 0.79} + 79%|███████▉ | 411/520 [36:30<06:37, 3.65s/it] 79%|███████▉ | 412/520 [36:33<06:33, 3.64s/it] {'loss': 1.506, 'grad_norm': 0.0005493916362266879, 'learning_rate': 0.0981758328893866, 'epoch': 0.79} + 79%|███████▉ | 412/520 [36:33<06:33, 3.64s/it] 79%|███████▉ | 413/520 [36:37<06:30, 3.65s/it] {'loss': 1.7338, 'grad_norm': 0.0006782757054212821, 'learning_rate': 0.09643379432225693, 'epoch': 0.79} + 79%|███████▉ | 413/520 [36:37<06:30, 3.65s/it] 80%|███████▉ | 414/520 [36:41<06:27, 3.66s/it] {'loss': 1.4411, 'grad_norm': 0.0006511179119246637, 'learning_rate': 0.09470549326489411, 'epoch': 0.8} + 80%|███████▉ | 414/520 [36:41<06:27, 3.66s/it] 80%|███████▉ | 415/520 [36:44<06:23, 3.66s/it] {'loss': 1.4505, 'grad_norm': 0.0007861369404472746, 'learning_rate': 0.09299099686894423, 'epoch': 0.8} + 80%|███████▉ | 415/520 [36:44<06:23, 3.66s/it] 80%|████████ | 416/520 [36:48<06:19, 3.65s/it] {'loss': 1.3755, 'grad_norm': 0.000590244518344095, 'learning_rate': 0.09129037174968503, 'epoch': 0.8} + 80%|████████ | 416/520 [36:48<06:19, 3.65s/it] 80%|████████ | 417/520 [36:52<06:15, 3.65s/it] {'loss': 1.5536, 'grad_norm': 0.0007961679417935082, 'learning_rate': 0.08960368398343747, 'epoch': 0.8} + 80%|████████ | 417/520 [36:52<06:15, 3.65s/it] 80%|████████ | 418/520 [36:55<06:11, 3.65s/it] {'loss': 1.5546, 'grad_norm': 0.0004854480928008809, 'learning_rate': 0.08793099910499924, 'epoch': 0.8} + 80%|████████ | 418/520 [36:55<06:11, 3.65s/it] 81%|████████ | 419/520 [36:59<06:06, 3.63s/it] {'loss': 1.5296, 'grad_norm': 0.0005035653279996663, 'learning_rate': 0.08627238210509765, 'epoch': 0.81} + 81%|████████ | 419/520 [36:59<06:06, 3.63s/it] 81%|████████ | 420/520 [37:02<06:02, 3.62s/it] {'loss': 1.3954, 'grad_norm': 0.0006658375732088655, 'learning_rate': 0.08462789742786457, 'epoch': 0.81} + 81%|████████ | 420/520 [37:02<06:02, 3.62s/it] 81%|████████ | 421/520 [37:06<05:58, 3.62s/it] {'loss': 1.3141, 'grad_norm': 0.0006561608204671801, 'learning_rate': 0.08299760896833293, 'epoch': 0.81} + 81%|████████ | 421/520 [37:06<05:58, 3.62s/it] 81%|████████ | 422/520 [37:10<05:55, 3.62s/it] {'loss': 1.4669, 'grad_norm': 0.0005560612838053569, 'learning_rate': 0.08138158006995365, 'epoch': 0.81} + 81%|████████ | 422/520 [37:10<05:55, 3.62s/it] 81%|████████▏ | 423/520 [37:13<05:51, 3.63s/it] {'loss': 1.4853, 'grad_norm': 0.0008093225581681115, 'learning_rate': 0.07977987352213499, 'epoch': 0.81} + 81%|████████▏ | 423/520 [37:13<05:51, 3.63s/it] 82%|████████▏ | 424/520 [37:17<05:48, 3.64s/it] {'loss': 1.7713, 'grad_norm': 0.0006796356721862113, 'learning_rate': 0.0781925515578024, 'epoch': 0.82} + 82%|████████▏ | 424/520 [37:17<05:48, 3.64s/it] 82%|████████▏ | 425/520 [37:21<05:46, 3.64s/it] {'loss': 1.4348, 'grad_norm': 0.0008414073313069072, 'learning_rate': 0.07661967585098063, 'epoch': 0.82} + 82%|████████▏ | 425/520 [37:21<05:46, 3.64s/it] 82%|████████▏ | 426/520 [37:24<05:41, 3.63s/it] {'loss': 1.5339, 'grad_norm': 0.0009746011780218057, 'learning_rate': 0.07506130751439803, 'epoch': 0.82} + 82%|████████▏ | 426/520 [37:24<05:41, 3.63s/it] 82%|████████▏ | 427/520 [37:28<05:37, 3.63s/it] {'loss': 1.3802, 'grad_norm': 0.0007619234047046429, 'learning_rate': 0.07351750709711112, 'epoch': 0.82} + 82%|████████▏ | 427/520 [37:28<05:37, 3.63s/it] 82%|████████▏ | 428/520 [37:32<05:33, 3.63s/it] {'loss': 1.3509, 'grad_norm': 0.0007303092710455358, 'learning_rate': 0.07198833458215287, 'epoch': 0.82} + 82%|████████▏ | 428/520 [37:32<05:33, 3.63s/it] 82%|████████▎ | 429/520 [37:35<05:31, 3.64s/it] {'loss': 1.4948, 'grad_norm': 0.0007317831820948439, 'learning_rate': 0.07047384938420152, 'epoch': 0.82} + 82%|████████▎ | 429/520 [37:35<05:31, 3.64s/it]Token indices sequence length is longer than the specified maximum sequence length for this model (2076 > 2048). Running this sequence through the model will result in indexing errors + 83%|████████▎ | 430/520 [37:39<05:27, 3.64s/it] {'loss': 1.4647, 'grad_norm': 0.000572745235951222, 'learning_rate': 0.06897411034727215, 'epoch': 0.83} + 83%|████████▎ | 430/520 [37:39<05:27, 3.64s/it] 83%|████████▎ | 431/520 [37:43<05:24, 3.64s/it] {'loss': 1.6572, 'grad_norm': 0.0007967827490871081, 'learning_rate': 0.06748917574243089, 'epoch': 0.83} + 83%|████████▎ | 431/520 [37:43<05:24, 3.64s/it] 83%|████████▎ | 432/520 [37:46<05:20, 3.65s/it] {'loss': 1.3733, 'grad_norm': 0.0006501310248973304, 'learning_rate': 0.06601910326552998, 'epoch': 0.83} + 83%|████████▎ | 432/520 [37:46<05:20, 3.65s/it] 83%|████████▎ | 433/520 [37:50<05:17, 3.65s/it] {'loss': 1.5349, 'grad_norm': 0.0005914334810387453, 'learning_rate': 0.0645639500349669, 'epoch': 0.83} + 83%|████████▎ | 433/520 [37:50<05:17, 3.65s/it] 83%|████████▎ | 434/520 [37:53<05:13, 3.65s/it] {'loss': 1.2484, 'grad_norm': 0.0006918334134158682, 'learning_rate': 0.06312377258946437, 'epoch': 0.83} + 83%|████████▎ | 434/520 [37:53<05:13, 3.65s/it] 84%|████████▎ | 435/520 [37:57<05:10, 3.65s/it] {'loss': 1.5981, 'grad_norm': 0.0006920449833436292, 'learning_rate': 0.06169862688587413, 'epoch': 0.84} + 84%|████████▎ | 435/520 [37:57<05:10, 3.65s/it] 84%|████████▍ | 436/520 [38:01<05:07, 3.66s/it] {'loss': 1.3398, 'grad_norm': 0.0007644863384810821, 'learning_rate': 0.06028856829700258, 'epoch': 0.84} + 84%|████████▍ | 436/520 [38:01<05:07, 3.66s/it] 84%|████████▍ | 437/520 [38:04<05:04, 3.67s/it] {'loss': 1.6148, 'grad_norm': 0.0006076905718767369, 'learning_rate': 0.05889365160945912, 'epoch': 0.84} + 84%|████████▍ | 437/520 [38:04<05:04, 3.67s/it] 84%|████████▍ | 438/520 [38:08<05:01, 3.67s/it] {'loss': 1.3667, 'grad_norm': 0.0005869853243429444, 'learning_rate': 0.0575139310215276, 'epoch': 0.84} + 84%|████████▍ | 438/520 [38:08<05:01, 3.67s/it] 84%|████████▍ | 439/520 [38:12<04:57, 3.67s/it] {'loss': 1.6076, 'grad_norm': 0.0005800495600989468, 'learning_rate': 0.05614946014106084, 'epoch': 0.84} + 84%|████████▍ | 439/520 [38:12<04:57, 3.67s/it] 85%|████████▍ | 440/520 [38:15<04:53, 3.66s/it] {'loss': 1.4472, 'grad_norm': 0.000570071453334759, 'learning_rate': 0.05480029198339711, 'epoch': 0.85} + 85%|████████▍ | 440/520 [38:15<04:53, 3.66s/it] 85%|████████▍ | 441/520 [38:19<04:51, 3.69s/it] {'loss': 1.6511, 'grad_norm': 0.0005965979631130819, 'learning_rate': 0.05346647896930092, 'epoch': 0.85} + 85%|████████▍ | 441/520 [38:19<04:51, 3.69s/it] 85%|████████▌ | 442/520 [38:23<04:46, 3.67s/it] {'loss': 1.4962, 'grad_norm': 0.0006420073103604559, 'learning_rate': 0.052148072922925656, 'epoch': 0.85} + 85%|████████▌ | 442/520 [38:23<04:46, 3.67s/it] 85%|████████▌ | 443/520 [38:26<04:42, 3.67s/it] {'loss': 1.5243, 'grad_norm': 0.0005194017932413136, 'learning_rate': 0.05084512506980023, 'epoch': 0.85} + 85%|████████▌ | 443/520 [38:27<04:42, 3.67s/it] 85%|████████▌ | 444/520 [38:30<04:39, 3.67s/it] {'loss': 1.4735, 'grad_norm': 0.000494648977781324, 'learning_rate': 0.049557686034839156, 'epoch': 0.85} + 85%|████████▌ | 444/520 [38:30<04:39, 3.67s/it] 86%|████████▌ | 445/520 [38:34<04:34, 3.66s/it] {'loss': 1.3903, 'grad_norm': 0.0005485772306457254, 'learning_rate': 0.04828580584037491, 'epoch': 0.86} + 86%|████████▌ | 445/520 [38:34<04:34, 3.66s/it] 86%|████████▌ | 446/520 [38:37<04:31, 3.66s/it] {'loss': 1.7176, 'grad_norm': 0.0005861874221534984, 'learning_rate': 0.047029533904214584, 'epoch': 0.86} + 86%|████████▌ | 446/520 [38:37<04:31, 3.66s/it] 86%|████████▌ | 447/520 [38:41<04:27, 3.67s/it] {'loss': 1.5012, 'grad_norm': 0.0007034817700623852, 'learning_rate': 0.045788919037720185, 'epoch': 0.86} + 86%|████████▌ | 447/520 [38:41<04:27, 3.67s/it] 86%|████████▌ | 448/520 [38:45<04:23, 3.66s/it] {'loss': 1.4568, 'grad_norm': 0.0005361590493432624, 'learning_rate': 0.04456400944391144, 'epoch': 0.86} + 86%|████████▌ | 448/520 [38:45<04:23, 3.66s/it] 86%|████████▋ | 449/520 [38:48<04:20, 3.66s/it] {'loss': 1.7019, 'grad_norm': 0.0007193813914876828, 'learning_rate': 0.043354852715593584, 'epoch': 0.86} + 86%|████████▋ | 449/520 [38:48<04:20, 3.66s/it] 87%|████████▋ | 450/520 [38:52<04:15, 3.65s/it] {'loss': 1.547, 'grad_norm': 0.000595321296042126, 'learning_rate': 0.04216149583350753, 'epoch': 0.87} + 87%|████████▋ | 450/520 [38:52<04:15, 3.65s/it] 87%|████████▋ | 451/520 [38:56<04:12, 3.66s/it] {'loss': 1.5458, 'grad_norm': 0.0005940878658020129, 'learning_rate': 0.04098398516450508, 'epoch': 0.87} + 87%|████████▋ | 451/520 [38:56<04:12, 3.66s/it] 87%|████████▋ | 452/520 [39:00<04:10, 3.68s/it] {'loss': 1.7282, 'grad_norm': 0.0006741848015595106, 'learning_rate': 0.03982236645974709, 'epoch': 0.87} + 87%|████████▋ | 452/520 [39:00<04:10, 3.68s/it] 87%|████████▋ | 453/520 [39:03<04:06, 3.68s/it] {'loss': 1.7021, 'grad_norm': 0.0006094242803616008, 'learning_rate': 0.03867668485292565, 'epoch': 0.87} + 87%|████████▋ | 453/520 [39:03<04:06, 3.68s/it] 87%|████████▋ | 454/520 [39:07<04:03, 3.68s/it] {'loss': 1.4149, 'grad_norm': 0.0006648454082819014, 'learning_rate': 0.03754698485851071, 'epoch': 0.87} + 87%|████████▋ | 454/520 [39:07<04:03, 3.68s/it] 88%|████████▊ | 455/520 [39:11<04:00, 3.70s/it] {'loss': 1.5724, 'grad_norm': 0.0007136377672504805, 'learning_rate': 0.036433310370020705, 'epoch': 0.88} + 88%|████████▊ | 455/520 [39:11<04:00, 3.70s/it] 88%|████████▊ | 456/520 [39:14<03:59, 3.75s/it] {'loss': 1.4786, 'grad_norm': 0.0005168139955581727, 'learning_rate': 0.03533570465831652, 'epoch': 0.88} + 88%|████████▊ | 456/520 [39:14<03:59, 3.75s/it] 88%|████████▊ | 457/520 [39:18<03:57, 3.78s/it] {'loss': 1.7797, 'grad_norm': 0.000670309389172752, 'learning_rate': 0.03425421036992097, 'epoch': 0.88} + 88%|████████▊ | 457/520 [39:18<03:57, 3.78s/it] 88%|████████▊ | 458/520 [39:22<03:54, 3.79s/it] {'loss': 1.6616, 'grad_norm': 0.0008176414271791524, 'learning_rate': 0.03318886952536111, 'epoch': 0.88} + 88%|████████▊ | 458/520 [39:22<03:54, 3.79s/it] 88%|████████▊ | 459/520 [39:26<03:52, 3.81s/it] {'loss': 1.5484, 'grad_norm': 0.0005732189346214647, 'learning_rate': 0.032139723517535905, 'epoch': 0.88} + 88%|████████▊ | 459/520 [39:26<03:52, 3.81s/it] 88%|████████▊ | 460/520 [39:30<03:49, 3.83s/it] {'loss': 1.4029, 'grad_norm': 0.0007380716475620509, 'learning_rate': 0.03110681311010814, 'epoch': 0.88} + 88%|████████▊ | 460/520 [39:30<03:49, 3.83s/it] 89%|████████▊ | 461/520 [39:34<03:46, 3.84s/it] {'loss': 1.8503, 'grad_norm': 0.0005126316137112819, 'learning_rate': 0.030090178435920073, 'epoch': 0.89} + 89%|████████▊ | 461/520 [39:34<03:46, 3.84s/it] 89%|████████▉ | 462/520 [39:38<03:42, 3.84s/it] {'loss': 1.7767, 'grad_norm': 0.0010588704246131313, 'learning_rate': 0.029089858995434703, 'epoch': 0.89} + 89%|████████▉ | 462/520 [39:38<03:42, 3.84s/it] 89%|████████▉ | 463/520 [39:41<03:38, 3.84s/it] {'loss': 1.3649, 'grad_norm': 0.0006737473605438966, 'learning_rate': 0.02810589365520041, 'epoch': 0.89} + 89%|████████▉ | 463/520 [39:41<03:38, 3.84s/it] 89%|████████▉ | 464/520 [39:45<03:35, 3.85s/it] {'loss': 1.5631, 'grad_norm': 0.0005758287123759046, 'learning_rate': 0.02713832064634126, 'epoch': 0.89} + 89%|████████▉ | 464/520 [39:45<03:35, 3.85s/it] 89%|████████▉ | 465/520 [39:49<03:32, 3.86s/it] {'loss': 1.6723, 'grad_norm': 0.0009150102862248338, 'learning_rate': 0.02618717756307144, 'epoch': 0.89} + 89%|████████▉ | 465/520 [39:49<03:32, 3.86s/it] 90%|████████▉ | 466/520 [39:53<03:28, 3.86s/it] {'loss': 1.5466, 'grad_norm': 0.0004802189595014571, 'learning_rate': 0.02525250136123459, 'epoch': 0.9} + 90%|████████▉ | 466/520 [39:53<03:28, 3.86s/it] 90%|████████▉ | 467/520 [39:57<03:24, 3.86s/it] {'loss': 1.6306, 'grad_norm': 0.0006513778765036908, 'learning_rate': 0.02433432835686779, 'epoch': 0.9} + 90%|████████▉ | 467/520 [39:57<03:24, 3.86s/it] 90%|█████████ | 468/520 [40:01<03:20, 3.86s/it] {'loss': 1.5211, 'grad_norm': 0.0006309067484583225, 'learning_rate': 0.023432694224790735, 'epoch': 0.9} + 90%|█████████ | 468/520 [40:01<03:20, 3.86s/it] 90%|█████████ | 469/520 [40:05<03:16, 3.86s/it] {'loss': 1.5847, 'grad_norm': 0.0007452489872100019, 'learning_rate': 0.0225476339972193, 'epoch': 0.9} + 90%|█████████ | 469/520 [40:05<03:16, 3.86s/it] 90%|█████████ | 470/520 [40:08<03:12, 3.85s/it] {'loss': 1.4083, 'grad_norm': 0.0005026846563912535, 'learning_rate': 0.02167918206240494, 'epoch': 0.9} + 90%|█████████ | 470/520 [40:08<03:12, 3.85s/it] 91%|█████████ | 471/520 [40:12<03:08, 3.84s/it] {'loss': 1.4759, 'grad_norm': 0.0006476282188392479, 'learning_rate': 0.02082737216329793, 'epoch': 0.91} + 91%|█████████ | 471/520 [40:12<03:08, 3.84s/it] 91%|█████████ | 472/520 [40:16<03:04, 3.85s/it] {'loss': 1.409, 'grad_norm': 0.0006703799956698421, 'learning_rate': 0.019992237396236647, 'epoch': 0.91} + 91%|█████████ | 472/520 [40:16<03:04, 3.85s/it] 91%|█████████ | 473/520 [40:20<03:01, 3.86s/it] {'loss': 1.4984, 'grad_norm': 0.0008935463316066084, 'learning_rate': 0.019173810209661867, 'epoch': 0.91} + 91%|█████████ | 473/520 [40:20<03:01, 3.86s/it] 91%|█████████ | 474/520 [40:24<02:57, 3.86s/it] {'loss': 1.6952, 'grad_norm': 0.0006575247442486742, 'learning_rate': 0.018372122402855507, 'epoch': 0.91} + 91%|█████████ | 474/520 [40:24<02:57, 3.86s/it] 91%|█████████▏| 475/520 [40:28<02:53, 3.85s/it] {'loss': 1.5772, 'grad_norm': 0.000627797685625244, 'learning_rate': 0.01758720512470523, 'epoch': 0.91} + 91%|█████████▏| 475/520 [40:28<02:53, 3.85s/it] 92%|█████████▏| 476/520 [40:32<02:49, 3.84s/it] {'loss': 1.4639, 'grad_norm': 0.0007487429401743017, 'learning_rate': 0.016819088872494586, 'epoch': 0.92} + 92%|█████████▏| 476/520 [40:32<02:49, 3.84s/it] 92%|█████████▏| 477/520 [40:35<02:45, 3.84s/it] {'loss': 1.461, 'grad_norm': 0.0006116663223392737, 'learning_rate': 0.016067803490717552, 'epoch': 0.92} + 92%|█████████▏| 477/520 [40:35<02:45, 3.84s/it] 92%|█████████▏| 478/520 [40:39<02:41, 3.84s/it] {'loss': 1.4013, 'grad_norm': 0.0005715999945035081, 'learning_rate': 0.01533337816991931, 'epoch': 0.92} + 92%|█████████▏| 478/520 [40:39<02:41, 3.84s/it] 92%|█████████▏| 479/520 [40:43<02:37, 3.84s/it] {'loss': 1.6765, 'grad_norm': 0.0009468855161572117, 'learning_rate': 0.01461584144556175, 'epoch': 0.92} + 92%|█████████▏| 479/520 [40:43<02:37, 3.84s/it] 92%|█████████▏| 480/520 [40:47<02:32, 3.81s/it] {'loss': 1.6717, 'grad_norm': 0.0006193628391595606, 'learning_rate': 0.013915221196914967, 'epoch': 0.92} + 92%|█████████▏| 480/520 [40:47<02:32, 3.81s/it] 92%|█████████▎| 481/520 [40:51<02:27, 3.79s/it] {'loss': 1.6993, 'grad_norm': 0.0006307572198957515, 'learning_rate': 0.013231544645974069, 'epoch': 0.93} + 92%|█████████▎| 481/520 [40:51<02:27, 3.79s/it] 93%|█████████▎| 482/520 [40:54<02:23, 3.78s/it] {'loss': 1.7281, 'grad_norm': 0.0007890327306293562, 'learning_rate': 0.012564838356401476, 'epoch': 0.93} + 93%|█████████▎| 482/520 [40:54<02:23, 3.78s/it] 93%|█████████▎| 483/520 [40:58<02:23, 3.88s/it] {'loss': 1.5041, 'grad_norm': 0.0006128175230917648, 'learning_rate': 0.011915128232494493, 'epoch': 0.93} + 93%|█████████▎| 483/520 [40:58<02:23, 3.88s/it] 93%|█████████▎| 484/520 [41:03<02:22, 3.97s/it] {'loss': 1.5032, 'grad_norm': 0.0006626811152313032, 'learning_rate': 0.011282439518179371, 'epoch': 0.93} + 93%|█████████▎| 484/520 [41:03<02:22, 3.97s/it] 93%|█████████▎| 485/520 [41:07<02:19, 3.99s/it] {'loss': 1.4426, 'grad_norm': 0.0008242558584604612, 'learning_rate': 0.010666796796029988, 'epoch': 0.93} + 93%|█████████▎| 485/520 [41:07<02:19, 3.99s/it] 93%|█████████▎| 486/520 [41:10<02:14, 3.95s/it] {'loss': 1.5794, 'grad_norm': 0.0005800900259410677, 'learning_rate': 0.010068223986312958, 'epoch': 0.93} + 93%|█████████▎| 486/520 [41:11<02:14, 3.95s/it] 94%|█████████▎| 487/520 [41:14<02:09, 3.92s/it] {'loss': 1.4018, 'grad_norm': 0.0005754383563501386, 'learning_rate': 0.009486744346058234, 'epoch': 0.94} + 94%|█████████▎| 487/520 [41:14<02:09, 3.92s/it] 94%|█████████▍| 488/520 [41:18<02:03, 3.87s/it] {'loss': 1.3586, 'grad_norm': 0.0008521702696073009, 'learning_rate': 0.008922380468155277, 'epoch': 0.94} + 94%|█████████▍| 488/520 [41:18<02:03, 3.87s/it] 94%|█████████▍| 489/520 [41:22<01:59, 3.85s/it] {'loss': 1.6696, 'grad_norm': 0.0005561122472092477, 'learning_rate': 0.008375154280475555, 'epoch': 0.94} + 94%|█████████▍| 489/520 [41:22<01:59, 3.85s/it] 94%|█████████▍| 490/520 [41:26<01:54, 3.83s/it] {'loss': 1.5078, 'grad_norm': 0.0005856078227440827, 'learning_rate': 0.007845087045020278, 'epoch': 0.94} + 94%|█████████▍| 490/520 [41:26<01:54, 3.83s/it] 94%|█████████▍| 491/520 [41:29<01:49, 3.78s/it] {'loss': 1.4442, 'grad_norm': 0.000857306422669658, 'learning_rate': 0.007332199357094405, 'epoch': 0.94} + 94%|█████████▍| 491/520 [41:29<01:49, 3.78s/it] 95%|█████████▍| 492/520 [41:33<01:44, 3.75s/it] {'loss': 1.592, 'grad_norm': 0.0007014606764035637, 'learning_rate': 0.006836511144506391, 'epoch': 0.95} + 95%|█████████▍| 492/520 [41:33<01:44, 3.75s/it] 95%|█████████▍| 493/520 [41:37<01:40, 3.73s/it] {'loss': 1.7708, 'grad_norm': 0.0006591314807187436, 'learning_rate': 0.006358041666793851, 'epoch': 0.95} + 95%|█████████▍| 493/520 [41:37<01:40, 3.73s/it] 95%|█████████▌| 494/520 [41:40<01:36, 3.71s/it] {'loss': 1.5002, 'grad_norm': 0.000839550319874172, 'learning_rate': 0.005896809514475509, 'epoch': 0.95} + 95%|█████████▌| 494/520 [41:40<01:36, 3.71s/it] 95%|█████████▌| 495/520 [41:44<01:32, 3.72s/it] {'loss': 1.4628, 'grad_norm': 0.0006070877972945726, 'learning_rate': 0.0054528326083283785, 'epoch': 0.95} + 95%|█████████▌| 495/520 [41:44<01:32, 3.72s/it] 95%|█████████▌| 496/520 [41:48<01:29, 3.71s/it] {'loss': 1.3902, 'grad_norm': 0.0006128957868752219, 'learning_rate': 0.005026128198692165, 'epoch': 0.95} + 95%|█████████▌| 496/520 [41:48<01:29, 3.71s/it] 96%|█████████▌| 497/520 [41:51<01:25, 3.71s/it] {'loss': 1.5997, 'grad_norm': 0.0005705163557937471, 'learning_rate': 0.004616712864798306, 'epoch': 0.96} + 96%|█████████▌| 497/520 [41:51<01:25, 3.71s/it] 96%|█████████▌| 498/520 [41:55<01:21, 3.70s/it] {'loss': 1.4625, 'grad_norm': 0.0007098838167978247, 'learning_rate': 0.0042246025141262356, 'epoch': 0.96} + 96%|█████████▌| 498/520 [41:55<01:21, 3.70s/it] 96%|█████████▌| 499/520 [41:59<01:17, 3.70s/it] {'loss': 1.7624, 'grad_norm': 0.0006015592491550736, 'learning_rate': 0.003849812381785328, 'epoch': 0.96} + 96%|█████████▌| 499/520 [41:59<01:17, 3.70s/it] 96%|█████████▌| 500/520 [42:03<01:13, 3.68s/it] {'loss': 1.6121, 'grad_norm': 0.0008720800591614745, 'learning_rate': 0.0034923570299225716, 'epoch': 0.96} + 96%|█████████▌| 500/520 [42:03<01:13, 3.68s/it] 96%|█████████▋| 501/520 [42:06<01:09, 3.68s/it] {'loss': 1.6755, 'grad_norm': 0.0010608917435951577, 'learning_rate': 0.0031522503471571707, 'epoch': 0.96} + 96%|█████████▋| 501/520 [42:06<01:09, 3.68s/it] 97%|█████████▋| 502/520 [42:10<01:06, 3.68s/it] {'loss': 1.4897, 'grad_norm': 0.0006351834456728442, 'learning_rate': 0.0028295055480408284, 'epoch': 0.97} + 97%|█████████▋| 502/520 [42:10<01:06, 3.68s/it] 97%|█████████▋| 503/520 [42:14<01:02, 3.67s/it] {'loss': 1.6311, 'grad_norm': 0.0006994689221392009, 'learning_rate': 0.0025241351725441064, 'epoch': 0.97} + 97%|█████████▋| 503/520 [42:14<01:02, 3.67s/it] 97%|█████████▋| 504/520 [42:17<00:58, 3.67s/it] {'loss': 1.5015, 'grad_norm': 0.0006277446883706768, 'learning_rate': 0.002236151085569366, 'epoch': 0.97} + 97%|█████████▋| 504/520 [42:17<00:58, 3.67s/it] 97%|█████████▋| 505/520 [42:21<00:55, 3.67s/it] {'loss': 1.5461, 'grad_norm': 0.0005477353007678515, 'learning_rate': 0.001965564476489784, 'epoch': 0.97} + 97%|█████████▋| 505/520 [42:21<00:55, 3.67s/it] 97%|█████████▋| 506/520 [42:25<00:51, 3.67s/it] {'loss': 1.4422, 'grad_norm': 0.0006524065429864107, 'learning_rate': 0.0017123858587145047, 'epoch': 0.97} + 97%|█████████▋| 506/520 [42:25<00:51, 3.67s/it] 98%|█████████▊| 507/520 [42:28<00:47, 3.67s/it] {'loss': 1.8218, 'grad_norm': 0.0007749139407754404, 'learning_rate': 0.001476625069280213, 'epoch': 0.97} + 98%|█████████▊| 507/520 [42:28<00:47, 3.67s/it] 98%|█████████▊| 508/520 [42:32<00:43, 3.65s/it] {'loss': 1.5827, 'grad_norm': 0.0005959281101268578, 'learning_rate': 0.0012582912684689419, 'epoch': 0.98} + 98%|█████████▊| 508/520 [42:32<00:43, 3.65s/it] 98%|█████████▊| 509/520 [42:35<00:40, 3.65s/it] {'loss': 1.5273, 'grad_norm': 0.0007055702875626032, 'learning_rate': 0.0010573929394520065, 'epoch': 0.98} + 98%|█████████▊| 509/520 [42:35<00:40, 3.65s/it] 98%|█████████▊| 510/520 [42:39<00:37, 3.71s/it] {'loss': 1.4841, 'grad_norm': 0.0010404782237270103, 'learning_rate': 0.0008739378879606685, 'epoch': 0.98} + 98%|█████████▊| 510/520 [42:39<00:37, 3.71s/it] 98%|█████████▊| 511/520 [42:43<00:33, 3.74s/it] {'loss': 1.4482, 'grad_norm': 0.0006933077467143631, 'learning_rate': 0.000707933241982528, 'epoch': 0.98} + 98%|█████████▊| 511/520 [42:43<00:33, 3.74s/it] 98%|█████████▊| 512/520 [42:47<00:30, 3.78s/it] {'loss': 1.3272, 'grad_norm': 0.0005864395832574333, 'learning_rate': 0.000559385451484945, 'epoch': 0.98} + 98%|█████████▊| 512/520 [42:47<00:30, 3.78s/it] 99%|█████████▊| 513/520 [42:51<00:26, 3.76s/it] {'loss': 1.5646, 'grad_norm': 0.000852559666321359, 'learning_rate': 0.0004283002881639908, 'epoch': 0.99} + 99%|█████████▊| 513/520 [42:51<00:26, 3.76s/it] 99%|█████████▉| 514/520 [42:54<00:22, 3.72s/it] {'loss': 1.5267, 'grad_norm': 0.0006964778245995108, 'learning_rate': 0.0003146828452206263, 'epoch': 0.99} + 99%|█████████▉| 514/520 [42:54<00:22, 3.72s/it] 99%|█████████▉| 515/520 [42:58<00:18, 3.70s/it] {'loss': 1.618, 'grad_norm': 0.0009147872933850774, 'learning_rate': 0.00021853753716256086, 'epoch': 0.99} + 99%|█████████▉| 515/520 [42:58<00:18, 3.70s/it] 99%|█████████▉| 516/520 [43:02<00:14, 3.69s/it] {'loss': 1.4576, 'grad_norm': 0.000974389632120011, 'learning_rate': 0.00013986809963268955, 'epoch': 0.99} + 99%|█████████▉| 516/520 [43:02<00:14, 3.69s/it] 99%|█████████▉| 517/520 [43:05<00:10, 3.66s/it] {'loss': 1.7225, 'grad_norm': 0.0006037325930229138, 'learning_rate': 7.867758926410895e-05, 'epoch': 0.99} + 99%|█████████▉| 517/520 [43:05<00:10, 3.66s/it] 100%|█████████▉| 518/520 [43:09<00:07, 3.63s/it] {'loss': 1.5162, 'grad_norm': 0.0005131977384595448, 'learning_rate': 3.496838356131238e-05, 'epoch': 1.0} + 100%|█████████▉| 518/520 [43:09<00:07, 3.63s/it] 100%|█████████▉| 519/520 [43:12<00:03, 3.62s/it] {'loss': 1.6666, 'grad_norm': 0.0006127284589347593, 'learning_rate': 8.742180807813637e-06, 'epoch': 1.0} + 100%|█████████▉| 519/520 [43:12<00:03, 3.62s/it] 100%|██████████| 520/520 [43:17<00:00, 3.89s/it] {'loss': 1.7659, 'grad_norm': 0.0006225841094729328, 'learning_rate': 0.0, 'epoch': 1.0} + 100%|██████████| 520/520 [43:17<00:00, 3.89s/it] {'train_runtime': 2597.4202, 'train_samples_per_second': 25.613, 'train_steps_per_second': 0.2, 'train_loss': 1.6619158375721712, 'epoch': 1.0} + 100%|██████████| 520/520 [43:17<00:00, 3.89s/it] 100%|██████████| 520/520 [43:17<00:00, 5.00s/it] +[2025-10-12 09:15:31,866] [INFO] [launch.py:348:main] Process 531658 exits successfully. +[2025-10-12 09:15:31,866] [INFO] [launch.py:348:main] Process 531654 exits successfully. +[2025-10-12 09:15:31,867] [INFO] [launch.py:348:main] Process 531656 exits successfully. +[2025-10-12 09:15:31,867] [INFO] [launch.py:348:main] Process 531657 exits successfully. +[2025-10-12 09:15:32,869] [INFO] [launch.py:348:main] Process 531652 exits successfully. +[2025-10-12 09:15:32,869] [INFO] [launch.py:348:main] Process 531653 exits successfully. +[2025-10-12 09:15:32,869] [INFO] [launch.py:348:main] Process 531655 exits successfully. +[2025-10-12 09:15:35,872] [INFO] [launch.py:348:main] Process 531651 exits successfully. +==== EXPERIMENT COMPLETED: qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_9e-1_connector-3.0_0.5_9e-1_ablation ==== +Log File: qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_9e-1_connector-3.0_0.5_9e-1_ablation_20251012_081416.log +Timestamp: 2025-10-12 09:15:38 +===================================== diff --git a/logs_oct11/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_9e-2_connector-3.0_0.5_9e-2_ablation_20251012_142902.log b/logs_oct11/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_9e-2_connector-3.0_0.5_9e-2_ablation_20251012_142902.log new file mode 100644 index 0000000000000000000000000000000000000000..96d0b9d30ae9974e822268b3ed87e205a987d0e6 --- /dev/null +++ b/logs_oct11/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_9e-2_connector-3.0_0.5_9e-2_ablation_20251012_142902.log @@ -0,0 +1,2312 @@ +==== STARTING EXPERIMENT: qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_9e-2_connector-3.0_0.5_9e-2_ablation ==== +Log File: qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_9e-2_connector-3.0_0.5_9e-2_ablation_20251012_142902.log +Timestamp: 2025-10-12 14:29:02 +===================================== +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 14:29:05,049] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 14:29:07,758] [WARNING] [runner.py:202:fetch_hostfile] Unable to find hostfile, will proceed with training with local resources only. +[2025-10-12 14:29:07,760] [INFO] [runner.py:568:main] cmd = /opt/conda/envs/tinyllava/bin/python3.10 -u -m deepspeed.launcher.launch --world_info=eyJsb2NhbGhvc3QiOiBbMCwgMSwgMiwgMywgNCwgNSwgNiwgN119 --master_addr=127.0.0.1 --master_port=29501 --enable_each_rank_log=None tinyllava/train/train.py --deepspeed ./scripts/zero3.json --data_path /s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json --image_folder /s3-code/ywang29/datasets/tinyllava --is_multimodal True --conv_version qwen2_base --model_name_or_path Qwen/Qwen2.5-0.5B --vision_tower google/siglip-so400m-patch14-384 --vision_tower2 --connector_type mlp2x_gelu --mm_vision_select_layer -2 --image_aspect_ratio square --attn_implementation flash_attention_2 --bf16 True --training_recipe common --tune_type_llm full --tune_type_vision_tower frozen --tune_vision_tower_from_layer 0 --tune_type_connector full --group_by_modality_length True --pretrained_model_path /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain --output_dir /nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_9e-2_connector-3.0_0.5_9e-2_ablation --num_train_epochs 1 --per_device_train_batch_size 4 --per_device_eval_batch_size 4 --gradient_accumulation_steps 4 --evaluation_strategy no --learning_rate 9e-2 --weight_decay 0. --warmup_ratio 0.03 --lr_scheduler_type cosine --logging_steps 1 --tf32 False --model_max_length 2048 --gradient_checkpointing True --dataloader_num_workers 8 --lazy_preprocess True --report_to tensorboard --tokenizer_use_fast False --run_name tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune --subnet_mode_text both --subnet_type_text None --mask_type_text soft --init_mean_text 3.0 --temperature_attn_text 0.5 --temperature_mlp_text 0.5 --backward_type_text normal --masked_layers_text all --subnet_mode_vision both --subnet_type_vision None --mask_type_vision soft --init_mean_vision 3.0 --temperature_attn_vision 0.5 --temperature_mlp_vision 0.5 --backward_type_vision normal --masked_layers_vision all --subnet_type_connector global --mask_type_connector soft --init_mean_connector 3.0 --temperature_connector 0.5 --backward_type_connector normal --mm_projector_lr 9e-2 --seed 42 --mask_model llm-connector --save_strategy steps --save_steps 50000 --save_total_limit 1 --train_data_ratio 0.1 +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 14:29:10,335] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 14:29:11,416] [INFO] [launch.py:138:main] 0 NCCL_VERSION=2.21.5 +[2025-10-12 14:29:11,416] [INFO] [launch.py:138:main] 0 NCCL_SOCKET_IFNAME=eth +[2025-10-12 14:29:11,416] [INFO] [launch.py:145:main] WORLD INFO DICT: {'localhost': [0, 1, 2, 3, 4, 5, 6, 7]} +[2025-10-12 14:29:11,416] [INFO] [launch.py:151:main] nnodes=1, num_local_procs=8, node_rank=0 +[2025-10-12 14:29:11,417] [INFO] [launch.py:162:main] global_rank_mapping=defaultdict(, {'localhost': [0, 1, 2, 3, 4, 5, 6, 7]}) +[2025-10-12 14:29:11,417] [INFO] [launch.py:163:main] dist_world_size=8 +[2025-10-12 14:29:11,417] [INFO] [launch.py:165:main] Setting CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 +[2025-10-12 14:29:11,419] [INFO] [launch.py:253:main] process 782087 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=0', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_9e-2_connector-3.0_0.5_9e-2_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '9e-2', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '9e-2', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 14:29:11,421] [INFO] [launch.py:253:main] process 782088 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=1', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_9e-2_connector-3.0_0.5_9e-2_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '9e-2', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '9e-2', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 14:29:11,423] [INFO] [launch.py:253:main] process 782089 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=2', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_9e-2_connector-3.0_0.5_9e-2_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '9e-2', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '9e-2', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 14:29:11,424] [INFO] [launch.py:253:main] process 782090 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=3', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_9e-2_connector-3.0_0.5_9e-2_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '9e-2', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '9e-2', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 14:29:11,426] [INFO] [launch.py:253:main] process 782091 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=4', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_9e-2_connector-3.0_0.5_9e-2_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '9e-2', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '9e-2', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 14:29:11,428] [INFO] [launch.py:253:main] process 782092 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=5', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_9e-2_connector-3.0_0.5_9e-2_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '9e-2', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '9e-2', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 14:29:11,430] [INFO] [launch.py:253:main] process 782093 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=6', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_9e-2_connector-3.0_0.5_9e-2_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '9e-2', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '9e-2', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 14:29:11,432] [INFO] [launch.py:253:main] process 782094 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=7', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_9e-2_connector-3.0_0.5_9e-2_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '9e-2', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.5', '--temperature_mlp_text', '0.5', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.5', '--temperature_mlp_vision', '0.5', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.5', '--backward_type_connector', 'normal', '--mm_projector_lr', '9e-2', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 14:29:18,117] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 14:29:18,345] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 14:29:18,345] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 14:29:18,375] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 14:29:18,411] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 14:29:18,411] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 14:29:18,443] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 14:29:18,443] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 14:29:18,538] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 14:29:18,753] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 14:29:18,753] [INFO] [comm.py:668:init_distributed] Initializing TorchBackend in DeepSpeed with backend nccl +[2025-10-12 14:29:18,757] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 14:29:18,784] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 14:29:18,822] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 14:29:18,825] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 14:29:18,854] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 14:29:18,856] [INFO] [comm.py:637:init_distributed] cdb=None +Apply masks for the following modules: ['llm', 'connector'] +Apply masks for the following modules: ['llm', 'connector'] +{'llm': {'model_name_or_path': 'Qwen/Qwen2.5-0.5B', 'cache_dir': None, 'attn_implementation': 'flash_attention_2', 'subnet_mode': 'both', 'subnet_type': 'None', 'sparsity_attn': None, 'sparsity_mlp': None, 'threshold_attn': None, 'threshold_mlp': None, 'temperature_attn': 0.5, 'temperature_mlp': 0.5, 'masked_layers': 'all', 'mask_type': 'soft', 'backward_type': 'normal'}, 'vision_tower': {'model_name_or_path': 'google/siglip-so400m-patch14-384'}, 'connector': {'connector_type': 'mlp2x_gelu', 'subnet_type': 'global', 'threshold': None, 'sparsity': None, 'temperature': 0.5, 'mask_type': 'soft', 'backward_type': 'normal'}} +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +TinyLlavaConfig { + "backward_type_connector": "normal", + "cache_dir": null, + "connector_type": "mlp2x_gelu", + "hidden_size": 896, + "ignore_index": -100, + "image_aspect_ratio": "square", + "image_token_index": -200, + "llm_model_name_or_path": "Qwen/Qwen2.5-0.5B", + "mask_model": [ + "llm", + "connector" + ], + "mask_type_connector": "soft", + "model_type": "tinyllava", + "num_queries": 128, + "num_resampler_layers": 3, + "pad_token": null, + "resampler_hidden_size": 768, + "sparsity_connector": null, + "subnet_type_connector": "global", + "temperature_connector": 0.5, + "text_config": { + "_name_or_path": "Qwen/Qwen2.5-0.5B", + "architectures": [ + "Qwen2ForCausalLM" + ], + "backward_type": "normal", + "bos_token_id": 151643, + "eos_token_id": 151643, + "hidden_size": 896, + "intermediate_size": 4864, + "mask_type": "soft", + "masked_layers": "all", + "max_position_embeddings": 32768, + "max_window_layers": 24, + "model_type": "qwen2", + "num_attention_heads": 14, + "num_hidden_layers": 24, + "num_key_value_heads": 2, + "rope_theta": 1000000.0, + "sliding_window": 32768, + "subnet_mode": "both", + "subnet_type": "None", + "temperature_attn": 0.5, + "temperature_mlp": 0.5, + "tie_word_embeddings": true, + "torch_dtype": "bfloat16", + "use_mrope": false, + "use_sliding_window": false, + "vocab_size": 151936 + }, + "threshold_connector": null, + "tokenizer_model_max_length": 2048, + "tokenizer_name_or_path": "Qwen/Qwen2.5-0.5B", + "tokenizer_padding_side": "right", + "tokenizer_use_fast": false, + "transformers_version": "4.40.1", + "tune_type_connector": "frozen", + "tune_type_llm": "frozen", + "tune_type_vision_tower": "frozen", + "tune_vision_tower_from_layer": -1, + "use_cache": false, + "vision_config": { + "hidden_act": "gelu_pytorch_tanh", + "hidden_size": 1152, + "image_size": 384, + "intermediate_size": 4304, + "layer_norm_eps": 1e-06, + "model_name_or_path": "google/siglip-so400m-patch14-384", + "model_name_or_path2": "", + "model_type": "siglip_vision_model", + "num_attention_heads": 16, + "num_hidden_layers": 27, + "patch_size": 14 + }, + "vision_feature_layer": -2, + "vision_feature_select_strategy": "patch", + "vision_hidden_size": 1152, + "vision_model_name_or_path": "google/siglip-so400m-patch14-384", + "vision_model_name_or_path2": "", + "vocab_size": 151936 +} + +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test2-worker-0:782087:782087 [0] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:782087:782087 [0] NCCL INFO Bootstrap : Using eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:782087:782087 [0] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test2-worker-0:782087:782087 [0] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test2-worker-0:782087:782087 [0] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test2-worker-0:782087:782087 [0] NCCL INFO cudaDriverVersion 12040 +NCCL version 2.21.5+cuda12.1 +ywang29-vrdb-test2-worker-0:782091:782091 [4] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test2-worker-0:782091:782091 [4] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:782091:782091 [4] NCCL INFO Bootstrap : Using eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:782091:782091 [4] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test2-worker-0:782091:782091 [4] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test2-worker-0:782091:782091 [4] NCCL INFO NET/Plugin: Using internal network plugin. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test2-worker-0:782093:782093 [6] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test2-worker-0:782093:782093 [6] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:782093:782093 [6] NCCL INFO Bootstrap : Using eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:782093:782093 [6] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test2-worker-0:782093:782093 [6] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test2-worker-0:782093:782093 [6] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test2-worker-0:782087:783669 [0] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:782087:783669 [0] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test2-worker-0:782087:783669 [0] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:782087:783669 [0] NCCL INFO NET/Socket : Using [0]eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:782087:783669 [0] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test2-worker-0:782087:783669 [0] NCCL INFO Using network Socket +ywang29-vrdb-test2-worker-0:782091:783670 [4] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:782091:783670 [4] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test2-worker-0:782091:783670 [4] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:782091:783670 [4] NCCL INFO NET/Socket : Using [0]eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:782091:783670 [4] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test2-worker-0:782091:783670 [4] NCCL INFO Using network Socket +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test2-worker-0:782090:782090 [3] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test2-worker-0:782090:782090 [3] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:782090:782090 [3] NCCL INFO Bootstrap : Using eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:782090:782090 [3] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test2-worker-0:782090:782090 [3] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test2-worker-0:782090:782090 [3] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test2-worker-0:782093:783687 [6] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:782093:783687 [6] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test2-worker-0:782093:783687 [6] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:782093:783687 [6] NCCL INFO NET/Socket : Using [0]eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:782093:783687 [6] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test2-worker-0:782093:783687 [6] NCCL INFO Using network Socket +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test2-worker-0:782092:782092 [5] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test2-worker-0:782092:782092 [5] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:782092:782092 [5] NCCL INFO Bootstrap : Using eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:782089:782089 [2] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test2-worker-0:782089:782089 [2] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:782089:782089 [2] NCCL INFO Bootstrap : Using eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:782092:782092 [5] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test2-worker-0:782092:782092 [5] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test2-worker-0:782092:782092 [5] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test2-worker-0:782089:782089 [2] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test2-worker-0:782089:782089 [2] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test2-worker-0:782089:782089 [2] NCCL INFO NET/Plugin: Using internal network plugin. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test2-worker-0:782090:783688 [3] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:782090:783688 [3] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test2-worker-0:782090:783688 [3] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:782090:783688 [3] NCCL INFO NET/Socket : Using [0]eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:782090:783688 [3] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test2-worker-0:782090:783688 [3] NCCL INFO Using network Socket +ywang29-vrdb-test2-worker-0:782088:782088 [1] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test2-worker-0:782088:782088 [1] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:782088:782088 [1] NCCL INFO Bootstrap : Using eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:782088:782088 [1] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test2-worker-0:782088:782088 [1] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test2-worker-0:782088:782088 [1] NCCL INFO NET/Plugin: Using internal network plugin. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test2-worker-0:782094:782094 [7] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test2-worker-0:782094:782094 [7] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:782094:782094 [7] NCCL INFO Bootstrap : Using eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:782094:782094 [7] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test2-worker-0:782094:782094 [7] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test2-worker-0:782094:782094 [7] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test2-worker-0:782092:783689 [5] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:782089:783690 [2] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:782092:783689 [5] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test2-worker-0:782089:783690 [2] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test2-worker-0:782089:783690 [2] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:782092:783689 [5] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:782089:783690 [2] NCCL INFO NET/Socket : Using [0]eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:782092:783689 [5] NCCL INFO NET/Socket : Using [0]eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:782089:783690 [2] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test2-worker-0:782089:783690 [2] NCCL INFO Using network Socket +ywang29-vrdb-test2-worker-0:782092:783689 [5] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test2-worker-0:782092:783689 [5] NCCL INFO Using network Socket +ywang29-vrdb-test2-worker-0:782088:783691 [1] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:782088:783691 [1] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test2-worker-0:782088:783691 [1] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:782088:783691 [1] NCCL INFO NET/Socket : Using [0]eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:782088:783691 [1] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test2-worker-0:782088:783691 [1] NCCL INFO Using network Socket +ywang29-vrdb-test2-worker-0:782094:783692 [7] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:782094:783692 [7] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test2-worker-0:782094:783692 [7] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test2-worker-0:782094:783692 [7] NCCL INFO NET/Socket : Using [0]eth0:10.200.152.48<0> +ywang29-vrdb-test2-worker-0:782094:783692 [7] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test2-worker-0:782094:783692 [7] NCCL INFO Using network Socket +ywang29-vrdb-test2-worker-0:782088:783691 [1] NCCL INFO ncclCommInitRank comm 0x562d761a6b30 rank 1 nranks 8 cudaDev 1 nvmlDev 1 busId 101d0 commId 0x5d3556e4537bc329 - Init START +ywang29-vrdb-test2-worker-0:782089:783690 [2] NCCL INFO ncclCommInitRank comm 0x555629a50910 rank 2 nranks 8 cudaDev 2 nvmlDev 2 busId 201c0 commId 0x5d3556e4537bc329 - Init START +ywang29-vrdb-test2-worker-0:782094:783692 [7] NCCL INFO ncclCommInitRank comm 0x55b3c9826fd0 rank 7 nranks 8 cudaDev 7 nvmlDev 7 busId a01d0 commId 0x5d3556e4537bc329 - Init START +ywang29-vrdb-test2-worker-0:782090:783688 [3] NCCL INFO ncclCommInitRank comm 0x55995d49b4f0 rank 3 nranks 8 cudaDev 3 nvmlDev 3 busId 201d0 commId 0x5d3556e4537bc329 - Init START +ywang29-vrdb-test2-worker-0:782092:783689 [5] NCCL INFO ncclCommInitRank comm 0x5616d445c0d0 rank 5 nranks 8 cudaDev 5 nvmlDev 5 busId 901d0 commId 0x5d3556e4537bc329 - Init START +ywang29-vrdb-test2-worker-0:782087:783669 [0] NCCL INFO ncclCommInitRank comm 0x55a04621adb0 rank 0 nranks 8 cudaDev 0 nvmlDev 0 busId 101c0 commId 0x5d3556e4537bc329 - Init START +ywang29-vrdb-test2-worker-0:782091:783670 [4] NCCL INFO ncclCommInitRank comm 0x56275fc79830 rank 4 nranks 8 cudaDev 4 nvmlDev 4 busId 901c0 commId 0x5d3556e4537bc329 - Init START +ywang29-vrdb-test2-worker-0:782093:783687 [6] NCCL INFO ncclCommInitRank comm 0x557eff3d5e00 rank 6 nranks 8 cudaDev 6 nvmlDev 6 busId a01c0 commId 0x5d3556e4537bc329 - Init START +ywang29-vrdb-test2-worker-0:782089:783690 [2] NCCL INFO Setting affinity for GPU 2 to ff,ffff0000,00ffffff +ywang29-vrdb-test2-worker-0:782089:783690 [2] NCCL INFO NVLS multicast support is not available on dev 2 +ywang29-vrdb-test2-worker-0:782092:783689 [5] NCCL INFO Setting affinity for GPU 5 to ffffff00,0000ffff,ff000000 +ywang29-vrdb-test2-worker-0:782092:783689 [5] NCCL INFO NVLS multicast support is not available on dev 5 +ywang29-vrdb-test2-worker-0:782091:783670 [4] NCCL INFO Setting affinity for GPU 4 to ffffff00,0000ffff,ff000000 +ywang29-vrdb-test2-worker-0:782091:783670 [4] NCCL INFO NVLS multicast support is not available on dev 4 +ywang29-vrdb-test2-worker-0:782088:783691 [1] NCCL INFO Setting affinity for GPU 1 to ff,ffff0000,00ffffff +ywang29-vrdb-test2-worker-0:782088:783691 [1] NCCL INFO NVLS multicast support is not available on dev 1 +ywang29-vrdb-test2-worker-0:782093:783687 [6] NCCL INFO Setting affinity for GPU 6 to ffffff00,0000ffff,ff000000 +ywang29-vrdb-test2-worker-0:782093:783687 [6] NCCL INFO NVLS multicast support is not available on dev 6 +ywang29-vrdb-test2-worker-0:782090:783688 [3] NCCL INFO Setting affinity for GPU 3 to ff,ffff0000,00ffffff +ywang29-vrdb-test2-worker-0:782090:783688 [3] NCCL INFO NVLS multicast support is not available on dev 3 +ywang29-vrdb-test2-worker-0:782094:783692 [7] NCCL INFO Setting affinity for GPU 7 to ffffff00,0000ffff,ff000000 +ywang29-vrdb-test2-worker-0:782087:783669 [0] NCCL INFO Setting affinity for GPU 0 to ff,ffff0000,00ffffff +ywang29-vrdb-test2-worker-0:782087:783669 [0] NCCL INFO NVLS multicast support is not available on dev 0 +ywang29-vrdb-test2-worker-0:782094:783692 [7] NCCL INFO NVLS multicast support is not available on dev 7 +ywang29-vrdb-test2-worker-0:782093:783687 [6] NCCL INFO comm 0x557eff3d5e00 rank 6 nRanks 8 nNodes 1 localRanks 8 localRank 6 MNNVL 0 +ywang29-vrdb-test2-worker-0:782091:783670 [4] NCCL INFO comm 0x56275fc79830 rank 4 nRanks 8 nNodes 1 localRanks 8 localRank 4 MNNVL 0 +ywang29-vrdb-test2-worker-0:782090:783688 [3] NCCL INFO comm 0x55995d49b4f0 rank 3 nRanks 8 nNodes 1 localRanks 8 localRank 3 MNNVL 0 +ywang29-vrdb-test2-worker-0:782092:783689 [5] NCCL INFO comm 0x5616d445c0d0 rank 5 nRanks 8 nNodes 1 localRanks 8 localRank 5 MNNVL 0 +ywang29-vrdb-test2-worker-0:782089:783690 [2] NCCL INFO comm 0x555629a50910 rank 2 nRanks 8 nNodes 1 localRanks 8 localRank 2 MNNVL 0 +ywang29-vrdb-test2-worker-0:782088:783691 [1] NCCL INFO comm 0x562d761a6b30 rank 1 nRanks 8 nNodes 1 localRanks 8 localRank 1 MNNVL 0 +ywang29-vrdb-test2-worker-0:782087:783669 [0] NCCL INFO comm 0x55a04621adb0 rank 0 nRanks 8 nNodes 1 localRanks 8 localRank 0 MNNVL 0 +ywang29-vrdb-test2-worker-0:782094:783692 [7] NCCL INFO comm 0x55b3c9826fd0 rank 7 nRanks 8 nNodes 1 localRanks 8 localRank 7 MNNVL 0 +ywang29-vrdb-test2-worker-0:782093:783687 [6] NCCL INFO Trees [0] 7/-1/-1->6->5 [1] 7/-1/-1->6->5 [2] 7/-1/-1->6->5 [3] 7/-1/-1->6->5 [4] 7/-1/-1->6->5 [5] 7/-1/-1->6->5 [6] 7/-1/-1->6->5 [7] 7/-1/-1->6->5 [8] 7/-1/-1->6->5 [9] 7/-1/-1->6->5 [10] 7/-1/-1->6->5 [11] 7/-1/-1->6->5 [12] 7/-1/-1->6->5 [13] 7/-1/-1->6->5 [14] 7/-1/-1->6->5 [15] 7/-1/-1->6->5 [16] 7/-1/-1->6->5 [17] 7/-1/-1->6->5 [18] 7/-1/-1->6->5 [19] 7/-1/-1->6->5 [20] 7/-1/-1->6->5 [21] 7/-1/-1->6->5 [22] 7/-1/-1->6->5 [23] 7/-1/-1->6->5 +ywang29-vrdb-test2-worker-0:782087:783669 [0] NCCL INFO Channel 00/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:782087:783669 [0] NCCL INFO Channel 01/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:782093:783687 [6] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:782087:783669 [0] NCCL INFO Channel 02/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:782087:783669 [0] NCCL INFO Channel 03/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:782091:783670 [4] NCCL INFO Trees [0] 5/-1/-1->4->3 [1] 5/-1/-1->4->3 [2] 5/-1/-1->4->3 [3] 5/-1/-1->4->3 [4] 5/-1/-1->4->3 [5] 5/-1/-1->4->3 [6] 5/-1/-1->4->3 [7] 5/-1/-1->4->3 [8] 5/-1/-1->4->3 [9] 5/-1/-1->4->3 [10] 5/-1/-1->4->3 [11] 5/-1/-1->4->3 [12] 5/-1/-1->4->3 [13] 5/-1/-1->4->3 [14] 5/-1/-1->4->3 [15] 5/-1/-1->4->3 [16] 5/-1/-1->4->3 [17] 5/-1/-1->4->3 [18] 5/-1/-1->4->3 [19] 5/-1/-1->4->3 [20] 5/-1/-1->4->3 [21] 5/-1/-1->4->3 [22] 5/-1/-1->4->3 [23] 5/-1/-1->4->3 +ywang29-vrdb-test2-worker-0:782087:783669 [0] NCCL INFO Channel 04/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:782087:783669 [0] NCCL INFO Channel 05/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:782090:783688 [3] NCCL INFO Trees [0] 4/-1/-1->3->2 [1] 4/-1/-1->3->2 [2] 4/-1/-1->3->2 [3] 4/-1/-1->3->2 [4] 4/-1/-1->3->2 [5] 4/-1/-1->3->2 [6] 4/-1/-1->3->2 [7] 4/-1/-1->3->2 [8] 4/-1/-1->3->2 [9] 4/-1/-1->3->2 [10] 4/-1/-1->3->2 [11] 4/-1/-1->3->2 [12] 4/-1/-1->3->2 [13] 4/-1/-1->3->2 [14] 4/-1/-1->3->2 [15] 4/-1/-1->3->2 [16] 4/-1/-1->3->2 [17] 4/-1/-1->3->2 [18] 4/-1/-1->3->2 [19] 4/-1/-1->3->2 [20] 4/-1/-1->3->2 [21] 4/-1/-1->3->2 [22] 4/-1/-1->3->2 [23] 4/-1/-1->3->2 +ywang29-vrdb-test2-worker-0:782091:783670 [4] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:782087:783669 [0] NCCL INFO Channel 06/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:782089:783690 [2] NCCL INFO Trees [0] 3/-1/-1->2->1 [1] 3/-1/-1->2->1 [2] 3/-1/-1->2->1 [3] 3/-1/-1->2->1 [4] 3/-1/-1->2->1 [5] 3/-1/-1->2->1 [6] 3/-1/-1->2->1 [7] 3/-1/-1->2->1 [8] 3/-1/-1->2->1 [9] 3/-1/-1->2->1 [10] 3/-1/-1->2->1 [11] 3/-1/-1->2->1 [12] 3/-1/-1->2->1 [13] 3/-1/-1->2->1 [14] 3/-1/-1->2->1 [15] 3/-1/-1->2->1 [16] 3/-1/-1->2->1 [17] 3/-1/-1->2->1 [18] 3/-1/-1->2->1 [19] 3/-1/-1->2->1 [20] 3/-1/-1->2->1 [21] 3/-1/-1->2->1 [22] 3/-1/-1->2->1 [23] 3/-1/-1->2->1 +ywang29-vrdb-test2-worker-0:782087:783669 [0] NCCL INFO Channel 07/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:782090:783688 [3] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:782087:783669 [0] NCCL INFO Channel 08/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:782088:783691 [1] NCCL INFO Trees [0] 2/-1/-1->1->0 [1] 2/-1/-1->1->0 [2] 2/-1/-1->1->0 [3] 2/-1/-1->1->0 [4] 2/-1/-1->1->0 [5] 2/-1/-1->1->0 [6] 2/-1/-1->1->0 [7] 2/-1/-1->1->0 [8] 2/-1/-1->1->0 [9] 2/-1/-1->1->0 [10] 2/-1/-1->1->0 [11] 2/-1/-1->1->0 [12] 2/-1/-1->1->0 [13] 2/-1/-1->1->0 [14] 2/-1/-1->1->0 [15] 2/-1/-1->1->0 [16] 2/-1/-1->1->0 [17] 2/-1/-1->1->0 [18] 2/-1/-1->1->0 [19] 2/-1/-1->1->0 [20] 2/-1/-1->1->0 [21] 2/-1/-1->1->0 [22] 2/-1/-1->1->0 [23] 2/-1/-1->1->0 +ywang29-vrdb-test2-worker-0:782087:783669 [0] NCCL INFO Channel 09/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:782089:783690 [2] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:782087:783669 [0] NCCL INFO Channel 10/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:782088:783691 [1] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:782087:783669 [0] NCCL INFO Channel 11/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:782087:783669 [0] NCCL INFO Channel 12/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:782087:783669 [0] NCCL INFO Channel 13/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:782087:783669 [0] NCCL INFO Channel 14/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:782092:783689 [5] NCCL INFO Trees [0] 6/-1/-1->5->4 [1] 6/-1/-1->5->4 [2] 6/-1/-1->5->4 [3] 6/-1/-1->5->4 [4] 6/-1/-1->5->4 [5] 6/-1/-1->5->4 [6] 6/-1/-1->5->4 [7] 6/-1/-1->5->4 [8] 6/-1/-1->5->4 [9] 6/-1/-1->5->4 [10] 6/-1/-1->5->4 [11] 6/-1/-1->5->4 [12] 6/-1/-1->5->4 [13] 6/-1/-1->5->4 [14] 6/-1/-1->5->4 [15] 6/-1/-1->5->4 [16] 6/-1/-1->5->4 [17] 6/-1/-1->5->4 [18] 6/-1/-1->5->4 [19] 6/-1/-1->5->4 [20] 6/-1/-1->5->4 [21] 6/-1/-1->5->4 [22] 6/-1/-1->5->4 [23] 6/-1/-1->5->4 +ywang29-vrdb-test2-worker-0:782087:783669 [0] NCCL INFO Channel 15/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:782087:783669 [0] NCCL INFO Channel 16/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:782087:783669 [0] NCCL INFO Channel 17/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:782094:783692 [7] NCCL INFO Trees [0] -1/-1/-1->7->6 [1] -1/-1/-1->7->6 [2] -1/-1/-1->7->6 [3] -1/-1/-1->7->6 [4] -1/-1/-1->7->6 [5] -1/-1/-1->7->6 [6] -1/-1/-1->7->6 [7] -1/-1/-1->7->6 [8] -1/-1/-1->7->6 [9] -1/-1/-1->7->6 [10] -1/-1/-1->7->6 [11] -1/-1/-1->7->6 [12] -1/-1/-1->7->6 [13] -1/-1/-1->7->6 [14] -1/-1/-1->7->6 [15] -1/-1/-1->7->6 [16] -1/-1/-1->7->6 [17] -1/-1/-1->7->6 [18] -1/-1/-1->7->6 [19] -1/-1/-1->7->6 [20] -1/-1/-1->7->6 [21] -1/-1/-1->7->6 [22] -1/-1/-1->7->6 [23] -1/-1/-1->7->6 +ywang29-vrdb-test2-worker-0:782092:783689 [5] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:782087:783669 [0] NCCL INFO Channel 18/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:782087:783669 [0] NCCL INFO Channel 19/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:782087:783669 [0] NCCL INFO Channel 20/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:782087:783669 [0] NCCL INFO Channel 21/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:782094:783692 [7] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:782087:783669 [0] NCCL INFO Channel 22/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:782087:783669 [0] NCCL INFO Channel 23/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:782087:783669 [0] NCCL INFO Trees [0] 1/-1/-1->0->-1 [1] 1/-1/-1->0->-1 [2] 1/-1/-1->0->-1 [3] 1/-1/-1->0->-1 [4] 1/-1/-1->0->-1 [5] 1/-1/-1->0->-1 [6] 1/-1/-1->0->-1 [7] 1/-1/-1->0->-1 [8] 1/-1/-1->0->-1 [9] 1/-1/-1->0->-1 [10] 1/-1/-1->0->-1 [11] 1/-1/-1->0->-1 [12] 1/-1/-1->0->-1 [13] 1/-1/-1->0->-1 [14] 1/-1/-1->0->-1 [15] 1/-1/-1->0->-1 [16] 1/-1/-1->0->-1 [17] 1/-1/-1->0->-1 [18] 1/-1/-1->0->-1 [19] 1/-1/-1->0->-1 [20] 1/-1/-1->0->-1 [21] 1/-1/-1->0->-1 [22] 1/-1/-1->0->-1 [23] 1/-1/-1->0->-1 +ywang29-vrdb-test2-worker-0:782087:783669 [0] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:782093:783687 [6] NCCL INFO Channel 00/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:783690 [2] NCCL INFO Channel 00/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:783670 [4] NCCL INFO Channel 00/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:783688 [3] NCCL INFO Channel 00/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:783691 [1] NCCL INFO Channel 00/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:783687 [6] NCCL INFO Channel 01/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:783692 [7] NCCL INFO Channel 00/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782087:783669 [0] NCCL INFO Channel 00/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:783690 [2] NCCL INFO Channel 01/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:783670 [4] NCCL INFO Channel 01/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:783688 [3] NCCL INFO Channel 01/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:783691 [1] NCCL INFO Channel 01/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:783687 [6] NCCL INFO Channel 02/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:783692 [7] NCCL INFO Channel 01/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782087:783669 [0] NCCL INFO Channel 01/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:783690 [2] NCCL INFO Channel 02/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:783670 [4] NCCL INFO Channel 02/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:783688 [3] NCCL INFO Channel 02/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:783691 [1] NCCL INFO Channel 02/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:783687 [6] NCCL INFO Channel 03/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:783692 [7] NCCL INFO Channel 02/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782087:783669 [0] NCCL INFO Channel 02/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:783690 [2] NCCL INFO Channel 03/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:783670 [4] NCCL INFO Channel 03/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:783688 [3] NCCL INFO Channel 03/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:783691 [1] NCCL INFO Channel 03/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:783687 [6] NCCL INFO Channel 04/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:783692 [7] NCCL INFO Channel 03/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782087:783669 [0] NCCL INFO Channel 03/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:783670 [4] NCCL INFO Channel 04/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:783690 [2] NCCL INFO Channel 04/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:783688 [3] NCCL INFO Channel 04/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:783691 [1] NCCL INFO Channel 04/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:783687 [6] NCCL INFO Channel 05/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:783692 [7] NCCL INFO Channel 04/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782087:783669 [0] NCCL INFO Channel 04/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:783670 [4] NCCL INFO Channel 05/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:783690 [2] NCCL INFO Channel 05/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:783688 [3] NCCL INFO Channel 05/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:783691 [1] NCCL INFO Channel 05/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:783687 [6] NCCL INFO Channel 06/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:783692 [7] NCCL INFO Channel 05/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782087:783669 [0] NCCL INFO Channel 05/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:783670 [4] NCCL INFO Channel 06/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:783690 [2] NCCL INFO Channel 06/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:783688 [3] NCCL INFO Channel 06/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:783691 [1] NCCL INFO Channel 06/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:783687 [6] NCCL INFO Channel 07/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:783692 [7] NCCL INFO Channel 06/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782087:783669 [0] NCCL INFO Channel 06/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:783670 [4] NCCL INFO Channel 07/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:783690 [2] NCCL INFO Channel 07/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:783688 [3] NCCL INFO Channel 07/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:783691 [1] NCCL INFO Channel 07/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:783687 [6] NCCL INFO Channel 08/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:783692 [7] NCCL INFO Channel 07/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782087:783669 [0] NCCL INFO Channel 07/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:783670 [4] NCCL INFO Channel 08/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:783690 [2] NCCL INFO Channel 08/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:783688 [3] NCCL INFO Channel 08/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:783691 [1] NCCL INFO Channel 08/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:783687 [6] NCCL INFO Channel 09/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:783692 [7] NCCL INFO Channel 08/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782087:783669 [0] NCCL INFO Channel 08/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:783670 [4] NCCL INFO Channel 09/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:783690 [2] NCCL INFO Channel 09/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:783688 [3] NCCL INFO Channel 09/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:783691 [1] NCCL INFO Channel 09/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:783687 [6] NCCL INFO Channel 10/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782087:783669 [0] NCCL INFO Channel 09/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:783692 [7] NCCL INFO Channel 09/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:783670 [4] NCCL INFO Channel 10/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:783690 [2] NCCL INFO Channel 10/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:783688 [3] NCCL INFO Channel 10/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:783687 [6] NCCL INFO Channel 11/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:783691 [1] NCCL INFO Channel 10/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782087:783669 [0] NCCL INFO Channel 10/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:783692 [7] NCCL INFO Channel 10/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:783670 [4] NCCL INFO Channel 11/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:783690 [2] NCCL INFO Channel 11/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:783688 [3] NCCL INFO Channel 11/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:783687 [6] NCCL INFO Channel 12/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:783691 [1] NCCL INFO Channel 11/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782087:783669 [0] NCCL INFO Channel 11/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:783689 [5] NCCL INFO Channel 00/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:783692 [7] NCCL INFO Channel 11/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:783670 [4] NCCL INFO Channel 12/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:783690 [2] NCCL INFO Channel 12/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:783688 [3] NCCL INFO Channel 12/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:783687 [6] NCCL INFO Channel 13/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:783691 [1] NCCL INFO Channel 12/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782087:783669 [0] NCCL INFO Channel 12/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:783689 [5] NCCL INFO Channel 01/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:783692 [7] NCCL INFO Channel 12/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:783670 [4] NCCL INFO Channel 13/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:783690 [2] NCCL INFO Channel 13/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:783688 [3] NCCL INFO Channel 13/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:783687 [6] NCCL INFO Channel 14/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:783691 [1] NCCL INFO Channel 13/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782087:783669 [0] NCCL INFO Channel 13/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:783689 [5] NCCL INFO Channel 02/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:783692 [7] NCCL INFO Channel 13/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:783670 [4] NCCL INFO Channel 14/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:783690 [2] NCCL INFO Channel 14/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:783688 [3] NCCL INFO Channel 14/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:783687 [6] NCCL INFO Channel 15/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:783691 [1] NCCL INFO Channel 14/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782087:783669 [0] NCCL INFO Channel 14/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:783670 [4] NCCL INFO Channel 15/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:783689 [5] NCCL INFO Channel 03/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:783692 [7] NCCL INFO Channel 14/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:783690 [2] NCCL INFO Channel 15/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:783688 [3] NCCL INFO Channel 15/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:783687 [6] NCCL INFO Channel 16/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:783691 [1] NCCL INFO Channel 15/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782087:783669 [0] NCCL INFO Channel 15/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:783670 [4] NCCL INFO Channel 16/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:783689 [5] NCCL INFO Channel 04/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:783692 [7] NCCL INFO Channel 15/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:783690 [2] NCCL INFO Channel 16/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:783688 [3] NCCL INFO Channel 16/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:783687 [6] NCCL INFO Channel 17/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:783691 [1] NCCL INFO Channel 16/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782087:783669 [0] NCCL INFO Channel 16/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:783670 [4] NCCL INFO Channel 17/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:783689 [5] NCCL INFO Channel 05/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:783692 [7] NCCL INFO Channel 16/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:783690 [2] NCCL INFO Channel 17/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:783688 [3] NCCL INFO Channel 17/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:783687 [6] NCCL INFO Channel 18/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:783691 [1] NCCL INFO Channel 17/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782087:783669 [0] NCCL INFO Channel 17/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:783670 [4] NCCL INFO Channel 18/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:783689 [5] NCCL INFO Channel 06/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:783692 [7] NCCL INFO Channel 17/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:783688 [3] NCCL INFO Channel 18/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:783690 [2] NCCL INFO Channel 18/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:783687 [6] NCCL INFO Channel 19/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:783691 [1] NCCL INFO Channel 18/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:783670 [4] NCCL INFO Channel 19/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782087:783669 [0] NCCL INFO Channel 18/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:783689 [5] NCCL INFO Channel 07/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:783692 [7] NCCL INFO Channel 18/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:783688 [3] NCCL INFO Channel 19/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:783690 [2] NCCL INFO Channel 19/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:783691 [1] NCCL INFO Channel 19/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782087:783669 [0] NCCL INFO Channel 19/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:783687 [6] NCCL INFO Channel 20/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:783670 [4] NCCL INFO Channel 20/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:783689 [5] NCCL INFO Channel 08/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:783692 [7] NCCL INFO Channel 19/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:783688 [3] NCCL INFO Channel 20/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:783690 [2] NCCL INFO Channel 20/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:783691 [1] NCCL INFO Channel 20/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782087:783669 [0] NCCL INFO Channel 20/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:783687 [6] NCCL INFO Channel 21/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:783670 [4] NCCL INFO Channel 21/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:783689 [5] NCCL INFO Channel 09/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:783692 [7] NCCL INFO Channel 20/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:783688 [3] NCCL INFO Channel 21/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:783690 [2] NCCL INFO Channel 21/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:783691 [1] NCCL INFO Channel 21/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782087:783669 [0] NCCL INFO Channel 21/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:783687 [6] NCCL INFO Channel 22/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:783670 [4] NCCL INFO Channel 22/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:783689 [5] NCCL INFO Channel 10/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:783692 [7] NCCL INFO Channel 21/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:783688 [3] NCCL INFO Channel 22/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:783690 [2] NCCL INFO Channel 22/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:783691 [1] NCCL INFO Channel 22/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782087:783669 [0] NCCL INFO Channel 22/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:783687 [6] NCCL INFO Channel 23/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:783670 [4] NCCL INFO Channel 23/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:783689 [5] NCCL INFO Channel 11/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:783692 [7] NCCL INFO Channel 22/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:783688 [3] NCCL INFO Channel 23/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:783690 [2] NCCL INFO Channel 23/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:783691 [1] NCCL INFO Channel 23/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782087:783669 [0] NCCL INFO Channel 23/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:783689 [5] NCCL INFO Channel 12/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:783692 [7] NCCL INFO Channel 23/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:783689 [5] NCCL INFO Channel 13/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:783689 [5] NCCL INFO Channel 14/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:783689 [5] NCCL INFO Channel 15/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:783689 [5] NCCL INFO Channel 16/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:783689 [5] NCCL INFO Channel 17/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:783689 [5] NCCL INFO Channel 18/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:783689 [5] NCCL INFO Channel 19/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:783689 [5] NCCL INFO Channel 20/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:783689 [5] NCCL INFO Channel 21/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:783689 [5] NCCL INFO Channel 22/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:783689 [5] NCCL INFO Channel 23/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:783690 [2] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:782088:783691 [1] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:782087:783669 [0] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:782089:783690 [2] NCCL INFO Channel 00/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:783690 [2] NCCL INFO Channel 01/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:783690 [2] NCCL INFO Channel 02/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:783690 [2] NCCL INFO Channel 03/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:783690 [2] NCCL INFO Channel 04/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:783690 [2] NCCL INFO Channel 05/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:783690 [2] NCCL INFO Channel 06/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:783688 [3] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:782089:783690 [2] NCCL INFO Channel 07/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:783691 [1] NCCL INFO Channel 00/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:783690 [2] NCCL INFO Channel 08/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:783690 [2] NCCL INFO Channel 09/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:783691 [1] NCCL INFO Channel 01/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:783690 [2] NCCL INFO Channel 10/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:783691 [1] NCCL INFO Channel 02/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:783690 [2] NCCL INFO Channel 11/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:783691 [1] NCCL INFO Channel 03/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:783690 [2] NCCL INFO Channel 12/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:783691 [1] NCCL INFO Channel 04/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:783691 [1] NCCL INFO Channel 05/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:783692 [7] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:782094:783692 [7] NCCL INFO Channel 00/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:783691 [1] NCCL INFO Channel 06/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:783690 [2] NCCL INFO Channel 13/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:783691 [1] NCCL INFO Channel 07/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:783690 [2] NCCL INFO Channel 14/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:783691 [1] NCCL INFO Channel 08/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:783690 [2] NCCL INFO Channel 15/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:783691 [1] NCCL INFO Channel 09/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:783690 [2] NCCL INFO Channel 16/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:783692 [7] NCCL INFO Channel 01/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:783691 [1] NCCL INFO Channel 10/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:783690 [2] NCCL INFO Channel 17/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:783692 [7] NCCL INFO Channel 02/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:783691 [1] NCCL INFO Channel 11/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:783690 [2] NCCL INFO Channel 18/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:783687 [6] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:782088:783691 [1] NCCL INFO Channel 12/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:783670 [4] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:782092:783689 [5] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:782089:783690 [2] NCCL INFO Channel 19/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:783691 [1] NCCL INFO Channel 13/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:783690 [2] NCCL INFO Channel 20/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:783691 [1] NCCL INFO Channel 14/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:783690 [2] NCCL INFO Channel 21/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:783691 [1] NCCL INFO Channel 15/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:783690 [2] NCCL INFO Channel 22/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:783691 [1] NCCL INFO Channel 16/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:783690 [2] NCCL INFO Channel 23/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:783691 [1] NCCL INFO Channel 17/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:783691 [1] NCCL INFO Channel 18/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:783691 [1] NCCL INFO Channel 19/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:783691 [1] NCCL INFO Channel 20/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:783691 [1] NCCL INFO Channel 21/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:783691 [1] NCCL INFO Channel 22/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:783691 [1] NCCL INFO Channel 23/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:783692 [7] NCCL INFO Channel 03/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:783688 [3] NCCL INFO Channel 00/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:783692 [7] NCCL INFO Channel 04/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:783688 [3] NCCL INFO Channel 01/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:783692 [7] NCCL INFO Channel 05/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:783688 [3] NCCL INFO Channel 02/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:783692 [7] NCCL INFO Channel 06/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:783688 [3] NCCL INFO Channel 03/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:783692 [7] NCCL INFO Channel 07/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:783689 [5] NCCL INFO Channel 00/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:783688 [3] NCCL INFO Channel 04/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:783689 [5] NCCL INFO Channel 01/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:783688 [3] NCCL INFO Channel 05/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:783689 [5] NCCL INFO Channel 02/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:783688 [3] NCCL INFO Channel 06/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:783689 [5] NCCL INFO Channel 03/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:783688 [3] NCCL INFO Channel 07/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:783689 [5] NCCL INFO Channel 04/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:783688 [3] NCCL INFO Channel 08/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:783689 [5] NCCL INFO Channel 05/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:783692 [7] NCCL INFO Channel 08/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:783689 [5] NCCL INFO Channel 06/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:783687 [6] NCCL INFO Channel 00/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:783692 [7] NCCL INFO Channel 09/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:783687 [6] NCCL INFO Channel 01/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:783688 [3] NCCL INFO Channel 09/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:783689 [5] NCCL INFO Channel 07/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:783692 [7] NCCL INFO Channel 10/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:783688 [3] NCCL INFO Channel 10/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:783692 [7] NCCL INFO Channel 11/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:783688 [3] NCCL INFO Channel 11/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:783692 [7] NCCL INFO Channel 12/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:783688 [3] NCCL INFO Channel 12/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:783670 [4] NCCL INFO Channel 00/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:783687 [6] NCCL INFO Channel 02/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:783670 [4] NCCL INFO Channel 01/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:783687 [6] NCCL INFO Channel 03/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:783689 [5] NCCL INFO Channel 08/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:783688 [3] NCCL INFO Channel 13/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:783687 [6] NCCL INFO Channel 04/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:783689 [5] NCCL INFO Channel 09/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:783692 [7] NCCL INFO Channel 13/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:783670 [4] NCCL INFO Channel 02/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:783692 [7] NCCL INFO Channel 14/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:783688 [3] NCCL INFO Channel 14/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:783670 [4] NCCL INFO Channel 03/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:783687 [6] NCCL INFO Channel 05/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:783692 [7] NCCL INFO Channel 15/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:783688 [3] NCCL INFO Channel 15/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:783670 [4] NCCL INFO Channel 04/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:783687 [6] NCCL INFO Channel 06/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:783692 [7] NCCL INFO Channel 16/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:783688 [3] NCCL INFO Channel 16/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:783670 [4] NCCL INFO Channel 05/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:783687 [6] NCCL INFO Channel 07/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:783692 [7] NCCL INFO Channel 17/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:783688 [3] NCCL INFO Channel 17/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:783670 [4] NCCL INFO Channel 06/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:783687 [6] NCCL INFO Channel 08/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:783692 [7] NCCL INFO Channel 18/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:783688 [3] NCCL INFO Channel 18/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:783670 [4] NCCL INFO Channel 07/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:783687 [6] NCCL INFO Channel 09/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:783692 [7] NCCL INFO Channel 19/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:783688 [3] NCCL INFO Channel 19/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:783670 [4] NCCL INFO Channel 08/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:783687 [6] NCCL INFO Channel 10/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:783692 [7] NCCL INFO Channel 20/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:783688 [3] NCCL INFO Channel 20/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:783687 [6] NCCL INFO Channel 11/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:783670 [4] NCCL INFO Channel 09/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:783692 [7] NCCL INFO Channel 21/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:783688 [3] NCCL INFO Channel 21/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:783687 [6] NCCL INFO Channel 12/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:783670 [4] NCCL INFO Channel 10/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:783692 [7] NCCL INFO Channel 22/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:783688 [3] NCCL INFO Channel 22/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:783687 [6] NCCL INFO Channel 13/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:783670 [4] NCCL INFO Channel 11/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:783692 [7] NCCL INFO Channel 23/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:783688 [3] NCCL INFO Channel 23/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:783687 [6] NCCL INFO Channel 14/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:783670 [4] NCCL INFO Channel 12/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:783687 [6] NCCL INFO Channel 15/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:783689 [5] NCCL INFO Channel 10/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:783670 [4] NCCL INFO Channel 13/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:783687 [6] NCCL INFO Channel 16/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:783687 [6] NCCL INFO Channel 17/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:783670 [4] NCCL INFO Channel 14/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:783689 [5] NCCL INFO Channel 11/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:783687 [6] NCCL INFO Channel 18/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:783689 [5] NCCL INFO Channel 12/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:783670 [4] NCCL INFO Channel 15/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:783687 [6] NCCL INFO Channel 19/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:783689 [5] NCCL INFO Channel 13/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:783670 [4] NCCL INFO Channel 16/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:783689 [5] NCCL INFO Channel 14/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:783687 [6] NCCL INFO Channel 20/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:783689 [5] NCCL INFO Channel 15/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:783687 [6] NCCL INFO Channel 21/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:783689 [5] NCCL INFO Channel 16/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:783670 [4] NCCL INFO Channel 17/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:783687 [6] NCCL INFO Channel 22/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:783689 [5] NCCL INFO Channel 17/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:783687 [6] NCCL INFO Channel 23/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:783689 [5] NCCL INFO Channel 18/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:783670 [4] NCCL INFO Channel 18/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:783670 [4] NCCL INFO Channel 19/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:783689 [5] NCCL INFO Channel 19/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:783670 [4] NCCL INFO Channel 20/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:783689 [5] NCCL INFO Channel 20/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:783670 [4] NCCL INFO Channel 21/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:783689 [5] NCCL INFO Channel 21/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:783670 [4] NCCL INFO Channel 22/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:783689 [5] NCCL INFO Channel 22/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:783670 [4] NCCL INFO Channel 23/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:783689 [5] NCCL INFO Channel 23/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782087:783669 [0] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:782087:783669 [0] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:782087:783669 [0] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:782088:783691 [1] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:782088:783691 [1] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:782088:783691 [1] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:782089:783690 [2] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:782089:783690 [2] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:782089:783690 [2] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:782090:783688 [3] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:782090:783688 [3] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:782090:783688 [3] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:782091:783670 [4] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:782091:783670 [4] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:782091:783670 [4] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:782092:783689 [5] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:782092:783689 [5] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:782092:783689 [5] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:782093:783687 [6] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:782093:783687 [6] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:782093:783687 [6] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:782094:783692 [7] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:782094:783692 [7] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:782094:783692 [7] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:782090:783688 [3] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test2-worker-0:782090:783688 [3] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test2-worker-0:782090:783688 [3] NCCL INFO ncclCommInitRank comm 0x55995d49b4f0 rank 3 nranks 8 cudaDev 3 nvmlDev 3 busId 201d0 commId 0x5d3556e4537bc329 - Init COMPLETE +ywang29-vrdb-test2-worker-0:782089:783690 [2] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test2-worker-0:782089:783690 [2] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test2-worker-0:782094:783692 [7] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test2-worker-0:782091:783670 [4] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test2-worker-0:782093:783687 [6] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test2-worker-0:782092:783689 [5] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test2-worker-0:782087:783669 [0] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test2-worker-0:782089:783690 [2] NCCL INFO ncclCommInitRank comm 0x555629a50910 rank 2 nranks 8 cudaDev 2 nvmlDev 2 busId 201c0 commId 0x5d3556e4537bc329 - Init COMPLETE +ywang29-vrdb-test2-worker-0:782094:783692 [7] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test2-worker-0:782091:783670 [4] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test2-worker-0:782093:783687 [6] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test2-worker-0:782092:783689 [5] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test2-worker-0:782087:783669 [0] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test2-worker-0:782094:783692 [7] NCCL INFO ncclCommInitRank comm 0x55b3c9826fd0 rank 7 nranks 8 cudaDev 7 nvmlDev 7 busId a01d0 commId 0x5d3556e4537bc329 - Init COMPLETE +ywang29-vrdb-test2-worker-0:782091:783670 [4] NCCL INFO ncclCommInitRank comm 0x56275fc79830 rank 4 nranks 8 cudaDev 4 nvmlDev 4 busId 901c0 commId 0x5d3556e4537bc329 - Init COMPLETE +ywang29-vrdb-test2-worker-0:782093:783687 [6] NCCL INFO ncclCommInitRank comm 0x557eff3d5e00 rank 6 nranks 8 cudaDev 6 nvmlDev 6 busId a01c0 commId 0x5d3556e4537bc329 - Init COMPLETE +ywang29-vrdb-test2-worker-0:782092:783689 [5] NCCL INFO ncclCommInitRank comm 0x5616d445c0d0 rank 5 nranks 8 cudaDev 5 nvmlDev 5 busId 901d0 commId 0x5d3556e4537bc329 - Init COMPLETE +ywang29-vrdb-test2-worker-0:782087:783669 [0] NCCL INFO ncclCommInitRank comm 0x55a04621adb0 rank 0 nranks 8 cudaDev 0 nvmlDev 0 busId 101c0 commId 0x5d3556e4537bc329 - Init COMPLETE +ywang29-vrdb-test2-worker-0:782088:783691 [1] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test2-worker-0:782088:783691 [1] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test2-worker-0:782088:783691 [1] NCCL INFO ncclCommInitRank comm 0x562d761a6b30 rank 1 nranks 8 cudaDev 1 nvmlDev 1 busId 101d0 commId 0x5d3556e4537bc329 - Init COMPLETE +[2025-10-12 14:30:05,106] [INFO] [partition_parameters.py:343:__exit__] finished initializing model - num_params = 459, num_elems = 0.99B +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +[2025-10-12 14:30:06,857] [INFO] [partition_parameters.py:343:__exit__] finished initializing model - num_params = 907, num_elems = 1.42B +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +TinyLlavaForConditionalGeneration( + (language_model): Qwen2ForCausalLM( + (model): Qwen2Model( + (embed_tokens): Embedding(151936, 896) + (layers): ModuleList( + (0-23): 24 x Qwen2DecoderLayer( + (self_attn): Qwen2FlashAttention2( + (q_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=True) + (k_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=128, bias=True) + (v_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=128, bias=True) + (o_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=False) + (rotary_emb): Qwen2RotaryEmbedding() + ) + (mlp): Qwen2MLP( + (gate_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=4864, bias=False) + (up_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=4864, bias=False) + (down_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=4864, out_features=896, bias=False) + (act_fn): SiLU() + ) + (input_layernorm): Qwen2RMSNorm() + (post_attention_layernorm): Qwen2RMSNorm() + ) + ) + (norm): Qwen2RMSNorm() + ) + (lm_head): Linear(in_features=896, out_features=151936, bias=False) + ) + (vision_tower): SIGLIPVisionTower( + (_vision_tower): SiglipVisionModel( + (vision_model): SiglipVisionTransformer( + (embeddings): SiglipVisionEmbeddings( + (patch_embedding): Conv2d(3, 1152, kernel_size=(14, 14), stride=(14, 14), padding=valid) + (position_embedding): Embedding(729, 1152) + ) + (encoder): SiglipEncoder( + (layers): ModuleList( + (0-26): 27 x SiglipEncoderLayer( + (self_attn): SiglipAttention( + (k_proj): Linear(in_features=1152, out_features=1152, bias=True) + (v_proj): Linear(in_features=1152, out_features=1152, bias=True) + (q_proj): Linear(in_features=1152, out_features=1152, bias=True) + (out_proj): Linear(in_features=1152, out_features=1152, bias=True) + ) + (layer_norm1): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + (layer_norm2): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + ) + ) + ) + (post_layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (head): SiglipMultiheadAttentionPoolingHead( + (attention): MultiheadAttention( + (out_proj): NonDynamicallyQuantizableLinear(in_features=1152, out_features=1152, bias=True) + ) + (layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + ) + ) + ) + ) + (connector): MLPConnector( + (_connector): Sequential( + (0): SupermaskLinearSparsity_SoftForward_Normal(in_features=1152, out_features=896, bias=True) + (1): GELU(approximate='none') + (2): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=True) + ) + ) +) +Pre-training init language_model.model.layers.0.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.0.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.0.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.0.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.0.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.0.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.0.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.mlp.down_proj.scores: Mean=3.000000 +Pre-training init connector._connector.0.scores: Mean=3.000005 +Pre-training init connector._connector.2.scores: Mean=2.999970 +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +2025-10-12 14:30:25,056 | INFO: Total Parameters: 1283756736, Total Trainable Parameters: 359661568 +2025-10-12 14:30:25,080 | INFO: Trainable Parameters: +language_model.model.layers.0.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.0.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.0.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.0.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.0.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.0.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.0.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.1.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.1.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.1.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.1.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.1.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.1.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.1.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.2.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.2.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.2.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.2.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.2.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.2.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.2.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.3.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.3.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.3.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.3.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.3.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.3.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.3.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.4.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.4.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.4.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.4.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.4.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.4.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.4.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.5.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.5.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.5.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.5.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.5.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.5.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.5.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.6.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.6.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.6.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.6.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.6.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.6.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.6.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.7.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.7.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.7.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.7.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.7.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.7.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.7.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.8.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.8.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.8.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.8.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.8.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.8.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.8.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.9.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.9.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.9.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.9.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.9.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.9.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.9.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.10.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.10.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.10.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.10.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.10.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.10.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.10.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.11.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.11.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.11.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.11.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.11.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.11.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.11.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.12.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.12.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.12.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.12.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.12.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.12.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.12.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.13.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.13.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.13.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.13.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.13.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.13.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.13.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.14.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.14.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.14.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.14.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.14.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.14.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.14.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.15.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.15.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.15.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.15.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.15.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.15.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.15.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.16.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.16.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.16.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.16.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.16.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.16.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.16.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.17.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.17.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.17.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.17.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.17.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.17.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.17.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.18.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.18.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.18.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.18.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.18.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.18.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.18.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.19.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.19.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.19.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.19.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.19.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.19.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.19.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.20.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.20.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.20.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.20.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.20.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.20.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.20.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.21.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.21.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.21.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.21.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.21.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.21.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.21.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.22.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.22.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.22.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.22.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.22.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.22.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.22.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.23.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.23.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.23.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.23.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.23.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.23.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.23.mlp.down_proj.scores: 4358144 parameters +connector._connector.0.scores: 1032192 parameters +connector._connector.2.scores: 802816 parameters +Parameter Offload: Total persistent parameters: 486464 in 403 params + 0%| | 0/520 [00:003->2 [1] 4/-1/-1->3->2 [2] 4/-1/-1->3->2 [3] 4/-1/-1->3->2 [4] 4/-1/-1->3->2 [5] 4/-1/-1->3->2 [6] 4/-1/-1->3->2 [7] 4/-1/-1->3->2 [8] 4/-1/-1->3->2 [9] 4/-1/-1->3->2 [10] 4/-1/-1->3->2 [11] 4/-1/-1->3->2 [12] 4/-1/-1->3->2 [13] 4/-1/-1->3->2 [14] 4/-1/-1->3->2 [15] 4/-1/-1->3->2 [16] 4/-1/-1->3->2 [17] 4/-1/-1->3->2 [18] 4/-1/-1->3->2 [19] 4/-1/-1->3->2 [20] 4/-1/-1->3->2 [21] 4/-1/-1->3->2 [22] 4/-1/-1->3->2 [23] 4/-1/-1->3->2 +ywang29-vrdb-test2-worker-0:782087:788616 [0] NCCL INFO Channel 02/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:782090:788622 [3] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:782087:788616 [0] NCCL INFO Channel 03/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:782087:788616 [0] NCCL INFO Channel 04/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:782087:788616 [0] NCCL INFO Channel 05/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:782088:788619 [1] NCCL INFO Trees [0] 2/-1/-1->1->0 [1] 2/-1/-1->1->0 [2] 2/-1/-1->1->0 [3] 2/-1/-1->1->0 [4] 2/-1/-1->1->0 [5] 2/-1/-1->1->0 [6] 2/-1/-1->1->0 [7] 2/-1/-1->1->0 [8] 2/-1/-1->1->0 [9] 2/-1/-1->1->0 [10] 2/-1/-1->1->0 [11] 2/-1/-1->1->0 [12] 2/-1/-1->1->0 [13] 2/-1/-1->1->0 [14] 2/-1/-1->1->0 [15] 2/-1/-1->1->0 [16] 2/-1/-1->1->0 [17] 2/-1/-1->1->0 [18] 2/-1/-1->1->0 [19] 2/-1/-1->1->0 [20] 2/-1/-1->1->0 [21] 2/-1/-1->1->0 [22] 2/-1/-1->1->0 [23] 2/-1/-1->1->0 +ywang29-vrdb-test2-worker-0:782089:788623 [2] NCCL INFO Trees [0] 3/-1/-1->2->1 [1] 3/-1/-1->2->1 [2] 3/-1/-1->2->1 [3] 3/-1/-1->2->1 [4] 3/-1/-1->2->1 [5] 3/-1/-1->2->1 [6] 3/-1/-1->2->1 [7] 3/-1/-1->2->1 [8] 3/-1/-1->2->1 [9] 3/-1/-1->2->1 [10] 3/-1/-1->2->1 [11] 3/-1/-1->2->1 [12] 3/-1/-1->2->1 [13] 3/-1/-1->2->1 [14] 3/-1/-1->2->1 [15] 3/-1/-1->2->1 [16] 3/-1/-1->2->1 [17] 3/-1/-1->2->1 [18] 3/-1/-1->2->1 [19] 3/-1/-1->2->1 [20] 3/-1/-1->2->1 [21] 3/-1/-1->2->1 [22] 3/-1/-1->2->1 [23] 3/-1/-1->2->1 +ywang29-vrdb-test2-worker-0:782087:788616 [0] NCCL INFO Channel 06/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:782088:788619 [1] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:782094:788618 [7] NCCL INFO Trees [0] -1/-1/-1->7->6 [1] -1/-1/-1->7->6 [2] -1/-1/-1->7->6 [3] -1/-1/-1->7->6 [4] -1/-1/-1->7->6 [5] -1/-1/-1->7->6 [6] -1/-1/-1->7->6 [7] -1/-1/-1->7->6 [8] -1/-1/-1->7->6 [9] -1/-1/-1->7->6 [10] -1/-1/-1->7->6 [11] -1/-1/-1->7->6 [12] -1/-1/-1->7->6 [13] -1/-1/-1->7->6 [14] -1/-1/-1->7->6 [15] -1/-1/-1->7->6 [16] -1/-1/-1->7->6 [17] -1/-1/-1->7->6 [18] -1/-1/-1->7->6 [19] -1/-1/-1->7->6 [20] -1/-1/-1->7->6 [21] -1/-1/-1->7->6 [22] -1/-1/-1->7->6 [23] -1/-1/-1->7->6 +ywang29-vrdb-test2-worker-0:782089:788623 [2] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:782087:788616 [0] NCCL INFO Channel 07/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:782094:788618 [7] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:782087:788616 [0] NCCL INFO Channel 08/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:782092:788617 [5] NCCL INFO Trees [0] 6/-1/-1->5->4 [1] 6/-1/-1->5->4 [2] 6/-1/-1->5->4 [3] 6/-1/-1->5->4 [4] 6/-1/-1->5->4 [5] 6/-1/-1->5->4 [6] 6/-1/-1->5->4 [7] 6/-1/-1->5->4 [8] 6/-1/-1->5->4 [9] 6/-1/-1->5->4 [10] 6/-1/-1->5->4 [11] 6/-1/-1->5->4 [12] 6/-1/-1->5->4 [13] 6/-1/-1->5->4 [14] 6/-1/-1->5->4 [15] 6/-1/-1->5->4 [16] 6/-1/-1->5->4 [17] 6/-1/-1->5->4 [18] 6/-1/-1->5->4 [19] 6/-1/-1->5->4 [20] 6/-1/-1->5->4 [21] 6/-1/-1->5->4 [22] 6/-1/-1->5->4 [23] 6/-1/-1->5->4 +ywang29-vrdb-test2-worker-0:782087:788616 [0] NCCL INFO Channel 09/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:782092:788617 [5] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:782087:788616 [0] NCCL INFO Channel 10/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:782087:788616 [0] NCCL INFO Channel 11/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:782087:788616 [0] NCCL INFO Channel 12/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:782087:788616 [0] NCCL INFO Channel 13/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:782087:788616 [0] NCCL INFO Channel 14/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:782087:788616 [0] NCCL INFO Channel 15/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:782087:788616 [0] NCCL INFO Channel 16/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:782087:788616 [0] NCCL INFO Channel 17/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:782087:788616 [0] NCCL INFO Channel 18/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:782087:788616 [0] NCCL INFO Channel 19/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:782087:788616 [0] NCCL INFO Channel 20/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:782093:788621 [6] NCCL INFO Trees [0] 7/-1/-1->6->5 [1] 7/-1/-1->6->5 [2] 7/-1/-1->6->5 [3] 7/-1/-1->6->5 [4] 7/-1/-1->6->5 [5] 7/-1/-1->6->5 [6] 7/-1/-1->6->5 [7] 7/-1/-1->6->5 [8] 7/-1/-1->6->5 [9] 7/-1/-1->6->5 [10] 7/-1/-1->6->5 [11] 7/-1/-1->6->5 [12] 7/-1/-1->6->5 [13] 7/-1/-1->6->5 [14] 7/-1/-1->6->5 [15] 7/-1/-1->6->5 [16] 7/-1/-1->6->5 [17] 7/-1/-1->6->5 [18] 7/-1/-1->6->5 [19] 7/-1/-1->6->5 [20] 7/-1/-1->6->5 [21] 7/-1/-1->6->5 [22] 7/-1/-1->6->5 [23] 7/-1/-1->6->5 +ywang29-vrdb-test2-worker-0:782087:788616 [0] NCCL INFO Channel 21/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:782093:788621 [6] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:782091:788620 [4] NCCL INFO Trees [0] 5/-1/-1->4->3 [1] 5/-1/-1->4->3 [2] 5/-1/-1->4->3 [3] 5/-1/-1->4->3 [4] 5/-1/-1->4->3 [5] 5/-1/-1->4->3 [6] 5/-1/-1->4->3 [7] 5/-1/-1->4->3 [8] 5/-1/-1->4->3 [9] 5/-1/-1->4->3 [10] 5/-1/-1->4->3 [11] 5/-1/-1->4->3 [12] 5/-1/-1->4->3 [13] 5/-1/-1->4->3 [14] 5/-1/-1->4->3 [15] 5/-1/-1->4->3 [16] 5/-1/-1->4->3 [17] 5/-1/-1->4->3 [18] 5/-1/-1->4->3 [19] 5/-1/-1->4->3 [20] 5/-1/-1->4->3 [21] 5/-1/-1->4->3 [22] 5/-1/-1->4->3 [23] 5/-1/-1->4->3 +ywang29-vrdb-test2-worker-0:782087:788616 [0] NCCL INFO Channel 22/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:782087:788616 [0] NCCL INFO Channel 23/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test2-worker-0:782091:788620 [4] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:782087:788616 [0] NCCL INFO Trees [0] 1/-1/-1->0->-1 [1] 1/-1/-1->0->-1 [2] 1/-1/-1->0->-1 [3] 1/-1/-1->0->-1 [4] 1/-1/-1->0->-1 [5] 1/-1/-1->0->-1 [6] 1/-1/-1->0->-1 [7] 1/-1/-1->0->-1 [8] 1/-1/-1->0->-1 [9] 1/-1/-1->0->-1 [10] 1/-1/-1->0->-1 [11] 1/-1/-1->0->-1 [12] 1/-1/-1->0->-1 [13] 1/-1/-1->0->-1 [14] 1/-1/-1->0->-1 [15] 1/-1/-1->0->-1 [16] 1/-1/-1->0->-1 [17] 1/-1/-1->0->-1 [18] 1/-1/-1->0->-1 [19] 1/-1/-1->0->-1 [20] 1/-1/-1->0->-1 [21] 1/-1/-1->0->-1 [22] 1/-1/-1->0->-1 [23] 1/-1/-1->0->-1 +ywang29-vrdb-test2-worker-0:782087:788616 [0] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test2-worker-0:782094:788618 [7] NCCL INFO Channel 00/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:788619 [1] NCCL INFO Channel 00/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782087:788616 [0] NCCL INFO Channel 00/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:788622 [3] NCCL INFO Channel 00/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:788618 [7] NCCL INFO Channel 01/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:788623 [2] NCCL INFO Channel 00/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:788617 [5] NCCL INFO Channel 00/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:788619 [1] NCCL INFO Channel 01/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782087:788616 [0] NCCL INFO Channel 01/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:788622 [3] NCCL INFO Channel 01/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:788618 [7] NCCL INFO Channel 02/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:788623 [2] NCCL INFO Channel 01/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:788617 [5] NCCL INFO Channel 01/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:788619 [1] NCCL INFO Channel 02/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782087:788616 [0] NCCL INFO Channel 02/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:788622 [3] NCCL INFO Channel 02/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:788618 [7] NCCL INFO Channel 03/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:788623 [2] NCCL INFO Channel 02/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:788617 [5] NCCL INFO Channel 02/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:788619 [1] NCCL INFO Channel 03/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:788620 [4] NCCL INFO Channel 00/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782087:788616 [0] NCCL INFO Channel 03/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:788622 [3] NCCL INFO Channel 03/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:788618 [7] NCCL INFO Channel 04/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:788623 [2] NCCL INFO Channel 03/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:788621 [6] NCCL INFO Channel 00/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:788617 [5] NCCL INFO Channel 03/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:788619 [1] NCCL INFO Channel 04/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:788620 [4] NCCL INFO Channel 01/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782087:788616 [0] NCCL INFO Channel 04/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:788622 [3] NCCL INFO Channel 04/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:788618 [7] NCCL INFO Channel 05/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:788623 [2] NCCL INFO Channel 04/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:788617 [5] NCCL INFO Channel 04/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:788621 [6] NCCL INFO Channel 01/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:788619 [1] NCCL INFO Channel 05/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782087:788616 [0] NCCL INFO Channel 05/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:788620 [4] NCCL INFO Channel 02/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:788622 [3] NCCL INFO Channel 05/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:788618 [7] NCCL INFO Channel 06/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:788623 [2] NCCL INFO Channel 05/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:788617 [5] NCCL INFO Channel 05/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:788621 [6] NCCL INFO Channel 02/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:788619 [1] NCCL INFO Channel 06/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782087:788616 [0] NCCL INFO Channel 06/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:788622 [3] NCCL INFO Channel 06/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:788620 [4] NCCL INFO Channel 03/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:788618 [7] NCCL INFO Channel 07/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:788623 [2] NCCL INFO Channel 06/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:788617 [5] NCCL INFO Channel 06/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:788621 [6] NCCL INFO Channel 03/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:788619 [1] NCCL INFO Channel 07/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782087:788616 [0] NCCL INFO Channel 07/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:788622 [3] NCCL INFO Channel 07/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:788620 [4] NCCL INFO Channel 04/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:788618 [7] NCCL INFO Channel 08/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:788623 [2] NCCL INFO Channel 07/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:788617 [5] NCCL INFO Channel 07/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:788619 [1] NCCL INFO Channel 08/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:788621 [6] NCCL INFO Channel 04/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:788622 [3] NCCL INFO Channel 08/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:788620 [4] NCCL INFO Channel 05/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:788618 [7] NCCL INFO Channel 09/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:788623 [2] NCCL INFO Channel 08/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:788617 [5] NCCL INFO Channel 08/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:788619 [1] NCCL INFO Channel 09/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:788621 [6] NCCL INFO Channel 05/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:788622 [3] NCCL INFO Channel 09/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:788620 [4] NCCL INFO Channel 06/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:788618 [7] NCCL INFO Channel 10/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:788623 [2] NCCL INFO Channel 09/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:788617 [5] NCCL INFO Channel 09/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:788619 [1] NCCL INFO Channel 10/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:788622 [3] NCCL INFO Channel 10/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:788621 [6] NCCL INFO Channel 06/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:788620 [4] NCCL INFO Channel 07/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:788618 [7] NCCL INFO Channel 11/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:788623 [2] NCCL INFO Channel 10/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:788619 [1] NCCL INFO Channel 11/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:788617 [5] NCCL INFO Channel 10/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:788622 [3] NCCL INFO Channel 11/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:788621 [6] NCCL INFO Channel 07/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:788620 [4] NCCL INFO Channel 08/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:788618 [7] NCCL INFO Channel 12/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:788623 [2] NCCL INFO Channel 11/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:788619 [1] NCCL INFO Channel 12/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:788617 [5] NCCL INFO Channel 11/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:788622 [3] NCCL INFO Channel 12/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:788618 [7] NCCL INFO Channel 13/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:788620 [4] NCCL INFO Channel 09/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:788621 [6] NCCL INFO Channel 08/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:788623 [2] NCCL INFO Channel 12/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:788619 [1] NCCL INFO Channel 13/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:788617 [5] NCCL INFO Channel 12/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:788622 [3] NCCL INFO Channel 13/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:788618 [7] NCCL INFO Channel 14/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:788620 [4] NCCL INFO Channel 10/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:788621 [6] NCCL INFO Channel 09/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:788623 [2] NCCL INFO Channel 13/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:788619 [1] NCCL INFO Channel 14/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:788617 [5] NCCL INFO Channel 13/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:788622 [3] NCCL INFO Channel 14/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:788618 [7] NCCL INFO Channel 15/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:788620 [4] NCCL INFO Channel 11/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:788623 [2] NCCL INFO Channel 14/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:788621 [6] NCCL INFO Channel 10/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:788619 [1] NCCL INFO Channel 15/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:788617 [5] NCCL INFO Channel 14/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:788622 [3] NCCL INFO Channel 15/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:788623 [2] NCCL INFO Channel 15/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:788620 [4] NCCL INFO Channel 12/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:788618 [7] NCCL INFO Channel 16/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:788619 [1] NCCL INFO Channel 16/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:788621 [6] NCCL INFO Channel 11/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:788617 [5] NCCL INFO Channel 15/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:788622 [3] NCCL INFO Channel 16/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:788620 [4] NCCL INFO Channel 13/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:788623 [2] NCCL INFO Channel 16/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:788618 [7] NCCL INFO Channel 17/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:788619 [1] NCCL INFO Channel 17/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:788621 [6] NCCL INFO Channel 12/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:788617 [5] NCCL INFO Channel 16/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:788622 [3] NCCL INFO Channel 17/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:788623 [2] NCCL INFO Channel 17/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:788620 [4] NCCL INFO Channel 14/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:788618 [7] NCCL INFO Channel 18/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:788619 [1] NCCL INFO Channel 18/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:788621 [6] NCCL INFO Channel 13/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:788617 [5] NCCL INFO Channel 17/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:788622 [3] NCCL INFO Channel 18/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:788623 [2] NCCL INFO Channel 18/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:788618 [7] NCCL INFO Channel 19/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:788620 [4] NCCL INFO Channel 15/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:788619 [1] NCCL INFO Channel 19/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:788621 [6] NCCL INFO Channel 14/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:788622 [3] NCCL INFO Channel 19/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:788617 [5] NCCL INFO Channel 18/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:788623 [2] NCCL INFO Channel 19/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:788618 [7] NCCL INFO Channel 20/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:788620 [4] NCCL INFO Channel 16/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:788619 [1] NCCL INFO Channel 20/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782087:788616 [0] NCCL INFO Channel 08/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:788622 [3] NCCL INFO Channel 20/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:788621 [6] NCCL INFO Channel 15/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:788617 [5] NCCL INFO Channel 19/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:788623 [2] NCCL INFO Channel 20/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:788618 [7] NCCL INFO Channel 21/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:788620 [4] NCCL INFO Channel 17/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:788619 [1] NCCL INFO Channel 21/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782087:788616 [0] NCCL INFO Channel 09/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:788622 [3] NCCL INFO Channel 21/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:788617 [5] NCCL INFO Channel 20/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:788623 [2] NCCL INFO Channel 21/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:788621 [6] NCCL INFO Channel 16/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:788618 [7] NCCL INFO Channel 22/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:788619 [1] NCCL INFO Channel 22/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:788620 [4] NCCL INFO Channel 18/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782087:788616 [0] NCCL INFO Channel 10/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:788622 [3] NCCL INFO Channel 22/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:788617 [5] NCCL INFO Channel 21/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:788623 [2] NCCL INFO Channel 22/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:788618 [7] NCCL INFO Channel 23/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:788621 [6] NCCL INFO Channel 17/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:788619 [1] NCCL INFO Channel 23/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:788620 [4] NCCL INFO Channel 19/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782087:788616 [0] NCCL INFO Channel 11/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:788622 [3] NCCL INFO Channel 23/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:788623 [2] NCCL INFO Channel 23/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:788617 [5] NCCL INFO Channel 22/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:788621 [6] NCCL INFO Channel 18/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:788620 [4] NCCL INFO Channel 20/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782087:788616 [0] NCCL INFO Channel 12/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:788617 [5] NCCL INFO Channel 23/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:788621 [6] NCCL INFO Channel 19/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:788620 [4] NCCL INFO Channel 21/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782087:788616 [0] NCCL INFO Channel 13/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:788621 [6] NCCL INFO Channel 20/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:788620 [4] NCCL INFO Channel 22/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782087:788616 [0] NCCL INFO Channel 14/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782087:788616 [0] NCCL INFO Channel 15/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:788621 [6] NCCL INFO Channel 21/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:788620 [4] NCCL INFO Channel 23/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782087:788616 [0] NCCL INFO Channel 16/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:788621 [6] NCCL INFO Channel 22/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782087:788616 [0] NCCL INFO Channel 17/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:788621 [6] NCCL INFO Channel 23/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782087:788616 [0] NCCL INFO Channel 18/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782087:788616 [0] NCCL INFO Channel 19/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782087:788616 [0] NCCL INFO Channel 20/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782087:788616 [0] NCCL INFO Channel 21/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782087:788616 [0] NCCL INFO Channel 22/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782087:788616 [0] NCCL INFO Channel 23/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:788623 [2] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:782090:788622 [3] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:782091:788620 [4] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:782089:788623 [2] NCCL INFO Channel 00/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:788623 [2] NCCL INFO Channel 01/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:788623 [2] NCCL INFO Channel 02/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782087:788616 [0] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:782089:788623 [2] NCCL INFO Channel 03/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:788623 [2] NCCL INFO Channel 04/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:788618 [7] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:782094:788618 [7] NCCL INFO Channel 00/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:788617 [5] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:782093:788621 [6] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:782089:788623 [2] NCCL INFO Channel 05/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:788618 [7] NCCL INFO Channel 01/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:788623 [2] NCCL INFO Channel 06/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:788618 [7] NCCL INFO Channel 02/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:788623 [2] NCCL INFO Channel 07/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:788619 [1] NCCL INFO Connected all rings +ywang29-vrdb-test2-worker-0:782094:788618 [7] NCCL INFO Channel 03/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:788623 [2] NCCL INFO Channel 08/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:788618 [7] NCCL INFO Channel 04/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:788623 [2] NCCL INFO Channel 09/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:788623 [2] NCCL INFO Channel 10/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:788623 [2] NCCL INFO Channel 11/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:788623 [2] NCCL INFO Channel 12/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:788623 [2] NCCL INFO Channel 13/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:788623 [2] NCCL INFO Channel 14/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:788623 [2] NCCL INFO Channel 15/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:788618 [7] NCCL INFO Channel 05/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:788623 [2] NCCL INFO Channel 16/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:788618 [7] NCCL INFO Channel 06/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:788623 [2] NCCL INFO Channel 17/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:788618 [7] NCCL INFO Channel 07/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:788620 [4] NCCL INFO Channel 00/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:788623 [2] NCCL INFO Channel 18/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:788618 [7] NCCL INFO Channel 08/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:788620 [4] NCCL INFO Channel 01/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:788623 [2] NCCL INFO Channel 19/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:788618 [7] NCCL INFO Channel 09/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:788620 [4] NCCL INFO Channel 02/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:788623 [2] NCCL INFO Channel 20/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:788618 [7] NCCL INFO Channel 10/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:788620 [4] NCCL INFO Channel 03/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:788623 [2] NCCL INFO Channel 21/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:788618 [7] NCCL INFO Channel 11/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:788620 [4] NCCL INFO Channel 04/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:788623 [2] NCCL INFO Channel 22/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:788620 [4] NCCL INFO Channel 05/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782089:788623 [2] NCCL INFO Channel 23/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:788620 [4] NCCL INFO Channel 06/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:788620 [4] NCCL INFO Channel 07/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:788620 [4] NCCL INFO Channel 08/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:788620 [4] NCCL INFO Channel 09/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:788620 [4] NCCL INFO Channel 10/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:788620 [4] NCCL INFO Channel 11/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:788618 [7] NCCL INFO Channel 12/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:788620 [4] NCCL INFO Channel 12/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:788618 [7] NCCL INFO Channel 13/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:788620 [4] NCCL INFO Channel 13/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:788621 [6] NCCL INFO Channel 00/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:788622 [3] NCCL INFO Channel 00/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:788618 [7] NCCL INFO Channel 14/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:788620 [4] NCCL INFO Channel 14/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:788621 [6] NCCL INFO Channel 01/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:788622 [3] NCCL INFO Channel 01/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:788618 [7] NCCL INFO Channel 15/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:788619 [1] NCCL INFO Channel 00/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:788620 [4] NCCL INFO Channel 15/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:788621 [6] NCCL INFO Channel 02/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:788622 [3] NCCL INFO Channel 02/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:788618 [7] NCCL INFO Channel 16/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:788619 [1] NCCL INFO Channel 01/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:788621 [6] NCCL INFO Channel 03/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:788620 [4] NCCL INFO Channel 16/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:788622 [3] NCCL INFO Channel 03/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:788618 [7] NCCL INFO Channel 17/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:788619 [1] NCCL INFO Channel 02/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:788620 [4] NCCL INFO Channel 17/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:788621 [6] NCCL INFO Channel 04/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:788622 [3] NCCL INFO Channel 04/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:788618 [7] NCCL INFO Channel 18/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:788620 [4] NCCL INFO Channel 18/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:788619 [1] NCCL INFO Channel 03/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:788621 [6] NCCL INFO Channel 05/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:788622 [3] NCCL INFO Channel 05/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:788618 [7] NCCL INFO Channel 19/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:788617 [5] NCCL INFO Channel 00/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:788619 [1] NCCL INFO Channel 04/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:788620 [4] NCCL INFO Channel 19/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:788621 [6] NCCL INFO Channel 06/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:788622 [3] NCCL INFO Channel 06/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:788618 [7] NCCL INFO Channel 20/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:788619 [1] NCCL INFO Channel 05/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:788617 [5] NCCL INFO Channel 01/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:788620 [4] NCCL INFO Channel 20/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:788621 [6] NCCL INFO Channel 07/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:788622 [3] NCCL INFO Channel 07/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:788618 [7] NCCL INFO Channel 21/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:788619 [1] NCCL INFO Channel 06/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:788617 [5] NCCL INFO Channel 02/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:788620 [4] NCCL INFO Channel 21/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:788621 [6] NCCL INFO Channel 08/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:788622 [3] NCCL INFO Channel 08/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:788618 [7] NCCL INFO Channel 22/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:788619 [1] NCCL INFO Channel 07/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:788620 [4] NCCL INFO Channel 22/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:788617 [5] NCCL INFO Channel 03/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:788621 [6] NCCL INFO Channel 09/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:788622 [3] NCCL INFO Channel 09/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782094:788618 [7] NCCL INFO Channel 23/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782091:788620 [4] NCCL INFO Channel 23/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:788619 [1] NCCL INFO Channel 08/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:788617 [5] NCCL INFO Channel 04/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:788621 [6] NCCL INFO Channel 10/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:788622 [3] NCCL INFO Channel 10/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:788619 [1] NCCL INFO Channel 09/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:788617 [5] NCCL INFO Channel 05/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:788621 [6] NCCL INFO Channel 11/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:788622 [3] NCCL INFO Channel 11/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:788619 [1] NCCL INFO Channel 10/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:788617 [5] NCCL INFO Channel 06/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:788621 [6] NCCL INFO Channel 12/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:788622 [3] NCCL INFO Channel 12/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:788619 [1] NCCL INFO Channel 11/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:788617 [5] NCCL INFO Channel 07/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:788622 [3] NCCL INFO Channel 13/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:788621 [6] NCCL INFO Channel 13/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:788619 [1] NCCL INFO Channel 12/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:788617 [5] NCCL INFO Channel 08/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:788622 [3] NCCL INFO Channel 14/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:788621 [6] NCCL INFO Channel 14/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:788619 [1] NCCL INFO Channel 13/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:788617 [5] NCCL INFO Channel 09/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:788622 [3] NCCL INFO Channel 15/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:788621 [6] NCCL INFO Channel 15/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:788619 [1] NCCL INFO Channel 14/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:788617 [5] NCCL INFO Channel 10/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:788622 [3] NCCL INFO Channel 16/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:788621 [6] NCCL INFO Channel 16/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:788619 [1] NCCL INFO Channel 15/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:788622 [3] NCCL INFO Channel 17/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:788617 [5] NCCL INFO Channel 11/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:788621 [6] NCCL INFO Channel 17/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:788619 [1] NCCL INFO Channel 16/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:788622 [3] NCCL INFO Channel 18/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:788617 [5] NCCL INFO Channel 12/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:788621 [6] NCCL INFO Channel 18/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:788619 [1] NCCL INFO Channel 17/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:788622 [3] NCCL INFO Channel 19/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:788617 [5] NCCL INFO Channel 13/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:788621 [6] NCCL INFO Channel 19/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:788619 [1] NCCL INFO Channel 18/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:788622 [3] NCCL INFO Channel 20/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:788617 [5] NCCL INFO Channel 14/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:788621 [6] NCCL INFO Channel 20/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:788619 [1] NCCL INFO Channel 19/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:788622 [3] NCCL INFO Channel 21/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:788617 [5] NCCL INFO Channel 15/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:788621 [6] NCCL INFO Channel 21/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:788619 [1] NCCL INFO Channel 20/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:788622 [3] NCCL INFO Channel 22/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:788617 [5] NCCL INFO Channel 16/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:788621 [6] NCCL INFO Channel 22/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:788619 [1] NCCL INFO Channel 21/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782090:788622 [3] NCCL INFO Channel 23/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:788617 [5] NCCL INFO Channel 17/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782093:788621 [6] NCCL INFO Channel 23/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:788619 [1] NCCL INFO Channel 22/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:788617 [5] NCCL INFO Channel 18/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782088:788619 [1] NCCL INFO Channel 23/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:788617 [5] NCCL INFO Channel 19/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:788617 [5] NCCL INFO Channel 20/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:788617 [5] NCCL INFO Channel 21/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:788617 [5] NCCL INFO Channel 22/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782092:788617 [5] NCCL INFO Channel 23/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test2-worker-0:782087:788616 [0] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:782087:788616 [0] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:782087:788616 [0] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:782088:788619 [1] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:782088:788619 [1] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:782088:788619 [1] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:782089:788623 [2] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:782089:788623 [2] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:782089:788623 [2] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:782090:788622 [3] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:782090:788622 [3] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:782090:788622 [3] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:782091:788620 [4] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:782091:788620 [4] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:782091:788620 [4] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:782092:788617 [5] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:782092:788617 [5] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:782092:788617 [5] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:782094:788618 [7] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:782094:788618 [7] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:782094:788618 [7] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:782093:788621 [6] NCCL INFO Connected all trees +ywang29-vrdb-test2-worker-0:782093:788621 [6] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test2-worker-0:782093:788621 [6] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test2-worker-0:782093:788621 [6] NCCL INFO ncclCommInitRank comm 0x7f841806a5b0 rank 6 nranks 8 cudaDev 6 nvmlDev 6 busId a01c0 commId 0x88ef12943ca706ca - Init COMPLETE +ywang29-vrdb-test2-worker-0:782091:788620 [4] NCCL INFO ncclCommInitRank comm 0x7f4fe406ada0 rank 4 nranks 8 cudaDev 4 nvmlDev 4 busId 901c0 commId 0x88ef12943ca706ca - Init COMPLETE +ywang29-vrdb-test2-worker-0:782087:788616 [0] NCCL INFO ncclCommInitRank comm 0x7f43b006adc0 rank 0 nranks 8 cudaDev 0 nvmlDev 0 busId 101c0 commId 0x88ef12943ca706ca - Init COMPLETE +ywang29-vrdb-test2-worker-0:782089:788623 [2] NCCL INFO ncclCommInitRank comm 0x7f7a8006ad00 rank 2 nranks 8 cudaDev 2 nvmlDev 2 busId 201c0 commId 0x88ef12943ca706ca - Init COMPLETE +ywang29-vrdb-test2-worker-0:782094:788618 [7] NCCL INFO ncclCommInitRank comm 0x7f65a8069ec0 rank 7 nranks 8 cudaDev 7 nvmlDev 7 busId a01d0 commId 0x88ef12943ca706ca - Init COMPLETE +ywang29-vrdb-test2-worker-0:782090:788622 [3] NCCL INFO ncclCommInitRank comm 0x7f058806abe0 rank 3 nranks 8 cudaDev 3 nvmlDev 3 busId 201d0 commId 0x88ef12943ca706ca - Init COMPLETE +ywang29-vrdb-test2-worker-0:782092:788617 [5] NCCL INFO ncclCommInitRank comm 0x7fd22806a520 rank 5 nranks 8 cudaDev 5 nvmlDev 5 busId 901d0 commId 0x88ef12943ca706ca - Init COMPLETE +ywang29-vrdb-test2-worker-0:782088:788619 [1] NCCL INFO ncclCommInitRank comm 0x7ff74c06ab10 rank 1 nranks 8 cudaDev 1 nvmlDev 1 busId 101d0 commId 0x88ef12943ca706ca - Init COMPLETE + 0%| | 1/520 [00:29<4:16:07, 29.61s/it] {'loss': 2.0453, 'grad_norm': 0.004834428236407264, 'learning_rate': 0.005625, 'epoch': 0.0} + 0%| | 1/520 [00:29<4:16:07, 29.61s/it] 0%| | 2/520 [00:33<2:04:30, 14.42s/it] {'loss': 2.0549, 'grad_norm': 0.005248799368144368, 'learning_rate': 0.01125, 'epoch': 0.0} + 0%| | 2/520 [00:33<2:04:30, 14.42s/it] 1%| | 3/520 [00:37<1:22:30, 9.58s/it] {'loss': 2.1899, 'grad_norm': 0.006007856531559498, 'learning_rate': 0.016875, 'epoch': 0.01} + 1%| | 3/520 [00:37<1:22:30, 9.58s/it] 1%| | 4/520 [00:40<1:02:07, 7.22s/it] {'loss': 2.0656, 'grad_norm': 0.004963850757251137, 'learning_rate': 0.0225, 'epoch': 0.01} + 1%| | 4/520 [00:40<1:02:07, 7.22s/it] 1%| | 5/520 [00:44<50:57, 5.94s/it] {'loss': 2.2333, 'grad_norm': 0.005481769746133786, 'learning_rate': 0.028124999999999997, 'epoch': 0.01} + 1%| | 5/520 [00:44<50:57, 5.94s/it] 1%| | 6/520 [00:48<44:12, 5.16s/it] {'loss': 1.6754, 'grad_norm': 0.0028027451229221782, 'learning_rate': 0.03375, 'epoch': 0.01} + 1%| | 6/520 [00:48<44:12, 5.16s/it] 1%|▏ | 7/520 [00:51<40:14, 4.71s/it] {'loss': 2.0776, 'grad_norm': 0.005414389127232943, 'learning_rate': 0.039375, 'epoch': 0.01} + 1%|▏ | 7/520 [00:51<40:14, 4.71s/it] 2%|▏ | 8/520 [00:56<39:05, 4.58s/it] {'loss': 1.6731, 'grad_norm': 0.0022902675361228943, 'learning_rate': 0.045, 'epoch': 0.02} + 2%|▏ | 8/520 [00:56<39:05, 4.58s/it] 2%|▏ | 9/520 [00:59<36:21, 4.27s/it] {'loss': 1.6801, 'grad_norm': 0.0009592633375846499, 'learning_rate': 0.050624999999999996, 'epoch': 0.02} + 2%|▏ | 9/520 [00:59<36:21, 4.27s/it] 2%|▏ | 10/520 [01:03<34:30, 4.06s/it] {'loss': 1.5368, 'grad_norm': 0.0009273406021711265, 'learning_rate': 0.056249999999999994, 'epoch': 0.02} + 2%|▏ | 10/520 [01:03<34:30, 4.06s/it] 2%|▏ | 11/520 [01:07<33:45, 3.98s/it] {'loss': 1.5371, 'grad_norm': 0.0006373521807742546, 'learning_rate': 0.061875, 'epoch': 0.02} + 2%|▏ | 11/520 [01:07<33:45, 3.98s/it] 2%|▏ | 12/520 [01:10<32:45, 3.87s/it] {'loss': 1.4031, 'grad_norm': 0.0004751995992756613, 'learning_rate': 0.0675, 'epoch': 0.02} + 2%|▏ | 12/520 [01:10<32:45, 3.87s/it][2025-10-12 14:31:45,988] [WARNING] [stage3.py:2069:step] 1 pytorch allocator cache flushes since last step. this happens when there is high memory pressure and is detrimental to performance. if this is happening frequently consider adjusting settings to reduce memory consumption. If you are unable to make the cache flushes go away consider adding get_accelerator().empty_cache() calls in your training loop to ensure that all ranks flush their caches at the same time + 2%|▎ | 13/520 [01:15<33:56, 4.02s/it] {'loss': 1.5034, 'grad_norm': 0.0005122147511763336, 'learning_rate': 0.073125, 'epoch': 0.03} + 2%|▎ | 13/520 [01:15<33:56, 4.02s/it] 3%|▎ | 14/520 [01:18<33:21, 3.95s/it] {'loss': 1.526, 'grad_norm': 0.00043912410095298045, 'learning_rate': 0.07875, 'epoch': 0.03} + 3%|▎ | 14/520 [01:18<33:21, 3.95s/it] 3%|▎ | 15/520 [01:22<32:54, 3.91s/it] {'loss': 1.4362, 'grad_norm': 0.00038517780873917546, 'learning_rate': 0.08437499999999999, 'epoch': 0.03} + 3%|▎ | 15/520 [01:22<32:54, 3.91s/it] 3%|▎ | 16/520 [01:26<32:36, 3.88s/it] {'loss': 1.3951, 'grad_norm': 0.00040585434921078955, 'learning_rate': 0.09, 'epoch': 0.03} + 3%|▎ | 16/520 [01:26<32:36, 3.88s/it] 3%|▎ | 17/520 [01:30<32:29, 3.88s/it] {'loss': 1.5361, 'grad_norm': 0.00041746780807785884, 'learning_rate': 0.08999912578191921, 'epoch': 0.03} + 3%|▎ | 17/520 [01:30<32:29, 3.88s/it] 3%|▎ | 18/520 [01:34<32:15, 3.85s/it] {'loss': 1.4003, 'grad_norm': 0.0005091267714472539, 'learning_rate': 0.08999650316164386, 'epoch': 0.03} + 3%|▎ | 18/520 [01:34<32:15, 3.85s/it] 4%|▎ | 19/520 [01:38<32:13, 3.86s/it] {'loss': 1.3808, 'grad_norm': 0.0004242089225954002, 'learning_rate': 0.08999213224107358, 'epoch': 0.04} + 4%|▎ | 19/520 [01:38<32:13, 3.86s/it] 4%|▍ | 20/520 [01:41<32:03, 3.85s/it] {'loss': 1.3692, 'grad_norm': 0.0005191765589155407, 'learning_rate': 0.08998601319003673, 'epoch': 0.04} + 4%|▍ | 20/520 [01:41<32:03, 3.85s/it] 4%|▍ | 21/520 [01:45<31:50, 3.83s/it] {'loss': 1.374, 'grad_norm': 0.0004991861563592257, 'learning_rate': 0.08997814624628374, 'epoch': 0.04} + 4%|▍ | 21/520 [01:45<31:50, 3.83s/it] 4%|▍ | 22/520 [01:49<31:32, 3.80s/it] {'loss': 1.4974, 'grad_norm': 0.0005234691113443424, 'learning_rate': 0.08996853171547793, 'epoch': 0.04} + 4%|▍ | 22/520 [01:49<31:32, 3.80s/it] 4%|▍ | 23/520 [01:53<31:23, 3.79s/it] {'loss': 1.4349, 'grad_norm': 0.0005067311493221449, 'learning_rate': 0.08995716997118361, 'epoch': 0.04} + 4%|▍ | 23/520 [01:53<31:23, 3.79s/it] 5%|▍ | 24/520 [01:56<31:00, 3.75s/it] {'loss': 1.3471, 'grad_norm': 0.000516932952761927, 'learning_rate': 0.0899440614548515, 'epoch': 0.05} + 5%|▍ | 24/520 [01:56<31:00, 3.75s/it] 5%|▍ | 25/520 [02:00<30:58, 3.76s/it] {'loss': 1.4497, 'grad_norm': 0.0006252935494117097, 'learning_rate': 0.08992920667580175, 'epoch': 0.05} + 5%|▍ | 25/520 [02:00<30:58, 3.76s/it] 5%|▌ | 26/520 [02:04<30:50, 3.75s/it] {'loss': 1.3733, 'grad_norm': 0.0005612215658408323, 'learning_rate': 0.08991260621120394, 'epoch': 0.05} + 5%|▌ | 26/520 [02:04<30:50, 3.75s/it] 5%|▌ | 27/520 [02:08<30:49, 3.75s/it] {'loss': 1.3027, 'grad_norm': 0.0005474608866418637, 'learning_rate': 0.0898942607060548, 'epoch': 0.05} + 5%|▌ | 27/520 [02:08<30:49, 3.75s/it] 5%|▌ | 28/520 [02:11<30:37, 3.74s/it] {'loss': 1.334, 'grad_norm': 0.0005875960772022944, 'learning_rate': 0.0898741708731531, 'epoch': 0.05} + 5%|▌ | 28/520 [02:11<30:37, 3.74s/it] 6%|▌ | 29/520 [02:15<30:44, 3.76s/it] {'loss': 1.3478, 'grad_norm': 0.0005855922000083657, 'learning_rate': 0.08985233749307198, 'epoch': 0.06} + 6%|▌ | 29/520 [02:15<30:44, 3.76s/it] 6%|▌ | 30/520 [02:19<30:31, 3.74s/it] {'loss': 1.403, 'grad_norm': 0.0005416727212317927, 'learning_rate': 0.08982876141412854, 'epoch': 0.06} + 6%|▌ | 30/520 [02:19<30:31, 3.74s/it] 6%|▌ | 31/520 [02:23<30:31, 3.75s/it] {'loss': 1.3061, 'grad_norm': 0.0005083151829313778, 'learning_rate': 0.08980344355235101, 'epoch': 0.06} + 6%|▌ | 31/520 [02:23<30:31, 3.75s/it] 6%|▌ | 32/520 [02:26<30:14, 3.72s/it] {'loss': 1.2186, 'grad_norm': 0.0005956309987067088, 'learning_rate': 0.08977638489144306, 'epoch': 0.06} + 6%|▌ | 32/520 [02:26<30:14, 3.72s/it] 6%|▋ | 33/520 [02:30<29:55, 3.69s/it] {'loss': 1.2982, 'grad_norm': 0.0006465499630382057, 'learning_rate': 0.08974758648274558, 'epoch': 0.06} + 6%|▋ | 33/520 [02:30<29:55, 3.69s/it] 7%|▋ | 34/520 [02:33<29:38, 3.66s/it] {'loss': 1.2947, 'grad_norm': 0.0006717363983518703, 'learning_rate': 0.08971704944519591, 'epoch': 0.07} + 7%|▋ | 34/520 [02:33<29:38, 3.66s/it] 7%|▋ | 35/520 [02:37<29:26, 3.64s/it] {'loss': 1.3088, 'grad_norm': 0.0007608489405387575, 'learning_rate': 0.08968477496528426, 'epoch': 0.07} + 7%|▋ | 35/520 [02:37<29:26, 3.64s/it] 7%|▋ | 36/520 [02:41<29:14, 3.62s/it] {'loss': 1.3978, 'grad_norm': 0.0006397257879554323, 'learning_rate': 0.08965076429700775, 'epoch': 0.07} + 7%|▋ | 36/520 [02:41<29:14, 3.62s/it] 7%|▋ | 37/520 [02:44<29:08, 3.62s/it] {'loss': 1.3662, 'grad_norm': 0.000626749655635802, 'learning_rate': 0.08961501876182147, 'epoch': 0.07} + 7%|▋ | 37/520 [02:44<29:08, 3.62s/it] 7%|▋ | 38/520 [02:48<29:12, 3.64s/it] {'loss': 1.4593, 'grad_norm': 0.0006604140319489749, 'learning_rate': 0.08957753974858737, 'epoch': 0.07} + 7%|▋ | 38/520 [02:48<29:12, 3.64s/it] 8%|▊ | 39/520 [02:52<28:58, 3.61s/it] {'loss': 1.333, 'grad_norm': 0.0008290521533998209, 'learning_rate': 0.08953832871352017, 'epoch': 0.07} + 8%|▊ | 39/520 [02:52<28:58, 3.61s/it] 8%|▊ | 40/520 [02:55<29:03, 3.63s/it] {'loss': 1.3563, 'grad_norm': 0.0006670145167020201, 'learning_rate': 0.08949738718013077, 'epoch': 0.08} + 8%|▊ | 40/520 [02:55<29:03, 3.63s/it] 8%|▊ | 41/520 [02:59<29:04, 3.64s/it] {'loss': 1.3321, 'grad_norm': 0.000710105040535405, 'learning_rate': 0.08945471673916715, 'epoch': 0.08} + 8%|▊ | 41/520 [02:59<29:04, 3.64s/it] 8%|▊ | 42/520 [03:03<29:15, 3.67s/it] {'loss': 1.327, 'grad_norm': 0.0009808052496686553, 'learning_rate': 0.08941031904855244, 'epoch': 0.08} + 8%|▊ | 42/520 [03:03<29:15, 3.67s/it] 8%|▊ | 43/520 [03:06<29:39, 3.73s/it] {'loss': 1.2566, 'grad_norm': 0.0007150781930540783, 'learning_rate': 0.0893641958333206, 'epoch': 0.08} + 8%|▊ | 43/520 [03:06<29:39, 3.73s/it] 8%|▊ | 44/520 [03:10<29:48, 3.76s/it] {'loss': 1.3499, 'grad_norm': 0.0008868814637570029, 'learning_rate': 0.08931634888554936, 'epoch': 0.08} + 8%|▊ | 44/520 [03:10<29:48, 3.76s/it] 9%|▊ | 45/520 [03:14<29:53, 3.78s/it] {'loss': 1.3515, 'grad_norm': 0.0008857750838257474, 'learning_rate': 0.08926678006429056, 'epoch': 0.09} + 9%|▊ | 45/520 [03:14<29:53, 3.78s/it] 9%|▉ | 46/520 [03:18<29:48, 3.77s/it] {'loss': 1.394, 'grad_norm': 0.0009277734537240646, 'learning_rate': 0.08921549129549797, 'epoch': 0.09} + 9%|▉ | 46/520 [03:18<29:48, 3.77s/it] 9%|▉ | 47/520 [03:22<29:44, 3.77s/it] {'loss': 1.3074, 'grad_norm': 0.0008659985076296659, 'learning_rate': 0.08916248457195244, 'epoch': 0.09} + 9%|▉ | 47/520 [03:22<29:44, 3.77s/it] 9%|▉ | 48/520 [03:25<29:41, 3.77s/it] {'loss': 1.3249, 'grad_norm': 0.0011378171073620424, 'learning_rate': 0.08910776195318447, 'epoch': 0.09} + 9%|▉ | 48/520 [03:25<29:41, 3.77s/it] 9%|▉ | 49/520 [03:29<29:40, 3.78s/it] {'loss': 1.3432, 'grad_norm': 0.0009039028988845788, 'learning_rate': 0.08905132556539418, 'epoch': 0.09} + 9%|▉ | 49/520 [03:29<29:40, 3.78s/it] 10%|▉ | 50/520 [03:33<29:35, 3.78s/it] {'loss': 1.3387, 'grad_norm': 0.0009130678653168482, 'learning_rate': 0.08899317760136871, 'epoch': 0.1} + 10%|▉ | 50/520 [03:33<29:35, 3.78s/it] 10%|▉ | 51/520 [03:37<29:28, 3.77s/it] {'loss': 1.2725, 'grad_norm': 0.0009854610090716033, 'learning_rate': 0.088933320320397, 'epoch': 0.1} + 10%|▉ | 51/520 [03:37<29:28, 3.77s/it] 10%|█ | 52/520 [03:40<29:16, 3.75s/it] {'loss': 1.4083, 'grad_norm': 0.00101967147079308, 'learning_rate': 0.08887175604818205, 'epoch': 0.1} + 10%|█ | 52/520 [03:40<29:16, 3.75s/it] 10%|█ | 53/520 [03:44<29:15, 3.76s/it] {'loss': 1.3775, 'grad_norm': 0.0009418695231800186, 'learning_rate': 0.08880848717675054, 'epoch': 0.1} + 10%|█ | 53/520 [03:44<29:15, 3.76s/it] 10%|█ | 54/520 [03:48<28:57, 3.73s/it] {'loss': 1.3186, 'grad_norm': 0.0009448298355502936, 'learning_rate': 0.08874351616435985, 'epoch': 0.1} + 10%|█ | 54/520 [03:48<28:57, 3.73s/it] 11%|█ | 55/520 [03:52<28:56, 3.73s/it] {'loss': 1.2643, 'grad_norm': 0.0010233636830717062, 'learning_rate': 0.08867684553540259, 'epoch': 0.11} + 11%|█ | 55/520 [03:52<28:56, 3.73s/it] 11%|█ | 56/520 [03:55<28:50, 3.73s/it] {'loss': 1.3923, 'grad_norm': 0.0009666986316487882, 'learning_rate': 0.08860847788030851, 'epoch': 0.11} + 11%|█ | 56/520 [03:55<28:50, 3.73s/it] 11%|█ | 57/520 [03:59<28:29, 3.69s/it] {'loss': 1.2652, 'grad_norm': 0.0011001430011324496, 'learning_rate': 0.08853841585544384, 'epoch': 0.11} + 11%|█ | 57/520 [03:59<28:29, 3.69s/it] 11%|█ | 58/520 [04:03<28:12, 3.66s/it] {'loss': 1.4045, 'grad_norm': 0.0008882868873877195, 'learning_rate': 0.08846666218300807, 'epoch': 0.11} + 11%|█ | 58/520 [04:03<28:12, 3.66s/it] 11%|█▏ | 59/520 [04:06<28:06, 3.66s/it] {'loss': 1.219, 'grad_norm': 0.0009968595499529878, 'learning_rate': 0.08839321965092825, 'epoch': 0.11} + 11%|█▏ | 59/520 [04:06<28:06, 3.66s/it] 12%|█▏ | 60/520 [04:10<28:06, 3.67s/it] {'loss': 1.319, 'grad_norm': 0.0009566560932202728, 'learning_rate': 0.08831809111275053, 'epoch': 0.12} + 12%|█▏ | 60/520 [04:10<28:06, 3.67s/it] 12%|█▏ | 61/520 [04:14<28:29, 3.73s/it] {'loss': 1.2935, 'grad_norm': 0.001036188910503026, 'learning_rate': 0.08824127948752948, 'epoch': 0.12} + 12%|█▏ | 61/520 [04:14<28:29, 3.73s/it] 12%|█▏ | 62/520 [04:18<28:40, 3.76s/it] {'loss': 1.302, 'grad_norm': 0.001154273940481832, 'learning_rate': 0.08816278775971445, 'epoch': 0.12} + 12%|█▏ | 62/520 [04:18<28:40, 3.76s/it] 12%|█▏ | 63/520 [04:21<28:46, 3.78s/it] {'loss': 1.3007, 'grad_norm': 0.0010765782301301377, 'learning_rate': 0.08808261897903381, 'epoch': 0.12} + 12%|█▏ | 63/520 [04:21<28:46, 3.78s/it] 12%|█▏ | 64/520 [04:25<28:33, 3.76s/it] {'loss': 1.3199, 'grad_norm': 0.0011089446793785488, 'learning_rate': 0.08800077626037633, 'epoch': 0.12} + 12%|█▏ | 64/520 [04:25<28:33, 3.76s/it] 12%|█▎ | 65/520 [04:29<28:35, 3.77s/it] {'loss': 1.3277, 'grad_norm': 0.0012648290595595657, 'learning_rate': 0.08791726278367021, 'epoch': 0.12} + 12%|█▎ | 65/520 [04:29<28:35, 3.77s/it] 13%|█▎ | 66/520 [04:33<28:55, 3.82s/it] {'loss': 1.2681, 'grad_norm': 0.001030861170438743, 'learning_rate': 0.08783208179375951, 'epoch': 0.13} + 13%|█▎ | 66/520 [04:33<28:55, 3.82s/it] 13%|█▎ | 67/520 [04:37<28:52, 3.82s/it] {'loss': 1.1915, 'grad_norm': 0.0011204270068607845, 'learning_rate': 0.08774523660027807, 'epoch': 0.13} + 13%|█▎ | 67/520 [04:37<28:52, 3.82s/it] 13%|█▎ | 68/520 [04:41<28:53, 3.83s/it] {'loss': 1.2572, 'grad_norm': 0.0010442101157166995, 'learning_rate': 0.08765673057752092, 'epoch': 0.13} + 13%|█▎ | 68/520 [04:41<28:53, 3.83s/it] 13%|█▎ | 69/520 [04:44<28:31, 3.80s/it] {'loss': 1.2363, 'grad_norm': 0.0011505120921622288, 'learning_rate': 0.08756656716431321, 'epoch': 0.13} + 13%|█▎ | 69/520 [04:44<28:31, 3.80s/it] 13%|█▎ | 70/520 [04:48<28:20, 3.78s/it] {'loss': 1.2661, 'grad_norm': 0.0012107053850471533, 'learning_rate': 0.08747474986387654, 'epoch': 0.13} + 13%|█▎ | 70/520 [04:48<28:20, 3.78s/it] 14%|█▎ | 71/520 [04:52<28:32, 3.81s/it] {'loss': 1.2084, 'grad_norm': 0.0010691970674529217, 'learning_rate': 0.08738128224369285, 'epoch': 0.14} + 14%|█▎ | 71/520 [04:52<28:32, 3.81s/it] 14%|█▍ | 72/520 [04:56<28:22, 3.80s/it] {'loss': 1.3556, 'grad_norm': 0.0011033118836674002, 'learning_rate': 0.08728616793536588, 'epoch': 0.14} + 14%|█▍ | 72/520 [04:56<28:22, 3.80s/it] 14%|█▍ | 73/520 [05:00<28:25, 3.82s/it] {'loss': 1.1849, 'grad_norm': 0.0012172368402162967, 'learning_rate': 0.08718941063447996, 'epoch': 0.14} + 14%|█▍ | 73/520 [05:00<28:25, 3.82s/it] 14%|█▍ | 74/520 [05:03<28:13, 3.80s/it] {'loss': 1.2857, 'grad_norm': 0.001259368113117326, 'learning_rate': 0.08709101410045653, 'epoch': 0.14} + 14%|█▍ | 74/520 [05:03<28:13, 3.80s/it] 14%|█▍ | 75/520 [05:07<28:18, 3.82s/it] {'loss': 1.2072, 'grad_norm': 0.0010768973642208616, 'learning_rate': 0.08699098215640798, 'epoch': 0.14} + 14%|█▍ | 75/520 [05:07<28:18, 3.82s/it] 15%|█▍ | 76/520 [05:11<28:05, 3.80s/it] {'loss': 1.3263, 'grad_norm': 0.0009676649381392729, 'learning_rate': 0.08688931868898919, 'epoch': 0.15} + 15%|█▍ | 76/520 [05:11<28:05, 3.80s/it] 15%|█▍ | 77/520 [05:15<27:41, 3.75s/it] {'loss': 1.1408, 'grad_norm': 0.0013263483268763363, 'learning_rate': 0.0867860276482464, 'epoch': 0.15} + 15%|█▍ | 77/520 [05:15<27:41, 3.75s/it] 15%|█▌ | 78/520 [05:18<27:31, 3.74s/it] {'loss': 1.243, 'grad_norm': 0.0011940504519385527, 'learning_rate': 0.08668111304746388, 'epoch': 0.15} + 15%|█▌ | 78/520 [05:18<27:31, 3.74s/it] 15%|█▌ | 79/520 [05:22<27:47, 3.78s/it] {'loss': 1.2239, 'grad_norm': 0.0011144064671001997, 'learning_rate': 0.0865745789630079, 'epoch': 0.15} + 15%|█▌ | 79/520 [05:22<27:47, 3.78s/it] 15%|█▌ | 80/520 [05:26<27:38, 3.77s/it] {'loss': 1.3065, 'grad_norm': 0.0011731690312654663, 'learning_rate': 0.08646642953416835, 'epoch': 0.15} + 15%|█▌ | 80/520 [05:26<27:38, 3.77s/it] 16%|█▌ | 81/520 [05:29<27:19, 3.73s/it] {'loss': 1.364, 'grad_norm': 0.0014974325273921936, 'learning_rate': 0.08635666896299792, 'epoch': 0.16} + 16%|█▌ | 81/520 [05:29<27:19, 3.73s/it] 16%|█▌ | 82/520 [05:33<27:16, 3.74s/it] {'loss': 1.2929, 'grad_norm': 0.0012100657022697313, 'learning_rate': 0.08624530151414893, 'epoch': 0.16} + 16%|█▌ | 82/520 [05:33<27:16, 3.74s/it] 16%|█▌ | 83/520 [05:37<27:07, 3.72s/it] {'loss': 1.304, 'grad_norm': 0.0012996839882032595, 'learning_rate': 0.08613233151470744, 'epoch': 0.16} + 16%|█▌ | 83/520 [05:37<27:07, 3.72s/it] 16%|█▌ | 84/520 [05:41<26:57, 3.71s/it] {'loss': 1.3169, 'grad_norm': 0.001316073039934338, 'learning_rate': 0.08601776335402529, 'epoch': 0.16} + 16%|█▌ | 84/520 [05:41<26:57, 3.71s/it] 16%|█▋ | 85/520 [05:44<26:52, 3.71s/it] {'loss': 1.3536, 'grad_norm': 0.0011977683267052244, 'learning_rate': 0.08590160148354949, 'epoch': 0.16} + 16%|█▋ | 85/520 [05:44<26:52, 3.71s/it] 17%|█▋ | 86/520 [05:48<26:43, 3.69s/it] {'loss': 1.3433, 'grad_norm': 0.00119894102351539, 'learning_rate': 0.08578385041664925, 'epoch': 0.17} + 17%|█▋ | 86/520 [05:48<26:43, 3.69s/it] 17%|█▋ | 87/520 [05:52<26:36, 3.69s/it] {'loss': 1.2601, 'grad_norm': 0.0011889653754787418, 'learning_rate': 0.08566451472844065, 'epoch': 0.17} + 17%|█▋ | 87/520 [05:52<26:36, 3.69s/it] 17%|█▋ | 88/520 [05:55<26:30, 3.68s/it] {'loss': 1.2001, 'grad_norm': 0.0009497526229324287, 'learning_rate': 0.08554359905560885, 'epoch': 0.17} + 17%|█▋ | 88/520 [05:55<26:30, 3.68s/it] 17%|█▋ | 89/520 [05:59<26:19, 3.66s/it] {'loss': 1.3026, 'grad_norm': 0.0012858867390367163, 'learning_rate': 0.08542110809622798, 'epoch': 0.17} + 17%|█▋ | 89/520 [05:59<26:19, 3.66s/it] 17%|█▋ | 90/520 [06:03<26:17, 3.67s/it] {'loss': 1.2387, 'grad_norm': 0.0012302321094498827, 'learning_rate': 0.08529704660957854, 'epoch': 0.17} + 17%|█▋ | 90/520 [06:03<26:17, 3.67s/it] 18%|█▊ | 91/520 [06:06<26:40, 3.73s/it] {'loss': 1.3047, 'grad_norm': 0.001136524273659427, 'learning_rate': 0.08517141941596251, 'epoch': 0.17} + 18%|█▊ | 91/520 [06:06<26:40, 3.73s/it] 18%|█▊ | 92/520 [06:10<26:38, 3.73s/it] {'loss': 1.2451, 'grad_norm': 0.0012785460048604447, 'learning_rate': 0.0850442313965161, 'epoch': 0.18} + 18%|█▊ | 92/520 [06:10<26:38, 3.73s/it] 18%|█▊ | 93/520 [06:14<26:29, 3.72s/it] {'loss': 1.2555, 'grad_norm': 0.001310439252320074, 'learning_rate': 0.08491548749301997, 'epoch': 0.18} + 18%|█▊ | 93/520 [06:14<26:29, 3.72s/it] 18%|█▊ | 94/520 [06:18<26:18, 3.71s/it] {'loss': 1.3341, 'grad_norm': 0.0012352189174429902, 'learning_rate': 0.08478519270770743, 'epoch': 0.18} + 18%|█▊ | 94/520 [06:18<26:18, 3.71s/it] 18%|█▊ | 95/520 [06:21<26:04, 3.68s/it] {'loss': 1.2375, 'grad_norm': 0.0014292154791196558, 'learning_rate': 0.0846533521030699, 'epoch': 0.18} + 18%|█▊ | 95/520 [06:21<26:04, 3.68s/it] 18%|█▊ | 96/520 [06:25<26:02, 3.68s/it] {'loss': 1.2597, 'grad_norm': 0.0011180380174879117, 'learning_rate': 0.08451997080166028, 'epoch': 0.18} + 18%|█▊ | 96/520 [06:25<26:02, 3.68s/it] 19%|█▊ | 97/520 [06:29<25:56, 3.68s/it] {'loss': 1.2317, 'grad_norm': 0.001446778405307539, 'learning_rate': 0.08438505398589392, 'epoch': 0.19} + 19%|█▊ | 97/520 [06:29<25:56, 3.68s/it] 19%|█▉ | 98/520 [06:32<25:52, 3.68s/it] {'loss': 1.2331, 'grad_norm': 0.0010469120932603143, 'learning_rate': 0.08424860689784724, 'epoch': 0.19} + 19%|█▉ | 98/520 [06:32<25:52, 3.68s/it] 19%|█▉ | 99/520 [06:36<25:44, 3.67s/it] {'loss': 1.2271, 'grad_norm': 0.0013062816516485928, 'learning_rate': 0.08411063483905408, 'epoch': 0.19} + 19%|█▉ | 99/520 [06:36<25:44, 3.67s/it] 19%|█▉ | 100/520 [06:40<25:39, 3.67s/it] {'loss': 1.2118, 'grad_norm': 0.001157899769289443, 'learning_rate': 0.08397114317029974, 'epoch': 0.19} + 19%|█▉ | 100/520 [06:40<25:39, 3.67s/it] 19%|█▉ | 101/520 [06:43<25:29, 3.65s/it] {'loss': 1.2537, 'grad_norm': 0.001301562356620008, 'learning_rate': 0.08383013731141259, 'epoch': 0.19} + 19%|█▉ | 101/520 [06:43<25:29, 3.65s/it] 20%|█▉ | 102/520 [06:47<25:37, 3.68s/it] {'loss': 1.2478, 'grad_norm': 0.001312006666502576, 'learning_rate': 0.08368762274105356, 'epoch': 0.2} + 20%|█▉ | 102/520 [06:47<25:37, 3.68s/it] 20%|█▉ | 103/520 [06:51<25:55, 3.73s/it] {'loss': 1.1914, 'grad_norm': 0.0011621380011128998, 'learning_rate': 0.08354360499650332, 'epoch': 0.2} + 20%|█▉ | 103/520 [06:51<25:55, 3.73s/it] 20%|██ | 104/520 [06:55<25:54, 3.74s/it] {'loss': 1.2605, 'grad_norm': 0.0013142932964250338, 'learning_rate': 0.083398089673447, 'epoch': 0.2} + 20%|██ | 104/520 [06:55<25:54, 3.74s/it] 20%|██ | 105/520 [06:58<25:47, 3.73s/it] {'loss': 1.2516, 'grad_norm': 0.0012178639853738125, 'learning_rate': 0.0832510824257569, 'epoch': 0.2} + 20%|██ | 105/520 [06:58<25:47, 3.73s/it] 20%|██ | 106/520 [07:02<25:42, 3.73s/it] {'loss': 1.2341, 'grad_norm': 0.0011041245394642148, 'learning_rate': 0.08310258896527278, 'epoch': 0.2} + 20%|██ | 106/520 [07:02<25:42, 3.73s/it] 21%|██ | 107/520 [07:06<25:36, 3.72s/it] {'loss': 1.2097, 'grad_norm': 0.0011932805081538436, 'learning_rate': 0.08295261506157986, 'epoch': 0.21} + 21%|██ | 107/520 [07:06<25:36, 3.72s/it] 21%|██ | 108/520 [07:09<25:33, 3.72s/it] {'loss': 1.2072, 'grad_norm': 0.0012958913459658853, 'learning_rate': 0.08280116654178472, 'epoch': 0.21} + 21%|██ | 108/520 [07:09<25:33, 3.72s/it] 21%|██ | 109/520 [07:13<25:23, 3.71s/it] {'loss': 1.1844, 'grad_norm': 0.0010905371801876811, 'learning_rate': 0.08264824929028888, 'epoch': 0.21} + 21%|██ | 109/520 [07:13<25:23, 3.71s/it] 21%|██ | 110/520 [07:17<25:07, 3.68s/it] {'loss': 1.379, 'grad_norm': 0.0013058385931613973, 'learning_rate': 0.0824938692485602, 'epoch': 0.21} + 21%|██ | 110/520 [07:17<25:07, 3.68s/it] 21%|██▏ | 111/520 [07:20<24:55, 3.66s/it] {'loss': 1.3718, 'grad_norm': 0.0013806940062280136, 'learning_rate': 0.08233803241490194, 'epoch': 0.21} + 21%|██▏ | 111/520 [07:20<24:55, 3.66s/it] 22%|██▏ | 112/520 [07:24<24:41, 3.63s/it] {'loss': 1.2701, 'grad_norm': 0.0012492346274198807, 'learning_rate': 0.08218074484421976, 'epoch': 0.22} + 22%|██▏ | 112/520 [07:24<24:41, 3.63s/it] 22%|██▏ | 113/520 [07:27<24:33, 3.62s/it] {'loss': 1.165, 'grad_norm': 0.0011980244109805066, 'learning_rate': 0.0820220126477865, 'epoch': 0.22} + 22%|██▏ | 113/520 [07:27<24:33, 3.62s/it] 22%|██▏ | 114/520 [07:31<24:26, 3.61s/it] {'loss': 1.2616, 'grad_norm': 0.0012605425281660265, 'learning_rate': 0.08186184199300463, 'epoch': 0.22} + 22%|██▏ | 114/520 [07:31<24:26, 3.61s/it] 22%|██▏ | 115/520 [07:35<24:26, 3.62s/it] {'loss': 1.3427, 'grad_norm': 0.0012627686342131152, 'learning_rate': 0.0817002391031667, 'epoch': 0.22} + 22%|██▏ | 115/520 [07:35<24:26, 3.62s/it] 22%|██▏ | 116/520 [07:38<24:20, 3.62s/it] {'loss': 1.3623, 'grad_norm': 0.0012156297996287764, 'learning_rate': 0.08153721025721355, 'epoch': 0.22} + 22%|██▏ | 116/520 [07:38<24:20, 3.62s/it] 22%|██▎ | 117/520 [07:42<24:15, 3.61s/it] {'loss': 1.3217, 'grad_norm': 0.0013097898095208345, 'learning_rate': 0.08137276178949024, 'epoch': 0.23} + 22%|██▎ | 117/520 [07:42<24:15, 3.61s/it] 23%|██▎ | 118/520 [07:45<24:12, 3.61s/it] {'loss': 1.2583, 'grad_norm': 0.0012205358140669753, 'learning_rate': 0.08120690008950007, 'epoch': 0.23} + 23%|██▎ | 118/520 [07:46<24:12, 3.61s/it] 23%|██▎ | 119/520 [07:49<24:07, 3.61s/it] {'loss': 1.2106, 'grad_norm': 0.001311948646175857, 'learning_rate': 0.08103963160165627, 'epoch': 0.23} + 23%|██▎ | 119/520 [07:49<24:07, 3.61s/it] 23%|██▎ | 120/520 [07:53<24:02, 3.61s/it] {'loss': 1.2189, 'grad_norm': 0.0014297934232023004, 'learning_rate': 0.0808709628250315, 'epoch': 0.23} + 23%|██▎ | 120/520 [07:53<24:02, 3.61s/it] 23%|██▎ | 121/520 [07:56<24:04, 3.62s/it] {'loss': 1.2645, 'grad_norm': 0.0013161189390362724, 'learning_rate': 0.08070090031310558, 'epoch': 0.23} + 23%|██▎ | 121/520 [07:56<24:04, 3.62s/it] 23%|██▎ | 122/520 [08:00<24:00, 3.62s/it] {'loss': 1.1872, 'grad_norm': 0.0012233245552801823, 'learning_rate': 0.0805294506735106, 'epoch': 0.23} + 23%|██▎ | 122/520 [08:00<24:00, 3.62s/it] 24%|██▎ | 123/520 [08:04<23:58, 3.62s/it] {'loss': 1.2719, 'grad_norm': 0.0012420555673891556, 'learning_rate': 0.08035662056777432, 'epoch': 0.24} + 24%|██▎ | 123/520 [08:04<23:58, 3.62s/it] 24%|██▍ | 124/520 [08:07<24:02, 3.64s/it] {'loss': 1.2319, 'grad_norm': 0.0013895766828962324, 'learning_rate': 0.08018241671106134, 'epoch': 0.24} + 24%|██▍ | 124/520 [08:07<24:02, 3.64s/it] 24%|██▍ | 125/520 [08:11<23:59, 3.64s/it] {'loss': 1.2344, 'grad_norm': 0.0012785757939418239, 'learning_rate': 0.08000684587191216, 'epoch': 0.24} + 24%|██▍ | 125/520 [08:11<23:59, 3.64s/it] 24%|██▍ | 126/520 [08:15<25:27, 3.88s/it] {'loss': 1.2048, 'grad_norm': 0.0010662648882711282, 'learning_rate': 0.07982991487198024, 'epoch': 0.24} + 24%|██▍ | 126/520 [08:15<25:27, 3.88s/it] 24%|██▍ | 127/520 [08:19<24:56, 3.81s/it] {'loss': 1.2229, 'grad_norm': 0.0014569709098771293, 'learning_rate': 0.07965163058576682, 'epoch': 0.24} + 24%|██▍ | 127/520 [08:19<24:56, 3.81s/it] 25%|██▍ | 128/520 [08:23<24:44, 3.79s/it] {'loss': 1.2562, 'grad_norm': 0.0012760891082461703, 'learning_rate': 0.079471999940354, 'epoch': 0.25} + 25%|██▍ | 128/520 [08:23<24:44, 3.79s/it] 25%|██▍ | 129/520 [08:26<24:24, 3.74s/it] {'loss': 1.2291, 'grad_norm': 0.0011918759820180285, 'learning_rate': 0.07929102991513548, 'epoch': 0.25} + 25%|██▍ | 129/520 [08:26<24:24, 3.74s/it] 25%|██▌ | 130/520 [08:30<24:05, 3.71s/it] {'loss': 1.2474, 'grad_norm': 0.0011683708928294208, 'learning_rate': 0.07910872754154538, 'epoch': 0.25} + 25%|██▌ | 130/520 [08:30<24:05, 3.71s/it] 25%|██▌ | 131/520 [08:34<23:49, 3.67s/it] {'loss': 1.1721, 'grad_norm': 0.001112705111545245, 'learning_rate': 0.0789250999027851, 'epoch': 0.25} + 25%|██▌ | 131/520 [08:34<23:49, 3.67s/it] 25%|██▌ | 132/520 [08:37<23:34, 3.65s/it] {'loss': 1.2997, 'grad_norm': 0.0014371868995740253, 'learning_rate': 0.07874015413354804, 'epoch': 0.25} + 25%|██▌ | 132/520 [08:37<23:34, 3.65s/it] 26%|██▌ | 133/520 [08:41<23:27, 3.64s/it] {'loss': 1.2172, 'grad_norm': 0.0014267238648757057, 'learning_rate': 0.07855389741974245, 'epoch': 0.26} + 26%|██▌ | 133/520 [08:41<23:27, 3.64s/it] 26%|██▌ | 134/520 [08:44<23:22, 3.63s/it] {'loss': 1.2837, 'grad_norm': 0.0012922114363126604, 'learning_rate': 0.0783663369982122, 'epoch': 0.26} + 26%|██▌ | 134/520 [08:44<23:22, 3.63s/it] 26%|██▌ | 135/520 [08:48<23:11, 3.61s/it] {'loss': 1.3393, 'grad_norm': 0.0013742858306566596, 'learning_rate': 0.07817748015645558, 'epoch': 0.26} + 26%|██▌ | 135/520 [08:48<23:11, 3.61s/it] 26%|██▌ | 136/520 [08:52<23:02, 3.60s/it] {'loss': 1.2958, 'grad_norm': 0.0013260173820893789, 'learning_rate': 0.07798733423234218, 'epoch': 0.26} + 26%|██▌ | 136/520 [08:52<23:02, 3.60s/it] 26%|██▋ | 137/520 [08:55<22:57, 3.60s/it] {'loss': 1.2109, 'grad_norm': 0.001467462485872576, 'learning_rate': 0.07779590661382778, 'epoch': 0.26} + 26%|██▋ | 137/520 [08:55<22:57, 3.60s/it] 27%|██▋ | 138/520 [08:59<22:52, 3.59s/it] {'loss': 1.2234, 'grad_norm': 0.0011958016496069908, 'learning_rate': 0.07760320473866726, 'epoch': 0.27} + 27%|██▋ | 138/520 [08:59<22:52, 3.59s/it] 27%|██▋ | 139/520 [09:02<22:52, 3.60s/it] {'loss': 1.0854, 'grad_norm': 0.0011359682669825958, 'learning_rate': 0.0774092360941257, 'epoch': 0.27} + 27%|██▋ | 139/520 [09:02<22:52, 3.60s/it] 27%|██▋ | 140/520 [09:06<22:58, 3.63s/it] {'loss': 1.2267, 'grad_norm': 0.0011341326253695736, 'learning_rate': 0.07721400821668732, 'epoch': 0.27} + 27%|██▋ | 140/520 [09:06<22:58, 3.63s/it] 27%|██▋ | 141/520 [09:10<22:52, 3.62s/it] {'loss': 1.3158, 'grad_norm': 0.0011926319652761216, 'learning_rate': 0.07701752869176286, 'epoch': 0.27} + 27%|██▋ | 141/520 [09:10<22:52, 3.62s/it] 27%|██▋ | 142/520 [09:13<22:43, 3.61s/it] {'loss': 1.2306, 'grad_norm': 0.0011226977831989393, 'learning_rate': 0.07681980515339464, 'epoch': 0.27} + 27%|██▋ | 142/520 [09:13<22:43, 3.61s/it] 28%|██▊ | 143/520 [09:17<22:42, 3.62s/it] {'loss': 1.2427, 'grad_norm': 0.001353206486101452, 'learning_rate': 0.07662084528396011, 'epoch': 0.28} + 28%|██▊ | 143/520 [09:17<22:42, 3.62s/it] 28%|██▊ | 144/520 [09:21<22:43, 3.63s/it] {'loss': 1.23, 'grad_norm': 0.0013962475748343908, 'learning_rate': 0.07642065681387326, 'epoch': 0.28} + 28%|██▊ | 144/520 [09:21<22:43, 3.63s/it] 28%|██▊ | 145/520 [09:24<23:11, 3.71s/it] {'loss': 1.152, 'grad_norm': 0.0012088310400576016, 'learning_rate': 0.07621924752128437, 'epoch': 0.28} + 28%|██▊ | 145/520 [09:24<23:11, 3.71s/it] 28%|██▊ | 146/520 [09:28<23:27, 3.76s/it] {'loss': 1.28, 'grad_norm': 0.001270731788038277, 'learning_rate': 0.07601662523177762, 'epoch': 0.28} + 28%|██▊ | 146/520 [09:28<23:27, 3.76s/it] 28%|██▊ | 147/520 [09:32<23:34, 3.79s/it] {'loss': 1.1976, 'grad_norm': 0.001325941065337015, 'learning_rate': 0.0758127978180672, 'epoch': 0.28} + 28%|██▊ | 147/520 [09:32<23:34, 3.79s/it] 28%|██▊ | 148/520 [09:36<23:34, 3.80s/it] {'loss': 1.2213, 'grad_norm': 0.0012246530068957733, 'learning_rate': 0.07560777319969136, 'epoch': 0.28} + 28%|██▊ | 148/520 [09:36<23:34, 3.80s/it] 29%|██▊ | 149/520 [09:40<23:38, 3.82s/it] {'loss': 1.1624, 'grad_norm': 0.0012508395340215662, 'learning_rate': 0.07540155934270472, 'epoch': 0.29} + 29%|██▊ | 149/520 [09:40<23:38, 3.82s/it] 29%|██▉ | 150/520 [09:44<23:33, 3.82s/it] {'loss': 1.3744, 'grad_norm': 0.001279563383126935, 'learning_rate': 0.07519416425936865, 'epoch': 0.29} + 29%|██▉ | 150/520 [09:44<23:33, 3.82s/it] 29%|██▉ | 151/520 [09:48<23:35, 3.84s/it] {'loss': 1.2132, 'grad_norm': 0.0013264673556815587, 'learning_rate': 0.07498559600784017, 'epoch': 0.29} + 29%|██▉ | 151/520 [09:48<23:35, 3.84s/it] 29%|██▉ | 152/520 [09:51<23:24, 3.82s/it] {'loss': 1.1821, 'grad_norm': 0.0013386704391022135, 'learning_rate': 0.07477586269185868, 'epoch': 0.29} + 29%|██▉ | 152/520 [09:51<23:24, 3.82s/it] 29%|██▉ | 153/520 [09:55<23:01, 3.77s/it] {'loss': 1.2097, 'grad_norm': 0.0012574905924577094, 'learning_rate': 0.07456497246043112, 'epoch': 0.29} + 29%|██▉ | 153/520 [09:55<23:01, 3.77s/it] 30%|██▉ | 154/520 [09:59<22:41, 3.72s/it] {'loss': 1.2919, 'grad_norm': 0.0012080465273849037, 'learning_rate': 0.07435293350751546, 'epoch': 0.3} + 30%|██▉ | 154/520 [09:59<22:41, 3.72s/it] 30%|██▉ | 155/520 [10:02<22:26, 3.69s/it] {'loss': 1.2087, 'grad_norm': 0.001357357368318336, 'learning_rate': 0.07413975407170216, 'epoch': 0.3} + 30%|██▉ | 155/520 [10:02<22:26, 3.69s/it] 30%|███ | 156/520 [10:06<22:20, 3.68s/it] {'loss': 1.2332, 'grad_norm': 0.001420956228383493, 'learning_rate': 0.07392544243589426, 'epoch': 0.3} + 30%|███ | 156/520 [10:06<22:20, 3.68s/it] 30%|███ | 157/520 [10:10<22:11, 3.67s/it] {'loss': 1.2737, 'grad_norm': 0.0012063527055862083, 'learning_rate': 0.07371000692698539, 'epoch': 0.3} + 30%|███ | 157/520 [10:10<22:11, 3.67s/it] 30%|███ | 158/520 [10:13<22:03, 3.66s/it] {'loss': 1.2182, 'grad_norm': 0.0013049498398044219, 'learning_rate': 0.0734934559155363, 'epoch': 0.3} + 30%|███ | 158/520 [10:13<22:03, 3.66s/it] 31%|███ | 159/520 [10:17<22:05, 3.67s/it] {'loss': 1.2643, 'grad_norm': 0.0013058609361910332, 'learning_rate': 0.07327579781544963, 'epoch': 0.31} + 31%|███ | 159/520 [10:17<22:05, 3.67s/it] 31%|███ | 160/520 [10:21<22:32, 3.76s/it] {'loss': 1.266, 'grad_norm': 0.001332800464391292, 'learning_rate': 0.07305704108364301, 'epoch': 0.31} + 31%|███ | 160/520 [10:21<22:32, 3.76s/it] 31%|███ | 161/520 [10:25<22:37, 3.78s/it] {'loss': 1.2419, 'grad_norm': 0.0012769431608013455, 'learning_rate': 0.07283719421972047, 'epoch': 0.31} + 31%|███ | 161/520 [10:25<22:37, 3.78s/it] 31%|███ | 162/520 [10:28<22:41, 3.80s/it] {'loss': 1.2172, 'grad_norm': 0.0012164098701714031, 'learning_rate': 0.07261626576564213, 'epoch': 0.31} + 31%|███ | 162/520 [10:28<22:41, 3.80s/it] 31%|███▏ | 163/520 [10:32<22:40, 3.81s/it] {'loss': 1.1513, 'grad_norm': 0.0014657884315559325, 'learning_rate': 0.07239426430539243, 'epoch': 0.31} + 31%|███▏ | 163/520 [10:32<22:40, 3.81s/it] 32%|███▏ | 164/520 [10:36<22:20, 3.77s/it] {'loss': 1.1096, 'grad_norm': 0.001220338668319855, 'learning_rate': 0.07217119846464648, 'epoch': 0.32} + 32%|███▏ | 164/520 [10:36<22:20, 3.77s/it] 32%|███▏ | 165/520 [10:40<21:56, 3.71s/it] {'loss': 1.2666, 'grad_norm': 0.0012476121356378316, 'learning_rate': 0.07194707691043502, 'epoch': 0.32} + 32%|███▏ | 165/520 [10:40<21:56, 3.71s/it] 32%|███▏ | 166/520 [10:43<21:40, 3.67s/it] {'loss': 1.2249, 'grad_norm': 0.0014050568051312455, 'learning_rate': 0.07172190835080756, 'epoch': 0.32} + 32%|███▏ | 166/520 [10:43<21:40, 3.67s/it] 32%|███▏ | 167/520 [10:47<21:31, 3.66s/it] {'loss': 1.2224, 'grad_norm': 0.0012948202612263768, 'learning_rate': 0.07149570153449421, 'epoch': 0.32} + 32%|███▏ | 167/520 [10:47<21:31, 3.66s/it] 32%|███▏ | 168/520 [10:50<21:28, 3.66s/it] {'loss': 1.1637, 'grad_norm': 0.0012544404709595616, 'learning_rate': 0.07126846525056554, 'epoch': 0.32} + 32%|███▏ | 168/520 [10:50<21:28, 3.66s/it] 32%|███▎ | 169/520 [10:54<21:18, 3.64s/it] {'loss': 1.2346, 'grad_norm': 0.0012236001235952844, 'learning_rate': 0.07104020832809127, 'epoch': 0.33} + 32%|███▎ | 169/520 [10:54<21:18, 3.64s/it] 33%|███▎ | 170/520 [10:58<21:16, 3.65s/it] {'loss': 1.1758, 'grad_norm': 0.0010770341161568267, 'learning_rate': 0.07081093963579707, 'epoch': 0.33} + 33%|███▎ | 170/520 [10:58<21:16, 3.65s/it] 33%|███▎ | 171/520 [11:01<21:05, 3.63s/it] {'loss': 1.1789, 'grad_norm': 0.0013674805612746259, 'learning_rate': 0.07058066808172016, 'epoch': 0.33} + 33%|███▎ | 171/520 [11:01<21:05, 3.63s/it] 33%|███▎ | 172/520 [11:05<20:57, 3.61s/it] {'loss': 1.2519, 'grad_norm': 0.0012763592296190826, 'learning_rate': 0.07034940261286299, 'epoch': 0.33} + 33%|███▎ | 172/520 [11:05<20:57, 3.61s/it] 33%|███▎ | 173/520 [11:08<20:52, 3.61s/it] {'loss': 1.184, 'grad_norm': 0.0012748583553449898, 'learning_rate': 0.07011715221484578, 'epoch': 0.33} + 33%|███▎ | 173/520 [11:08<20:52, 3.61s/it] 33%|███▎ | 174/520 [11:12<20:48, 3.61s/it] {'loss': 1.2337, 'grad_norm': 0.001301753331602706, 'learning_rate': 0.06988392591155727, 'epoch': 0.33} + 33%|███▎ | 174/520 [11:12<20:48, 3.61s/it] 34%|███▎ | 175/520 [11:16<20:45, 3.61s/it] {'loss': 1.1548, 'grad_norm': 0.0012208633185463932, 'learning_rate': 0.06964973276480421, 'epoch': 0.34} + 34%|███▎ | 175/520 [11:16<20:45, 3.61s/it] 34%|███▍ | 176/520 [11:19<20:42, 3.61s/it] {'loss': 1.2447, 'grad_norm': 0.0012654590549843346, 'learning_rate': 0.06941458187395917, 'epoch': 0.34} + 34%|███▍ | 176/520 [11:19<20:42, 3.61s/it] 34%|███▍ | 177/520 [11:23<21:00, 3.67s/it] {'loss': 1.1293, 'grad_norm': 0.0012695179742111503, 'learning_rate': 0.06917848237560707, 'epoch': 0.34} + 34%|███▍ | 177/520 [11:23<21:00, 3.67s/it] 34%|███▍ | 178/520 [11:27<21:10, 3.72s/it] {'loss': 1.2239, 'grad_norm': 0.0013790313658388537, 'learning_rate': 0.06894144344319014, 'epoch': 0.34} + 34%|███▍ | 178/520 [11:27<21:10, 3.72s/it] 34%|███▍ | 179/520 [11:31<21:03, 3.71s/it] {'loss': 1.2957, 'grad_norm': 0.0012381773868190248, 'learning_rate': 0.06870347428665152, 'epoch': 0.34} + 34%|███▍ | 179/520 [11:31<21:03, 3.71s/it] 35%|███▍ | 180/520 [11:34<21:13, 3.75s/it] {'loss': 1.211, 'grad_norm': 0.0013061522327952105, 'learning_rate': 0.0684645841520774, 'epoch': 0.35} + 35%|███▍ | 180/520 [11:34<21:13, 3.75s/it] 35%|███▍ | 181/520 [11:38<21:16, 3.77s/it] {'loss': 1.196, 'grad_norm': 0.0011479286337264334, 'learning_rate': 0.0682247823213378, 'epoch': 0.35} + 35%|███▍ | 181/520 [11:38<21:16, 3.77s/it] 35%|███▌ | 182/520 [11:42<21:20, 3.79s/it] {'loss': 1.2063, 'grad_norm': 0.0013412585634507848, 'learning_rate': 0.06798407811172585, 'epoch': 0.35} + 35%|███▌ | 182/520 [11:42<21:20, 3.79s/it] 35%|███▌ | 183/520 [11:46<21:19, 3.80s/it] {'loss': 1.2237, 'grad_norm': 0.0012445851687095592, 'learning_rate': 0.06774248087559588, 'epoch': 0.35} + 35%|███▌ | 183/520 [11:46<21:19, 3.80s/it] 35%|███▌ | 184/520 [11:50<21:21, 3.81s/it] {'loss': 1.1675, 'grad_norm': 0.001358954391016957, 'learning_rate': 0.0675, 'epoch': 0.35} + 35%|███▌ | 184/520 [11:50<21:21, 3.81s/it] 36%|███▌ | 185/520 [11:54<21:19, 3.82s/it] {'loss': 1.2911, 'grad_norm': 0.0012609675750338907, 'learning_rate': 0.06725664490632333, 'epoch': 0.36} + 36%|███▌ | 185/520 [11:54<21:19, 3.82s/it] 36%|███▌ | 186/520 [11:57<20:56, 3.76s/it] {'loss': 1.1889, 'grad_norm': 0.00133418884651319, 'learning_rate': 0.06701242504991801, 'epoch': 0.36} + 36%|███▌ | 186/520 [11:57<20:56, 3.76s/it] 36%|███▌ | 187/520 [12:01<20:42, 3.73s/it] {'loss': 1.1821, 'grad_norm': 0.0014533962284616702, 'learning_rate': 0.06676734991973579, 'epoch': 0.36} + 36%|███▌ | 187/520 [12:01<20:42, 3.73s/it] 36%|███▌ | 188/520 [12:04<20:24, 3.69s/it] {'loss': 1.2789, 'grad_norm': 0.0013468395095591536, 'learning_rate': 0.06652142903795932, 'epoch': 0.36} + 36%|███▌ | 188/520 [12:04<20:24, 3.69s/it] 36%|███▋ | 189/520 [12:08<20:15, 3.67s/it] {'loss': 1.2723, 'grad_norm': 0.001185852891408283, 'learning_rate': 0.06627467195963221, 'epoch': 0.36} + 36%|███▋ | 189/520 [12:08<20:15, 3.67s/it] 37%|███▋ | 190/520 [12:12<20:03, 3.65s/it] {'loss': 1.1975, 'grad_norm': 0.0013929635520083113, 'learning_rate': 0.06602708827228779, 'epoch': 0.37} + 37%|███▋ | 190/520 [12:12<20:03, 3.65s/it] 37%|███▋ | 191/520 [12:15<19:54, 3.63s/it] {'loss': 1.1603, 'grad_norm': 0.001250506300163297, 'learning_rate': 0.06577868759557653, 'epoch': 0.37} + 37%|███▋ | 191/520 [12:15<19:54, 3.63s/it] 37%|███▋ | 192/520 [12:19<19:49, 3.63s/it] {'loss': 1.2371, 'grad_norm': 0.0012383040771771147, 'learning_rate': 0.06552947958089234, 'epoch': 0.37} + 37%|███▋ | 192/520 [12:19<19:49, 3.63s/it] 37%|███▋ | 193/520 [12:23<19:46, 3.63s/it] {'loss': 1.1842, 'grad_norm': 0.001367542323271306, 'learning_rate': 0.06527947391099755, 'epoch': 0.37} + 37%|███▋ | 193/520 [12:23<19:46, 3.63s/it] 37%|███▋ | 194/520 [12:26<19:47, 3.64s/it] {'loss': 1.0769, 'grad_norm': 0.001135539734715216, 'learning_rate': 0.06502868029964665, 'epoch': 0.37} + 37%|███▋ | 194/520 [12:26<19:47, 3.64s/it] 38%|███▊ | 195/520 [12:30<19:46, 3.65s/it] {'loss': 1.2519, 'grad_norm': 0.0012693015149686725, 'learning_rate': 0.06477710849120902, 'epoch': 0.38} + 38%|███▊ | 195/520 [12:30<19:46, 3.65s/it] 38%|███▊ | 196/520 [12:34<19:41, 3.65s/it] {'loss': 1.2395, 'grad_norm': 0.001486223044235976, 'learning_rate': 0.0645247682602901, 'epoch': 0.38} + 38%|███▊ | 196/520 [12:34<19:41, 3.65s/it] 38%|███▊ | 197/520 [12:37<19:40, 3.65s/it] {'loss': 1.1822, 'grad_norm': 0.0013240450954574246, 'learning_rate': 0.06427166941135182, 'epoch': 0.38} + 38%|███▊ | 197/520 [12:37<19:40, 3.65s/it] 38%|███▊ | 198/520 [12:41<19:33, 3.65s/it] {'loss': 1.2492, 'grad_norm': 0.0014240550044525434, 'learning_rate': 0.06401782177833147, 'epoch': 0.38} + 38%|███▊ | 198/520 [12:41<19:33, 3.65s/it] 38%|███▊ | 199/520 [12:44<19:31, 3.65s/it] {'loss': 1.166, 'grad_norm': 0.0013181584950144431, 'learning_rate': 0.06376323522425976, 'epoch': 0.38} + 38%|███▊ | 199/520 [12:44<19:31, 3.65s/it] 38%|███▊ | 200/520 [12:48<19:25, 3.64s/it] {'loss': 1.1402, 'grad_norm': 0.0012675356611911504, 'learning_rate': 0.06350791964087753, 'epoch': 0.38} + 38%|███▊ | 200/520 [12:48<19:25, 3.64s/it] 39%|███▊ | 201/520 [12:52<19:23, 3.65s/it] {'loss': 1.1564, 'grad_norm': 0.0011363742716834241, 'learning_rate': 0.06325188494825137, 'epoch': 0.39} + 39%|███▊ | 201/520 [12:52<19:23, 3.65s/it] 39%|███▉ | 202/520 [12:55<19:20, 3.65s/it] {'loss': 1.1824, 'grad_norm': 0.001313911614913662, 'learning_rate': 0.06299514109438833, 'epoch': 0.39} + 39%|███▉ | 202/520 [12:55<19:20, 3.65s/it] 39%|███▉ | 203/520 [12:59<19:27, 3.68s/it] {'loss': 1.2209, 'grad_norm': 0.0013260095896345226, 'learning_rate': 0.06273769805484927, 'epoch': 0.39} + 39%|███▉ | 203/520 [12:59<19:27, 3.68s/it] 39%|███▉ | 204/520 [13:03<19:35, 3.72s/it] {'loss': 1.2356, 'grad_norm': 0.0013291759069220448, 'learning_rate': 0.062479565832361264, 'epoch': 0.39} + 39%|███▉ | 204/520 [13:03<19:35, 3.72s/it] 39%|███▉ | 205/520 [13:07<19:37, 3.74s/it] {'loss': 1.1561, 'grad_norm': 0.001189198981968616, 'learning_rate': 0.062220754456429044, 'epoch': 0.39} + 39%|███▉ | 205/520 [13:07<19:37, 3.74s/it] 40%|███▉ | 206/520 [13:11<19:38, 3.75s/it] {'loss': 1.2576, 'grad_norm': 0.001243675486703743, 'learning_rate': 0.061961273982945236, 'epoch': 0.4} + 40%|███▉ | 206/520 [13:11<19:38, 3.75s/it] 40%|███▉ | 207/520 [13:14<19:38, 3.77s/it] {'loss': 1.1207, 'grad_norm': 0.0010916718130851374, 'learning_rate': 0.06170113449379971, 'epoch': 0.4} + 40%|███▉ | 207/520 [13:14<19:38, 3.77s/it] 40%|████ | 208/520 [13:18<19:38, 3.78s/it] {'loss': 1.2612, 'grad_norm': 0.0014322146916536046, 'learning_rate': 0.06144034609648778, 'epoch': 0.4} + 40%|████ | 208/520 [13:18<19:38, 3.78s/it] 40%|████ | 209/520 [13:22<19:36, 3.78s/it] {'loss': 1.1769, 'grad_norm': 0.0012237621182562758, 'learning_rate': 0.06117891892371754, 'epoch': 0.4} + 40%|████ | 209/520 [13:22<19:36, 3.78s/it] 40%|████ | 210/520 [13:26<19:32, 3.78s/it] {'loss': 1.2422, 'grad_norm': 0.0012980664464183362, 'learning_rate': 0.06091686313301616, 'epoch': 0.4} + 40%|████ | 210/520 [13:26<19:32, 3.78s/it] 41%|████ | 211/520 [13:29<19:26, 3.77s/it] {'loss': 1.2409, 'grad_norm': 0.0011787658427979708, 'learning_rate': 0.06065418890633522, 'epoch': 0.41} + 41%|████ | 211/520 [13:29<19:26, 3.77s/it] 41%|████ | 212/520 [13:33<19:23, 3.78s/it] {'loss': 1.2447, 'grad_norm': 0.0012435604459464184, 'learning_rate': 0.0603909064496551, 'epoch': 0.41} + 41%|████ | 212/520 [13:33<19:23, 3.78s/it] 41%|████ | 213/520 [13:37<19:20, 3.78s/it] {'loss': 1.1939, 'grad_norm': 0.0014517805011846252, 'learning_rate': 0.06012702599258839, 'epoch': 0.41} + 41%|████ | 213/520 [13:37<19:20, 3.78s/it] 41%|████ | 214/520 [13:41<19:17, 3.78s/it] {'loss': 1.189, 'grad_norm': 0.0013092855179597527, 'learning_rate': 0.05986255778798252, 'epoch': 0.41} + 41%|████ | 214/520 [13:41<19:17, 3.78s/it] 41%|████▏ | 215/520 [13:45<19:15, 3.79s/it] {'loss': 1.0951, 'grad_norm': 0.001206802252592396, 'learning_rate': 0.05959751211152132, 'epoch': 0.41} + 41%|████▏ | 215/520 [13:45<19:15, 3.79s/it] 42%|████▏ | 216/520 [13:48<19:15, 3.80s/it] {'loss': 1.1066, 'grad_norm': 0.0012677085063414846, 'learning_rate': 0.0593318992613258, 'epoch': 0.42} + 42%|████▏ | 216/520 [13:48<19:15, 3.80s/it] 42%|████▏ | 217/520 [13:52<19:11, 3.80s/it] {'loss': 1.24, 'grad_norm': 0.0013408825170685792, 'learning_rate': 0.059065729557554, 'epoch': 0.42} + 42%|████▏ | 217/520 [13:52<19:11, 3.80s/it] 42%|████▏ | 218/520 [13:56<19:07, 3.80s/it] {'loss': 1.2118, 'grad_norm': 0.001451641401136286, 'learning_rate': 0.058799013342000044, 'epoch': 0.42} + 42%|████▏ | 218/520 [13:56<19:07, 3.80s/it] 42%|████▏ | 219/520 [14:00<19:00, 3.79s/it] {'loss': 1.2319, 'grad_norm': 0.0012206577897773785, 'learning_rate': 0.05853176097769228, 'epoch': 0.42} + 42%|████▏ | 219/520 [14:00<19:00, 3.79s/it] 42%|████▏ | 220/520 [14:04<18:56, 3.79s/it] {'loss': 1.1417, 'grad_norm': 0.0011805890808571106, 'learning_rate': 0.058263982848490686, 'epoch': 0.42} + 42%|████▏ | 220/520 [14:04<18:56, 3.79s/it] 42%|████▎ | 221/520 [14:07<18:53, 3.79s/it] {'loss': 1.224, 'grad_norm': 0.0013060519759259574, 'learning_rate': 0.05799568935868334, 'epoch': 0.42} + 42%|████▎ | 221/520 [14:07<18:53, 3.79s/it] 43%|████▎ | 222/520 [14:11<18:49, 3.79s/it] {'loss': 1.1684, 'grad_norm': 0.0012856076794857817, 'learning_rate': 0.05772689093258224, 'epoch': 0.43} + 43%|████▎ | 222/520 [14:11<18:49, 3.79s/it] 43%|████▎ | 223/520 [14:15<18:45, 3.79s/it] {'loss': 1.1665, 'grad_norm': 0.0012809904691313267, 'learning_rate': 0.05745759801411822, 'epoch': 0.43} + 43%|████▎ | 223/520 [14:15<18:45, 3.79s/it] 43%|████▎ | 224/520 [14:19<18:41, 3.79s/it] {'loss': 1.1865, 'grad_norm': 0.001118678245247266, 'learning_rate': 0.05718782106643523, 'epoch': 0.43} + 43%|████▎ | 224/520 [14:19<18:41, 3.79s/it] 43%|████▎ | 225/520 [14:23<18:42, 3.80s/it] {'loss': 1.1662, 'grad_norm': 0.0013273521000355638, 'learning_rate': 0.05691757057148372, 'epoch': 0.43} + 43%|████▎ | 225/520 [14:23<18:42, 3.80s/it] 43%|████▎ | 226/520 [14:26<18:37, 3.80s/it] {'loss': 1.2646, 'grad_norm': 0.0012530060051311248, 'learning_rate': 0.05664685702961343, 'epoch': 0.43} + 43%|████▎ | 226/520 [14:26<18:37, 3.80s/it] 44%|████▎ | 227/520 [14:30<18:33, 3.80s/it] {'loss': 1.2484, 'grad_norm': 0.001244760194697891, 'learning_rate': 0.05637569095916538, 'epoch': 0.44} + 44%|████▎ | 227/520 [14:30<18:33, 3.80s/it] 44%|████▍ | 228/520 [14:34<18:29, 3.80s/it] {'loss': 1.2299, 'grad_norm': 0.0012485117923773026, 'learning_rate': 0.0561040828960632, 'epoch': 0.44} + 44%|████▍ | 228/520 [14:34<18:29, 3.80s/it] 44%|████▍ | 229/520 [14:38<18:22, 3.79s/it] {'loss': 1.2224, 'grad_norm': 0.0011845154096690928, 'learning_rate': 0.05583204339340379, 'epoch': 0.44} + 44%|████▍ | 229/520 [14:38<18:22, 3.79s/it] 44%|████▍ | 230/520 [14:41<18:09, 3.76s/it] {'loss': 1.1161, 'grad_norm': 0.0012147471973737718, 'learning_rate': 0.05555958302104719, 'epoch': 0.44} + 44%|████▍ | 230/520 [14:41<18:09, 3.76s/it] 44%|████▍ | 231/520 [14:45<17:52, 3.71s/it] {'loss': 1.1844, 'grad_norm': 0.0012041694375389972, 'learning_rate': 0.055286712365206034, 'epoch': 0.44} + 44%|████▍ | 231/520 [14:45<17:52, 3.71s/it] 45%|████▍ | 232/520 [14:49<17:39, 3.68s/it] {'loss': 1.2678, 'grad_norm': 0.0012645207866867796, 'learning_rate': 0.05501344202803414, 'epoch': 0.45} + 45%|████▍ | 232/520 [14:49<17:39, 3.68s/it] 45%|████▍ | 233/520 [14:52<17:33, 3.67s/it] {'loss': 1.1632, 'grad_norm': 0.0013248978326349913, 'learning_rate': 0.05473978262721463, 'epoch': 0.45} + 45%|████▍ | 233/520 [14:52<17:33, 3.67s/it] 45%|████▌ | 234/520 [14:56<17:24, 3.65s/it] {'loss': 1.1435, 'grad_norm': 0.0013731487942316007, 'learning_rate': 0.05446574479554731, 'epoch': 0.45} + 45%|████▌ | 234/520 [14:56<17:24, 3.65s/it] 45%|████▌ | 235/520 [15:00<17:19, 3.65s/it] {'loss': 1.1911, 'grad_norm': 0.0013147077439773374, 'learning_rate': 0.05419133918053562, 'epoch': 0.45} + 45%|████▌ | 235/520 [15:00<17:19, 3.65s/it] 45%|████▌ | 236/520 [15:03<17:13, 3.64s/it] {'loss': 1.2515, 'grad_norm': 0.0013692445809145933, 'learning_rate': 0.05391657644397289, 'epoch': 0.45} + 45%|████▌ | 236/520 [15:03<17:13, 3.64s/it] 46%|████▌ | 237/520 [15:07<17:08, 3.63s/it] {'loss': 1.2583, 'grad_norm': 0.0012680462597671174, 'learning_rate': 0.05364146726152813, 'epoch': 0.46} + 46%|████▌ | 237/520 [15:07<17:08, 3.63s/it] 46%|████▌ | 238/520 [15:10<17:02, 3.62s/it] {'loss': 1.1907, 'grad_norm': 0.0013245207640956251, 'learning_rate': 0.05336602232233116, 'epoch': 0.46} + 46%|████▌ | 238/520 [15:10<17:02, 3.62s/it] 46%|████▌ | 239/520 [15:14<16:58, 3.62s/it] {'loss': 1.2528, 'grad_norm': 0.0013114541791750917, 'learning_rate': 0.053090252328557366, 'epoch': 0.46} + 46%|████▌ | 239/520 [15:14<16:58, 3.62s/it] 46%|████▌ | 240/520 [15:18<16:54, 3.62s/it] {'loss': 1.0877, 'grad_norm': 0.001195101273591371, 'learning_rate': 0.052814167995011874, 'epoch': 0.46} + 46%|████▌ | 240/520 [15:18<16:54, 3.62s/it] 46%|████▋ | 241/520 [15:21<17:01, 3.66s/it] {'loss': 1.168, 'grad_norm': 0.0012979808626626978, 'learning_rate': 0.05253778004871314, 'epoch': 0.46} + 46%|████▋ | 241/520 [15:21<17:01, 3.66s/it] 47%|████▋ | 242/520 [15:25<17:13, 3.72s/it] {'loss': 1.1749, 'grad_norm': 0.0012215532803125557, 'learning_rate': 0.0522610992284763, 'epoch': 0.47} + 47%|████▋ | 242/520 [15:25<17:13, 3.72s/it] 47%|████▋ | 243/520 [15:29<17:18, 3.75s/it] {'loss': 1.1692, 'grad_norm': 0.0012823428980540579, 'learning_rate': 0.051984136284495816, 'epoch': 0.47} + 47%|████▋ | 243/520 [15:29<17:18, 3.75s/it] 47%|████▋ | 244/520 [15:33<17:21, 3.77s/it] {'loss': 1.2648, 'grad_norm': 0.0012653020093498241, 'learning_rate': 0.05170690197792784, 'epoch': 0.47} + 47%|████▋ | 244/520 [15:33<17:21, 3.77s/it] 47%|████▋ | 245/520 [15:37<17:28, 3.81s/it] {'loss': 1.1537, 'grad_norm': 0.001321375094030421, 'learning_rate': 0.0514294070804721, 'epoch': 0.47} + 47%|████▋ | 245/520 [15:37<17:28, 3.81s/it] 47%|████▋ | 246/520 [15:41<17:31, 3.84s/it] {'loss': 1.2638, 'grad_norm': 0.001285322158863594, 'learning_rate': 0.051151662373953306, 'epoch': 0.47} + 47%|████▋ | 246/520 [15:41<17:31, 3.84s/it] 48%|████▊ | 247/520 [15:45<17:29, 3.84s/it] {'loss': 1.3237, 'grad_norm': 0.001330615481345049, 'learning_rate': 0.05087367864990233, 'epoch': 0.47} + 48%|████▊ | 247/520 [15:45<17:29, 3.84s/it] 48%|████▊ | 248/520 [15:48<17:22, 3.83s/it] {'loss': 1.1458, 'grad_norm': 0.001314785614842987, 'learning_rate': 0.05059546670913684, 'epoch': 0.48} + 48%|████▊ | 248/520 [15:48<17:22, 3.83s/it] 48%|████▊ | 249/520 [15:52<17:16, 3.82s/it] {'loss': 1.2366, 'grad_norm': 0.0013013773235235851, 'learning_rate': 0.05031703736134168, 'epoch': 0.48} + 48%|████▊ | 249/520 [15:52<17:16, 3.82s/it] 48%|████▊ | 250/520 [15:56<17:13, 3.83s/it] {'loss': 1.1889, 'grad_norm': 0.001349484733242477, 'learning_rate': 0.050038401424648854, 'epoch': 0.48} + 48%|████▊ | 250/520 [15:56<17:13, 3.83s/it] 48%|████▊ | 251/520 [16:00<17:13, 3.84s/it] {'loss': 1.2502, 'grad_norm': 0.0012243325939783236, 'learning_rate': 0.0497595697252172, 'epoch': 0.48} + 48%|████▊ | 251/520 [16:00<17:13, 3.84s/it] 48%|████▊ | 252/520 [16:04<17:10, 3.85s/it] {'loss': 1.1689, 'grad_norm': 0.0011892307290991221, 'learning_rate': 0.04948055309681175, 'epoch': 0.48} + 48%|████▊ | 252/520 [16:04<17:10, 3.85s/it] 49%|████▊ | 253/520 [16:07<16:59, 3.82s/it] {'loss': 1.2303, 'grad_norm': 0.0013780165674851935, 'learning_rate': 0.04920136238038277, 'epoch': 0.49} + 49%|████▊ | 253/520 [16:07<16:59, 3.82s/it] 49%|████▉ | 254/520 [16:11<16:56, 3.82s/it] {'loss': 1.1788, 'grad_norm': 0.0012380867854289996, 'learning_rate': 0.04892200842364462, 'epoch': 0.49} + 49%|████▉ | 254/520 [16:11<16:56, 3.82s/it] 49%|████▉ | 255/520 [16:15<16:53, 3.82s/it] {'loss': 1.1793, 'grad_norm': 0.0013422216720742552, 'learning_rate': 0.04864250208065415, 'epoch': 0.49} + 49%|████▉ | 255/520 [16:15<16:53, 3.82s/it] 49%|████▉ | 256/520 [16:19<16:49, 3.82s/it] {'loss': 1.2357, 'grad_norm': 0.0013545737471875544, 'learning_rate': 0.0483628542113891, 'epoch': 0.49} + 49%|████▉ | 256/520 [16:19<16:49, 3.82s/it] 49%|████▉ | 257/520 [16:23<16:44, 3.82s/it] {'loss': 1.2174, 'grad_norm': 0.001301823399819082, 'learning_rate': 0.04808307568132605, 'epoch': 0.49} + 49%|████▉ | 257/520 [16:23<16:44, 3.82s/it] 50%|████▉ | 258/520 [16:27<16:47, 3.84s/it] {'loss': 1.2212, 'grad_norm': 0.0011134467332315106, 'learning_rate': 0.04780317736101835, 'epoch': 0.5} + 50%|████▉ | 258/520 [16:27<16:47, 3.84s/it] 50%|████▉ | 259/520 [16:31<16:43, 3.84s/it] {'loss': 1.2953, 'grad_norm': 0.001430967823053669, 'learning_rate': 0.04752317012567363, 'epoch': 0.5} + 50%|████▉ | 259/520 [16:31<16:43, 3.84s/it] 50%|█████ | 260/520 [16:34<16:44, 3.86s/it] {'loss': 1.2122, 'grad_norm': 0.001018411304373632, 'learning_rate': 0.04724306485473137, 'epoch': 0.5} + 50%|█████ | 260/520 [16:34<16:44, 3.86s/it] 50%|█████ | 261/520 [16:38<16:41, 3.87s/it] {'loss': 1.1604, 'grad_norm': 0.0012268930914819253, 'learning_rate': 0.04696287243144012, 'epoch': 0.5} + 50%|█████ | 261/520 [16:38<16:41, 3.87s/it] 50%|█████ | 262/520 [16:42<16:18, 3.79s/it] {'loss': 1.1657, 'grad_norm': 0.0013095991692585872, 'learning_rate': 0.046682603742434665, 'epoch': 0.5} + 50%|█████ | 262/520 [16:42<16:18, 3.79s/it] 51%|█████ | 263/520 [16:46<16:02, 3.74s/it] {'loss': 1.1862, 'grad_norm': 0.0012491273896988848, 'learning_rate': 0.04640226967731299, 'epoch': 0.51} + 51%|█████ | 263/520 [16:46<16:02, 3.74s/it] 51%|█████ | 264/520 [16:49<15:49, 3.71s/it] {'loss': 1.2467, 'grad_norm': 0.0012227845817127788, 'learning_rate': 0.046121881128213284, 'epoch': 0.51} + 51%|█████ | 264/520 [16:49<15:49, 3.71s/it] 51%|█████ | 265/520 [16:53<15:42, 3.70s/it] {'loss': 1.1742, 'grad_norm': 0.0014457336556537382, 'learning_rate': 0.0458414489893906, 'epoch': 0.51} + 51%|█████ | 265/520 [16:53<15:42, 3.70s/it] 51%|█████ | 266/520 [16:56<15:33, 3.68s/it] {'loss': 1.0562, 'grad_norm': 0.0011265328020938646, 'learning_rate': 0.04556098415679368, 'epoch': 0.51} + 51%|█████ | 266/520 [16:56<15:33, 3.68s/it] 51%|█████▏ | 267/520 [17:00<15:27, 3.67s/it] {'loss': 1.1714, 'grad_norm': 0.0012400576615007157, 'learning_rate': 0.04528049752764151, 'epoch': 0.51} + 51%|█████▏ | 267/520 [17:00<15:27, 3.67s/it] 52%|█████▏ | 268/520 [17:04<15:23, 3.66s/it] {'loss': 1.2809, 'grad_norm': 0.001255019110770088, 'learning_rate': 0.045, 'epoch': 0.52} + 52%|█████▏ | 268/520 [17:04<15:23, 3.66s/it] 52%|█████▏ | 269/520 [17:07<15:18, 3.66s/it] {'loss': 1.2728, 'grad_norm': 0.00140184912062973, 'learning_rate': 0.04471950247235849, 'epoch': 0.52} + 52%|█████▏ | 269/520 [17:07<15:18, 3.66s/it] 52%|█████▏ | 270/520 [17:11<15:15, 3.66s/it] {'loss': 1.1269, 'grad_norm': 0.0012585319896588028, 'learning_rate': 0.04443901584320632, 'epoch': 0.52} + 52%|█████▏ | 270/520 [17:11<15:15, 3.66s/it] 52%|█████▏ | 271/520 [17:15<15:09, 3.65s/it] {'loss': 1.2476, 'grad_norm': 0.0013357998118152402, 'learning_rate': 0.044158551010609406, 'epoch': 0.52} + 52%|█████▏ | 271/520 [17:15<15:09, 3.65s/it] 52%|█████▏ | 272/520 [17:18<15:04, 3.65s/it] {'loss': 1.1418, 'grad_norm': 0.0012392876732124959, 'learning_rate': 0.04387811887178673, 'epoch': 0.52} + 52%|█████▏ | 272/520 [17:18<15:04, 3.65s/it] 52%|█████▎ | 273/520 [17:22<14:59, 3.64s/it] {'loss': 1.2392, 'grad_norm': 0.0012154131523221, 'learning_rate': 0.04359773032268702, 'epoch': 0.53} + 52%|█████▎ | 273/520 [17:22<14:59, 3.64s/it] 53%|█████▎ | 274/520 [17:26<14:55, 3.64s/it] {'loss': 1.2396, 'grad_norm': 0.001372140501807235, 'learning_rate': 0.04331739625756535, 'epoch': 0.53} + 53%|█████▎ | 274/520 [17:26<14:55, 3.64s/it] 53%|█████▎ | 275/520 [17:29<14:51, 3.64s/it] {'loss': 1.178, 'grad_norm': 0.0012914679197876505, 'learning_rate': 0.04303712756855988, 'epoch': 0.53} + 53%|█████▎ | 275/520 [17:29<14:51, 3.64s/it] 53%|█████▎ | 276/520 [17:33<14:47, 3.64s/it] {'loss': 1.235, 'grad_norm': 0.001392036752562398, 'learning_rate': 0.04275693514526862, 'epoch': 0.53} + 53%|█████▎ | 276/520 [17:33<14:47, 3.64s/it] 53%|█████▎ | 277/520 [17:37<14:44, 3.64s/it] {'loss': 1.2539, 'grad_norm': 0.0011665901388487184, 'learning_rate': 0.04247682987432636, 'epoch': 0.53} + 53%|█████▎ | 277/520 [17:37<14:44, 3.64s/it] 53%|█████▎ | 278/520 [17:40<14:39, 3.63s/it] {'loss': 1.1418, 'grad_norm': 0.001182139839323662, 'learning_rate': 0.042196822638981654, 'epoch': 0.53} + 53%|█████▎ | 278/520 [17:40<14:39, 3.63s/it] 54%|█████▎ | 279/520 [17:44<14:33, 3.62s/it] {'loss': 1.1306, 'grad_norm': 0.0012985450334028843, 'learning_rate': 0.04191692431867395, 'epoch': 0.54} + 54%|█████▎ | 279/520 [17:44<14:33, 3.62s/it] 54%|█████▍ | 280/520 [17:47<14:31, 3.63s/it] {'loss': 1.1697, 'grad_norm': 0.0013760656322946666, 'learning_rate': 0.041637145788610906, 'epoch': 0.54} + 54%|█████▍ | 280/520 [17:47<14:31, 3.63s/it] 54%|█████▍ | 281/520 [17:51<14:26, 3.63s/it] {'loss': 1.2713, 'grad_norm': 0.0013434286168740766, 'learning_rate': 0.04135749791934585, 'epoch': 0.54} + 54%|█████▍ | 281/520 [17:51<14:26, 3.63s/it] 54%|█████▍ | 282/520 [17:55<14:24, 3.63s/it] {'loss': 1.1528, 'grad_norm': 0.0012334268356154199, 'learning_rate': 0.041077991576355374, 'epoch': 0.54} + 54%|█████▍ | 282/520 [17:55<14:24, 3.63s/it] 54%|█████▍ | 283/520 [17:58<14:20, 3.63s/it] {'loss': 1.281, 'grad_norm': 0.0013984930717789014, 'learning_rate': 0.040798637619617224, 'epoch': 0.54} + 54%|█████▍ | 283/520 [17:58<14:20, 3.63s/it] 55%|█████▍ | 284/520 [18:02<14:17, 3.63s/it] {'loss': 1.1404, 'grad_norm': 0.0013282893271433893, 'learning_rate': 0.040519446903188254, 'epoch': 0.55} + 55%|█████▍ | 284/520 [18:02<14:17, 3.63s/it] 55%|█████▍ | 285/520 [18:06<14:12, 3.63s/it] {'loss': 1.1731, 'grad_norm': 0.0012784283925908965, 'learning_rate': 0.0402404302747828, 'epoch': 0.55} + 55%|█████▍ | 285/520 [18:06<14:12, 3.63s/it] 55%|█████▌ | 286/520 [18:09<14:16, 3.66s/it] {'loss': 1.0584, 'grad_norm': 0.0012918178634462112, 'learning_rate': 0.03996159857535114, 'epoch': 0.55} + 55%|█████▌ | 286/520 [18:09<14:16, 3.66s/it] 55%|█████▌ | 287/520 [18:13<14:16, 3.68s/it] {'loss': 1.2861, 'grad_norm': 0.0012709125140568174, 'learning_rate': 0.03968296263865832, 'epoch': 0.55} + 55%|█████▌ | 287/520 [18:13<14:16, 3.68s/it] 55%|█████▌ | 288/520 [18:17<14:09, 3.66s/it] {'loss': 1.3018, 'grad_norm': 0.0011984129970917514, 'learning_rate': 0.03940453329086318, 'epoch': 0.55} + 55%|█████▌ | 288/520 [18:17<14:09, 3.66s/it] 56%|█████▌ | 289/520 [18:20<14:08, 3.67s/it] {'loss': 1.1963, 'grad_norm': 0.0012397077126236485, 'learning_rate': 0.039126321350097686, 'epoch': 0.56} + 56%|█████▌ | 289/520 [18:20<14:08, 3.67s/it] 56%|█████▌ | 290/520 [18:24<14:01, 3.66s/it] {'loss': 1.1182, 'grad_norm': 0.001180742173420183, 'learning_rate': 0.038848337626046704, 'epoch': 0.56} + 56%|█████▌ | 290/520 [18:24<14:01, 3.66s/it] 56%|█████▌ | 291/520 [18:28<13:56, 3.65s/it] {'loss': 1.1517, 'grad_norm': 0.0012675936430314747, 'learning_rate': 0.038570592919527906, 'epoch': 0.56} + 56%|█████▌ | 291/520 [18:28<13:56, 3.65s/it] 56%|█████▌ | 292/520 [18:31<13:49, 3.64s/it] {'loss': 1.2122, 'grad_norm': 0.0012799037079767647, 'learning_rate': 0.03829309802207214, 'epoch': 0.56} + 56%|█████▌ | 292/520 [18:31<13:49, 3.64s/it] 56%|█████▋ | 293/520 [18:35<13:48, 3.65s/it] {'loss': 1.1601, 'grad_norm': 0.001339072906564285, 'learning_rate': 0.038015863715504174, 'epoch': 0.56} + 56%|█████▋ | 293/520 [18:35<13:48, 3.65s/it] 57%|█████▋ | 294/520 [18:39<13:43, 3.64s/it] {'loss': 1.1819, 'grad_norm': 0.0014208155175913428, 'learning_rate': 0.0377389007715237, 'epoch': 0.57} + 57%|█████▋ | 294/520 [18:39<13:43, 3.64s/it] 57%|█████▋ | 295/520 [18:42<13:39, 3.64s/it] {'loss': 1.1665, 'grad_norm': 0.0011422350800835142, 'learning_rate': 0.03746221995128687, 'epoch': 0.57} + 57%|█████▋ | 295/520 [18:42<13:39, 3.64s/it] 57%|█████▋ | 296/520 [18:46<13:33, 3.63s/it] {'loss': 1.1324, 'grad_norm': 0.001307789945077067, 'learning_rate': 0.03718583200498814, 'epoch': 0.57} + 57%|█████▋ | 296/520 [18:46<13:33, 3.63s/it] 57%|█████▋ | 297/520 [18:49<13:27, 3.62s/it] {'loss': 1.2607, 'grad_norm': 0.001388907586428929, 'learning_rate': 0.03690974767144263, 'epoch': 0.57} + 57%|█████▋ | 297/520 [18:49<13:27, 3.62s/it] 57%|█████▋ | 298/520 [18:53<13:22, 3.61s/it] {'loss': 1.2218, 'grad_norm': 0.0012082688269411614, 'learning_rate': 0.03663397767766884, 'epoch': 0.57} + 57%|█████▋ | 298/520 [18:53<13:22, 3.61s/it] 57%|█████▊ | 299/520 [18:57<13:19, 3.62s/it] {'loss': 1.222, 'grad_norm': 0.0012085299282510536, 'learning_rate': 0.03635853273847187, 'epoch': 0.57} + 57%|█████▊ | 299/520 [18:57<13:19, 3.62s/it] 58%|█████▊ | 300/520 [19:00<13:16, 3.62s/it] {'loss': 1.2684, 'grad_norm': 0.0012826504718960743, 'learning_rate': 0.03608342355602712, 'epoch': 0.58} + 58%|█████▊ | 300/520 [19:00<13:16, 3.62s/it] 58%|█████▊ | 301/520 [19:04<13:12, 3.62s/it] {'loss': 1.2604, 'grad_norm': 0.001301600674293776, 'learning_rate': 0.03580866081946439, 'epoch': 0.58} + 58%|█████▊ | 301/520 [19:04<13:12, 3.62s/it] 58%|█████▊ | 302/520 [19:07<13:10, 3.62s/it] {'loss': 1.2197, 'grad_norm': 0.0012057302084933677, 'learning_rate': 0.0355342552044527, 'epoch': 0.58} + 58%|█████▊ | 302/520 [19:07<13:10, 3.62s/it] 58%|█████▊ | 303/520 [19:11<13:06, 3.62s/it] {'loss': 1.1777, 'grad_norm': 0.0014132246368852544, 'learning_rate': 0.03526021737278537, 'epoch': 0.58} + 58%|█████▊ | 303/520 [19:11<13:06, 3.62s/it] 58%|█████▊ | 304/520 [19:15<13:05, 3.64s/it] {'loss': 1.1356, 'grad_norm': 0.0012829183461203909, 'learning_rate': 0.034986557971965856, 'epoch': 0.58} + 58%|█████▊ | 304/520 [19:15<13:05, 3.64s/it] 59%|█████▊ | 305/520 [19:18<13:04, 3.65s/it] {'loss': 1.2773, 'grad_norm': 0.001462904466320587, 'learning_rate': 0.03471328763479398, 'epoch': 0.59} + 59%|█████▊ | 305/520 [19:18<13:04, 3.65s/it] 59%|█████▉ | 306/520 [19:22<13:01, 3.65s/it] {'loss': 1.2267, 'grad_norm': 0.0013094814557059419, 'learning_rate': 0.03444041697895282, 'epoch': 0.59} + 59%|█████▉ | 306/520 [19:22<13:01, 3.65s/it] 59%|█████▉ | 307/520 [19:26<13:21, 3.76s/it] {'loss': 1.1707, 'grad_norm': 0.001252126362614757, 'learning_rate': 0.034167956606596224, 'epoch': 0.59} + 59%|█████▉ | 307/520 [19:26<13:21, 3.76s/it] 59%|█████▉ | 308/520 [19:30<13:10, 3.73s/it] {'loss': 1.2837, 'grad_norm': 0.0012447122760241634, 'learning_rate': 0.03389591710393678, 'epoch': 0.59} + 59%|█████▉ | 308/520 [19:30<13:10, 3.73s/it] 59%|█████▉ | 309/520 [19:33<12:59, 3.69s/it] {'loss': 1.177, 'grad_norm': 0.0012649516933373249, 'learning_rate': 0.033624309040834605, 'epoch': 0.59} + 59%|█████▉ | 309/520 [19:33<12:59, 3.69s/it] 60%|█████▉ | 310/520 [19:37<12:51, 3.67s/it] {'loss': 1.1531, 'grad_norm': 0.0012900968818693125, 'learning_rate': 0.03335314297038656, 'epoch': 0.6} + 60%|█████▉ | 310/520 [19:37<12:51, 3.67s/it] 60%|█████▉ | 311/520 [19:41<12:45, 3.66s/it] {'loss': 1.141, 'grad_norm': 0.0012500454203611046, 'learning_rate': 0.03308242942851627, 'epoch': 0.6} + 60%|█████▉ | 311/520 [19:41<12:45, 3.66s/it] 60%|██████ | 312/520 [19:44<12:38, 3.65s/it] {'loss': 1.1285, 'grad_norm': 0.0013506450088929807, 'learning_rate': 0.03281217893356477, 'epoch': 0.6} + 60%|██████ | 312/520 [19:44<12:38, 3.65s/it] 60%|██████ | 313/520 [19:48<12:34, 3.64s/it] {'loss': 1.1021, 'grad_norm': 0.0011591493902130508, 'learning_rate': 0.03254240198588178, 'epoch': 0.6} + 60%|██████ | 313/520 [19:48<12:34, 3.64s/it] 60%|██████ | 314/520 [19:52<12:57, 3.78s/it] {'loss': 1.1457, 'grad_norm': 0.0012589729800814775, 'learning_rate': 0.03227310906741776, 'epoch': 0.6} + 60%|██████ | 314/520 [19:52<12:57, 3.78s/it] 61%|██████ | 315/520 [19:56<12:44, 3.73s/it] {'loss': 1.1858, 'grad_norm': 0.0013724256041588306, 'learning_rate': 0.03200431064131666, 'epoch': 0.61} + 61%|██████ | 315/520 [19:56<12:44, 3.73s/it] 61%|██████ | 316/520 [20:00<13:02, 3.83s/it] {'loss': 1.1334, 'grad_norm': 0.0013365917729003202, 'learning_rate': 0.03173601715150931, 'epoch': 0.61} + 61%|██████ | 316/520 [20:00<13:02, 3.83s/it] 61%|██████ | 317/520 [20:03<12:46, 3.77s/it] {'loss': 1.1332, 'grad_norm': 0.0011443451089987468, 'learning_rate': 0.03146823902230771, 'epoch': 0.61} + 61%|██████ | 317/520 [20:03<12:46, 3.77s/it] 61%|██████ | 318/520 [20:07<12:33, 3.73s/it] {'loss': 1.2444, 'grad_norm': 0.0013633318670741247, 'learning_rate': 0.03120098665799996, 'epoch': 0.61} + 61%|██████ | 318/520 [20:07<12:33, 3.73s/it] 61%|██████▏ | 319/520 [20:11<12:43, 3.80s/it] {'loss': 1.1277, 'grad_norm': 0.0011518130972359545, 'learning_rate': 0.030934270442446003, 'epoch': 0.61} + 61%|██████▏ | 319/520 [20:11<12:43, 3.80s/it] 62%|██████▏ | 320/520 [20:14<12:29, 3.75s/it] {'loss': 1.0721, 'grad_norm': 0.0012669919242441225, 'learning_rate': 0.030668100738674205, 'epoch': 0.62} + 62%|██████▏ | 320/520 [20:14<12:29, 3.75s/it] 62%|██████▏ | 321/520 [20:18<12:19, 3.71s/it] {'loss': 1.2658, 'grad_norm': 0.001255562660325407, 'learning_rate': 0.030402487888478685, 'epoch': 0.62} + 62%|██████▏ | 321/520 [20:18<12:19, 3.71s/it] 62%|██████▏ | 322/520 [20:22<12:09, 3.69s/it] {'loss': 1.0843, 'grad_norm': 0.0012176894517936698, 'learning_rate': 0.030137442212017494, 'epoch': 0.62} + 62%|██████▏ | 322/520 [20:22<12:09, 3.69s/it] 62%|██████▏ | 323/520 [20:25<12:04, 3.68s/it] {'loss': 1.1577, 'grad_norm': 0.001284197634224607, 'learning_rate': 0.029872974007411623, 'epoch': 0.62} + 62%|██████▏ | 323/520 [20:25<12:04, 3.68s/it] 62%|██████▏ | 324/520 [20:29<11:58, 3.67s/it] {'loss': 1.213, 'grad_norm': 0.001305768865303939, 'learning_rate': 0.029609093550344908, 'epoch': 0.62} + 62%|██████▏ | 324/520 [20:29<11:58, 3.67s/it] 62%|██████▎ | 325/520 [20:33<11:55, 3.67s/it] {'loss': 1.2057, 'grad_norm': 0.0013371155360802733, 'learning_rate': 0.02934581109366477, 'epoch': 0.62} + 62%|██████▎ | 325/520 [20:33<11:55, 3.67s/it] 63%|██████▎ | 326/520 [20:36<11:51, 3.67s/it] {'loss': 1.2071, 'grad_norm': 0.0013673931417771139, 'learning_rate': 0.029083136866983838, 'epoch': 0.63} + 63%|██████▎ | 326/520 [20:36<11:51, 3.67s/it] 63%|██████▎ | 327/520 [20:40<11:46, 3.66s/it] {'loss': 1.1861, 'grad_norm': 0.001314740223839851, 'learning_rate': 0.02882108107628246, 'epoch': 0.63} + 63%|██████▎ | 327/520 [20:40<11:46, 3.66s/it] 63%|██████▎ | 328/520 [20:44<11:44, 3.67s/it] {'loss': 1.2471, 'grad_norm': 0.0013319906634417622, 'learning_rate': 0.02855965390351222, 'epoch': 0.63} + 63%|██████▎ | 328/520 [20:44<11:44, 3.67s/it] 63%|██████▎ | 329/520 [20:47<11:39, 3.66s/it] {'loss': 1.1301, 'grad_norm': 0.0011559276059334416, 'learning_rate': 0.028298865506200293, 'epoch': 0.63} + 63%|██████▎ | 329/520 [20:47<11:39, 3.66s/it] 63%|██████▎ | 330/520 [20:51<11:34, 3.66s/it] {'loss': 1.2074, 'grad_norm': 0.0012031384346928512, 'learning_rate': 0.028038726017054764, 'epoch': 0.63} + 63%|██████▎ | 330/520 [20:51<11:34, 3.66s/it] 64%|██████▎ | 331/520 [20:55<11:32, 3.66s/it] {'loss': 1.1653, 'grad_norm': 0.0013531346476375148, 'learning_rate': 0.02777924554357096, 'epoch': 0.64} + 64%|██████▎ | 331/520 [20:55<11:32, 3.66s/it] 64%|██████▍ | 332/520 [20:58<11:28, 3.66s/it] {'loss': 1.2155, 'grad_norm': 0.0011646517349287048, 'learning_rate': 0.02752043416763874, 'epoch': 0.64} + 64%|██████▍ | 332/520 [20:58<11:28, 3.66s/it] 64%|██████▍ | 333/520 [21:02<11:25, 3.66s/it] {'loss': 1.3002, 'grad_norm': 0.0013678902220915121, 'learning_rate': 0.027262301945150735, 'epoch': 0.64} + 64%|██████▍ | 333/520 [21:02<11:25, 3.66s/it] 64%|██████▍ | 334/520 [21:06<11:20, 3.66s/it] {'loss': 1.2102, 'grad_norm': 0.0013685481222375875, 'learning_rate': 0.02700485890561166, 'epoch': 0.64} + 64%|██████▍ | 334/520 [21:06<11:20, 3.66s/it] 64%|██████▍ | 335/520 [21:09<11:15, 3.65s/it] {'loss': 1.2102, 'grad_norm': 0.001247478905543384, 'learning_rate': 0.02674811505174863, 'epoch': 0.64} + 64%|██████▍ | 335/520 [21:09<11:15, 3.65s/it] 65%|██████▍ | 336/520 [21:13<11:12, 3.65s/it] {'loss': 1.1219, 'grad_norm': 0.0013952773752619713, 'learning_rate': 0.02649208035912249, 'epoch': 0.65} + 65%|██████▍ | 336/520 [21:13<11:12, 3.65s/it] 65%|██████▍ | 337/520 [21:17<11:08, 3.65s/it] {'loss': 1.1135, 'grad_norm': 0.0013104834976976074, 'learning_rate': 0.02623676477574025, 'epoch': 0.65} + 65%|██████▍ | 337/520 [21:17<11:08, 3.65s/it] 65%|██████▌ | 338/520 [21:20<11:05, 3.66s/it] {'loss': 1.2197, 'grad_norm': 0.001302632496656121, 'learning_rate': 0.025982178221668532, 'epoch': 0.65} + 65%|██████▌ | 338/520 [21:20<11:05, 3.66s/it] 65%|██████▌ | 339/520 [21:24<11:00, 3.65s/it] {'loss': 1.1623, 'grad_norm': 0.0013226474156909972, 'learning_rate': 0.02572833058864817, 'epoch': 0.65} + 65%|██████▌ | 339/520 [21:24<11:00, 3.65s/it] 65%|██████▌ | 340/520 [21:28<11:00, 3.67s/it] {'loss': 1.1493, 'grad_norm': 0.0012371264319879912, 'learning_rate': 0.025475231739709885, 'epoch': 0.65} + 65%|██████▌ | 340/520 [21:28<11:00, 3.67s/it] 66%|██████▌ | 341/520 [21:31<10:53, 3.65s/it] {'loss': 1.1809, 'grad_norm': 0.0013828790907618542, 'learning_rate': 0.025222891508790972, 'epoch': 0.66} + 66%|██████▌ | 341/520 [21:31<10:53, 3.65s/it] 66%|██████▌ | 342/520 [21:35<10:47, 3.64s/it] {'loss': 1.1944, 'grad_norm': 0.0014783499327637114, 'learning_rate': 0.02497131970035334, 'epoch': 0.66} + 66%|██████▌ | 342/520 [21:35<10:47, 3.64s/it] 66%|██████▌ | 343/520 [21:38<10:43, 3.64s/it] {'loss': 1.1418, 'grad_norm': 0.001048566553838941, 'learning_rate': 0.024720526089002454, 'epoch': 0.66} + 66%|██████▌ | 343/520 [21:38<10:43, 3.64s/it] 66%|██████▌ | 344/520 [21:42<10:41, 3.64s/it] {'loss': 1.1374, 'grad_norm': 0.0011743254934303484, 'learning_rate': 0.024470520419107664, 'epoch': 0.66} + 66%|██████▌ | 344/520 [21:42<10:41, 3.64s/it] 66%|██████▋ | 345/520 [21:46<10:39, 3.65s/it] {'loss': 1.2336, 'grad_norm': 0.0013149464789181972, 'learning_rate': 0.024221312404423483, 'epoch': 0.66} + 66%|██████▋ | 345/520 [21:46<10:39, 3.65s/it] 67%|██████▋ | 346/520 [21:49<10:36, 3.66s/it] {'loss': 1.1611, 'grad_norm': 0.0012342511430487463, 'learning_rate': 0.023972911727712206, 'epoch': 0.67} + 67%|██████▋ | 346/520 [21:49<10:36, 3.66s/it] 67%|██████▋ | 347/520 [21:53<10:31, 3.65s/it] {'loss': 1.1552, 'grad_norm': 0.0011783785997813655, 'learning_rate': 0.02372532804036779, 'epoch': 0.67} + 67%|██████▋ | 347/520 [21:53<10:31, 3.65s/it]Token indices sequence length is longer than the specified maximum sequence length for this model (2778 > 2048). Running this sequence through the model will result in indexing errors + 67%|██████▋ | 348/520 [21:57<10:28, 3.66s/it] {'loss': 1.1131, 'grad_norm': 0.0015029885006923587, 'learning_rate': 0.023478570962040694, 'epoch': 0.67} + 67%|██████▋ | 348/520 [21:57<10:28, 3.66s/it] 67%|██████▋ | 349/520 [22:00<10:26, 3.66s/it] {'loss': 1.1459, 'grad_norm': 0.0012995918141404763, 'learning_rate': 0.02323265008026421, 'epoch': 0.67} + 67%|██████▋ | 349/520 [22:00<10:26, 3.66s/it] 67%|██████▋ | 350/520 [22:04<10:23, 3.67s/it] {'loss': 1.1902, 'grad_norm': 0.0013230789060268769, 'learning_rate': 0.022987574950081996, 'epoch': 0.67} + 67%|██████▋ | 350/520 [22:04<10:23, 3.67s/it] 68%|██████▊ | 351/520 [22:08<10:21, 3.68s/it] {'loss': 1.1036, 'grad_norm': 0.0012222718951888359, 'learning_rate': 0.022743355093676663, 'epoch': 0.68} + 68%|██████▊ | 351/520 [22:08<10:21, 3.68s/it] 68%|██████▊ | 352/520 [22:12<10:16, 3.67s/it] {'loss': 1.2139, 'grad_norm': 0.001204060250760057, 'learning_rate': 0.02250000000000001, 'epoch': 0.68} + 68%|██████▊ | 352/520 [22:12<10:16, 3.67s/it] 68%|██████▊ | 353/520 [22:15<10:14, 3.68s/it] {'loss': 1.1377, 'grad_norm': 0.0010559078259056648, 'learning_rate': 0.02225751912440413, 'epoch': 0.68} + 68%|██████▊ | 353/520 [22:15<10:14, 3.68s/it] 68%|██████▊ | 354/520 [22:19<10:10, 3.68s/it] {'loss': 1.2276, 'grad_norm': 0.0011660926186119042, 'learning_rate': 0.02201592188827416, 'epoch': 0.68} + 68%|██████▊ | 354/520 [22:19<10:10, 3.68s/it] 68%|██████▊ | 355/520 [22:23<10:05, 3.67s/it] {'loss': 1.1667, 'grad_norm': 0.0012909432501336086, 'learning_rate': 0.021775217678662195, 'epoch': 0.68} + 68%|██████▊ | 355/520 [22:23<10:05, 3.67s/it] 68%|██████▊ | 356/520 [22:26<10:01, 3.67s/it] {'loss': 1.1652, 'grad_norm': 0.0013090999605994752, 'learning_rate': 0.02153541584792259, 'epoch': 0.68} + 68%|██████▊ | 356/520 [22:26<10:01, 3.67s/it] 69%|██████▊ | 357/520 [22:30<09:55, 3.65s/it] {'loss': 1.2019, 'grad_norm': 0.0012291539946033072, 'learning_rate': 0.021296525713348464, 'epoch': 0.69} + 69%|██████▊ | 357/520 [22:30<09:55, 3.65s/it] 69%|██████▉ | 358/520 [22:33<09:53, 3.66s/it] {'loss': 1.127, 'grad_norm': 0.0013134121793548644, 'learning_rate': 0.021058556556809858, 'epoch': 0.69} + 69%|██████▉ | 358/520 [22:33<09:53, 3.66s/it] 69%|██████▉ | 359/520 [22:37<09:48, 3.66s/it] {'loss': 1.1729, 'grad_norm': 0.0012662198362273327, 'learning_rate': 0.02082151762439292, 'epoch': 0.69} + 69%|██████▉ | 359/520 [22:37<09:48, 3.66s/it] 69%|██████▉ | 360/520 [22:41<09:44, 3.66s/it] {'loss': 1.1777, 'grad_norm': 0.0012347315131041822, 'learning_rate': 0.020585418126040828, 'epoch': 0.69} + 69%|██████▉ | 360/520 [22:41<09:44, 3.66s/it] 69%|██████▉ | 361/520 [22:44<09:40, 3.65s/it] {'loss': 1.1964, 'grad_norm': 0.0011238377658590601, 'learning_rate': 0.020350267235195795, 'epoch': 0.69} + 69%|██████▉ | 361/520 [22:44<09:40, 3.65s/it] 70%|██████▉ | 362/520 [22:48<09:34, 3.64s/it] {'loss': 1.1736, 'grad_norm': 0.0013986349598902363, 'learning_rate': 0.020116074088442723, 'epoch': 0.7} + 70%|██████▉ | 362/520 [22:48<09:34, 3.64s/it] 70%|██████▉ | 363/520 [22:52<09:29, 3.63s/it] {'loss': 1.2076, 'grad_norm': 0.0012817059762398836, 'learning_rate': 0.01988284778515423, 'epoch': 0.7} + 70%|██████▉ | 363/520 [22:52<09:29, 3.63s/it] 70%|███████ | 364/520 [22:55<09:27, 3.64s/it] {'loss': 1.2092, 'grad_norm': 0.0012773815752214598, 'learning_rate': 0.019650597387137005, 'epoch': 0.7} + 70%|███████ | 364/520 [22:55<09:27, 3.64s/it] 70%|███████ | 365/520 [22:59<09:22, 3.63s/it] {'loss': 1.258, 'grad_norm': 0.0013150647679775833, 'learning_rate': 0.01941933191827985, 'epoch': 0.7} + 70%|███████ | 365/520 [22:59<09:22, 3.63s/it] 70%|███████ | 366/520 [23:03<09:18, 3.63s/it] {'loss': 1.2252, 'grad_norm': 0.0012377335024862828, 'learning_rate': 0.019189060364202936, 'epoch': 0.7} + 70%|███████ | 366/520 [23:03<09:18, 3.63s/it] 71%|███████ | 367/520 [23:06<09:17, 3.65s/it] {'loss': 1.2236, 'grad_norm': 0.0013316946845547209, 'learning_rate': 0.01895979167190874, 'epoch': 0.71} + 71%|███████ | 367/520 [23:06<09:17, 3.65s/it] 71%|███████ | 368/520 [23:10<09:13, 3.64s/it] {'loss': 1.0768, 'grad_norm': 0.0012910009372143541, 'learning_rate': 0.018731534749434467, 'epoch': 0.71} + 71%|███████ | 368/520 [23:10<09:13, 3.64s/it] 71%|███████ | 369/520 [23:13<09:09, 3.64s/it] {'loss': 1.1725, 'grad_norm': 0.0011399857553527795, 'learning_rate': 0.01850429846550579, 'epoch': 0.71} + 71%|███████ | 369/520 [23:13<09:09, 3.64s/it] 71%|███████ | 370/520 [23:17<09:05, 3.64s/it] {'loss': 1.1388, 'grad_norm': 0.0012172746499795385, 'learning_rate': 0.018278091649192432, 'epoch': 0.71} + 71%|███████ | 370/520 [23:17<09:05, 3.64s/it] 71%|███████▏ | 371/520 [23:21<09:00, 3.63s/it] {'loss': 1.1228, 'grad_norm': 0.001352522052077922, 'learning_rate': 0.018052923089564986, 'epoch': 0.71} + 71%|███████▏ | 371/520 [23:21<09:00, 3.63s/it] 72%|███████▏ | 372/520 [23:24<08:57, 3.63s/it] {'loss': 1.2408, 'grad_norm': 0.0011343973593853532, 'learning_rate': 0.017828801535353506, 'epoch': 0.72} + 72%|███████▏ | 372/520 [23:24<08:57, 3.63s/it] 72%|███████▏ | 373/520 [23:28<08:53, 3.63s/it] {'loss': 1.1289, 'grad_norm': 0.0013446738415606823, 'learning_rate': 0.01760573569460757, 'epoch': 0.72} + 72%|███████▏ | 373/520 [23:28<08:53, 3.63s/it] 72%|███████▏ | 374/520 [23:32<08:48, 3.62s/it] {'loss': 1.2206, 'grad_norm': 0.0013778551524888428, 'learning_rate': 0.017383734234357875, 'epoch': 0.72} + 72%|███████▏ | 374/520 [23:32<08:48, 3.62s/it] 72%|███████▏ | 375/520 [23:35<08:45, 3.62s/it] {'loss': 1.1382, 'grad_norm': 0.0012965482096596067, 'learning_rate': 0.01716280578027953, 'epoch': 0.72} + 72%|███████▏ | 375/520 [23:35<08:45, 3.62s/it] 72%|███████▏ | 376/520 [23:39<08:41, 3.62s/it] {'loss': 1.2427, 'grad_norm': 0.001229288920522398, 'learning_rate': 0.01694295891635699, 'epoch': 0.72} + 72%|███████▏ | 376/520 [23:39<08:41, 3.62s/it] 72%|███████▎ | 377/520 [23:42<08:36, 3.61s/it] {'loss': 1.1741, 'grad_norm': 0.0013925816630739048, 'learning_rate': 0.016724202184550372, 'epoch': 0.72} + 72%|███████▎ | 377/520 [23:42<08:36, 3.61s/it] 73%|███████▎ | 378/520 [23:46<08:33, 3.62s/it] {'loss': 1.2375, 'grad_norm': 0.0012194222128422013, 'learning_rate': 0.016506544084463712, 'epoch': 0.73} + 73%|███████▎ | 378/520 [23:46<08:33, 3.62s/it] 73%|███████▎ | 379/520 [23:50<08:31, 3.62s/it] {'loss': 1.2031, 'grad_norm': 0.0011915808380809495, 'learning_rate': 0.016289993073014618, 'epoch': 0.73} + 73%|███████▎ | 379/520 [23:50<08:31, 3.62s/it] 73%|███████▎ | 380/520 [23:53<08:26, 3.62s/it] {'loss': 1.2186, 'grad_norm': 0.001288800148306863, 'learning_rate': 0.016074557564105727, 'epoch': 0.73} + 73%|███████▎ | 380/520 [23:53<08:26, 3.62s/it] 73%|███████▎ | 381/520 [23:57<08:22, 3.62s/it] {'loss': 1.2132, 'grad_norm': 0.0012073287774165959, 'learning_rate': 0.015860245928297836, 'epoch': 0.73} + 73%|███████▎ | 381/520 [23:57<08:22, 3.62s/it] 73%|███████▎ | 382/520 [24:01<08:21, 3.64s/it] {'loss': 1.1878, 'grad_norm': 0.0011606285262287916, 'learning_rate': 0.01564706649248456, 'epoch': 0.73} + 73%|███████▎ | 382/520 [24:01<08:21, 3.64s/it] 74%|███████▎ | 383/520 [24:04<08:21, 3.66s/it] {'loss': 1.0589, 'grad_norm': 0.0013985371074556052, 'learning_rate': 0.015435027539568883, 'epoch': 0.74} + 74%|███████▎ | 383/520 [24:04<08:21, 3.66s/it] 74%|███████▍ | 384/520 [24:08<08:17, 3.66s/it] {'loss': 1.21, 'grad_norm': 0.0010856532493908643, 'learning_rate': 0.015224137308141338, 'epoch': 0.74} + 74%|███████▍ | 384/520 [24:08<08:17, 3.66s/it] 74%|███████▍ | 385/520 [24:12<08:14, 3.66s/it] {'loss': 1.2008, 'grad_norm': 0.0012428719824675136, 'learning_rate': 0.015014403992159824, 'epoch': 0.74} + 74%|███████▍ | 385/520 [24:12<08:14, 3.66s/it] 74%|███████▍ | 386/520 [24:15<08:11, 3.67s/it] {'loss': 1.1505, 'grad_norm': 0.0010919153339260198, 'learning_rate': 0.014805835740631353, 'epoch': 0.74} + 74%|███████▍ | 386/520 [24:15<08:11, 3.67s/it] 74%|███████▍ | 387/520 [24:19<08:07, 3.67s/it] {'loss': 1.2372, 'grad_norm': 0.001219361718129805, 'learning_rate': 0.014598440657295288, 'epoch': 0.74} + 74%|███████▍ | 387/520 [24:19<08:07, 3.67s/it] 75%|███████▍ | 388/520 [24:23<08:04, 3.67s/it] {'loss': 1.1121, 'grad_norm': 0.0012152449972158504, 'learning_rate': 0.014392226800308619, 'epoch': 0.75} + 75%|███████▍ | 388/520 [24:23<08:04, 3.67s/it] 75%|███████▍ | 389/520 [24:26<07:59, 3.66s/it] {'loss': 1.1571, 'grad_norm': 0.001564317799277769, 'learning_rate': 0.014187202181932791, 'epoch': 0.75} + 75%|███████▍ | 389/520 [24:26<07:59, 3.66s/it] 75%|███████▌ | 390/520 [24:30<07:56, 3.66s/it] {'loss': 1.2272, 'grad_norm': 0.0012296663863237685, 'learning_rate': 0.013983374768222383, 'epoch': 0.75} + 75%|███████▌ | 390/520 [24:30<07:56, 3.66s/it] 75%|███████▌ | 391/520 [24:34<07:52, 3.66s/it] {'loss': 1.285, 'grad_norm': 0.0012826567954692988, 'learning_rate': 0.013780752478715625, 'epoch': 0.75} + 75%|███████▌ | 391/520 [24:34<07:52, 3.66s/it] 75%|███████▌ | 392/520 [24:37<07:48, 3.66s/it] {'loss': 1.113, 'grad_norm': 0.001251865553154953, 'learning_rate': 0.013579343186126726, 'epoch': 0.75} + 75%|███████▌ | 392/520 [24:37<07:48, 3.66s/it] 76%|███████▌ | 393/520 [24:41<07:44, 3.66s/it] {'loss': 1.1007, 'grad_norm': 0.0010857849591456286, 'learning_rate': 0.013379154716039888, 'epoch': 0.76} + 76%|███████▌ | 393/520 [24:41<07:44, 3.66s/it] 76%|███████▌ | 394/520 [24:45<07:43, 3.68s/it] {'loss': 1.181, 'grad_norm': 0.0013423094050021474, 'learning_rate': 0.013180194846605363, 'epoch': 0.76} + 76%|███████▌ | 394/520 [24:45<07:43, 3.68s/it] 76%|███████▌ | 395/520 [24:48<07:44, 3.72s/it] {'loss': 1.1502, 'grad_norm': 0.001376638767344858, 'learning_rate': 0.012982471308237153, 'epoch': 0.76} + 76%|███████▌ | 395/520 [24:48<07:44, 3.72s/it] 76%|███████▌ | 396/520 [24:52<07:43, 3.74s/it] {'loss': 1.2244, 'grad_norm': 0.0013633990724899285, 'learning_rate': 0.01278599178331267, 'epoch': 0.76} + 76%|███████▌ | 396/520 [24:52<07:43, 3.74s/it] 76%|███████▋ | 397/520 [24:56<07:43, 3.77s/it] {'loss': 1.1995, 'grad_norm': 0.0012569221629758734, 'learning_rate': 0.012590763905874313, 'epoch': 0.76} + 76%|███████▋ | 397/520 [24:56<07:43, 3.77s/it] 77%|███████▋ | 398/520 [25:00<07:41, 3.78s/it] {'loss': 1.1935, 'grad_norm': 0.001349747354312207, 'learning_rate': 0.01239679526133273, 'epoch': 0.77} + 77%|███████▋ | 398/520 [25:00<07:41, 3.78s/it] 77%|███████▋ | 399/520 [25:04<07:39, 3.80s/it] {'loss': 1.1315, 'grad_norm': 0.0012010110887628839, 'learning_rate': 0.012204093386172225, 'epoch': 0.77} + 77%|███████▋ | 399/520 [25:04<07:39, 3.80s/it] 77%|███████▋ | 400/520 [25:08<07:38, 3.82s/it] {'loss': 1.1634, 'grad_norm': 0.00115396815697924, 'learning_rate': 0.012012665767657825, 'epoch': 0.77} + 77%|███████▋ | 400/520 [25:08<07:38, 3.82s/it] 77%|███████▋ | 401/520 [25:11<07:30, 3.79s/it] {'loss': 1.0383, 'grad_norm': 0.0013695100508914913, 'learning_rate': 0.01182251984354442, 'epoch': 0.77} + 77%|███████▋ | 401/520 [25:11<07:30, 3.79s/it] 77%|███████▋ | 402/520 [25:15<07:21, 3.74s/it] {'loss': 1.1666, 'grad_norm': 0.0013177261518952856, 'learning_rate': 0.011633663001787796, 'epoch': 0.77} + 77%|███████▋ | 402/520 [25:15<07:21, 3.74s/it] 78%|███████▊ | 403/520 [25:19<07:12, 3.70s/it] {'loss': 1.1864, 'grad_norm': 0.001391772306549658, 'learning_rate': 0.011446102580257548, 'epoch': 0.78} + 78%|███████▊ | 403/520 [25:19<07:12, 3.70s/it] 78%|███████▊ | 404/520 [25:22<07:05, 3.67s/it] {'loss': 1.1016, 'grad_norm': 0.0014825257556678082, 'learning_rate': 0.011259845866451955, 'epoch': 0.78} + 78%|███████▊ | 404/520 [25:22<07:05, 3.67s/it] 78%|███████▊ | 405/520 [25:26<07:02, 3.67s/it] {'loss': 1.146, 'grad_norm': 0.0012408858372508435, 'learning_rate': 0.011074900097214908, 'epoch': 0.78} + 78%|███████▊ | 405/520 [25:26<07:02, 3.67s/it] 78%|███████▊ | 406/520 [25:29<06:57, 3.66s/it] {'loss': 1.0732, 'grad_norm': 0.001482935802518131, 'learning_rate': 0.010891272458454608, 'epoch': 0.78} + 78%|███████▊ | 406/520 [25:29<06:57, 3.66s/it] 78%|███████▊ | 407/520 [25:33<06:53, 3.66s/it] {'loss': 1.2665, 'grad_norm': 0.0013003638290999546, 'learning_rate': 0.010708970084864513, 'epoch': 0.78} + 78%|███████▊ | 407/520 [25:33<06:53, 3.66s/it] 78%|███████▊ | 408/520 [25:37<06:48, 3.65s/it] {'loss': 1.1803, 'grad_norm': 0.0014218062134506385, 'learning_rate': 0.010528000059645994, 'epoch': 0.78} + 78%|███████▊ | 408/520 [25:37<06:48, 3.65s/it] 79%|███████▊ | 409/520 [25:40<06:44, 3.64s/it] {'loss': 1.2974, 'grad_norm': 0.0013658454459943151, 'learning_rate': 0.010348369414233173, 'epoch': 0.79} + 79%|███████▊ | 409/520 [25:40<06:44, 3.64s/it] 79%|███████▉ | 410/520 [25:44<06:40, 3.64s/it] {'loss': 1.0398, 'grad_norm': 0.0012993760582202083, 'learning_rate': 0.010170085128019767, 'epoch': 0.79} + 79%|███████▉ | 410/520 [25:44<06:40, 3.64s/it] 79%|███████▉ | 411/520 [25:48<06:36, 3.63s/it] {'loss': 1.2775, 'grad_norm': 0.001375334548166793, 'learning_rate': 0.009993154128087836, 'epoch': 0.79} + 79%|███████▉ | 411/520 [25:48<06:36, 3.63s/it] 79%|███████▉ | 412/520 [25:51<06:31, 3.63s/it] {'loss': 1.1853, 'grad_norm': 0.0012636636847934302, 'learning_rate': 0.009817583288938659, 'epoch': 0.79} + 79%|███████▉ | 412/520 [25:51<06:31, 3.63s/it] 79%|███████▉ | 413/520 [25:55<06:28, 3.63s/it] {'loss': 1.1624, 'grad_norm': 0.0011913865702666477, 'learning_rate': 0.009643379432225692, 'epoch': 0.79} + 79%|███████▉ | 413/520 [25:55<06:28, 3.63s/it] 80%|███████▉ | 414/520 [25:59<06:25, 3.64s/it] {'loss': 0.9744, 'grad_norm': 0.0010506016430238648, 'learning_rate': 0.00947054932648941, 'epoch': 0.8} + 80%|███████▉ | 414/520 [25:59<06:25, 3.64s/it] 80%|███████▉ | 415/520 [26:02<06:21, 3.64s/it] {'loss': 1.1678, 'grad_norm': 0.0012181063505547448, 'learning_rate': 0.009299099686894421, 'epoch': 0.8} + 80%|███████▉ | 415/520 [26:02<06:21, 3.64s/it] 80%|████████ | 416/520 [26:06<06:18, 3.64s/it] {'loss': 1.0712, 'grad_norm': 0.0013952754319526712, 'learning_rate': 0.009129037174968503, 'epoch': 0.8} + 80%|████████ | 416/520 [26:06<06:18, 3.64s/it] 80%|████████ | 417/520 [26:09<06:14, 3.64s/it] {'loss': 1.2346, 'grad_norm': 0.0012733456204164848, 'learning_rate': 0.008960368398343747, 'epoch': 0.8} + 80%|████████ | 417/520 [26:09<06:14, 3.64s/it] 80%|████████ | 418/520 [26:13<06:11, 3.64s/it] {'loss': 1.2291, 'grad_norm': 0.0011917896603487379, 'learning_rate': 0.008793099910499924, 'epoch': 0.8} + 80%|████████ | 418/520 [26:13<06:11, 3.64s/it] 81%|████████ | 419/520 [26:17<06:08, 3.65s/it] {'loss': 1.2216, 'grad_norm': 0.0014056340588320068, 'learning_rate': 0.008627238210509765, 'epoch': 0.81} + 81%|████████ | 419/520 [26:17<06:08, 3.65s/it] 81%|████████ | 420/520 [26:20<06:04, 3.65s/it] {'loss': 1.1124, 'grad_norm': 0.0013324307406657938, 'learning_rate': 0.008462789742786457, 'epoch': 0.81} + 81%|████████ | 420/520 [26:20<06:04, 3.65s/it] 81%|████████ | 421/520 [26:24<06:01, 3.65s/it] {'loss': 1.0518, 'grad_norm': 0.0013404665117765719, 'learning_rate': 0.008299760896833291, 'epoch': 0.81} + 81%|████████ | 421/520 [26:24<06:01, 3.65s/it] 81%|████████ | 422/520 [26:28<05:56, 3.64s/it] {'loss': 1.1741, 'grad_norm': 0.0013575764116163789, 'learning_rate': 0.008138158006995364, 'epoch': 0.81} + 81%|████████ | 422/520 [26:28<05:56, 3.64s/it] 81%|████████▏ | 423/520 [26:31<05:52, 3.63s/it] {'loss': 1.1406, 'grad_norm': 0.0013857126690079854, 'learning_rate': 0.007977987352213499, 'epoch': 0.81} + 81%|████████▏ | 423/520 [26:31<05:52, 3.63s/it] 82%|████████▏ | 424/520 [26:35<05:49, 3.64s/it] {'loss': 1.246, 'grad_norm': 0.0011684729047259535, 'learning_rate': 0.007819255155780238, 'epoch': 0.82} + 82%|████████▏ | 424/520 [26:35<05:49, 3.64s/it] 82%|████████▏ | 425/520 [26:39<05:44, 3.63s/it] {'loss': 1.1588, 'grad_norm': 0.0012689432412996264, 'learning_rate': 0.0076619675850980624, 'epoch': 0.82} + 82%|████████▏ | 425/520 [26:39<05:44, 3.63s/it] 82%|████████▏ | 426/520 [26:42<05:41, 3.63s/it] {'loss': 1.1937, 'grad_norm': 0.0016998894956206608, 'learning_rate': 0.0075061307514398025, 'epoch': 0.82} + 82%|████████▏ | 426/520 [26:42<05:41, 3.63s/it] 82%|████████▏ | 427/520 [26:46<05:38, 3.63s/it] {'loss': 1.0912, 'grad_norm': 0.0012289399696905439, 'learning_rate': 0.007351750709711111, 'epoch': 0.82} + 82%|████████▏ | 427/520 [26:46<05:38, 3.63s/it] 82%|████████▏ | 428/520 [26:49<05:34, 3.63s/it] {'loss': 1.0863, 'grad_norm': 0.0013933980097123005, 'learning_rate': 0.007198833458215287, 'epoch': 0.82} + 82%|████████▏ | 428/520 [26:49<05:34, 3.63s/it] 82%|████████▎ | 429/520 [26:53<05:31, 3.64s/it] {'loss': 1.1827, 'grad_norm': 0.0012920768886835231, 'learning_rate': 0.007047384938420153, 'epoch': 0.82} + 82%|████████▎ | 429/520 [26:53<05:31, 3.64s/it]Token indices sequence length is longer than the specified maximum sequence length for this model (2076 > 2048). Running this sequence through the model will result in indexing errors + 83%|████████▎ | 430/520 [26:57<05:27, 3.64s/it] {'loss': 1.1813, 'grad_norm': 0.0012239011411575233, 'learning_rate': 0.006897411034727214, 'epoch': 0.83} + 83%|████████▎ | 430/520 [26:57<05:27, 3.64s/it] 83%|████████▎ | 431/520 [27:00<05:25, 3.66s/it] {'loss': 1.1357, 'grad_norm': 0.0012425190182162264, 'learning_rate': 0.006748917574243089, 'epoch': 0.83} + 83%|████████▎ | 431/520 [27:00<05:25, 3.66s/it] 83%|████████▎ | 432/520 [27:04<05:22, 3.67s/it] {'loss': 1.0887, 'grad_norm': 0.0013380241420592423, 'learning_rate': 0.006601910326552997, 'epoch': 0.83} + 83%|████████▎ | 432/520 [27:04<05:22, 3.67s/it] 83%|████████▎ | 433/520 [27:08<05:17, 3.65s/it] {'loss': 1.2233, 'grad_norm': 0.0012661296222993387, 'learning_rate': 0.00645639500349669, 'epoch': 0.83} + 83%|████████▎ | 433/520 [27:08<05:17, 3.65s/it] 83%|████████▎ | 434/520 [27:11<05:13, 3.64s/it] {'loss': 0.9761, 'grad_norm': 0.0013280415141483882, 'learning_rate': 0.0063123772589464364, 'epoch': 0.83} + 83%|████████▎ | 434/520 [27:11<05:13, 3.64s/it] 84%|████████▎ | 435/520 [27:15<05:08, 3.63s/it] {'loss': 1.2526, 'grad_norm': 0.0013801572555908871, 'learning_rate': 0.006169862688587413, 'epoch': 0.84} + 84%|████████▎ | 435/520 [27:15<05:08, 3.63s/it] 84%|████████▍ | 436/520 [27:19<05:04, 3.63s/it] {'loss': 1.0656, 'grad_norm': 0.0013639548707399893, 'learning_rate': 0.006028856829700258, 'epoch': 0.84} + 84%|████████▍ | 436/520 [27:19<05:04, 3.63s/it] 84%|████████▍ | 437/520 [27:22<05:00, 3.62s/it] {'loss': 1.2765, 'grad_norm': 0.0013138464125765654, 'learning_rate': 0.005889365160945912, 'epoch': 0.84} + 84%|████████▍ | 437/520 [27:22<05:00, 3.62s/it] 84%|████████▍ | 438/520 [27:26<04:56, 3.62s/it] {'loss': 1.0938, 'grad_norm': 0.001271250631253267, 'learning_rate': 0.00575139310215276, 'epoch': 0.84} + 84%|████████▍ | 438/520 [27:26<04:56, 3.62s/it] 84%|████████▍ | 439/520 [27:29<04:53, 3.62s/it] {'loss': 1.1193, 'grad_norm': 0.0010310767346916256, 'learning_rate': 0.005614946014106084, 'epoch': 0.84} + 84%|████████▍ | 439/520 [27:29<04:53, 3.62s/it] 85%|████████▍ | 440/520 [27:33<04:48, 3.61s/it] {'loss': 1.1331, 'grad_norm': 0.0013449110247739497, 'learning_rate': 0.00548002919833971, 'epoch': 0.85} + 85%|████████▍ | 440/520 [27:33<04:48, 3.61s/it] 85%|████████▍ | 441/520 [27:37<04:46, 3.63s/it] {'loss': 1.1315, 'grad_norm': 0.0012952873610193357, 'learning_rate': 0.005346647896930091, 'epoch': 0.85} + 85%|████████▍ | 441/520 [27:37<04:46, 3.63s/it] 85%|████████▌ | 442/520 [27:40<04:42, 3.63s/it] {'loss': 1.1941, 'grad_norm': 0.0014312864616740904, 'learning_rate': 0.005214807292292565, 'epoch': 0.85} + 85%|████████▌ | 442/520 [27:40<04:42, 3.63s/it] 85%|████████▌ | 443/520 [27:44<04:39, 3.64s/it] {'loss': 1.2018, 'grad_norm': 0.0012676174815932167, 'learning_rate': 0.005084512506980022, 'epoch': 0.85} + 85%|████████▌ | 443/520 [27:44<04:39, 3.64s/it] 85%|████████▌ | 444/520 [27:48<04:36, 3.64s/it] {'loss': 1.172, 'grad_norm': 0.0011497757838080857, 'learning_rate': 0.004955768603483915, 'epoch': 0.85} + 85%|████████▌ | 444/520 [27:48<04:36, 3.64s/it] 86%|████████▌ | 445/520 [27:51<04:32, 3.63s/it] {'loss': 1.0992, 'grad_norm': 0.0012259486544054518, 'learning_rate': 0.00482858058403749, 'epoch': 0.86} + 86%|████████▌ | 445/520 [27:51<04:32, 3.63s/it] 86%|████████▌ | 446/520 [27:55<04:29, 3.64s/it] {'loss': 1.2078, 'grad_norm': 0.0011268949553359757, 'learning_rate': 0.004702953390421458, 'epoch': 0.86} + 86%|████████▌ | 446/520 [27:55<04:29, 3.64s/it] 86%|████████▌ | 447/520 [27:59<04:27, 3.66s/it] {'loss': 1.1648, 'grad_norm': 0.0012654382262247985, 'learning_rate': 0.004578891903772018, 'epoch': 0.86} + 86%|████████▌ | 447/520 [27:59<04:27, 3.66s/it] 86%|████████▌ | 448/520 [28:02<04:23, 3.66s/it] {'loss': 1.1665, 'grad_norm': 0.001418477890983188, 'learning_rate': 0.0044564009443911435, 'epoch': 0.86} + 86%|████████▌ | 448/520 [28:02<04:23, 3.66s/it] 86%|████████▋ | 449/520 [28:06<04:20, 3.67s/it] {'loss': 1.1694, 'grad_norm': 0.0012474602948013517, 'learning_rate': 0.004335485271559358, 'epoch': 0.86} + 86%|████████▋ | 449/520 [28:06<04:20, 3.67s/it] 87%|████████▋ | 450/520 [28:10<04:16, 3.66s/it] {'loss': 1.1924, 'grad_norm': 0.0013106387468138497, 'learning_rate': 0.004216149583350753, 'epoch': 0.87} + 87%|████████▋ | 450/520 [28:10<04:16, 3.66s/it] 87%|████████▋ | 451/520 [28:13<04:13, 3.68s/it] {'loss': 1.1979, 'grad_norm': 0.0013416982646933776, 'learning_rate': 0.004098398516450508, 'epoch': 0.87} + 87%|████████▋ | 451/520 [28:13<04:13, 3.68s/it] 87%|████████▋ | 452/520 [28:17<04:09, 3.67s/it] {'loss': 1.2168, 'grad_norm': 0.001176638573816191, 'learning_rate': 0.003982236645974709, 'epoch': 0.87} + 87%|████████▋ | 452/520 [28:17<04:09, 3.67s/it] 87%|████████▋ | 453/520 [28:21<04:05, 3.67s/it] {'loss': 1.1913, 'grad_norm': 0.0011937455119698332, 'learning_rate': 0.0038676684852925647, 'epoch': 0.87} + 87%|████████▋ | 453/520 [28:21<04:05, 3.67s/it] 87%|████████▋ | 454/520 [28:24<04:02, 3.67s/it] {'loss': 1.1043, 'grad_norm': 0.001284733665224088, 'learning_rate': 0.003754698485851071, 'epoch': 0.87} + 87%|████████▋ | 454/520 [28:24<04:02, 3.67s/it] 88%|████████▊ | 455/520 [28:28<03:57, 3.65s/it] {'loss': 1.243, 'grad_norm': 0.0012884939206680675, 'learning_rate': 0.0036433310370020703, 'epoch': 0.88} + 88%|████████▊ | 455/520 [28:28<03:57, 3.65s/it] 88%|████████▊ | 456/520 [28:32<03:52, 3.64s/it] {'loss': 1.1784, 'grad_norm': 0.001314118526040304, 'learning_rate': 0.0035335704658316517, 'epoch': 0.88} + 88%|████████▊ | 456/520 [28:32<03:52, 3.64s/it] 88%|████████▊ | 457/520 [28:35<03:50, 3.65s/it] {'loss': 1.0746, 'grad_norm': 0.0010969619307770466, 'learning_rate': 0.0034254210369920966, 'epoch': 0.88} + 88%|████████▊ | 457/520 [28:35<03:50, 3.65s/it] 88%|████████▊ | 458/520 [28:39<03:46, 3.65s/it] {'loss': 1.2949, 'grad_norm': 0.0013880633324222778, 'learning_rate': 0.003318886952536111, 'epoch': 0.88} + 88%|████████▊ | 458/520 [28:39<03:46, 3.65s/it] 88%|████████▊ | 459/520 [28:42<03:42, 3.64s/it] {'loss': 1.2257, 'grad_norm': 0.00126090210806565, 'learning_rate': 0.00321397235175359, 'epoch': 0.88} + 88%|████████▊ | 459/520 [28:42<03:42, 3.64s/it] 88%|████████▊ | 460/520 [28:46<03:38, 3.64s/it] {'loss': 1.1153, 'grad_norm': 0.0012628296331972269, 'learning_rate': 0.0031106813110108136, 'epoch': 0.88} + 88%|████████▊ | 460/520 [28:46<03:38, 3.64s/it] 89%|████████▊ | 461/520 [28:50<03:34, 3.64s/it] {'loss': 1.1587, 'grad_norm': 0.0009252808255364296, 'learning_rate': 0.003009017843592007, 'epoch': 0.89} + 89%|████████▊ | 461/520 [28:50<03:34, 3.64s/it] 89%|████████▉ | 462/520 [28:53<03:31, 3.64s/it] {'loss': 1.2558, 'grad_norm': 0.0012088698984748522, 'learning_rate': 0.0029089858995434703, 'epoch': 0.89} + 89%|████████▉ | 462/520 [28:53<03:31, 3.64s/it] 89%|████████▉ | 463/520 [28:57<03:27, 3.65s/it] {'loss': 1.092, 'grad_norm': 0.001338021779684646, 'learning_rate': 0.002810589365520041, 'epoch': 0.89} + 89%|████████▉ | 463/520 [28:57<03:27, 3.65s/it] 89%|████████▉ | 464/520 [29:01<03:24, 3.65s/it] {'loss': 1.2089, 'grad_norm': 0.0013027541281740223, 'learning_rate': 0.0027138320646341255, 'epoch': 0.89} + 89%|████████▉ | 464/520 [29:01<03:24, 3.65s/it] 89%|████████▉ | 465/520 [29:04<03:21, 3.66s/it] {'loss': 1.3114, 'grad_norm': 0.0013334402540176005, 'learning_rate': 0.002618717756307144, 'epoch': 0.89} + 89%|████████▉ | 465/520 [29:04<03:21, 3.66s/it] 90%|████████▉ | 466/520 [29:08<03:17, 3.66s/it] {'loss': 1.2104, 'grad_norm': 0.0011705020323442522, 'learning_rate': 0.002525250136123459, 'epoch': 0.9} + 90%|████████▉ | 466/520 [29:08<03:17, 3.66s/it] 90%|████████▉ | 467/520 [29:12<03:14, 3.67s/it] {'loss': 1.1469, 'grad_norm': 0.0011796332072901564, 'learning_rate': 0.002433432835686779, 'epoch': 0.9} + 90%|████████▉ | 467/520 [29:12<03:14, 3.67s/it] 90%|█████████ | 468/520 [29:15<03:10, 3.65s/it] {'loss': 1.1716, 'grad_norm': 0.0014610892519786507, 'learning_rate': 0.0023432694224790732, 'epoch': 0.9} + 90%|█████████ | 468/520 [29:15<03:10, 3.65s/it] 90%|█████████ | 469/520 [29:19<03:05, 3.64s/it] {'loss': 1.2428, 'grad_norm': 0.001375242362146661, 'learning_rate': 0.00225476339972193, 'epoch': 0.9} + 90%|█████████ | 469/520 [29:19<03:05, 3.64s/it] 90%|█████████ | 470/520 [29:23<03:03, 3.67s/it] {'loss': 1.1154, 'grad_norm': 0.001188510339824273, 'learning_rate': 0.002167918206240494, 'epoch': 0.9} + 90%|█████████ | 470/520 [29:23<03:03, 3.67s/it] 91%|█████████ | 471/520 [29:27<03:03, 3.74s/it] {'loss': 1.1441, 'grad_norm': 0.0013981026026048437, 'learning_rate': 0.002082737216329793, 'epoch': 0.91} + 91%|█████████ | 471/520 [29:27<03:03, 3.74s/it] 91%|█████████ | 472/520 [29:30<03:00, 3.76s/it] {'loss': 1.1129, 'grad_norm': 0.0012249410717347312, 'learning_rate': 0.0019992237396236645, 'epoch': 0.91} + 91%|█████████ | 472/520 [29:30<03:00, 3.76s/it] 91%|█████████ | 473/520 [29:34<02:55, 3.73s/it] {'loss': 1.1857, 'grad_norm': 0.001335023746097261, 'learning_rate': 0.0019173810209661868, 'epoch': 0.91} + 91%|█████████ | 473/520 [29:34<02:55, 3.73s/it] 91%|█████████ | 474/520 [29:38<02:50, 3.70s/it] {'loss': 1.18, 'grad_norm': 0.0011797744518014378, 'learning_rate': 0.0018372122402855505, 'epoch': 0.91} + 91%|█████████ | 474/520 [29:38<02:50, 3.70s/it] 91%|█████████▏| 475/520 [29:41<02:45, 3.69s/it] {'loss': 1.0962, 'grad_norm': 0.0011797647380544546, 'learning_rate': 0.001758720512470523, 'epoch': 0.91} + 91%|█████████▏| 475/520 [29:41<02:45, 3.69s/it] 92%|█████████▏| 476/520 [29:45<02:41, 3.66s/it] {'loss': 1.1675, 'grad_norm': 0.00135065471753282, 'learning_rate': 0.0016819088872494586, 'epoch': 0.92} + 92%|█████████▏| 476/520 [29:45<02:41, 3.66s/it] 92%|█████████▏| 477/520 [29:49<02:36, 3.64s/it] {'loss': 1.1649, 'grad_norm': 0.001407473597406844, 'learning_rate': 0.0016067803490717552, 'epoch': 0.92} + 92%|█████████▏| 477/520 [29:49<02:36, 3.64s/it] 92%|█████████▏| 478/520 [29:52<02:32, 3.63s/it] {'loss': 1.1045, 'grad_norm': 0.0012547111340480682, 'learning_rate': 0.001533337816991931, 'epoch': 0.92} + 92%|█████████▏| 478/520 [29:52<02:32, 3.63s/it] 92%|█████████▏| 479/520 [29:56<02:31, 3.69s/it] {'loss': 1.1538, 'grad_norm': 0.0013394074257541523, 'learning_rate': 0.001461584144556175, 'epoch': 0.92} + 92%|█████████▏| 479/520 [29:56<02:31, 3.69s/it] 92%|█████████▏| 480/520 [30:00<02:29, 3.73s/it] {'loss': 1.1686, 'grad_norm': 0.0011874885213157865, 'learning_rate': 0.0013915221196914968, 'epoch': 0.92} + 92%|█████████▏| 480/520 [30:00<02:29, 3.73s/it] 92%|█████████▎| 481/520 [30:03<02:24, 3.70s/it] {'loss': 1.1563, 'grad_norm': 0.001114214273026453, 'learning_rate': 0.001323154464597407, 'epoch': 0.93} + 92%|█████████▎| 481/520 [30:03<02:24, 3.70s/it] 93%|█████████▎| 482/520 [30:07<02:20, 3.69s/it] {'loss': 1.1771, 'grad_norm': 0.001157280397025905, 'learning_rate': 0.0012564838356401474, 'epoch': 0.93} + 93%|█████████▎| 482/520 [30:07<02:20, 3.69s/it] 93%|█████████▎| 483/520 [30:11<02:16, 3.69s/it] {'loss': 1.1748, 'grad_norm': 0.0012672427503351103, 'learning_rate': 0.0011915128232494493, 'epoch': 0.93} + 93%|█████████▎| 483/520 [30:11<02:16, 3.69s/it] 93%|█████████▎| 484/520 [30:14<02:12, 3.68s/it] {'loss': 1.1855, 'grad_norm': 0.0013293280752086332, 'learning_rate': 0.001128243951817937, 'epoch': 0.93} + 93%|█████████▎| 484/520 [30:14<02:12, 3.68s/it] 93%|█████████▎| 485/520 [30:18<02:10, 3.73s/it] {'loss': 1.1354, 'grad_norm': 0.0012473248026638403, 'learning_rate': 0.0010666796796029987, 'epoch': 0.93} + 93%|█████████▎| 485/520 [30:18<02:10, 3.73s/it] 93%|█████████▎| 486/520 [30:22<02:07, 3.74s/it] {'loss': 1.2552, 'grad_norm': 0.0013133975839038962, 'learning_rate': 0.0010068223986312956, 'epoch': 0.93} + 93%|█████████▎| 486/520 [30:22<02:07, 3.74s/it] 94%|█████████▎| 487/520 [30:26<02:04, 3.77s/it] {'loss': 1.1198, 'grad_norm': 0.0012404620316747344, 'learning_rate': 0.0009486744346058234, 'epoch': 0.94} + 94%|█████████▎| 487/520 [30:26<02:04, 3.77s/it] 94%|█████████▍| 488/520 [30:30<02:01, 3.79s/it] {'loss': 1.0621, 'grad_norm': 0.0013120122477696915, 'learning_rate': 0.0008922380468155278, 'epoch': 0.94} + 94%|█████████▍| 488/520 [30:30<02:01, 3.79s/it] 94%|█████████▍| 489/520 [30:34<01:57, 3.80s/it] {'loss': 1.1791, 'grad_norm': 0.001078211302989296, 'learning_rate': 0.0008375154280475555, 'epoch': 0.94} + 94%|█████████▍| 489/520 [30:34<01:57, 3.80s/it] 94%|█████████▍| 490/520 [30:37<01:54, 3.82s/it] {'loss': 1.1799, 'grad_norm': 0.001305270802280329, 'learning_rate': 0.0007845087045020276, 'epoch': 0.94} + 94%|█████████▍| 490/520 [30:37<01:54, 3.82s/it] 94%|█████████▍| 491/520 [30:41<01:50, 3.83s/it] {'loss': 1.1453, 'grad_norm': 0.0013644499364905969, 'learning_rate': 0.0007332199357094404, 'epoch': 0.94} + 94%|█████████▍| 491/520 [30:41<01:50, 3.83s/it] 95%|█████████▍| 492/520 [30:45<01:47, 3.82s/it] {'loss': 1.2568, 'grad_norm': 0.0013130367277938858, 'learning_rate': 0.000683651114450639, 'epoch': 0.95} + 95%|█████████▍| 492/520 [30:45<01:47, 3.82s/it] 95%|█████████▍| 493/520 [30:49<01:43, 3.83s/it] {'loss': 1.1726, 'grad_norm': 0.001336342963026261, 'learning_rate': 0.000635804166679385, 'epoch': 0.95} + 95%|█████████▍| 493/520 [30:49<01:43, 3.83s/it] 95%|█████████▌| 494/520 [30:53<01:39, 3.84s/it] {'loss': 1.1837, 'grad_norm': 0.0011752087831870671, 'learning_rate': 0.0005896809514475509, 'epoch': 0.95} + 95%|█████████▌| 494/520 [30:53<01:39, 3.84s/it] 95%|█████████▌| 495/520 [30:57<01:35, 3.84s/it] {'loss': 1.169, 'grad_norm': 0.0013987195848385293, 'learning_rate': 0.0005452832608328378, 'epoch': 0.95} + 95%|█████████▌| 495/520 [30:57<01:35, 3.84s/it] 95%|█████████▌| 496/520 [31:01<01:32, 3.84s/it] {'loss': 1.0973, 'grad_norm': 0.0013689643382655081, 'learning_rate': 0.0005026128198692164, 'epoch': 0.95} + 95%|█████████▌| 496/520 [31:01<01:32, 3.84s/it] 96%|█████████▌| 497/520 [31:04<01:28, 3.85s/it] {'loss': 1.1086, 'grad_norm': 0.0010838630189646712, 'learning_rate': 0.00046167128647983056, 'epoch': 0.96} + 96%|█████████▌| 497/520 [31:04<01:28, 3.85s/it] 96%|█████████▌| 498/520 [31:08<01:24, 3.85s/it] {'loss': 1.1584, 'grad_norm': 0.0012963482581111263, 'learning_rate': 0.00042246025141262357, 'epoch': 0.96} + 96%|█████████▌| 498/520 [31:08<01:24, 3.85s/it] 96%|█████████▌| 499/520 [31:12<01:20, 3.85s/it] {'loss': 1.2467, 'grad_norm': 0.001185593974038838, 'learning_rate': 0.0003849812381785328, 'epoch': 0.96} + 96%|█████████▌| 499/520 [31:12<01:20, 3.85s/it] 96%|█████████▌| 500/520 [31:16<01:16, 3.85s/it] {'loss': 1.2809, 'grad_norm': 0.0014661310628399796, 'learning_rate': 0.0003492357029922571, 'epoch': 0.96} + 96%|█████████▌| 500/520 [31:16<01:16, 3.85s/it] 96%|█████████▋| 501/520 [31:20<01:13, 3.85s/it] {'loss': 1.1492, 'grad_norm': 0.0012975519544514715, 'learning_rate': 0.00031522503471571704, 'epoch': 0.96} + 96%|█████████▋| 501/520 [31:20<01:13, 3.85s/it] 97%|█████████▋| 502/520 [31:24<01:09, 3.85s/it] {'loss': 1.1925, 'grad_norm': 0.0012337279438533026, 'learning_rate': 0.0002829505548040828, 'epoch': 0.97} + 97%|█████████▋| 502/520 [31:24<01:09, 3.85s/it] 97%|█████████▋| 503/520 [31:28<01:05, 3.85s/it] {'loss': 1.1365, 'grad_norm': 0.001221946291006144, 'learning_rate': 0.00025241351725441063, 'epoch': 0.97} + 97%|█████████▋| 503/520 [31:28<01:05, 3.85s/it] 97%|█████████▋| 504/520 [31:31<01:01, 3.84s/it] {'loss': 1.1834, 'grad_norm': 0.0014627227729673972, 'learning_rate': 0.00022361510855693656, 'epoch': 0.97} + 97%|█████████▋| 504/520 [31:31<01:01, 3.84s/it] 97%|█████████▋| 505/520 [31:35<00:57, 3.84s/it] {'loss': 1.213, 'grad_norm': 0.0012838009339582205, 'learning_rate': 0.0001965564476489784, 'epoch': 0.97} + 97%|█████████▋| 505/520 [31:35<00:57, 3.84s/it] 97%|█████████▋| 506/520 [31:39<00:53, 3.82s/it] {'loss': 1.1486, 'grad_norm': 0.0013727743261301175, 'learning_rate': 0.00017123858587145046, 'epoch': 0.97} + 97%|█████████▋| 506/520 [31:39<00:53, 3.82s/it] 98%|█████████▊| 507/520 [31:43<00:49, 3.77s/it] {'loss': 1.2796, 'grad_norm': 0.00113273375263867, 'learning_rate': 0.0001476625069280213, 'epoch': 0.97} + 98%|█████████▊| 507/520 [31:43<00:49, 3.77s/it] 98%|█████████▊| 508/520 [31:46<00:44, 3.73s/it] {'loss': 1.2669, 'grad_norm': 0.0013237072178729527, 'learning_rate': 0.00012582912684689417, 'epoch': 0.98} + 98%|█████████▊| 508/520 [31:46<00:44, 3.73s/it] 98%|█████████▊| 509/520 [31:50<00:40, 3.70s/it] {'loss': 1.2363, 'grad_norm': 0.0012302693214559712, 'learning_rate': 0.00010573929394520064, 'epoch': 0.98} + 98%|█████████▊| 509/520 [31:50<00:40, 3.70s/it] 98%|█████████▊| 510/520 [31:53<00:36, 3.68s/it] {'loss': 1.1872, 'grad_norm': 0.0012929666087092065, 'learning_rate': 8.739378879606684e-05, 'epoch': 0.98} + 98%|█████████▊| 510/520 [31:53<00:36, 3.68s/it] 98%|█████████▊| 511/520 [31:57<00:33, 3.67s/it] {'loss': 1.147, 'grad_norm': 0.0012563313006731633, 'learning_rate': 7.07933241982528e-05, 'epoch': 0.98} + 98%|█████████▊| 511/520 [31:57<00:33, 3.67s/it] 98%|█████████▊| 512/520 [32:01<00:29, 3.66s/it] {'loss': 1.0428, 'grad_norm': 0.0012489917411830417, 'learning_rate': 5.59385451484945e-05, 'epoch': 0.98} + 98%|█████████▊| 512/520 [32:01<00:29, 3.66s/it] 99%|█████████▊| 513/520 [32:04<00:25, 3.66s/it] {'loss': 1.2371, 'grad_norm': 0.0014301287192650823, 'learning_rate': 4.283002881639908e-05, 'epoch': 0.99} + 99%|█████████▊| 513/520 [32:04<00:25, 3.66s/it] 99%|█████████▉| 514/520 [32:08<00:21, 3.65s/it] {'loss': 1.2047, 'grad_norm': 0.0011815963920917935, 'learning_rate': 3.1468284522062626e-05, 'epoch': 0.99} + 99%|█████████▉| 514/520 [32:08<00:21, 3.65s/it] 99%|█████████▉| 515/520 [32:12<00:18, 3.64s/it] {'loss': 1.2596, 'grad_norm': 0.0015155912389102086, 'learning_rate': 2.1853753716256084e-05, 'epoch': 0.99} + 99%|█████████▉| 515/520 [32:12<00:18, 3.64s/it] 99%|█████████▉| 516/520 [32:15<00:14, 3.67s/it] {'loss': 1.173, 'grad_norm': 0.0012902357451675536, 'learning_rate': 1.3986809963268954e-05, 'epoch': 0.99} + 99%|█████████▉| 516/520 [32:15<00:14, 3.67s/it] 99%|█████████▉| 517/520 [32:19<00:10, 3.64s/it] {'loss': 1.1777, 'grad_norm': 0.0011931902428065427, 'learning_rate': 7.867758926410895e-06, 'epoch': 0.99} + 99%|█████████▉| 517/520 [32:19<00:10, 3.64s/it] 100%|█████████▉| 518/520 [32:23<00:07, 3.62s/it] {'loss': 1.1753, 'grad_norm': 0.0013463898866574087, 'learning_rate': 3.4968383561312375e-06, 'epoch': 1.0} + 100%|█████████▉| 518/520 [32:23<00:07, 3.62s/it] 100%|█████████▉| 519/520 [32:26<00:03, 3.61s/it] {'loss': 1.1552, 'grad_norm': 0.0012433796406840514, 'learning_rate': 8.742180807813637e-07, 'epoch': 1.0} + 100%|█████████▉| 519/520 [32:26<00:03, 3.61s/it] 100%|██████████| 520/520 [32:31<00:00, 3.85s/it] {'loss': 1.1426, 'grad_norm': 0.001099004755880345, 'learning_rate': 0.0, 'epoch': 1.0} + 100%|██████████| 520/520 [32:31<00:00, 3.85s/it] {'train_runtime': 1951.315, 'train_samples_per_second': 34.094, 'train_steps_per_second': 0.266, 'train_loss': 1.2290185458385028, 'epoch': 1.0} + 100%|██████████| 520/520 [32:31<00:00, 3.85s/it] 100%|██████████| 520/520 [32:31<00:00, 3.75s/it] +[2025-10-12 15:03:07,602] [INFO] [launch.py:348:main] Process 782089 exits successfully. +[2025-10-12 15:03:08,604] [INFO] [launch.py:348:main] Process 782093 exits successfully. +[2025-10-12 15:03:08,604] [INFO] [launch.py:348:main] Process 782090 exits successfully. +[2025-10-12 15:03:08,604] [INFO] [launch.py:348:main] Process 782094 exits successfully. +[2025-10-12 15:03:08,605] [INFO] [launch.py:348:main] Process 782088 exits successfully. +[2025-10-12 15:03:08,605] [INFO] [launch.py:348:main] Process 782091 exits successfully. +[2025-10-12 15:03:08,605] [INFO] [launch.py:348:main] Process 782092 exits successfully. +[2025-10-12 15:03:12,610] [INFO] [launch.py:348:main] Process 782087 exits successfully. +==== EXPERIMENT COMPLETED: qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_9e-2_connector-3.0_0.5_9e-2_ablation ==== +Log File: qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.5_9e-2_connector-3.0_0.5_9e-2_ablation_20251012_142902.log +Timestamp: 2025-10-12 15:03:15 +===================================== diff --git a/logs_oct11/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.7_2e-1_connector-3.0_0.7_2e-1_ablation_20251012_133702.log b/logs_oct11/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.7_2e-1_connector-3.0_0.7_2e-1_ablation_20251012_133702.log new file mode 100644 index 0000000000000000000000000000000000000000..3c6774f883e3dbe84d4581ace2e68cb01b2b147b --- /dev/null +++ b/logs_oct11/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.7_2e-1_connector-3.0_0.7_2e-1_ablation_20251012_133702.log @@ -0,0 +1,2312 @@ +==== STARTING EXPERIMENT: qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.7_2e-1_connector-3.0_0.7_2e-1_ablation ==== +Log File: qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.7_2e-1_connector-3.0_0.7_2e-1_ablation_20251012_133702.log +Timestamp: 2025-10-12 13:37:02 +===================================== +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 13:37:05,513] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 13:37:08,344] [WARNING] [runner.py:202:fetch_hostfile] Unable to find hostfile, will proceed with training with local resources only. +[2025-10-12 13:37:08,346] [INFO] [runner.py:568:main] cmd = /opt/conda/envs/tinyllava/bin/python3.10 -u -m deepspeed.launcher.launch --world_info=eyJsb2NhbGhvc3QiOiBbMCwgMSwgMiwgMywgNCwgNSwgNiwgN119 --master_addr=127.0.0.1 --master_port=29501 --enable_each_rank_log=None tinyllava/train/train.py --deepspeed ./scripts/zero3.json --data_path /s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json --image_folder /s3-code/ywang29/datasets/tinyllava --is_multimodal True --conv_version qwen2_base --model_name_or_path Qwen/Qwen2.5-0.5B --vision_tower google/siglip-so400m-patch14-384 --vision_tower2 --connector_type mlp2x_gelu --mm_vision_select_layer -2 --image_aspect_ratio square --attn_implementation flash_attention_2 --bf16 True --training_recipe common --tune_type_llm full --tune_type_vision_tower frozen --tune_vision_tower_from_layer 0 --tune_type_connector full --group_by_modality_length True --pretrained_model_path /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain --output_dir /nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.7_2e-1_connector-3.0_0.7_2e-1_ablation --num_train_epochs 1 --per_device_train_batch_size 4 --per_device_eval_batch_size 4 --gradient_accumulation_steps 4 --evaluation_strategy no --learning_rate 2e-1 --weight_decay 0. --warmup_ratio 0.03 --lr_scheduler_type cosine --logging_steps 1 --tf32 False --model_max_length 2048 --gradient_checkpointing True --dataloader_num_workers 8 --lazy_preprocess True --report_to tensorboard --tokenizer_use_fast False --run_name tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune --subnet_mode_text both --subnet_type_text None --mask_type_text soft --init_mean_text 3.0 --temperature_attn_text 0.7 --temperature_mlp_text 0.7 --backward_type_text normal --masked_layers_text all --subnet_mode_vision both --subnet_type_vision None --mask_type_vision soft --init_mean_vision 3.0 --temperature_attn_vision 0.7 --temperature_mlp_vision 0.7 --backward_type_vision normal --masked_layers_vision all --subnet_type_connector global --mask_type_connector soft --init_mean_connector 3.0 --temperature_connector 0.7 --backward_type_connector normal --mm_projector_lr 2e-1 --seed 42 --mask_model llm-connector --save_strategy steps --save_steps 50000 --save_total_limit 1 --train_data_ratio 0.1 +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 13:37:10,927] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 13:37:11,940] [INFO] [launch.py:138:main] 0 NCCL_VERSION=2.21.5 +[2025-10-12 13:37:11,941] [INFO] [launch.py:138:main] 0 NCCL_SOCKET_IFNAME=eth +[2025-10-12 13:37:11,941] [INFO] [launch.py:145:main] WORLD INFO DICT: {'localhost': [0, 1, 2, 3, 4, 5, 6, 7]} +[2025-10-12 13:37:11,941] [INFO] [launch.py:151:main] nnodes=1, num_local_procs=8, node_rank=0 +[2025-10-12 13:37:11,941] [INFO] [launch.py:162:main] global_rank_mapping=defaultdict(, {'localhost': [0, 1, 2, 3, 4, 5, 6, 7]}) +[2025-10-12 13:37:11,941] [INFO] [launch.py:163:main] dist_world_size=8 +[2025-10-12 13:37:11,941] [INFO] [launch.py:165:main] Setting CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 +[2025-10-12 13:37:11,943] [INFO] [launch.py:253:main] process 379798 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=0', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.7_2e-1_connector-3.0_0.7_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.7', '--temperature_mlp_text', '0.7', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.7', '--temperature_mlp_vision', '0.7', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.7', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 13:37:11,945] [INFO] [launch.py:253:main] process 379799 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=1', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.7_2e-1_connector-3.0_0.7_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.7', '--temperature_mlp_text', '0.7', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.7', '--temperature_mlp_vision', '0.7', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.7', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 13:37:11,948] [INFO] [launch.py:253:main] process 379800 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=2', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.7_2e-1_connector-3.0_0.7_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.7', '--temperature_mlp_text', '0.7', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.7', '--temperature_mlp_vision', '0.7', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.7', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 13:37:11,950] [INFO] [launch.py:253:main] process 379801 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=3', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.7_2e-1_connector-3.0_0.7_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.7', '--temperature_mlp_text', '0.7', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.7', '--temperature_mlp_vision', '0.7', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.7', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 13:37:11,952] [INFO] [launch.py:253:main] process 379802 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=4', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.7_2e-1_connector-3.0_0.7_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.7', '--temperature_mlp_text', '0.7', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.7', '--temperature_mlp_vision', '0.7', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.7', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 13:37:11,954] [INFO] [launch.py:253:main] process 379803 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=5', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.7_2e-1_connector-3.0_0.7_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.7', '--temperature_mlp_text', '0.7', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.7', '--temperature_mlp_vision', '0.7', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.7', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 13:37:11,956] [INFO] [launch.py:253:main] process 379804 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=6', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.7_2e-1_connector-3.0_0.7_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.7', '--temperature_mlp_text', '0.7', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.7', '--temperature_mlp_vision', '0.7', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.7', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 13:37:11,959] [INFO] [launch.py:253:main] process 379805 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=7', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.7_2e-1_connector-3.0_0.7_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.7', '--temperature_mlp_text', '0.7', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.7', '--temperature_mlp_vision', '0.7', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.7', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 13:37:18,589] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 13:37:18,772] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 13:37:18,772] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 13:37:18,775] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 13:37:18,775] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 13:37:18,776] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 13:37:18,779] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 13:37:18,813] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 13:37:18,991] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 13:37:19,173] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 13:37:19,173] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 13:37:19,174] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 13:37:19,175] [INFO] [comm.py:668:init_distributed] Initializing TorchBackend in DeepSpeed with backend nccl +[2025-10-12 13:37:19,175] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 13:37:19,177] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 13:37:19,179] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 13:37:19,215] [INFO] [comm.py:637:init_distributed] cdb=None +Apply masks for the following modules: ['llm', 'connector'] +Apply masks for the following modules: ['llm', 'connector'] +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +{'llm': {'model_name_or_path': 'Qwen/Qwen2.5-0.5B', 'cache_dir': None, 'attn_implementation': 'flash_attention_2', 'subnet_mode': 'both', 'subnet_type': 'None', 'sparsity_attn': None, 'sparsity_mlp': None, 'threshold_attn': None, 'threshold_mlp': None, 'temperature_attn': 0.7, 'temperature_mlp': 0.7, 'masked_layers': 'all', 'mask_type': 'soft', 'backward_type': 'normal'}, 'vision_tower': {'model_name_or_path': 'google/siglip-so400m-patch14-384'}, 'connector': {'connector_type': 'mlp2x_gelu', 'subnet_type': 'global', 'threshold': None, 'sparsity': None, 'temperature': 0.7, 'mask_type': 'soft', 'backward_type': 'normal'}} +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +TinyLlavaConfig { + "backward_type_connector": "normal", + "cache_dir": null, + "connector_type": "mlp2x_gelu", + "hidden_size": 896, + "ignore_index": -100, + "image_aspect_ratio": "square", + "image_token_index": -200, + "llm_model_name_or_path": "Qwen/Qwen2.5-0.5B", + "mask_model": [ + "llm", + "connector" + ], + "mask_type_connector": "soft", + "model_type": "tinyllava", + "num_queries": 128, + "num_resampler_layers": 3, + "pad_token": null, + "resampler_hidden_size": 768, + "sparsity_connector": null, + "subnet_type_connector": "global", + "temperature_connector": 0.7, + "text_config": { + "_name_or_path": "Qwen/Qwen2.5-0.5B", + "architectures": [ + "Qwen2ForCausalLM" + ], + "backward_type": "normal", + "bos_token_id": 151643, + "eos_token_id": 151643, + "hidden_size": 896, + "intermediate_size": 4864, + "mask_type": "soft", + "masked_layers": "all", + "max_position_embeddings": 32768, + "max_window_layers": 24, + "model_type": "qwen2", + "num_attention_heads": 14, + "num_hidden_layers": 24, + "num_key_value_heads": 2, + "rope_theta": 1000000.0, + "sliding_window": 32768, + "subnet_mode": "both", + "subnet_type": "None", + "temperature_attn": 0.7, + "temperature_mlp": 0.7, + "tie_word_embeddings": true, + "torch_dtype": "bfloat16", + "use_mrope": false, + "use_sliding_window": false, + "vocab_size": 151936 + }, + "threshold_connector": null, + "tokenizer_model_max_length": 2048, + "tokenizer_name_or_path": "Qwen/Qwen2.5-0.5B", + "tokenizer_padding_side": "right", + "tokenizer_use_fast": false, + "transformers_version": "4.40.1", + "tune_type_connector": "frozen", + "tune_type_llm": "frozen", + "tune_type_vision_tower": "frozen", + "tune_vision_tower_from_layer": -1, + "use_cache": false, + "vision_config": { + "hidden_act": "gelu_pytorch_tanh", + "hidden_size": 1152, + "image_size": 384, + "intermediate_size": 4304, + "layer_norm_eps": 1e-06, + "model_name_or_path": "google/siglip-so400m-patch14-384", + "model_name_or_path2": "", + "model_type": "siglip_vision_model", + "num_attention_heads": 16, + "num_hidden_layers": 27, + "patch_size": 14 + }, + "vision_feature_layer": -2, + "vision_feature_select_strategy": "patch", + "vision_hidden_size": 1152, + "vision_model_name_or_path": "google/siglip-so400m-patch14-384", + "vision_model_name_or_path2": "", + "vocab_size": 151936 +} + +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test1-worker-0:379798:379798 [0] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:379798:379798 [0] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:379798:379798 [0] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:379798:379798 [0] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:379798:379798 [0] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test1-worker-0:379798:379798 [0] NCCL INFO cudaDriverVersion 12040 +NCCL version 2.21.5+cuda12.1 +ywang29-vrdb-test1-worker-0:379801:379801 [3] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test1-worker-0:379801:379801 [3] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:379801:379801 [3] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:379801:379801 [3] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:379801:379801 [3] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:379801:379801 [3] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test1-worker-0:379799:379799 [1] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test1-worker-0:379799:379799 [1] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:379799:379799 [1] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:379799:379799 [1] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:379799:379799 [1] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:379799:379799 [1] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test1-worker-0:379803:379803 [5] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test1-worker-0:379803:379803 [5] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:379803:379803 [5] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:379803:379803 [5] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:379803:379803 [5] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:379803:379803 [5] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test1-worker-0:379804:379804 [6] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test1-worker-0:379804:379804 [6] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:379804:379804 [6] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:379804:379804 [6] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:379804:379804 [6] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:379804:379804 [6] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test1-worker-0:379798:381368 [0] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:379798:381368 [0] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:379798:381368 [0] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:379798:381368 [0] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:379798:381368 [0] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:379798:381368 [0] NCCL INFO Using network Socket +ywang29-vrdb-test1-worker-0:379801:381369 [3] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:379801:381369 [3] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:379801:381369 [3] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:379801:381369 [3] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:379801:381369 [3] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:379801:381369 [3] NCCL INFO Using network Socket +ywang29-vrdb-test1-worker-0:379803:381371 [5] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:379803:381371 [5] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:379803:381371 [5] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:379803:381371 [5] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:379803:381371 [5] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:379803:381371 [5] NCCL INFO Using network Socket +ywang29-vrdb-test1-worker-0:379804:381372 [6] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:379804:381372 [6] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:379804:381372 [6] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:379804:381372 [6] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:379804:381372 [6] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:379804:381372 [6] NCCL INFO Using network Socket +ywang29-vrdb-test1-worker-0:379799:381370 [1] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:379799:381370 [1] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:379799:381370 [1] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:379799:381370 [1] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:379799:381370 [1] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:379799:381370 [1] NCCL INFO Using network Socket +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test1-worker-0:379805:379805 [7] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test1-worker-0:379805:379805 [7] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:379805:379805 [7] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:379805:379805 [7] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:379805:379805 [7] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:379805:379805 [7] NCCL INFO NET/Plugin: Using internal network plugin. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test1-worker-0:379802:379802 [4] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test1-worker-0:379802:379802 [4] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:379802:379802 [4] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:379802:379802 [4] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:379802:379802 [4] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:379802:379802 [4] NCCL INFO NET/Plugin: Using internal network plugin. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test1-worker-0:379800:379800 [2] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test1-worker-0:379800:379800 [2] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:379800:379800 [2] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:379800:379800 [2] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:379800:379800 [2] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:379800:379800 [2] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test1-worker-0:379805:381373 [7] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:379805:381373 [7] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:379805:381373 [7] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:379805:381373 [7] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:379805:381373 [7] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:379805:381373 [7] NCCL INFO Using network Socket +ywang29-vrdb-test1-worker-0:379802:381374 [4] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:379802:381374 [4] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:379802:381374 [4] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:379802:381374 [4] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:379802:381374 [4] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:379802:381374 [4] NCCL INFO Using network Socket +ywang29-vrdb-test1-worker-0:379800:381376 [2] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:379800:381376 [2] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:379800:381376 [2] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:379800:381376 [2] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:379800:381376 [2] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:379800:381376 [2] NCCL INFO Using network Socket +ywang29-vrdb-test1-worker-0:379800:381376 [2] NCCL INFO ncclCommInitRank comm 0x55632dbd85f0 rank 2 nranks 8 cudaDev 2 nvmlDev 2 busId 201c0 commId 0x1c6b6315416d1d4e - Init START +ywang29-vrdb-test1-worker-0:379805:381373 [7] NCCL INFO ncclCommInitRank comm 0x557ebfa40200 rank 7 nranks 8 cudaDev 7 nvmlDev 7 busId a01d0 commId 0x1c6b6315416d1d4e - Init START +ywang29-vrdb-test1-worker-0:379799:381370 [1] NCCL INFO ncclCommInitRank comm 0x5568e5c40ec0 rank 1 nranks 8 cudaDev 1 nvmlDev 1 busId 101d0 commId 0x1c6b6315416d1d4e - Init START +ywang29-vrdb-test1-worker-0:379798:381368 [0] NCCL INFO ncclCommInitRank comm 0x564dbaaaee80 rank 0 nranks 8 cudaDev 0 nvmlDev 0 busId 101c0 commId 0x1c6b6315416d1d4e - Init START +ywang29-vrdb-test1-worker-0:379804:381372 [6] NCCL INFO ncclCommInitRank comm 0x564dae08ab40 rank 6 nranks 8 cudaDev 6 nvmlDev 6 busId a01c0 commId 0x1c6b6315416d1d4e - Init START +ywang29-vrdb-test1-worker-0:379802:381374 [4] NCCL INFO ncclCommInitRank comm 0x562df8b72bb0 rank 4 nranks 8 cudaDev 4 nvmlDev 4 busId 901c0 commId 0x1c6b6315416d1d4e - Init START +ywang29-vrdb-test1-worker-0:379803:381371 [5] NCCL INFO ncclCommInitRank comm 0x563ba9c11900 rank 5 nranks 8 cudaDev 5 nvmlDev 5 busId 901d0 commId 0x1c6b6315416d1d4e - Init START +ywang29-vrdb-test1-worker-0:379801:381369 [3] NCCL INFO ncclCommInitRank comm 0x5557321fb5b0 rank 3 nranks 8 cudaDev 3 nvmlDev 3 busId 201d0 commId 0x1c6b6315416d1d4e - Init START +ywang29-vrdb-test1-worker-0:379805:381373 [7] NCCL INFO Setting affinity for GPU 7 to ffffff00,0000ffff,ff000000 +ywang29-vrdb-test1-worker-0:379805:381373 [7] NCCL INFO NVLS multicast support is not available on dev 7 +ywang29-vrdb-test1-worker-0:379798:381368 [0] NCCL INFO Setting affinity for GPU 0 to ff,ffff0000,00ffffff +ywang29-vrdb-test1-worker-0:379798:381368 [0] NCCL INFO NVLS multicast support is not available on dev 0 +ywang29-vrdb-test1-worker-0:379799:381370 [1] NCCL INFO Setting affinity for GPU 1 to ff,ffff0000,00ffffff +ywang29-vrdb-test1-worker-0:379799:381370 [1] NCCL INFO NVLS multicast support is not available on dev 1 +ywang29-vrdb-test1-worker-0:379800:381376 [2] NCCL INFO Setting affinity for GPU 2 to ff,ffff0000,00ffffff +ywang29-vrdb-test1-worker-0:379800:381376 [2] NCCL INFO NVLS multicast support is not available on dev 2 +ywang29-vrdb-test1-worker-0:379802:381374 [4] NCCL INFO Setting affinity for GPU 4 to ffffff00,0000ffff,ff000000 +ywang29-vrdb-test1-worker-0:379802:381374 [4] NCCL INFO NVLS multicast support is not available on dev 4 +ywang29-vrdb-test1-worker-0:379801:381369 [3] NCCL INFO Setting affinity for GPU 3 to ff,ffff0000,00ffffff +ywang29-vrdb-test1-worker-0:379801:381369 [3] NCCL INFO NVLS multicast support is not available on dev 3 +ywang29-vrdb-test1-worker-0:379803:381371 [5] NCCL INFO Setting affinity for GPU 5 to ffffff00,0000ffff,ff000000 +ywang29-vrdb-test1-worker-0:379803:381371 [5] NCCL INFO NVLS multicast support is not available on dev 5 +ywang29-vrdb-test1-worker-0:379804:381372 [6] NCCL INFO Setting affinity for GPU 6 to ffffff00,0000ffff,ff000000 +ywang29-vrdb-test1-worker-0:379804:381372 [6] NCCL INFO NVLS multicast support is not available on dev 6 +ywang29-vrdb-test1-worker-0:379800:381376 [2] NCCL INFO comm 0x55632dbd85f0 rank 2 nRanks 8 nNodes 1 localRanks 8 localRank 2 MNNVL 0 +ywang29-vrdb-test1-worker-0:379799:381370 [1] NCCL INFO comm 0x5568e5c40ec0 rank 1 nRanks 8 nNodes 1 localRanks 8 localRank 1 MNNVL 0 +ywang29-vrdb-test1-worker-0:379804:381372 [6] NCCL INFO comm 0x564dae08ab40 rank 6 nRanks 8 nNodes 1 localRanks 8 localRank 6 MNNVL 0 +ywang29-vrdb-test1-worker-0:379798:381368 [0] NCCL INFO comm 0x564dbaaaee80 rank 0 nRanks 8 nNodes 1 localRanks 8 localRank 0 MNNVL 0 +ywang29-vrdb-test1-worker-0:379803:381371 [5] NCCL INFO comm 0x563ba9c11900 rank 5 nRanks 8 nNodes 1 localRanks 8 localRank 5 MNNVL 0 +ywang29-vrdb-test1-worker-0:379805:381373 [7] NCCL INFO comm 0x557ebfa40200 rank 7 nRanks 8 nNodes 1 localRanks 8 localRank 7 MNNVL 0 +ywang29-vrdb-test1-worker-0:379801:381369 [3] NCCL INFO comm 0x5557321fb5b0 rank 3 nRanks 8 nNodes 1 localRanks 8 localRank 3 MNNVL 0 +ywang29-vrdb-test1-worker-0:379802:381374 [4] NCCL INFO comm 0x562df8b72bb0 rank 4 nRanks 8 nNodes 1 localRanks 8 localRank 4 MNNVL 0 +ywang29-vrdb-test1-worker-0:379798:381368 [0] NCCL INFO Channel 00/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:379798:381368 [0] NCCL INFO Channel 01/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:379798:381368 [0] NCCL INFO Channel 02/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:379798:381368 [0] NCCL INFO Channel 03/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:379798:381368 [0] NCCL INFO Channel 04/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:379798:381368 [0] NCCL INFO Channel 05/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:379804:381372 [6] NCCL INFO Trees [0] 7/-1/-1->6->5 [1] 7/-1/-1->6->5 [2] 7/-1/-1->6->5 [3] 7/-1/-1->6->5 [4] 7/-1/-1->6->5 [5] 7/-1/-1->6->5 [6] 7/-1/-1->6->5 [7] 7/-1/-1->6->5 [8] 7/-1/-1->6->5 [9] 7/-1/-1->6->5 [10] 7/-1/-1->6->5 [11] 7/-1/-1->6->5 [12] 7/-1/-1->6->5 [13] 7/-1/-1->6->5 [14] 7/-1/-1->6->5 [15] 7/-1/-1->6->5 [16] 7/-1/-1->6->5 [17] 7/-1/-1->6->5 [18] 7/-1/-1->6->5 [19] 7/-1/-1->6->5 [20] 7/-1/-1->6->5 [21] 7/-1/-1->6->5 [22] 7/-1/-1->6->5 [23] 7/-1/-1->6->5 +ywang29-vrdb-test1-worker-0:379798:381368 [0] NCCL INFO Channel 06/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:379803:381371 [5] NCCL INFO Trees [0] 6/-1/-1->5->4 [1] 6/-1/-1->5->4 [2] 6/-1/-1->5->4 [3] 6/-1/-1->5->4 [4] 6/-1/-1->5->4 [5] 6/-1/-1->5->4 [6] 6/-1/-1->5->4 [7] 6/-1/-1->5->4 [8] 6/-1/-1->5->4 [9] 6/-1/-1->5->4 [10] 6/-1/-1->5->4 [11] 6/-1/-1->5->4 [12] 6/-1/-1->5->4 [13] 6/-1/-1->5->4 [14] 6/-1/-1->5->4 [15] 6/-1/-1->5->4 [16] 6/-1/-1->5->4 [17] 6/-1/-1->5->4 [18] 6/-1/-1->5->4 [19] 6/-1/-1->5->4 [20] 6/-1/-1->5->4 [21] 6/-1/-1->5->4 [22] 6/-1/-1->5->4 [23] 6/-1/-1->5->4 +ywang29-vrdb-test1-worker-0:379798:381368 [0] NCCL INFO Channel 07/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:379800:381376 [2] NCCL INFO Trees [0] 3/-1/-1->2->1 [1] 3/-1/-1->2->1 [2] 3/-1/-1->2->1 [3] 3/-1/-1->2->1 [4] 3/-1/-1->2->1 [5] 3/-1/-1->2->1 [6] 3/-1/-1->2->1 [7] 3/-1/-1->2->1 [8] 3/-1/-1->2->1 [9] 3/-1/-1->2->1 [10] 3/-1/-1->2->1 [11] 3/-1/-1->2->1 [12] 3/-1/-1->2->1 [13] 3/-1/-1->2->1 [14] 3/-1/-1->2->1 [15] 3/-1/-1->2->1 [16] 3/-1/-1->2->1 [17] 3/-1/-1->2->1 [18] 3/-1/-1->2->1 [19] 3/-1/-1->2->1 [20] 3/-1/-1->2->1 [21] 3/-1/-1->2->1 [22] 3/-1/-1->2->1 [23] 3/-1/-1->2->1 +ywang29-vrdb-test1-worker-0:379804:381372 [6] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:379805:381373 [7] NCCL INFO Trees [0] -1/-1/-1->7->6 [1] -1/-1/-1->7->6 [2] -1/-1/-1->7->6 [3] -1/-1/-1->7->6 [4] -1/-1/-1->7->6 [5] -1/-1/-1->7->6 [6] -1/-1/-1->7->6 [7] -1/-1/-1->7->6 [8] -1/-1/-1->7->6 [9] -1/-1/-1->7->6 [10] -1/-1/-1->7->6 [11] -1/-1/-1->7->6 [12] -1/-1/-1->7->6 [13] -1/-1/-1->7->6 [14] -1/-1/-1->7->6 [15] -1/-1/-1->7->6 [16] -1/-1/-1->7->6 [17] -1/-1/-1->7->6 [18] -1/-1/-1->7->6 [19] -1/-1/-1->7->6 [20] -1/-1/-1->7->6 [21] -1/-1/-1->7->6 [22] -1/-1/-1->7->6 [23] -1/-1/-1->7->6 +ywang29-vrdb-test1-worker-0:379803:381371 [5] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:379798:381368 [0] NCCL INFO Channel 08/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:379799:381370 [1] NCCL INFO Trees [0] 2/-1/-1->1->0 [1] 2/-1/-1->1->0 [2] 2/-1/-1->1->0 [3] 2/-1/-1->1->0 [4] 2/-1/-1->1->0 [5] 2/-1/-1->1->0 [6] 2/-1/-1->1->0 [7] 2/-1/-1->1->0 [8] 2/-1/-1->1->0 [9] 2/-1/-1->1->0 [10] 2/-1/-1->1->0 [11] 2/-1/-1->1->0 [12] 2/-1/-1->1->0 [13] 2/-1/-1->1->0 [14] 2/-1/-1->1->0 [15] 2/-1/-1->1->0 [16] 2/-1/-1->1->0 [17] 2/-1/-1->1->0 [18] 2/-1/-1->1->0 [19] 2/-1/-1->1->0 [20] 2/-1/-1->1->0 [21] 2/-1/-1->1->0 [22] 2/-1/-1->1->0 [23] 2/-1/-1->1->0 +ywang29-vrdb-test1-worker-0:379801:381369 [3] NCCL INFO Trees [0] 4/-1/-1->3->2 [1] 4/-1/-1->3->2 [2] 4/-1/-1->3->2 [3] 4/-1/-1->3->2 [4] 4/-1/-1->3->2 [5] 4/-1/-1->3->2 [6] 4/-1/-1->3->2 [7] 4/-1/-1->3->2 [8] 4/-1/-1->3->2 [9] 4/-1/-1->3->2 [10] 4/-1/-1->3->2 [11] 4/-1/-1->3->2 [12] 4/-1/-1->3->2 [13] 4/-1/-1->3->2 [14] 4/-1/-1->3->2 [15] 4/-1/-1->3->2 [16] 4/-1/-1->3->2 [17] 4/-1/-1->3->2 [18] 4/-1/-1->3->2 [19] 4/-1/-1->3->2 [20] 4/-1/-1->3->2 [21] 4/-1/-1->3->2 [22] 4/-1/-1->3->2 [23] 4/-1/-1->3->2 +ywang29-vrdb-test1-worker-0:379798:381368 [0] NCCL INFO Channel 09/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:379800:381376 [2] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:379802:381374 [4] NCCL INFO Trees [0] 5/-1/-1->4->3 [1] 5/-1/-1->4->3 [2] 5/-1/-1->4->3 [3] 5/-1/-1->4->3 [4] 5/-1/-1->4->3 [5] 5/-1/-1->4->3 [6] 5/-1/-1->4->3 [7] 5/-1/-1->4->3 [8] 5/-1/-1->4->3 [9] 5/-1/-1->4->3 [10] 5/-1/-1->4->3 [11] 5/-1/-1->4->3 [12] 5/-1/-1->4->3 [13] 5/-1/-1->4->3 [14] 5/-1/-1->4->3 [15] 5/-1/-1->4->3 [16] 5/-1/-1->4->3 [17] 5/-1/-1->4->3 [18] 5/-1/-1->4->3 [19] 5/-1/-1->4->3 [20] 5/-1/-1->4->3 [21] 5/-1/-1->4->3 [22] 5/-1/-1->4->3 [23] 5/-1/-1->4->3 +ywang29-vrdb-test1-worker-0:379805:381373 [7] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:379798:381368 [0] NCCL INFO Channel 10/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:379799:381370 [1] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:379801:381369 [3] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:379798:381368 [0] NCCL INFO Channel 11/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:379802:381374 [4] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:379798:381368 [0] NCCL INFO Channel 12/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:379798:381368 [0] NCCL INFO Channel 13/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:379798:381368 [0] NCCL INFO Channel 14/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:379798:381368 [0] NCCL INFO Channel 15/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:379798:381368 [0] NCCL INFO Channel 16/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:379798:381368 [0] NCCL INFO Channel 17/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:379798:381368 [0] NCCL INFO Channel 18/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:379798:381368 [0] NCCL INFO Channel 19/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:379798:381368 [0] NCCL INFO Channel 20/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:379798:381368 [0] NCCL INFO Channel 21/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:379798:381368 [0] NCCL INFO Channel 22/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:379798:381368 [0] NCCL INFO Channel 23/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:379798:381368 [0] NCCL INFO Trees [0] 1/-1/-1->0->-1 [1] 1/-1/-1->0->-1 [2] 1/-1/-1->0->-1 [3] 1/-1/-1->0->-1 [4] 1/-1/-1->0->-1 [5] 1/-1/-1->0->-1 [6] 1/-1/-1->0->-1 [7] 1/-1/-1->0->-1 [8] 1/-1/-1->0->-1 [9] 1/-1/-1->0->-1 [10] 1/-1/-1->0->-1 [11] 1/-1/-1->0->-1 [12] 1/-1/-1->0->-1 [13] 1/-1/-1->0->-1 [14] 1/-1/-1->0->-1 [15] 1/-1/-1->0->-1 [16] 1/-1/-1->0->-1 [17] 1/-1/-1->0->-1 [18] 1/-1/-1->0->-1 [19] 1/-1/-1->0->-1 [20] 1/-1/-1->0->-1 [21] 1/-1/-1->0->-1 [22] 1/-1/-1->0->-1 [23] 1/-1/-1->0->-1 +ywang29-vrdb-test1-worker-0:379798:381368 [0] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:379800:381376 [2] NCCL INFO Channel 00/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:381371 [5] NCCL INFO Channel 00/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:381373 [7] NCCL INFO Channel 00/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379798:381368 [0] NCCL INFO Channel 00/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:381369 [3] NCCL INFO Channel 00/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:381372 [6] NCCL INFO Channel 00/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:381370 [1] NCCL INFO Channel 00/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:381374 [4] NCCL INFO Channel 00/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:381376 [2] NCCL INFO Channel 01/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:381371 [5] NCCL INFO Channel 01/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:381373 [7] NCCL INFO Channel 01/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379798:381368 [0] NCCL INFO Channel 01/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:381369 [3] NCCL INFO Channel 01/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:381372 [6] NCCL INFO Channel 01/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:381370 [1] NCCL INFO Channel 01/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:381376 [2] NCCL INFO Channel 02/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:381374 [4] NCCL INFO Channel 01/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:381371 [5] NCCL INFO Channel 02/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:381373 [7] NCCL INFO Channel 02/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379798:381368 [0] NCCL INFO Channel 02/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:381369 [3] NCCL INFO Channel 02/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:381372 [6] NCCL INFO Channel 02/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:381370 [1] NCCL INFO Channel 02/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:381376 [2] NCCL INFO Channel 03/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:381374 [4] NCCL INFO Channel 02/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:381371 [5] NCCL INFO Channel 03/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:381373 [7] NCCL INFO Channel 03/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379798:381368 [0] NCCL INFO Channel 03/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:381369 [3] NCCL INFO Channel 03/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:381372 [6] NCCL INFO Channel 03/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:381370 [1] NCCL INFO Channel 03/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:381376 [2] NCCL INFO Channel 04/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:381374 [4] NCCL INFO Channel 03/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:381371 [5] NCCL INFO Channel 04/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:381373 [7] NCCL INFO Channel 04/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379798:381368 [0] NCCL INFO Channel 04/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:381369 [3] NCCL INFO Channel 04/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:381372 [6] NCCL INFO Channel 04/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:381370 [1] NCCL INFO Channel 04/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:381376 [2] NCCL INFO Channel 05/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:381374 [4] NCCL INFO Channel 04/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:381371 [5] NCCL INFO Channel 05/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:381373 [7] NCCL INFO Channel 05/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379798:381368 [0] NCCL INFO Channel 05/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:381369 [3] NCCL INFO Channel 05/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:381370 [1] NCCL INFO Channel 05/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:381372 [6] NCCL INFO Channel 05/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:381376 [2] NCCL INFO Channel 06/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:381374 [4] NCCL INFO Channel 05/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:381371 [5] NCCL INFO Channel 06/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:381373 [7] NCCL INFO Channel 06/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379798:381368 [0] NCCL INFO Channel 06/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:381369 [3] NCCL INFO Channel 06/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:381370 [1] NCCL INFO Channel 06/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:381372 [6] NCCL INFO Channel 06/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:381374 [4] NCCL INFO Channel 06/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:381376 [2] NCCL INFO Channel 07/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:381371 [5] NCCL INFO Channel 07/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:381373 [7] NCCL INFO Channel 07/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379798:381368 [0] NCCL INFO Channel 07/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:381369 [3] NCCL INFO Channel 07/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:381370 [1] NCCL INFO Channel 07/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:381372 [6] NCCL INFO Channel 07/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:381374 [4] NCCL INFO Channel 07/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:381376 [2] NCCL INFO Channel 08/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:381371 [5] NCCL INFO Channel 08/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:381373 [7] NCCL INFO Channel 08/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379798:381368 [0] NCCL INFO Channel 08/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:381369 [3] NCCL INFO Channel 08/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:381370 [1] NCCL INFO Channel 08/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:381372 [6] NCCL INFO Channel 08/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:381374 [4] NCCL INFO Channel 08/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:381376 [2] NCCL INFO Channel 09/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:381371 [5] NCCL INFO Channel 09/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:381373 [7] NCCL INFO Channel 09/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379798:381368 [0] NCCL INFO Channel 09/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:381369 [3] NCCL INFO Channel 09/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:381370 [1] NCCL INFO Channel 09/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:381372 [6] NCCL INFO Channel 09/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:381374 [4] NCCL INFO Channel 09/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:381376 [2] NCCL INFO Channel 10/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:381371 [5] NCCL INFO Channel 10/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:381373 [7] NCCL INFO Channel 10/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379798:381368 [0] NCCL INFO Channel 10/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:381369 [3] NCCL INFO Channel 10/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:381370 [1] NCCL INFO Channel 10/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:381372 [6] NCCL INFO Channel 10/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:381374 [4] NCCL INFO Channel 10/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:381376 [2] NCCL INFO Channel 11/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:381371 [5] NCCL INFO Channel 11/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:381373 [7] NCCL INFO Channel 11/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379798:381368 [0] NCCL INFO Channel 11/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:381369 [3] NCCL INFO Channel 11/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:381370 [1] NCCL INFO Channel 11/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:381372 [6] NCCL INFO Channel 11/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:381374 [4] NCCL INFO Channel 11/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:381376 [2] NCCL INFO Channel 12/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:381371 [5] NCCL INFO Channel 12/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:381373 [7] NCCL INFO Channel 12/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379798:381368 [0] NCCL INFO Channel 12/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:381369 [3] NCCL INFO Channel 12/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:381370 [1] NCCL INFO Channel 12/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:381372 [6] NCCL INFO Channel 12/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:381374 [4] NCCL INFO Channel 12/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:381376 [2] NCCL INFO Channel 13/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:381371 [5] NCCL INFO Channel 13/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:381373 [7] NCCL INFO Channel 13/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379798:381368 [0] NCCL INFO Channel 13/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:381369 [3] NCCL INFO Channel 13/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:381370 [1] NCCL INFO Channel 13/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:381372 [6] NCCL INFO Channel 13/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:381374 [4] NCCL INFO Channel 13/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:381376 [2] NCCL INFO Channel 14/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:381371 [5] NCCL INFO Channel 14/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:381373 [7] NCCL INFO Channel 14/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379798:381368 [0] NCCL INFO Channel 14/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:381369 [3] NCCL INFO Channel 14/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:381370 [1] NCCL INFO Channel 14/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:381372 [6] NCCL INFO Channel 14/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:381374 [4] NCCL INFO Channel 14/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:381376 [2] NCCL INFO Channel 15/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:381371 [5] NCCL INFO Channel 15/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:381373 [7] NCCL INFO Channel 15/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379798:381368 [0] NCCL INFO Channel 15/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:381369 [3] NCCL INFO Channel 15/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:381370 [1] NCCL INFO Channel 15/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:381374 [4] NCCL INFO Channel 15/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:381372 [6] NCCL INFO Channel 15/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:381376 [2] NCCL INFO Channel 16/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:381371 [5] NCCL INFO Channel 16/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:381373 [7] NCCL INFO Channel 16/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379798:381368 [0] NCCL INFO Channel 16/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:381369 [3] NCCL INFO Channel 16/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:381370 [1] NCCL INFO Channel 16/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:381374 [4] NCCL INFO Channel 16/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:381372 [6] NCCL INFO Channel 16/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:381376 [2] NCCL INFO Channel 17/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:381371 [5] NCCL INFO Channel 17/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:381373 [7] NCCL INFO Channel 17/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379798:381368 [0] NCCL INFO Channel 17/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:381369 [3] NCCL INFO Channel 17/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:381370 [1] NCCL INFO Channel 17/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:381374 [4] NCCL INFO Channel 17/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:381372 [6] NCCL INFO Channel 17/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:381371 [5] NCCL INFO Channel 18/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:381376 [2] NCCL INFO Channel 18/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:381373 [7] NCCL INFO Channel 18/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379798:381368 [0] NCCL INFO Channel 18/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:381369 [3] NCCL INFO Channel 18/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:381370 [1] NCCL INFO Channel 18/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:381374 [4] NCCL INFO Channel 18/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:381372 [6] NCCL INFO Channel 18/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:381371 [5] NCCL INFO Channel 19/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:381376 [2] NCCL INFO Channel 19/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:381373 [7] NCCL INFO Channel 19/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379798:381368 [0] NCCL INFO Channel 19/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:381369 [3] NCCL INFO Channel 19/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:381370 [1] NCCL INFO Channel 19/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:381374 [4] NCCL INFO Channel 19/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:381372 [6] NCCL INFO Channel 19/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:381371 [5] NCCL INFO Channel 20/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:381376 [2] NCCL INFO Channel 20/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:381373 [7] NCCL INFO Channel 20/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379798:381368 [0] NCCL INFO Channel 20/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:381369 [3] NCCL INFO Channel 20/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:381370 [1] NCCL INFO Channel 20/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:381374 [4] NCCL INFO Channel 20/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:381372 [6] NCCL INFO Channel 20/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:381371 [5] NCCL INFO Channel 21/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:381376 [2] NCCL INFO Channel 21/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:381373 [7] NCCL INFO Channel 21/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379798:381368 [0] NCCL INFO Channel 21/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:381369 [3] NCCL INFO Channel 21/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:381370 [1] NCCL INFO Channel 21/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:381374 [4] NCCL INFO Channel 21/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:381372 [6] NCCL INFO Channel 21/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:381371 [5] NCCL INFO Channel 22/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:381376 [2] NCCL INFO Channel 22/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:381373 [7] NCCL INFO Channel 22/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379798:381368 [0] NCCL INFO Channel 22/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:381369 [3] NCCL INFO Channel 22/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:381370 [1] NCCL INFO Channel 22/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:381374 [4] NCCL INFO Channel 22/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:381372 [6] NCCL INFO Channel 22/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:381371 [5] NCCL INFO Channel 23/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:381376 [2] NCCL INFO Channel 23/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:381373 [7] NCCL INFO Channel 23/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379798:381368 [0] NCCL INFO Channel 23/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:381369 [3] NCCL INFO Channel 23/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:381370 [1] NCCL INFO Channel 23/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:381374 [4] NCCL INFO Channel 23/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:381372 [6] NCCL INFO Channel 23/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:381376 [2] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:379801:381369 [3] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:379800:381376 [2] NCCL INFO Channel 00/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:381370 [1] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:379798:381368 [0] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:379800:381376 [2] NCCL INFO Channel 01/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:381376 [2] NCCL INFO Channel 02/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:381376 [2] NCCL INFO Channel 03/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:381376 [2] NCCL INFO Channel 04/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:381376 [2] NCCL INFO Channel 05/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:381376 [2] NCCL INFO Channel 06/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:381376 [2] NCCL INFO Channel 07/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:381373 [7] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:379805:381373 [7] NCCL INFO Channel 00/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:381371 [5] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:379800:381376 [2] NCCL INFO Channel 08/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:381372 [6] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:379802:381374 [4] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:379805:381373 [7] NCCL INFO Channel 01/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:381376 [2] NCCL INFO Channel 09/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:381373 [7] NCCL INFO Channel 02/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:381376 [2] NCCL INFO Channel 10/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:381376 [2] NCCL INFO Channel 11/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:381373 [7] NCCL INFO Channel 03/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:381369 [3] NCCL INFO Channel 00/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:381369 [3] NCCL INFO Channel 01/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:381369 [3] NCCL INFO Channel 02/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:381376 [2] NCCL INFO Channel 12/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:381376 [2] NCCL INFO Channel 13/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:381376 [2] NCCL INFO Channel 14/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:381376 [2] NCCL INFO Channel 15/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:381376 [2] NCCL INFO Channel 16/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:381376 [2] NCCL INFO Channel 17/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:381376 [2] NCCL INFO Channel 18/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:381376 [2] NCCL INFO Channel 19/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:381376 [2] NCCL INFO Channel 20/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:381369 [3] NCCL INFO Channel 03/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:381376 [2] NCCL INFO Channel 21/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:381373 [7] NCCL INFO Channel 04/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:381369 [3] NCCL INFO Channel 04/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:381376 [2] NCCL INFO Channel 22/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:381373 [7] NCCL INFO Channel 05/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:381369 [3] NCCL INFO Channel 05/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:381376 [2] NCCL INFO Channel 23/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:381373 [7] NCCL INFO Channel 06/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:381369 [3] NCCL INFO Channel 06/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:381373 [7] NCCL INFO Channel 07/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:381369 [3] NCCL INFO Channel 07/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:381373 [7] NCCL INFO Channel 08/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:381369 [3] NCCL INFO Channel 08/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:381372 [6] NCCL INFO Channel 00/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:381373 [7] NCCL INFO Channel 09/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:381369 [3] NCCL INFO Channel 09/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:381372 [6] NCCL INFO Channel 01/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:381373 [7] NCCL INFO Channel 10/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:381374 [4] NCCL INFO Channel 00/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:381369 [3] NCCL INFO Channel 10/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:381372 [6] NCCL INFO Channel 02/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:381373 [7] NCCL INFO Channel 11/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:381374 [4] NCCL INFO Channel 01/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:381369 [3] NCCL INFO Channel 11/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:381372 [6] NCCL INFO Channel 03/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:381373 [7] NCCL INFO Channel 12/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:381374 [4] NCCL INFO Channel 02/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:381369 [3] NCCL INFO Channel 12/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:381372 [6] NCCL INFO Channel 04/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:381373 [7] NCCL INFO Channel 13/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:381374 [4] NCCL INFO Channel 03/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:381369 [3] NCCL INFO Channel 13/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:381372 [6] NCCL INFO Channel 05/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:381370 [1] NCCL INFO Channel 00/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:381373 [7] NCCL INFO Channel 14/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:381374 [4] NCCL INFO Channel 04/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:381369 [3] NCCL INFO Channel 14/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:381372 [6] NCCL INFO Channel 06/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:381370 [1] NCCL INFO Channel 01/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:381373 [7] NCCL INFO Channel 15/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:381374 [4] NCCL INFO Channel 05/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:381369 [3] NCCL INFO Channel 15/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:381372 [6] NCCL INFO Channel 07/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:381374 [4] NCCL INFO Channel 06/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:381370 [1] NCCL INFO Channel 02/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:381373 [7] NCCL INFO Channel 16/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:381369 [3] NCCL INFO Channel 16/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:381372 [6] NCCL INFO Channel 08/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:381371 [5] NCCL INFO Channel 00/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:381374 [4] NCCL INFO Channel 07/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:381370 [1] NCCL INFO Channel 03/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:381373 [7] NCCL INFO Channel 17/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:381369 [3] NCCL INFO Channel 17/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:381372 [6] NCCL INFO Channel 09/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:381371 [5] NCCL INFO Channel 01/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:381374 [4] NCCL INFO Channel 08/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:381370 [1] NCCL INFO Channel 04/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:381373 [7] NCCL INFO Channel 18/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:381369 [3] NCCL INFO Channel 18/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:381372 [6] NCCL INFO Channel 10/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:381371 [5] NCCL INFO Channel 02/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:381374 [4] NCCL INFO Channel 09/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:381370 [1] NCCL INFO Channel 05/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:381373 [7] NCCL INFO Channel 19/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:381369 [3] NCCL INFO Channel 19/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:381372 [6] NCCL INFO Channel 11/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:381371 [5] NCCL INFO Channel 03/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:381374 [4] NCCL INFO Channel 10/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:381370 [1] NCCL INFO Channel 06/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:381373 [7] NCCL INFO Channel 20/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:381369 [3] NCCL INFO Channel 20/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:381372 [6] NCCL INFO Channel 12/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:381371 [5] NCCL INFO Channel 04/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:381374 [4] NCCL INFO Channel 11/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:381370 [1] NCCL INFO Channel 07/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:381373 [7] NCCL INFO Channel 21/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:381371 [5] NCCL INFO Channel 05/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:381374 [4] NCCL INFO Channel 12/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:381369 [3] NCCL INFO Channel 21/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:381372 [6] NCCL INFO Channel 13/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:381373 [7] NCCL INFO Channel 22/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:381371 [5] NCCL INFO Channel 06/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:381374 [4] NCCL INFO Channel 13/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:381369 [3] NCCL INFO Channel 22/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:381372 [6] NCCL INFO Channel 14/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:381373 [7] NCCL INFO Channel 23/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:381371 [5] NCCL INFO Channel 07/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:381374 [4] NCCL INFO Channel 14/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:381369 [3] NCCL INFO Channel 23/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:381372 [6] NCCL INFO Channel 15/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:381371 [5] NCCL INFO Channel 08/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:381374 [4] NCCL INFO Channel 15/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:381372 [6] NCCL INFO Channel 16/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:381371 [5] NCCL INFO Channel 09/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:381374 [4] NCCL INFO Channel 16/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:381372 [6] NCCL INFO Channel 17/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:381370 [1] NCCL INFO Channel 08/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:381371 [5] NCCL INFO Channel 10/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:381374 [4] NCCL INFO Channel 17/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:381372 [6] NCCL INFO Channel 18/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:381371 [5] NCCL INFO Channel 11/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:381374 [4] NCCL INFO Channel 18/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:381370 [1] NCCL INFO Channel 09/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:381372 [6] NCCL INFO Channel 19/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:381371 [5] NCCL INFO Channel 12/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:381374 [4] NCCL INFO Channel 19/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:381370 [1] NCCL INFO Channel 10/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:381372 [6] NCCL INFO Channel 20/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:381371 [5] NCCL INFO Channel 13/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:381374 [4] NCCL INFO Channel 20/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:381370 [1] NCCL INFO Channel 11/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:381372 [6] NCCL INFO Channel 21/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:381371 [5] NCCL INFO Channel 14/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:381374 [4] NCCL INFO Channel 21/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:381370 [1] NCCL INFO Channel 12/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:381372 [6] NCCL INFO Channel 22/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:381371 [5] NCCL INFO Channel 15/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:381374 [4] NCCL INFO Channel 22/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:381372 [6] NCCL INFO Channel 23/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:381370 [1] NCCL INFO Channel 13/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:381371 [5] NCCL INFO Channel 16/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:381374 [4] NCCL INFO Channel 23/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:381371 [5] NCCL INFO Channel 17/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:381370 [1] NCCL INFO Channel 14/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:381371 [5] NCCL INFO Channel 18/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:381370 [1] NCCL INFO Channel 15/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:381371 [5] NCCL INFO Channel 19/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:381370 [1] NCCL INFO Channel 16/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:381371 [5] NCCL INFO Channel 20/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:381371 [5] NCCL INFO Channel 21/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:381370 [1] NCCL INFO Channel 17/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:381371 [5] NCCL INFO Channel 22/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:381370 [1] NCCL INFO Channel 18/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:381371 [5] NCCL INFO Channel 23/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:381370 [1] NCCL INFO Channel 19/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:381370 [1] NCCL INFO Channel 20/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:381370 [1] NCCL INFO Channel 21/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:381370 [1] NCCL INFO Channel 22/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:381370 [1] NCCL INFO Channel 23/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379798:381368 [0] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:379798:381368 [0] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:379798:381368 [0] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:379800:381376 [2] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:379800:381376 [2] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:379800:381376 [2] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:379801:381369 [3] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:379801:381369 [3] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:379801:381369 [3] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:379799:381370 [1] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:379799:381370 [1] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:379799:381370 [1] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:379805:381373 [7] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:379805:381373 [7] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:379805:381373 [7] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:379802:381374 [4] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:379802:381374 [4] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:379802:381374 [4] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:379804:381372 [6] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:379803:381371 [5] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:379804:381372 [6] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:379804:381372 [6] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:379803:381371 [5] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:379803:381371 [5] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:379804:381372 [6] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:379803:381371 [5] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:379805:381373 [7] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:379802:381374 [4] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:379803:381371 [5] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:379805:381373 [7] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:379802:381374 [4] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:379803:381371 [5] NCCL INFO ncclCommInitRank comm 0x563ba9c11900 rank 5 nranks 8 cudaDev 5 nvmlDev 5 busId 901d0 commId 0x1c6b6315416d1d4e - Init COMPLETE +ywang29-vrdb-test1-worker-0:379805:381373 [7] NCCL INFO ncclCommInitRank comm 0x557ebfa40200 rank 7 nranks 8 cudaDev 7 nvmlDev 7 busId a01d0 commId 0x1c6b6315416d1d4e - Init COMPLETE +ywang29-vrdb-test1-worker-0:379804:381372 [6] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:379802:381374 [4] NCCL INFO ncclCommInitRank comm 0x562df8b72bb0 rank 4 nranks 8 cudaDev 4 nvmlDev 4 busId 901c0 commId 0x1c6b6315416d1d4e - Init COMPLETE +ywang29-vrdb-test1-worker-0:379804:381372 [6] NCCL INFO ncclCommInitRank comm 0x564dae08ab40 rank 6 nranks 8 cudaDev 6 nvmlDev 6 busId a01c0 commId 0x1c6b6315416d1d4e - Init COMPLETE +ywang29-vrdb-test1-worker-0:379800:381376 [2] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:379801:381369 [3] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:379799:381370 [1] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:379800:381376 [2] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:379801:381369 [3] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:379799:381370 [1] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:379800:381376 [2] NCCL INFO ncclCommInitRank comm 0x55632dbd85f0 rank 2 nranks 8 cudaDev 2 nvmlDev 2 busId 201c0 commId 0x1c6b6315416d1d4e - Init COMPLETE +ywang29-vrdb-test1-worker-0:379801:381369 [3] NCCL INFO ncclCommInitRank comm 0x5557321fb5b0 rank 3 nranks 8 cudaDev 3 nvmlDev 3 busId 201d0 commId 0x1c6b6315416d1d4e - Init COMPLETE +ywang29-vrdb-test1-worker-0:379798:381368 [0] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:379799:381370 [1] NCCL INFO ncclCommInitRank comm 0x5568e5c40ec0 rank 1 nranks 8 cudaDev 1 nvmlDev 1 busId 101d0 commId 0x1c6b6315416d1d4e - Init COMPLETE +ywang29-vrdb-test1-worker-0:379798:381368 [0] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:379798:381368 [0] NCCL INFO ncclCommInitRank comm 0x564dbaaaee80 rank 0 nranks 8 cudaDev 0 nvmlDev 0 busId 101c0 commId 0x1c6b6315416d1d4e - Init COMPLETE +[2025-10-12 13:38:05,279] [INFO] [partition_parameters.py:343:__exit__] finished initializing model - num_params = 459, num_elems = 0.99B +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +[2025-10-12 13:38:06,998] [INFO] [partition_parameters.py:343:__exit__] finished initializing model - num_params = 907, num_elems = 1.42B +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +TinyLlavaForConditionalGeneration( + (language_model): Qwen2ForCausalLM( + (model): Qwen2Model( + (embed_tokens): Embedding(151936, 896) + (layers): ModuleList( + (0-23): 24 x Qwen2DecoderLayer( + (self_attn): Qwen2FlashAttention2( + (q_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=True) + (k_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=128, bias=True) + (v_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=128, bias=True) + (o_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=False) + (rotary_emb): Qwen2RotaryEmbedding() + ) + (mlp): Qwen2MLP( + (gate_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=4864, bias=False) + (up_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=4864, bias=False) + (down_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=4864, out_features=896, bias=False) + (act_fn): SiLU() + ) + (input_layernorm): Qwen2RMSNorm() + (post_attention_layernorm): Qwen2RMSNorm() + ) + ) + (norm): Qwen2RMSNorm() + ) + (lm_head): Linear(in_features=896, out_features=151936, bias=False) + ) + (vision_tower): SIGLIPVisionTower( + (_vision_tower): SiglipVisionModel( + (vision_model): SiglipVisionTransformer( + (embeddings): SiglipVisionEmbeddings( + (patch_embedding): Conv2d(3, 1152, kernel_size=(14, 14), stride=(14, 14), padding=valid) + (position_embedding): Embedding(729, 1152) + ) + (encoder): SiglipEncoder( + (layers): ModuleList( + (0-26): 27 x SiglipEncoderLayer( + (self_attn): SiglipAttention( + (k_proj): Linear(in_features=1152, out_features=1152, bias=True) + (v_proj): Linear(in_features=1152, out_features=1152, bias=True) + (q_proj): Linear(in_features=1152, out_features=1152, bias=True) + (out_proj): Linear(in_features=1152, out_features=1152, bias=True) + ) + (layer_norm1): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + (layer_norm2): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + ) + ) + ) + (post_layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (head): SiglipMultiheadAttentionPoolingHead( + (attention): MultiheadAttention( + (out_proj): NonDynamicallyQuantizableLinear(in_features=1152, out_features=1152, bias=True) + ) + (layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + ) + ) + ) + ) + (connector): MLPConnector( + (_connector): Sequential( + (0): SupermaskLinearSparsity_SoftForward_Normal(in_features=1152, out_features=896, bias=True) + (1): GELU(approximate='none') + (2): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=True) + ) + ) +) +Pre-training init language_model.model.layers.0.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.0.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.0.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.0.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.0.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.0.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.0.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.mlp.down_proj.scores: Mean=3.000000 +Pre-training init connector._connector.0.scores: Mean=3.000005 +Pre-training init connector._connector.2.scores: Mean=2.999970 +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +2025-10-12 14:08:13,777 | INFO: Total Parameters: 1283756736, Total Trainable Parameters: 359661568 +2025-10-12 14:08:13,784 | INFO: Trainable Parameters: +language_model.model.layers.0.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.0.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.0.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.0.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.0.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.0.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.0.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.1.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.1.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.1.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.1.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.1.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.1.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.1.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.2.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.2.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.2.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.2.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.2.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.2.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.2.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.3.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.3.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.3.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.3.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.3.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.3.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.3.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.4.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.4.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.4.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.4.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.4.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.4.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.4.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.5.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.5.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.5.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.5.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.5.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.5.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.5.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.6.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.6.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.6.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.6.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.6.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.6.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.6.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.7.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.7.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.7.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.7.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.7.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.7.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.7.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.8.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.8.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.8.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.8.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.8.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.8.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.8.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.9.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.9.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.9.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.9.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.9.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.9.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.9.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.10.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.10.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.10.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.10.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.10.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.10.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.10.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.11.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.11.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.11.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.11.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.11.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.11.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.11.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.12.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.12.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.12.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.12.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.12.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.12.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.12.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.13.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.13.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.13.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.13.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.13.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.13.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.13.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.14.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.14.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.14.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.14.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.14.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.14.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.14.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.15.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.15.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.15.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.15.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.15.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.15.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.15.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.16.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.16.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.16.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.16.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.16.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.16.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.16.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.17.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.17.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.17.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.17.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.17.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.17.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.17.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.18.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.18.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.18.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.18.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.18.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.18.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.18.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.19.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.19.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.19.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.19.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.19.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.19.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.19.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.20.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.20.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.20.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.20.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.20.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.20.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.20.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.21.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.21.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.21.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.21.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.21.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.21.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.21.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.22.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.22.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.22.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.22.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.22.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.22.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.22.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.23.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.23.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.23.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.23.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.23.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.23.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.23.mlp.down_proj.scores: 4358144 parameters +connector._connector.0.scores: 1032192 parameters +connector._connector.2.scores: 802816 parameters +Parameter Offload: Total persistent parameters: 486464 in 403 params + 0%| | 0/520 [00:002->1 [1] 3/-1/-1->2->1 [2] 3/-1/-1->2->1 [3] 3/-1/-1->2->1 [4] 3/-1/-1->2->1 [5] 3/-1/-1->2->1 [6] 3/-1/-1->2->1 [7] 3/-1/-1->2->1 [8] 3/-1/-1->2->1 [9] 3/-1/-1->2->1 [10] 3/-1/-1->2->1 [11] 3/-1/-1->2->1 [12] 3/-1/-1->2->1 [13] 3/-1/-1->2->1 [14] 3/-1/-1->2->1 [15] 3/-1/-1->2->1 [16] 3/-1/-1->2->1 [17] 3/-1/-1->2->1 [18] 3/-1/-1->2->1 [19] 3/-1/-1->2->1 [20] 3/-1/-1->2->1 [21] 3/-1/-1->2->1 [22] 3/-1/-1->2->1 [23] 3/-1/-1->2->1 +ywang29-vrdb-test1-worker-0:379798:386772 [0] NCCL INFO Channel 04/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:379800:386773 [2] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:379798:386772 [0] NCCL INFO Channel 05/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:379798:386772 [0] NCCL INFO Channel 06/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:379804:386775 [6] NCCL INFO Trees [0] 7/-1/-1->6->5 [1] 7/-1/-1->6->5 [2] 7/-1/-1->6->5 [3] 7/-1/-1->6->5 [4] 7/-1/-1->6->5 [5] 7/-1/-1->6->5 [6] 7/-1/-1->6->5 [7] 7/-1/-1->6->5 [8] 7/-1/-1->6->5 [9] 7/-1/-1->6->5 [10] 7/-1/-1->6->5 [11] 7/-1/-1->6->5 [12] 7/-1/-1->6->5 [13] 7/-1/-1->6->5 [14] 7/-1/-1->6->5 [15] 7/-1/-1->6->5 [16] 7/-1/-1->6->5 [17] 7/-1/-1->6->5 [18] 7/-1/-1->6->5 [19] 7/-1/-1->6->5 [20] 7/-1/-1->6->5 [21] 7/-1/-1->6->5 [22] 7/-1/-1->6->5 [23] 7/-1/-1->6->5 +ywang29-vrdb-test1-worker-0:379803:386777 [5] NCCL INFO Trees [0] 6/-1/-1->5->4 [1] 6/-1/-1->5->4 [2] 6/-1/-1->5->4 [3] 6/-1/-1->5->4 [4] 6/-1/-1->5->4 [5] 6/-1/-1->5->4 [6] 6/-1/-1->5->4 [7] 6/-1/-1->5->4 [8] 6/-1/-1->5->4 [9] 6/-1/-1->5->4 [10] 6/-1/-1->5->4 [11] 6/-1/-1->5->4 [12] 6/-1/-1->5->4 [13] 6/-1/-1->5->4 [14] 6/-1/-1->5->4 [15] 6/-1/-1->5->4 [16] 6/-1/-1->5->4 [17] 6/-1/-1->5->4 [18] 6/-1/-1->5->4 [19] 6/-1/-1->5->4 [20] 6/-1/-1->5->4 [21] 6/-1/-1->5->4 [22] 6/-1/-1->5->4 [23] 6/-1/-1->5->4 +ywang29-vrdb-test1-worker-0:379801:386778 [3] NCCL INFO Trees [0] 4/-1/-1->3->2 [1] 4/-1/-1->3->2 [2] 4/-1/-1->3->2 [3] 4/-1/-1->3->2 [4] 4/-1/-1->3->2 [5] 4/-1/-1->3->2 [6] 4/-1/-1->3->2 [7] 4/-1/-1->3->2 [8] 4/-1/-1->3->2 [9] 4/-1/-1->3->2 [10] 4/-1/-1->3->2 [11] 4/-1/-1->3->2 [12] 4/-1/-1->3->2 [13] 4/-1/-1->3->2 [14] 4/-1/-1->3->2 [15] 4/-1/-1->3->2 [16] 4/-1/-1->3->2 [17] 4/-1/-1->3->2 [18] 4/-1/-1->3->2 [19] 4/-1/-1->3->2 [20] 4/-1/-1->3->2 [21] 4/-1/-1->3->2 [22] 4/-1/-1->3->2 [23] 4/-1/-1->3->2 +ywang29-vrdb-test1-worker-0:379798:386772 [0] NCCL INFO Channel 07/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:379805:386776 [7] NCCL INFO Trees [0] -1/-1/-1->7->6 [1] -1/-1/-1->7->6 [2] -1/-1/-1->7->6 [3] -1/-1/-1->7->6 [4] -1/-1/-1->7->6 [5] -1/-1/-1->7->6 [6] -1/-1/-1->7->6 [7] -1/-1/-1->7->6 [8] -1/-1/-1->7->6 [9] -1/-1/-1->7->6 [10] -1/-1/-1->7->6 [11] -1/-1/-1->7->6 [12] -1/-1/-1->7->6 [13] -1/-1/-1->7->6 [14] -1/-1/-1->7->6 [15] -1/-1/-1->7->6 [16] -1/-1/-1->7->6 [17] -1/-1/-1->7->6 [18] -1/-1/-1->7->6 [19] -1/-1/-1->7->6 [20] -1/-1/-1->7->6 [21] -1/-1/-1->7->6 [22] -1/-1/-1->7->6 [23] -1/-1/-1->7->6 +ywang29-vrdb-test1-worker-0:379801:386778 [3] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:379799:386779 [1] NCCL INFO Trees [0] 2/-1/-1->1->0 [1] 2/-1/-1->1->0 [2] 2/-1/-1->1->0 [3] 2/-1/-1->1->0 [4] 2/-1/-1->1->0 [5] 2/-1/-1->1->0 [6] 2/-1/-1->1->0 [7] 2/-1/-1->1->0 [8] 2/-1/-1->1->0 [9] 2/-1/-1->1->0 [10] 2/-1/-1->1->0 [11] 2/-1/-1->1->0 [12] 2/-1/-1->1->0 [13] 2/-1/-1->1->0 [14] 2/-1/-1->1->0 [15] 2/-1/-1->1->0 [16] 2/-1/-1->1->0 [17] 2/-1/-1->1->0 [18] 2/-1/-1->1->0 [19] 2/-1/-1->1->0 [20] 2/-1/-1->1->0 [21] 2/-1/-1->1->0 [22] 2/-1/-1->1->0 [23] 2/-1/-1->1->0 +ywang29-vrdb-test1-worker-0:379804:386775 [6] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:379803:386777 [5] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:379802:386774 [4] NCCL INFO Trees [0] 5/-1/-1->4->3 [1] 5/-1/-1->4->3 [2] 5/-1/-1->4->3 [3] 5/-1/-1->4->3 [4] 5/-1/-1->4->3 [5] 5/-1/-1->4->3 [6] 5/-1/-1->4->3 [7] 5/-1/-1->4->3 [8] 5/-1/-1->4->3 [9] 5/-1/-1->4->3 [10] 5/-1/-1->4->3 [11] 5/-1/-1->4->3 [12] 5/-1/-1->4->3 [13] 5/-1/-1->4->3 [14] 5/-1/-1->4->3 [15] 5/-1/-1->4->3 [16] 5/-1/-1->4->3 [17] 5/-1/-1->4->3 [18] 5/-1/-1->4->3 [19] 5/-1/-1->4->3 [20] 5/-1/-1->4->3 [21] 5/-1/-1->4->3 [22] 5/-1/-1->4->3 [23] 5/-1/-1->4->3 +ywang29-vrdb-test1-worker-0:379798:386772 [0] NCCL INFO Channel 08/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:379805:386776 [7] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:379799:386779 [1] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:379802:386774 [4] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:379798:386772 [0] NCCL INFO Channel 09/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:379798:386772 [0] NCCL INFO Channel 10/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:379798:386772 [0] NCCL INFO Channel 11/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:379798:386772 [0] NCCL INFO Channel 12/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:379798:386772 [0] NCCL INFO Channel 13/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:379798:386772 [0] NCCL INFO Channel 14/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:379798:386772 [0] NCCL INFO Channel 15/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:379798:386772 [0] NCCL INFO Channel 16/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:379798:386772 [0] NCCL INFO Channel 17/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:379798:386772 [0] NCCL INFO Channel 18/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:379798:386772 [0] NCCL INFO Channel 19/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:379798:386772 [0] NCCL INFO Channel 20/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:379798:386772 [0] NCCL INFO Channel 21/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:379798:386772 [0] NCCL INFO Channel 22/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:379798:386772 [0] NCCL INFO Channel 23/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:379798:386772 [0] NCCL INFO Trees [0] 1/-1/-1->0->-1 [1] 1/-1/-1->0->-1 [2] 1/-1/-1->0->-1 [3] 1/-1/-1->0->-1 [4] 1/-1/-1->0->-1 [5] 1/-1/-1->0->-1 [6] 1/-1/-1->0->-1 [7] 1/-1/-1->0->-1 [8] 1/-1/-1->0->-1 [9] 1/-1/-1->0->-1 [10] 1/-1/-1->0->-1 [11] 1/-1/-1->0->-1 [12] 1/-1/-1->0->-1 [13] 1/-1/-1->0->-1 [14] 1/-1/-1->0->-1 [15] 1/-1/-1->0->-1 [16] 1/-1/-1->0->-1 [17] 1/-1/-1->0->-1 [18] 1/-1/-1->0->-1 [19] 1/-1/-1->0->-1 [20] 1/-1/-1->0->-1 [21] 1/-1/-1->0->-1 [22] 1/-1/-1->0->-1 [23] 1/-1/-1->0->-1 +ywang29-vrdb-test1-worker-0:379798:386772 [0] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:379800:386773 [2] NCCL INFO Channel 00/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:386776 [7] NCCL INFO Channel 00/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:386777 [5] NCCL INFO Channel 00/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:386775 [6] NCCL INFO Channel 00/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:386773 [2] NCCL INFO Channel 01/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:386778 [3] NCCL INFO Channel 00/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:386779 [1] NCCL INFO Channel 00/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:386774 [4] NCCL INFO Channel 00/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379798:386772 [0] NCCL INFO Channel 00/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:386776 [7] NCCL INFO Channel 01/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:386777 [5] NCCL INFO Channel 01/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:386775 [6] NCCL INFO Channel 01/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:386773 [2] NCCL INFO Channel 02/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:386778 [3] NCCL INFO Channel 01/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:386779 [1] NCCL INFO Channel 01/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:386774 [4] NCCL INFO Channel 01/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379798:386772 [0] NCCL INFO Channel 01/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:386776 [7] NCCL INFO Channel 02/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:386777 [5] NCCL INFO Channel 02/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:386775 [6] NCCL INFO Channel 02/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:386773 [2] NCCL INFO Channel 03/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:386778 [3] NCCL INFO Channel 02/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:386779 [1] NCCL INFO Channel 02/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:386774 [4] NCCL INFO Channel 02/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379798:386772 [0] NCCL INFO Channel 02/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:386776 [7] NCCL INFO Channel 03/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:386777 [5] NCCL INFO Channel 03/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:386775 [6] NCCL INFO Channel 03/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:386773 [2] NCCL INFO Channel 04/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:386778 [3] NCCL INFO Channel 03/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:386779 [1] NCCL INFO Channel 03/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:386774 [4] NCCL INFO Channel 03/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379798:386772 [0] NCCL INFO Channel 03/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:386776 [7] NCCL INFO Channel 04/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:386777 [5] NCCL INFO Channel 04/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:386775 [6] NCCL INFO Channel 04/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:386773 [2] NCCL INFO Channel 05/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:386778 [3] NCCL INFO Channel 04/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:386779 [1] NCCL INFO Channel 04/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:386774 [4] NCCL INFO Channel 04/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379798:386772 [0] NCCL INFO Channel 04/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:386776 [7] NCCL INFO Channel 05/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:386777 [5] NCCL INFO Channel 05/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:386775 [6] NCCL INFO Channel 05/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:386773 [2] NCCL INFO Channel 06/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:386778 [3] NCCL INFO Channel 05/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:386779 [1] NCCL INFO Channel 05/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:386774 [4] NCCL INFO Channel 05/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379798:386772 [0] NCCL INFO Channel 05/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:386776 [7] NCCL INFO Channel 06/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:386777 [5] NCCL INFO Channel 06/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:386775 [6] NCCL INFO Channel 06/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:386773 [2] NCCL INFO Channel 07/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:386778 [3] NCCL INFO Channel 06/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:386779 [1] NCCL INFO Channel 06/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:386774 [4] NCCL INFO Channel 06/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379798:386772 [0] NCCL INFO Channel 06/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:386776 [7] NCCL INFO Channel 07/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:386777 [5] NCCL INFO Channel 07/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:386775 [6] NCCL INFO Channel 07/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:386778 [3] NCCL INFO Channel 07/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379798:386772 [0] NCCL INFO Channel 07/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:386779 [1] NCCL INFO Channel 07/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:386774 [4] NCCL INFO Channel 07/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:386773 [2] NCCL INFO Channel 08/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:386776 [7] NCCL INFO Channel 08/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:386777 [5] NCCL INFO Channel 08/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:386775 [6] NCCL INFO Channel 08/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:386778 [3] NCCL INFO Channel 08/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379798:386772 [0] NCCL INFO Channel 08/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:386779 [1] NCCL INFO Channel 08/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:386774 [4] NCCL INFO Channel 08/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:386773 [2] NCCL INFO Channel 09/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:386777 [5] NCCL INFO Channel 09/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:386776 [7] NCCL INFO Channel 09/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:386775 [6] NCCL INFO Channel 09/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379798:386772 [0] NCCL INFO Channel 09/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:386778 [3] NCCL INFO Channel 09/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:386779 [1] NCCL INFO Channel 09/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:386774 [4] NCCL INFO Channel 09/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:386773 [2] NCCL INFO Channel 10/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:386777 [5] NCCL INFO Channel 10/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:386776 [7] NCCL INFO Channel 10/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:386775 [6] NCCL INFO Channel 10/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379798:386772 [0] NCCL INFO Channel 10/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:386778 [3] NCCL INFO Channel 10/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:386779 [1] NCCL INFO Channel 10/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:386774 [4] NCCL INFO Channel 10/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:386773 [2] NCCL INFO Channel 11/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:386777 [5] NCCL INFO Channel 11/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:386776 [7] NCCL INFO Channel 11/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:386775 [6] NCCL INFO Channel 11/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379798:386772 [0] NCCL INFO Channel 11/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:386778 [3] NCCL INFO Channel 11/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:386779 [1] NCCL INFO Channel 11/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:386773 [2] NCCL INFO Channel 12/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:386774 [4] NCCL INFO Channel 11/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:386777 [5] NCCL INFO Channel 12/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:386776 [7] NCCL INFO Channel 12/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:386775 [6] NCCL INFO Channel 12/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379798:386772 [0] NCCL INFO Channel 12/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:386778 [3] NCCL INFO Channel 12/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:386779 [1] NCCL INFO Channel 12/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:386774 [4] NCCL INFO Channel 12/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:386776 [7] NCCL INFO Channel 13/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:386777 [5] NCCL INFO Channel 13/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:386775 [6] NCCL INFO Channel 13/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379798:386772 [0] NCCL INFO Channel 13/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:386778 [3] NCCL INFO Channel 13/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:386779 [1] NCCL INFO Channel 13/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:386774 [4] NCCL INFO Channel 13/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:386776 [7] NCCL INFO Channel 14/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:386777 [5] NCCL INFO Channel 14/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:386775 [6] NCCL INFO Channel 14/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379798:386772 [0] NCCL INFO Channel 14/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:386778 [3] NCCL INFO Channel 14/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:386779 [1] NCCL INFO Channel 14/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:386774 [4] NCCL INFO Channel 14/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:386776 [7] NCCL INFO Channel 15/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:386777 [5] NCCL INFO Channel 15/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:386775 [6] NCCL INFO Channel 15/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379798:386772 [0] NCCL INFO Channel 15/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:386778 [3] NCCL INFO Channel 15/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:386779 [1] NCCL INFO Channel 15/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:386774 [4] NCCL INFO Channel 15/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:386776 [7] NCCL INFO Channel 16/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:386777 [5] NCCL INFO Channel 16/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:386775 [6] NCCL INFO Channel 16/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379798:386772 [0] NCCL INFO Channel 16/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:386778 [3] NCCL INFO Channel 16/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:386779 [1] NCCL INFO Channel 16/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:386774 [4] NCCL INFO Channel 16/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:386776 [7] NCCL INFO Channel 17/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:386777 [5] NCCL INFO Channel 17/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:386775 [6] NCCL INFO Channel 17/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379798:386772 [0] NCCL INFO Channel 17/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:386778 [3] NCCL INFO Channel 17/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:386779 [1] NCCL INFO Channel 17/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:386774 [4] NCCL INFO Channel 17/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:386776 [7] NCCL INFO Channel 18/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:386777 [5] NCCL INFO Channel 18/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:386775 [6] NCCL INFO Channel 18/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379798:386772 [0] NCCL INFO Channel 18/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:386778 [3] NCCL INFO Channel 18/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:386779 [1] NCCL INFO Channel 18/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:386774 [4] NCCL INFO Channel 18/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:386776 [7] NCCL INFO Channel 19/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:386777 [5] NCCL INFO Channel 19/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:386775 [6] NCCL INFO Channel 19/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379798:386772 [0] NCCL INFO Channel 19/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:386778 [3] NCCL INFO Channel 19/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:386779 [1] NCCL INFO Channel 19/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:386774 [4] NCCL INFO Channel 19/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:386776 [7] NCCL INFO Channel 20/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:386777 [5] NCCL INFO Channel 20/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:386775 [6] NCCL INFO Channel 20/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379798:386772 [0] NCCL INFO Channel 20/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:386778 [3] NCCL INFO Channel 20/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:386779 [1] NCCL INFO Channel 20/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:386774 [4] NCCL INFO Channel 20/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:386776 [7] NCCL INFO Channel 21/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:386777 [5] NCCL INFO Channel 21/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:386775 [6] NCCL INFO Channel 21/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379798:386772 [0] NCCL INFO Channel 21/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:386778 [3] NCCL INFO Channel 21/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:386779 [1] NCCL INFO Channel 21/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:386774 [4] NCCL INFO Channel 21/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:386776 [7] NCCL INFO Channel 22/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:386777 [5] NCCL INFO Channel 22/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:386775 [6] NCCL INFO Channel 22/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379798:386772 [0] NCCL INFO Channel 22/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:386779 [1] NCCL INFO Channel 22/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:386774 [4] NCCL INFO Channel 22/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:386778 [3] NCCL INFO Channel 22/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:386776 [7] NCCL INFO Channel 23/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:386777 [5] NCCL INFO Channel 23/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:386775 [6] NCCL INFO Channel 23/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379798:386772 [0] NCCL INFO Channel 23/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:386779 [1] NCCL INFO Channel 23/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:386774 [4] NCCL INFO Channel 23/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:386778 [3] NCCL INFO Channel 23/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:386773 [2] NCCL INFO Channel 13/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:386773 [2] NCCL INFO Channel 14/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:386773 [2] NCCL INFO Channel 15/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:386773 [2] NCCL INFO Channel 16/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:386773 [2] NCCL INFO Channel 17/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:386773 [2] NCCL INFO Channel 18/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:386773 [2] NCCL INFO Channel 19/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:386773 [2] NCCL INFO Channel 20/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:386773 [2] NCCL INFO Channel 21/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:386773 [2] NCCL INFO Channel 22/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:386773 [2] NCCL INFO Channel 23/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:386779 [1] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:379798:386772 [0] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:379802:386774 [4] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:379800:386773 [2] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:379801:386778 [3] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:379805:386776 [7] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:379805:386776 [7] NCCL INFO Channel 00/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:386775 [6] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:379803:386777 [5] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:379805:386776 [7] NCCL INFO Channel 01/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:386776 [7] NCCL INFO Channel 02/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:386776 [7] NCCL INFO Channel 03/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:386776 [7] NCCL INFO Channel 04/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:386776 [7] NCCL INFO Channel 05/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:386776 [7] NCCL INFO Channel 06/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:386776 [7] NCCL INFO Channel 07/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:386776 [7] NCCL INFO Channel 08/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:386776 [7] NCCL INFO Channel 09/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:386776 [7] NCCL INFO Channel 10/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:386776 [7] NCCL INFO Channel 11/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:386776 [7] NCCL INFO Channel 12/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:386776 [7] NCCL INFO Channel 13/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:386776 [7] NCCL INFO Channel 14/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:386776 [7] NCCL INFO Channel 15/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:386776 [7] NCCL INFO Channel 16/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:386779 [1] NCCL INFO Channel 00/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:386776 [7] NCCL INFO Channel 17/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:386778 [3] NCCL INFO Channel 00/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:386779 [1] NCCL INFO Channel 01/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:386776 [7] NCCL INFO Channel 18/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:386778 [3] NCCL INFO Channel 01/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:386779 [1] NCCL INFO Channel 02/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:386776 [7] NCCL INFO Channel 19/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:386778 [3] NCCL INFO Channel 02/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:386779 [1] NCCL INFO Channel 03/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:386774 [4] NCCL INFO Channel 00/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:386778 [3] NCCL INFO Channel 03/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:386779 [1] NCCL INFO Channel 04/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:386774 [4] NCCL INFO Channel 01/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:386778 [3] NCCL INFO Channel 04/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:386779 [1] NCCL INFO Channel 05/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:386775 [6] NCCL INFO Channel 00/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:386774 [4] NCCL INFO Channel 02/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:386778 [3] NCCL INFO Channel 05/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:386779 [1] NCCL INFO Channel 06/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:386775 [6] NCCL INFO Channel 01/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:386774 [4] NCCL INFO Channel 03/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:386779 [1] NCCL INFO Channel 07/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:386778 [3] NCCL INFO Channel 06/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:386775 [6] NCCL INFO Channel 02/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:386774 [4] NCCL INFO Channel 04/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:386779 [1] NCCL INFO Channel 08/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:386778 [3] NCCL INFO Channel 07/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:386775 [6] NCCL INFO Channel 03/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:386774 [4] NCCL INFO Channel 05/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:386779 [1] NCCL INFO Channel 09/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:386773 [2] NCCL INFO Channel 00/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:386778 [3] NCCL INFO Channel 08/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:386775 [6] NCCL INFO Channel 04/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:386774 [4] NCCL INFO Channel 06/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:386779 [1] NCCL INFO Channel 10/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:386773 [2] NCCL INFO Channel 01/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:386778 [3] NCCL INFO Channel 09/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:386775 [6] NCCL INFO Channel 05/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:386774 [4] NCCL INFO Channel 07/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:386779 [1] NCCL INFO Channel 11/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:386778 [3] NCCL INFO Channel 10/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:386773 [2] NCCL INFO Channel 02/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:386775 [6] NCCL INFO Channel 06/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:386774 [4] NCCL INFO Channel 08/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:386779 [1] NCCL INFO Channel 12/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:386778 [3] NCCL INFO Channel 11/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:386773 [2] NCCL INFO Channel 03/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:386775 [6] NCCL INFO Channel 07/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:386779 [1] NCCL INFO Channel 13/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:386774 [4] NCCL INFO Channel 09/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:386778 [3] NCCL INFO Channel 12/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:386773 [2] NCCL INFO Channel 04/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:386775 [6] NCCL INFO Channel 08/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:386776 [7] NCCL INFO Channel 20/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:386778 [3] NCCL INFO Channel 13/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:386773 [2] NCCL INFO Channel 05/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:386775 [6] NCCL INFO Channel 09/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:386778 [3] NCCL INFO Channel 14/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:386776 [7] NCCL INFO Channel 21/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:386773 [2] NCCL INFO Channel 06/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:386775 [6] NCCL INFO Channel 10/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:386774 [4] NCCL INFO Channel 10/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:386778 [3] NCCL INFO Channel 15/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:386776 [7] NCCL INFO Channel 22/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:386773 [2] NCCL INFO Channel 07/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:386775 [6] NCCL INFO Channel 11/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:386774 [4] NCCL INFO Channel 11/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:386779 [1] NCCL INFO Channel 14/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:386778 [3] NCCL INFO Channel 16/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379805:386776 [7] NCCL INFO Channel 23/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:386773 [2] NCCL INFO Channel 08/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:386775 [6] NCCL INFO Channel 12/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:386774 [4] NCCL INFO Channel 12/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:386779 [1] NCCL INFO Channel 15/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:386778 [3] NCCL INFO Channel 17/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:386773 [2] NCCL INFO Channel 09/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:386774 [4] NCCL INFO Channel 13/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:386775 [6] NCCL INFO Channel 13/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:386779 [1] NCCL INFO Channel 16/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:386778 [3] NCCL INFO Channel 18/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:386773 [2] NCCL INFO Channel 10/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:386774 [4] NCCL INFO Channel 14/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:386775 [6] NCCL INFO Channel 14/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:386779 [1] NCCL INFO Channel 17/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:386778 [3] NCCL INFO Channel 19/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:386773 [2] NCCL INFO Channel 11/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:386774 [4] NCCL INFO Channel 15/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:386775 [6] NCCL INFO Channel 15/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:386779 [1] NCCL INFO Channel 18/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:386778 [3] NCCL INFO Channel 20/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:386775 [6] NCCL INFO Channel 16/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:386773 [2] NCCL INFO Channel 12/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:386774 [4] NCCL INFO Channel 16/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:386779 [1] NCCL INFO Channel 19/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:386778 [3] NCCL INFO Channel 21/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:386777 [5] NCCL INFO Channel 00/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:386775 [6] NCCL INFO Channel 17/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:386773 [2] NCCL INFO Channel 13/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:386774 [4] NCCL INFO Channel 17/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:386779 [1] NCCL INFO Channel 20/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:386778 [3] NCCL INFO Channel 22/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:386777 [5] NCCL INFO Channel 01/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:386773 [2] NCCL INFO Channel 14/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:386775 [6] NCCL INFO Channel 18/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:386774 [4] NCCL INFO Channel 18/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:386779 [1] NCCL INFO Channel 21/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379801:386778 [3] NCCL INFO Channel 23/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:386777 [5] NCCL INFO Channel 02/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:386773 [2] NCCL INFO Channel 15/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:386775 [6] NCCL INFO Channel 19/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:386779 [1] NCCL INFO Channel 22/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:386774 [4] NCCL INFO Channel 19/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:386777 [5] NCCL INFO Channel 03/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:386773 [2] NCCL INFO Channel 16/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:386775 [6] NCCL INFO Channel 20/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379799:386779 [1] NCCL INFO Channel 23/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:386774 [4] NCCL INFO Channel 20/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:386777 [5] NCCL INFO Channel 04/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:386773 [2] NCCL INFO Channel 17/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:386775 [6] NCCL INFO Channel 21/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:386774 [4] NCCL INFO Channel 21/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:386777 [5] NCCL INFO Channel 05/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:386773 [2] NCCL INFO Channel 18/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:386775 [6] NCCL INFO Channel 22/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:386774 [4] NCCL INFO Channel 22/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:386777 [5] NCCL INFO Channel 06/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:386773 [2] NCCL INFO Channel 19/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379804:386775 [6] NCCL INFO Channel 23/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379802:386774 [4] NCCL INFO Channel 23/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:386777 [5] NCCL INFO Channel 07/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:386773 [2] NCCL INFO Channel 20/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:386777 [5] NCCL INFO Channel 08/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:386773 [2] NCCL INFO Channel 21/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:386777 [5] NCCL INFO Channel 09/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:386773 [2] NCCL INFO Channel 22/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:386777 [5] NCCL INFO Channel 10/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379800:386773 [2] NCCL INFO Channel 23/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:386777 [5] NCCL INFO Channel 11/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:386777 [5] NCCL INFO Channel 12/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:386777 [5] NCCL INFO Channel 13/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:386777 [5] NCCL INFO Channel 14/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:386777 [5] NCCL INFO Channel 15/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:386777 [5] NCCL INFO Channel 16/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:386777 [5] NCCL INFO Channel 17/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:386777 [5] NCCL INFO Channel 18/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:386777 [5] NCCL INFO Channel 19/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:386777 [5] NCCL INFO Channel 20/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:386777 [5] NCCL INFO Channel 21/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:386777 [5] NCCL INFO Channel 22/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379803:386777 [5] NCCL INFO Channel 23/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:379798:386772 [0] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:379798:386772 [0] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:379798:386772 [0] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:379799:386779 [1] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:379799:386779 [1] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:379799:386779 [1] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:379800:386773 [2] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:379800:386773 [2] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:379800:386773 [2] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:379801:386778 [3] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:379801:386778 [3] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:379801:386778 [3] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:379802:386774 [4] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:379802:386774 [4] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:379802:386774 [4] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:379803:386777 [5] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:379803:386777 [5] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:379803:386777 [5] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:379805:386776 [7] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:379804:386775 [6] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:379805:386776 [7] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:379805:386776 [7] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:379804:386775 [6] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:379804:386775 [6] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:379798:386772 [0] NCCL INFO ncclCommInitRank comm 0x7fa22406afb0 rank 0 nranks 8 cudaDev 0 nvmlDev 0 busId 101c0 commId 0xe8014001d19baaba - Init COMPLETE +ywang29-vrdb-test1-worker-0:379802:386774 [4] NCCL INFO ncclCommInitRank comm 0x7f0cdc06b490 rank 4 nranks 8 cudaDev 4 nvmlDev 4 busId 901c0 commId 0xe8014001d19baaba - Init COMPLETE +ywang29-vrdb-test1-worker-0:379803:386777 [5] NCCL INFO ncclCommInitRank comm 0x7f99a406af60 rank 5 nranks 8 cudaDev 5 nvmlDev 5 busId 901d0 commId 0xe8014001d19baaba - Init COMPLETE +ywang29-vrdb-test1-worker-0:379799:386779 [1] NCCL INFO ncclCommInitRank comm 0x7fb7cc06a5f0 rank 1 nranks 8 cudaDev 1 nvmlDev 1 busId 101d0 commId 0xe8014001d19baaba - Init COMPLETE +ywang29-vrdb-test1-worker-0:379805:386776 [7] NCCL INFO ncclCommInitRank comm 0x7fc44406a980 rank 7 nranks 8 cudaDev 7 nvmlDev 7 busId a01d0 commId 0xe8014001d19baaba - Init COMPLETE +ywang29-vrdb-test1-worker-0:379801:386778 [3] NCCL INFO ncclCommInitRank comm 0x7feeb406afd0 rank 3 nranks 8 cudaDev 3 nvmlDev 3 busId 201d0 commId 0xe8014001d19baaba - Init COMPLETE +ywang29-vrdb-test1-worker-0:379804:386775 [6] NCCL INFO ncclCommInitRank comm 0x7fc19406aab0 rank 6 nranks 8 cudaDev 6 nvmlDev 6 busId a01c0 commId 0xe8014001d19baaba - Init COMPLETE +ywang29-vrdb-test1-worker-0:379800:386773 [2] NCCL INFO ncclCommInitRank comm 0x7fd5dc06a8a0 rank 2 nranks 8 cudaDev 2 nvmlDev 2 busId 201c0 commId 0xe8014001d19baaba - Init COMPLETE + 0%| | 1/520 [00:13<2:00:33, 13.94s/it] {'loss': 2.0428, 'grad_norm': 0.011451927653349517, 'learning_rate': 0.0125, 'epoch': 0.0} + 0%| | 1/520 [00:13<2:00:33, 13.94s/it] 0%| | 2/520 [00:17<1:09:34, 8.06s/it] {'loss': 2.051, 'grad_norm': 0.012418001849800088, 'learning_rate': 0.025, 'epoch': 0.0} + 0%| | 2/520 [00:17<1:09:34, 8.06s/it] 1%| | 3/520 [00:21<52:59, 6.15s/it] {'loss': 1.6848, 'grad_norm': 0.006518234458816051, 'learning_rate': 0.037500000000000006, 'epoch': 0.01} + 1%| | 3/520 [00:21<52:59, 6.15s/it] 1%| | 4/520 [00:25<44:44, 5.20s/it] {'loss': 1.5745, 'grad_norm': 0.002143777381553114, 'learning_rate': 0.05, 'epoch': 0.01} + 1%| | 4/520 [00:25<44:44, 5.20s/it] 1%| | 5/520 [00:29<39:51, 4.64s/it] {'loss': 1.6041, 'grad_norm': 0.0017166760604213254, 'learning_rate': 0.0625, 'epoch': 0.01} + 1%| | 5/520 [00:29<39:51, 4.64s/it] 1%| | 6/520 [00:32<37:12, 4.34s/it] {'loss': 1.3548, 'grad_norm': 0.0010063328451630363, 'learning_rate': 0.07500000000000001, 'epoch': 0.01} + 1%| | 6/520 [00:32<37:12, 4.34s/it] 1%|▏ | 7/520 [00:36<35:40, 4.17s/it] {'loss': 1.4311, 'grad_norm': 0.0012388585847609769, 'learning_rate': 0.08750000000000001, 'epoch': 0.01} + 1%|▏ | 7/520 [00:36<35:40, 4.17s/it] 2%|▏ | 8/520 [00:41<36:18, 4.26s/it] {'loss': 1.4508, 'grad_norm': 0.001195600898448798, 'learning_rate': 0.1, 'epoch': 0.02} + 2%|▏ | 8/520 [00:41<36:18, 4.26s/it] 2%|▏ | 9/520 [00:45<36:29, 4.29s/it] {'loss': 1.5328, 'grad_norm': 0.0016787088987737528, 'learning_rate': 0.1125, 'epoch': 0.02} + 2%|▏ | 9/520 [00:45<36:29, 4.29s/it] 2%|▏ | 10/520 [00:49<35:12, 4.14s/it] {'loss': 1.3801, 'grad_norm': 0.0016547987171835286, 'learning_rate': 0.125, 'epoch': 0.02} + 2%|▏ | 10/520 [00:49<35:12, 4.14s/it] 2%|▏ | 11/520 [00:53<34:31, 4.07s/it] {'loss': 1.4389, 'grad_norm': 0.0013851000482788475, 'learning_rate': 0.1375, 'epoch': 0.02} + 2%|▏ | 11/520 [00:53<34:31, 4.07s/it] 2%|▏ | 12/520 [00:57<33:50, 4.00s/it] {'loss': 1.3197, 'grad_norm': 0.0012910584251054104, 'learning_rate': 0.15000000000000002, 'epoch': 0.02} + 2%|▏ | 12/520 [00:57<33:50, 4.00s/it][2025-10-12 14:09:19,705] [WARNING] [stage3.py:2069:step] 1 pytorch allocator cache flushes since last step. this happens when there is high memory pressure and is detrimental to performance. if this is happening frequently consider adjusting settings to reduce memory consumption. If you are unable to make the cache flushes go away consider adding get_accelerator().empty_cache() calls in your training loop to ensure that all ranks flush their caches at the same time + 2%|▎ | 13/520 [01:01<34:58, 4.14s/it] {'loss': 1.3752, 'grad_norm': 0.0015326799664218134, 'learning_rate': 0.1625, 'epoch': 0.03} + 2%|▎ | 13/520 [01:01<34:58, 4.14s/it] 3%|▎ | 14/520 [01:05<33:36, 3.98s/it] {'loss': 1.4251, 'grad_norm': 0.0015025921207448378, 'learning_rate': 0.17500000000000002, 'epoch': 0.03} + 3%|▎ | 14/520 [01:05<33:36, 3.98s/it] 3%|▎ | 15/520 [01:08<32:46, 3.90s/it] {'loss': 1.3597, 'grad_norm': 0.0010396141944622227, 'learning_rate': 0.1875, 'epoch': 0.03} + 3%|▎ | 15/520 [01:08<32:46, 3.90s/it] 3%|▎ | 16/520 [01:12<32:08, 3.83s/it] {'loss': 1.3193, 'grad_norm': 0.001145648963808095, 'learning_rate': 0.2, 'epoch': 0.03} + 3%|▎ | 16/520 [01:12<32:08, 3.83s/it] 3%|▎ | 17/520 [01:16<31:38, 3.77s/it] {'loss': 1.441, 'grad_norm': 0.0011818931760029835, 'learning_rate': 0.1999980572931538, 'epoch': 0.03} + 3%|▎ | 17/520 [01:16<31:38, 3.77s/it] 3%|▎ | 18/520 [01:19<31:12, 3.73s/it] {'loss': 1.2973, 'grad_norm': 0.0012850359840013402, 'learning_rate': 0.19999222924809748, 'epoch': 0.03} + 3%|▎ | 18/520 [01:19<31:12, 3.73s/it] 4%|▎ | 19/520 [01:23<30:54, 3.70s/it] {'loss': 1.3179, 'grad_norm': 0.001118709171926159, 'learning_rate': 0.19998251609127465, 'epoch': 0.04} + 4%|▎ | 19/520 [01:23<30:54, 3.70s/it] 4%|▍ | 20/520 [01:27<30:45, 3.69s/it] {'loss': 1.2828, 'grad_norm': 0.0014210242955228999, 'learning_rate': 0.19996891820008164, 'epoch': 0.04} + 4%|▍ | 20/520 [01:27<30:45, 3.69s/it] 4%|▍ | 21/520 [01:30<30:47, 3.70s/it] {'loss': 1.3351, 'grad_norm': 0.0020948671204595185, 'learning_rate': 0.19995143610285276, 'epoch': 0.04} + 4%|▍ | 21/520 [01:30<30:47, 3.70s/it] 4%|▍ | 22/520 [01:34<30:46, 3.71s/it] {'loss': 1.4279, 'grad_norm': 0.0012398093735216176, 'learning_rate': 0.19993007047883987, 'epoch': 0.04} + 4%|▍ | 22/520 [01:34<30:46, 3.71s/it] 4%|▍ | 23/520 [01:38<30:42, 3.71s/it] {'loss': 1.3818, 'grad_norm': 0.0013359626335291731, 'learning_rate': 0.1999048221581858, 'epoch': 0.04} + 4%|▍ | 23/520 [01:38<30:42, 3.71s/it] 5%|▍ | 24/520 [01:41<30:39, 3.71s/it] {'loss': 1.3046, 'grad_norm': 0.0013259407596770674, 'learning_rate': 0.19987569212189224, 'epoch': 0.05} + 5%|▍ | 24/520 [01:41<30:39, 3.71s/it] 5%|▍ | 25/520 [01:45<30:33, 3.70s/it] {'loss': 1.3713, 'grad_norm': 0.0014370656144570345, 'learning_rate': 0.19984268150178167, 'epoch': 0.05} + 5%|▍ | 25/520 [01:45<30:33, 3.70s/it] 5%|▌ | 26/520 [01:49<30:25, 3.70s/it] {'loss': 1.3355, 'grad_norm': 0.0012017939438748058, 'learning_rate': 0.1998057915804532, 'epoch': 0.05} + 5%|▌ | 26/520 [01:49<30:25, 3.70s/it] 5%|▌ | 27/520 [01:53<30:19, 3.69s/it] {'loss': 1.2741, 'grad_norm': 0.0013358338569505618, 'learning_rate': 0.1997650237912329, 'epoch': 0.05} + 5%|▌ | 27/520 [01:53<30:19, 3.69s/it] 5%|▌ | 28/520 [01:56<30:21, 3.70s/it] {'loss': 1.2848, 'grad_norm': 0.0013852799982664026, 'learning_rate': 0.199720379718118, 'epoch': 0.05} + 5%|▌ | 28/520 [01:56<30:21, 3.70s/it] 6%|▌ | 29/520 [02:00<30:14, 3.70s/it] {'loss': 1.3136, 'grad_norm': 0.0014442752260721212, 'learning_rate': 0.19967186109571552, 'epoch': 0.06} + 6%|▌ | 29/520 [02:00<30:14, 3.70s/it] 6%|▌ | 30/520 [02:04<30:09, 3.69s/it] {'loss': 1.3845, 'grad_norm': 0.0012672357359624102, 'learning_rate': 0.19961946980917455, 'epoch': 0.06} + 6%|▌ | 30/520 [02:04<30:09, 3.69s/it] 6%|▌ | 31/520 [02:07<30:06, 3.69s/it] {'loss': 1.2885, 'grad_norm': 0.0012191987026680623, 'learning_rate': 0.1995632078941134, 'epoch': 0.06} + 6%|▌ | 31/520 [02:07<30:06, 3.69s/it] 6%|▌ | 32/520 [02:11<30:00, 3.69s/it] {'loss': 1.2168, 'grad_norm': 0.0012590510606934062, 'learning_rate': 0.19950307753654017, 'epoch': 0.06} + 6%|▌ | 32/520 [02:11<30:00, 3.69s/it] 6%|▋ | 33/520 [02:15<30:18, 3.73s/it] {'loss': 1.2866, 'grad_norm': 0.0014689087515596182, 'learning_rate': 0.19943908107276798, 'epoch': 0.06} + 6%|▋ | 33/520 [02:15<30:18, 3.73s/it] 7%|▋ | 34/520 [02:19<30:28, 3.76s/it] {'loss': 1.2914, 'grad_norm': 0.0016009771773709987, 'learning_rate': 0.19937122098932428, 'epoch': 0.07} + 7%|▋ | 34/520 [02:19<30:28, 3.76s/it] 7%|▋ | 35/520 [02:22<30:33, 3.78s/it] {'loss': 1.291, 'grad_norm': 0.0015881026930456858, 'learning_rate': 0.19929949992285395, 'epoch': 0.07} + 7%|▋ | 35/520 [02:22<30:33, 3.78s/it] 7%|▋ | 36/520 [02:26<30:41, 3.81s/it] {'loss': 1.3861, 'grad_norm': 0.0013521908662923272, 'learning_rate': 0.19922392066001723, 'epoch': 0.07} + 7%|▋ | 36/520 [02:26<30:41, 3.81s/it] 7%|▋ | 37/520 [02:30<30:41, 3.81s/it] {'loss': 1.3689, 'grad_norm': 0.0012731028496666516, 'learning_rate': 0.19914448613738106, 'epoch': 0.07} + 7%|▋ | 37/520 [02:30<30:41, 3.81s/it] 7%|▋ | 38/520 [02:34<30:41, 3.82s/it] {'loss': 1.4664, 'grad_norm': 0.0014247347894970226, 'learning_rate': 0.1990611994413053, 'epoch': 0.07} + 7%|▋ | 38/520 [02:34<30:41, 3.82s/it] 8%|▊ | 39/520 [02:38<30:37, 3.82s/it] {'loss': 1.3245, 'grad_norm': 0.001646173437102172, 'learning_rate': 0.19897406380782262, 'epoch': 0.07} + 8%|▊ | 39/520 [02:38<30:37, 3.82s/it] 8%|▊ | 40/520 [02:42<30:35, 3.82s/it] {'loss': 1.3541, 'grad_norm': 0.001350887237570176, 'learning_rate': 0.19888308262251286, 'epoch': 0.08} + 8%|▊ | 40/520 [02:42<30:35, 3.82s/it] 8%|▊ | 41/520 [02:46<30:33, 3.83s/it] {'loss': 1.3288, 'grad_norm': 0.0014140246275613135, 'learning_rate': 0.19878825942037148, 'epoch': 0.08} + 8%|▊ | 41/520 [02:46<30:33, 3.83s/it] 8%|▊ | 42/520 [02:49<30:23, 3.82s/it] {'loss': 1.3405, 'grad_norm': 0.0018886324942413743, 'learning_rate': 0.19868959788567211, 'epoch': 0.08} + 8%|▊ | 42/520 [02:49<30:23, 3.82s/it] 8%|▊ | 43/520 [02:53<30:19, 3.81s/it] {'loss': 1.2733, 'grad_norm': 0.0012619029691278616, 'learning_rate': 0.1985871018518236, 'epoch': 0.08} + 8%|▊ | 43/520 [02:53<30:19, 3.81s/it] 8%|▊ | 44/520 [02:57<30:13, 3.81s/it] {'loss': 1.3685, 'grad_norm': 0.0014058199317530725, 'learning_rate': 0.19848077530122082, 'epoch': 0.08} + 8%|▊ | 44/520 [02:57<30:13, 3.81s/it] 9%|▊ | 45/520 [03:01<30:08, 3.81s/it] {'loss': 1.3559, 'grad_norm': 0.001526150914652272, 'learning_rate': 0.19837062236509015, 'epoch': 0.09} + 9%|▊ | 45/520 [03:01<30:08, 3.81s/it] 9%|▉ | 46/520 [03:05<30:03, 3.81s/it] {'loss': 1.4214, 'grad_norm': 0.001425205486510709, 'learning_rate': 0.19825664732332884, 'epoch': 0.09} + 9%|▉ | 46/520 [03:05<30:03, 3.81s/it] 9%|▉ | 47/520 [03:08<29:41, 3.77s/it] {'loss': 1.3374, 'grad_norm': 0.0015400204021245958, 'learning_rate': 0.19813885460433878, 'epoch': 0.09} + 9%|▉ | 47/520 [03:08<29:41, 3.77s/it] 9%|▉ | 48/520 [03:12<29:25, 3.74s/it] {'loss': 1.3117, 'grad_norm': 0.0016017782787142085, 'learning_rate': 0.19801724878485438, 'epoch': 0.09} + 9%|▉ | 48/520 [03:12<29:25, 3.74s/it] 9%|▉ | 49/520 [03:16<29:17, 3.73s/it] {'loss': 1.369, 'grad_norm': 0.0016151086256347843, 'learning_rate': 0.19789183458976486, 'epoch': 0.09} + 9%|▉ | 49/520 [03:16<29:17, 3.73s/it] 10%|▉ | 50/520 [03:19<29:02, 3.71s/it] {'loss': 1.3609, 'grad_norm': 0.0015552840882555633, 'learning_rate': 0.19776261689193048, 'epoch': 0.1} + 10%|▉ | 50/520 [03:19<29:02, 3.71s/it] 10%|▉ | 51/520 [03:23<29:12, 3.74s/it] {'loss': 1.2933, 'grad_norm': 0.0017093575528902417, 'learning_rate': 0.19762960071199334, 'epoch': 0.1} + 10%|▉ | 51/520 [03:23<29:12, 3.74s/it] 10%|█ | 52/520 [03:27<29:20, 3.76s/it] {'loss': 1.4205, 'grad_norm': 0.00172932853927916, 'learning_rate': 0.19749279121818236, 'epoch': 0.1} + 10%|█ | 52/520 [03:27<29:20, 3.76s/it] 10%|█ | 53/520 [03:31<29:21, 3.77s/it] {'loss': 1.4019, 'grad_norm': 0.0015564854926093336, 'learning_rate': 0.19735219372611235, 'epoch': 0.1} + 10%|█ | 53/520 [03:31<29:21, 3.77s/it] 10%|█ | 54/520 [03:34<29:21, 3.78s/it] {'loss': 1.3253, 'grad_norm': 0.001465846639529024, 'learning_rate': 0.19720781369857746, 'epoch': 0.1} + 10%|█ | 54/520 [03:34<29:21, 3.78s/it] 11%|█ | 55/520 [03:38<29:16, 3.78s/it] {'loss': 1.2967, 'grad_norm': 0.0016180374326297791, 'learning_rate': 0.1970596567453391, 'epoch': 0.11} + 11%|█ | 55/520 [03:38<29:16, 3.78s/it] 11%|█ | 56/520 [03:42<29:12, 3.78s/it] {'loss': 1.428, 'grad_norm': 0.0015451452866951068, 'learning_rate': 0.1969077286229078, 'epoch': 0.11} + 11%|█ | 56/520 [03:42<29:12, 3.78s/it] 11%|█ | 57/520 [03:46<29:27, 3.82s/it] {'loss': 1.2903, 'grad_norm': 0.001683990100811615, 'learning_rate': 0.19675203523431964, 'epoch': 0.11} + 11%|█ | 57/520 [03:46<29:27, 3.82s/it] 11%|█ | 58/520 [03:50<29:23, 3.82s/it] {'loss': 1.4515, 'grad_norm': 0.0013876854784671238, 'learning_rate': 0.19659258262890683, 'epoch': 0.11} + 11%|█ | 58/520 [03:50<29:23, 3.82s/it] 11%|█▏ | 59/520 [03:54<29:15, 3.81s/it] {'loss': 1.2647, 'grad_norm': 0.00150790526131653, 'learning_rate': 0.19642937700206278, 'epoch': 0.11} + 11%|█▏ | 59/520 [03:54<29:15, 3.81s/it] 12%|█▏ | 60/520 [03:57<29:12, 3.81s/it] {'loss': 1.3609, 'grad_norm': 0.0014272219223491242, 'learning_rate': 0.19626242469500121, 'epoch': 0.12} + 12%|█▏ | 60/520 [03:57<29:12, 3.81s/it] 12%|█▏ | 61/520 [04:01<29:04, 3.80s/it] {'loss': 1.3399, 'grad_norm': 0.0015117823185922359, 'learning_rate': 0.19609173219450998, 'epoch': 0.12} + 12%|█▏ | 61/520 [04:01<29:04, 3.80s/it] 12%|█▏ | 62/520 [04:05<29:00, 3.80s/it] {'loss': 1.3423, 'grad_norm': 0.0015848391245289425, 'learning_rate': 0.19591730613269878, 'epoch': 0.12} + 12%|█▏ | 62/520 [04:05<29:00, 3.80s/it] 12%|█▏ | 63/520 [04:09<28:57, 3.80s/it] {'loss': 1.3256, 'grad_norm': 0.0015114763173368366, 'learning_rate': 0.19573915328674182, 'epoch': 0.12} + 12%|█▏ | 63/520 [04:09<28:57, 3.80s/it] 12%|█▏ | 64/520 [04:13<28:53, 3.80s/it] {'loss': 1.3611, 'grad_norm': 0.0015548393511548835, 'learning_rate': 0.1955572805786141, 'epoch': 0.12} + 12%|█▏ | 64/520 [04:13<28:53, 3.80s/it] 12%|█▎ | 65/520 [04:16<28:49, 3.80s/it] {'loss': 1.3705, 'grad_norm': 0.001819450703473056, 'learning_rate': 0.1953716950748227, 'epoch': 0.12} + 12%|█▎ | 65/520 [04:16<28:49, 3.80s/it] 13%|█▎ | 66/520 [04:20<28:47, 3.80s/it] {'loss': 1.3224, 'grad_norm': 0.0013886777619905935, 'learning_rate': 0.19518240398613226, 'epoch': 0.13} + 13%|█▎ | 66/520 [04:20<28:47, 3.80s/it] 13%|█▎ | 67/520 [04:24<28:49, 3.82s/it] {'loss': 1.2256, 'grad_norm': 0.0014892620350957669, 'learning_rate': 0.1949894146672846, 'epoch': 0.13} + 13%|█▎ | 67/520 [04:24<28:49, 3.82s/it] 13%|█▎ | 68/520 [04:28<28:52, 3.83s/it] {'loss': 1.2863, 'grad_norm': 0.0014554690074641873, 'learning_rate': 0.1947927346167132, 'epoch': 0.13} + 13%|█▎ | 68/520 [04:28<28:52, 3.83s/it] 13%|█▎ | 69/520 [04:32<28:42, 3.82s/it] {'loss': 1.2643, 'grad_norm': 0.0015571765671132038, 'learning_rate': 0.1945923714762516, 'epoch': 0.13} + 13%|█▎ | 69/520 [04:32<28:42, 3.82s/it] 13%|█▎ | 70/520 [04:35<28:33, 3.81s/it] {'loss': 1.3059, 'grad_norm': 0.0016002360501239563, 'learning_rate': 0.19438833303083677, 'epoch': 0.13} + 13%|█▎ | 70/520 [04:35<28:33, 3.81s/it] 14%|█▎ | 71/520 [04:39<28:27, 3.80s/it] {'loss': 1.2452, 'grad_norm': 0.0014689262411406983, 'learning_rate': 0.19418062720820636, 'epoch': 0.14} + 14%|█▎ | 71/520 [04:39<28:27, 3.80s/it] 14%|█▍ | 72/520 [04:43<28:19, 3.79s/it] {'loss': 1.3915, 'grad_norm': 0.0015561007018707936, 'learning_rate': 0.19396926207859086, 'epoch': 0.14} + 14%|█▍ | 72/520 [04:43<28:19, 3.79s/it] 14%|█▍ | 73/520 [04:47<28:18, 3.80s/it] {'loss': 1.2257, 'grad_norm': 0.0014963579431217348, 'learning_rate': 0.19375424585439993, 'epoch': 0.14} + 14%|█▍ | 73/520 [04:47<28:18, 3.80s/it] 14%|█▍ | 74/520 [04:51<28:16, 3.80s/it] {'loss': 1.3317, 'grad_norm': 0.0015902097919539622, 'learning_rate': 0.1935355868899034, 'epoch': 0.14} + 14%|█▍ | 74/520 [04:51<28:16, 3.80s/it] 14%|█▍ | 75/520 [04:54<28:09, 3.80s/it] {'loss': 1.2448, 'grad_norm': 0.001380893979409509, 'learning_rate': 0.19331329368090666, 'epoch': 0.14} + 14%|█▍ | 75/520 [04:54<28:09, 3.80s/it] 15%|█▍ | 76/520 [04:58<28:05, 3.80s/it] {'loss': 1.3986, 'grad_norm': 0.001360646933515734, 'learning_rate': 0.19308737486442043, 'epoch': 0.15} + 15%|█▍ | 76/520 [04:58<28:05, 3.80s/it] 15%|█▍ | 77/520 [05:02<27:59, 3.79s/it] {'loss': 1.1706, 'grad_norm': 0.0016461857020297458, 'learning_rate': 0.19285783921832536, 'epoch': 0.15} + 15%|█▍ | 77/520 [05:02<27:59, 3.79s/it] 15%|█▌ | 78/520 [05:06<27:57, 3.80s/it] {'loss': 1.283, 'grad_norm': 0.0015502924760223307, 'learning_rate': 0.19262469566103088, 'epoch': 0.15} + 15%|█▌ | 78/520 [05:06<27:57, 3.80s/it] 15%|█▌ | 79/520 [05:10<27:52, 3.79s/it] {'loss': 1.2716, 'grad_norm': 0.0014871570838551017, 'learning_rate': 0.19238795325112867, 'epoch': 0.15} + 15%|█▌ | 79/520 [05:10<27:52, 3.79s/it] 15%|█▌ | 80/520 [05:13<27:43, 3.78s/it] {'loss': 1.3885, 'grad_norm': 0.0015418323931605746, 'learning_rate': 0.19214762118704076, 'epoch': 0.15} + 15%|█▌ | 80/520 [05:13<27:43, 3.78s/it] 16%|█▌ | 81/520 [05:17<27:38, 3.78s/it] {'loss': 1.4238, 'grad_norm': 0.0019432163442023134, 'learning_rate': 0.19190370880666208, 'epoch': 0.16} + 16%|█▌ | 81/520 [05:17<27:38, 3.78s/it] 16%|█▌ | 82/520 [05:21<27:40, 3.79s/it] {'loss': 1.3458, 'grad_norm': 0.0014845295783221415, 'learning_rate': 0.19165622558699763, 'epoch': 0.16} + 16%|█▌ | 82/520 [05:21<27:40, 3.79s/it] 16%|█▌ | 83/520 [05:25<27:33, 3.78s/it] {'loss': 1.3528, 'grad_norm': 0.0015534430032330987, 'learning_rate': 0.19140518114379435, 'epoch': 0.16} + 16%|█▌ | 83/520 [05:25<27:33, 3.78s/it] 16%|█▌ | 84/520 [05:28<27:26, 3.78s/it] {'loss': 1.367, 'grad_norm': 0.001532948151080889, 'learning_rate': 0.19115058523116735, 'epoch': 0.16} + 16%|█▌ | 84/520 [05:28<27:26, 3.78s/it] 16%|█▋ | 85/520 [05:32<27:23, 3.78s/it] {'loss': 1.4088, 'grad_norm': 0.0014585297294413475, 'learning_rate': 0.1908924477412211, 'epoch': 0.16} + 16%|█▋ | 85/520 [05:32<27:23, 3.78s/it] 17%|█▋ | 86/520 [05:36<27:22, 3.79s/it] {'loss': 1.4017, 'grad_norm': 0.0014639426532829752, 'learning_rate': 0.19063077870366502, 'epoch': 0.17} + 17%|█▋ | 86/520 [05:36<27:22, 3.79s/it] 17%|█▋ | 87/520 [05:40<27:16, 3.78s/it] {'loss': 1.342, 'grad_norm': 0.0014867762216193982, 'learning_rate': 0.1903655882854237, 'epoch': 0.17} + 17%|█▋ | 87/520 [05:40<27:16, 3.78s/it] 17%|█▋ | 88/520 [05:44<27:15, 3.79s/it] {'loss': 1.2798, 'grad_norm': 0.0011925749949511894, 'learning_rate': 0.19009688679024192, 'epoch': 0.17} + 17%|█▋ | 88/520 [05:44<27:15, 3.79s/it] 17%|█▋ | 89/520 [05:47<27:07, 3.78s/it] {'loss': 1.3571, 'grad_norm': 0.0014868486535573608, 'learning_rate': 0.18982468465828442, 'epoch': 0.17} + 17%|█▋ | 89/520 [05:47<27:07, 3.78s/it] 17%|█▋ | 90/520 [05:51<27:04, 3.78s/it] {'loss': 1.2918, 'grad_norm': 0.001428694194383474, 'learning_rate': 0.1895489924657301, 'epoch': 0.17} + 17%|█▋ | 90/520 [05:51<27:04, 3.78s/it] 18%|█▊ | 91/520 [05:55<27:01, 3.78s/it] {'loss': 1.3635, 'grad_norm': 0.0013611530528144552, 'learning_rate': 0.18926982092436118, 'epoch': 0.17} + 18%|█▊ | 91/520 [05:55<27:01, 3.78s/it] 18%|█▊ | 92/520 [05:59<26:59, 3.78s/it] {'loss': 1.2946, 'grad_norm': 0.001499161215054262, 'learning_rate': 0.18898718088114688, 'epoch': 0.18} + 18%|█▊ | 92/520 [05:59<26:59, 3.78s/it] 18%|█▊ | 93/520 [06:02<26:54, 3.78s/it] {'loss': 1.3145, 'grad_norm': 0.001473079643159267, 'learning_rate': 0.18870108331782218, 'epoch': 0.18} + 18%|█▊ | 93/520 [06:02<26:54, 3.78s/it] 18%|█▊ | 94/520 [06:06<26:58, 3.80s/it] {'loss': 1.3912, 'grad_norm': 0.0014070332379454821, 'learning_rate': 0.18841153935046098, 'epoch': 0.18} + 18%|█▊ | 94/520 [06:06<26:58, 3.80s/it] 18%|█▊ | 95/520 [06:10<26:51, 3.79s/it] {'loss': 1.2883, 'grad_norm': 0.0016526684450726295, 'learning_rate': 0.18811856022904425, 'epoch': 0.18} + 18%|█▊ | 95/520 [06:10<26:51, 3.79s/it] 18%|█▊ | 96/520 [06:14<26:51, 3.80s/it] {'loss': 1.3011, 'grad_norm': 0.0012540371762355938, 'learning_rate': 0.18782215733702287, 'epoch': 0.18} + 18%|█▊ | 96/520 [06:14<26:51, 3.80s/it] 19%|█▊ | 97/520 [06:18<26:45, 3.80s/it] {'loss': 1.267, 'grad_norm': 0.0015351299455229612, 'learning_rate': 0.18752234219087538, 'epoch': 0.19} + 19%|█▊ | 97/520 [06:18<26:45, 3.80s/it] 19%|█▉ | 98/520 [06:21<26:42, 3.80s/it] {'loss': 1.2682, 'grad_norm': 0.001198572085974785, 'learning_rate': 0.18721912643966054, 'epoch': 0.19} + 19%|█▉ | 98/520 [06:21<26:42, 3.80s/it] 19%|█▉ | 99/520 [06:25<26:40, 3.80s/it] {'loss': 1.2805, 'grad_norm': 0.0014410120821238585, 'learning_rate': 0.18691252186456464, 'epoch': 0.19} + 19%|█▉ | 99/520 [06:25<26:40, 3.80s/it] 19%|█▉ | 100/520 [06:29<26:33, 3.79s/it] {'loss': 1.275, 'grad_norm': 0.0013389067138681362, 'learning_rate': 0.1866025403784439, 'epoch': 0.19} + 19%|█▉ | 100/520 [06:29<26:33, 3.79s/it] 19%|█▉ | 101/520 [06:33<26:27, 3.79s/it] {'loss': 1.2852, 'grad_norm': 0.00141166030106168, 'learning_rate': 0.18628919402536132, 'epoch': 0.19} + 19%|█▉ | 101/520 [06:33<26:27, 3.79s/it] 20%|█▉ | 102/520 [06:37<26:23, 3.79s/it] {'loss': 1.2936, 'grad_norm': 0.0014914758691351462, 'learning_rate': 0.18597249498011903, 'epoch': 0.2} + 20%|█▉ | 102/520 [06:37<26:23, 3.79s/it] 20%|█▉ | 103/520 [06:40<26:15, 3.78s/it] {'loss': 1.2241, 'grad_norm': 0.001302341609572692, 'learning_rate': 0.18565245554778517, 'epoch': 0.2} + 20%|█▉ | 103/520 [06:40<26:15, 3.78s/it] 20%|██ | 104/520 [06:44<25:56, 3.74s/it] {'loss': 1.2984, 'grad_norm': 0.0013905332003873233, 'learning_rate': 0.18532908816321558, 'epoch': 0.2} + 20%|██ | 104/520 [06:44<25:56, 3.74s/it] 20%|██ | 105/520 [06:48<25:47, 3.73s/it] {'loss': 1.2905, 'grad_norm': 0.0012987430386871034, 'learning_rate': 0.18500240539057092, 'epoch': 0.2} + 20%|██ | 105/520 [06:48<25:47, 3.73s/it] 20%|██ | 106/520 [06:52<25:49, 3.74s/it] {'loss': 1.3052, 'grad_norm': 0.001271987415628399, 'learning_rate': 0.18467241992282843, 'epoch': 0.2} + 20%|██ | 106/520 [06:52<25:49, 3.74s/it] 21%|██ | 107/520 [06:55<25:38, 3.72s/it] {'loss': 1.2749, 'grad_norm': 0.0013425305615233129, 'learning_rate': 0.18433914458128858, 'epoch': 0.21} + 21%|██ | 107/520 [06:55<25:38, 3.72s/it] 21%|██ | 108/520 [06:59<25:42, 3.74s/it] {'loss': 1.2529, 'grad_norm': 0.0013991724778647768, 'learning_rate': 0.18400259231507718, 'epoch': 0.21} + 21%|██ | 108/520 [06:59<25:42, 3.74s/it] 21%|██ | 109/520 [07:03<25:57, 3.79s/it] {'loss': 1.2594, 'grad_norm': 0.001232047915628087, 'learning_rate': 0.18366277620064198, 'epoch': 0.21} + 21%|██ | 109/520 [07:03<25:57, 3.79s/it] 21%|██ | 110/520 [07:07<26:00, 3.81s/it] {'loss': 1.4393, 'grad_norm': 0.0014272068727302926, 'learning_rate': 0.1833197094412449, 'epoch': 0.21} + 21%|██ | 110/520 [07:07<26:00, 3.81s/it] 21%|██▏ | 111/520 [07:11<26:03, 3.82s/it] {'loss': 1.4334, 'grad_norm': 0.0014843337742714922, 'learning_rate': 0.18297340536644877, 'epoch': 0.21} + 21%|██▏ | 111/520 [07:11<26:03, 3.82s/it] 22%|██▏ | 112/520 [07:14<26:00, 3.82s/it] {'loss': 1.3125, 'grad_norm': 0.0013016240512886033, 'learning_rate': 0.1826238774315995, 'epoch': 0.22} + 22%|██▏ | 112/520 [07:14<26:00, 3.82s/it] 22%|██▏ | 113/520 [07:18<26:00, 3.83s/it] {'loss': 1.204, 'grad_norm': 0.0012568208999897419, 'learning_rate': 0.18227113921730334, 'epoch': 0.22} + 22%|██▏ | 113/520 [07:18<26:00, 3.83s/it] 22%|██▏ | 114/520 [07:22<25:57, 3.84s/it] {'loss': 1.299, 'grad_norm': 0.0012757185983215732, 'learning_rate': 0.1819152044288992, 'epoch': 0.22} + 22%|██▏ | 114/520 [07:22<25:57, 3.84s/it] 22%|██▏ | 115/520 [07:26<26:00, 3.85s/it] {'loss': 1.4069, 'grad_norm': 0.0013010237979865357, 'learning_rate': 0.18155608689592603, 'epoch': 0.22} + 22%|██▏ | 115/520 [07:26<26:00, 3.85s/it] 22%|██▏ | 116/520 [07:30<25:46, 3.83s/it] {'loss': 1.4103, 'grad_norm': 0.0012537763737873844, 'learning_rate': 0.18119380057158568, 'epoch': 0.22} + 22%|██▏ | 116/520 [07:30<25:46, 3.83s/it] 22%|██▎ | 117/520 [07:33<25:22, 3.78s/it] {'loss': 1.3815, 'grad_norm': 0.001392187705783452, 'learning_rate': 0.18082835953220056, 'epoch': 0.23} + 22%|██▎ | 117/520 [07:33<25:22, 3.78s/it] 23%|██▎ | 118/520 [07:37<25:07, 3.75s/it] {'loss': 1.2871, 'grad_norm': 0.0012273794526999817, 'learning_rate': 0.18045977797666685, 'epoch': 0.23} + 23%|██▎ | 118/520 [07:37<25:07, 3.75s/it] 23%|██▎ | 119/520 [07:41<24:55, 3.73s/it] {'loss': 1.2437, 'grad_norm': 0.0013039337427061622, 'learning_rate': 0.1800880702259028, 'epoch': 0.23} + 23%|██▎ | 119/520 [07:41<24:55, 3.73s/it] 23%|██▎ | 120/520 [07:44<24:45, 3.71s/it] {'loss': 1.253, 'grad_norm': 0.0014965209477694297, 'learning_rate': 0.17971325072229227, 'epoch': 0.23} + 23%|██▎ | 120/520 [07:44<24:45, 3.71s/it] 23%|██▎ | 121/520 [07:48<24:35, 3.70s/it] {'loss': 1.3145, 'grad_norm': 0.0014107985858582692, 'learning_rate': 0.17933533402912352, 'epoch': 0.23} + 23%|██▎ | 121/520 [07:48<24:35, 3.70s/it] 23%|██▎ | 122/520 [07:52<24:31, 3.70s/it] {'loss': 1.2197, 'grad_norm': 0.0012278271946581877, 'learning_rate': 0.17895433483002354, 'epoch': 0.23} + 23%|██▎ | 122/520 [07:52<24:31, 3.70s/it] 24%|██▎ | 123/520 [07:56<24:29, 3.70s/it] {'loss': 1.3476, 'grad_norm': 0.0013041742692246148, 'learning_rate': 0.17857026792838737, 'epoch': 0.24} + 24%|██▎ | 123/520 [07:56<24:29, 3.70s/it] 24%|██▍ | 124/520 [07:59<24:24, 3.70s/it] {'loss': 1.2819, 'grad_norm': 0.001422918652487946, 'learning_rate': 0.178183148246803, 'epoch': 0.24} + 24%|██▍ | 124/520 [07:59<24:24, 3.70s/it] 24%|██▍ | 125/520 [08:03<25:04, 3.81s/it] {'loss': 1.2747, 'grad_norm': 0.0013261227370817753, 'learning_rate': 0.1777929908264715, 'epoch': 0.24} + 24%|██▍ | 125/520 [08:03<25:04, 3.81s/it] 24%|██▍ | 126/520 [08:08<26:25, 4.02s/it] {'loss': 1.2722, 'grad_norm': 0.0011243438159655801, 'learning_rate': 0.17739981082662276, 'epoch': 0.24} + 24%|██▍ | 126/520 [08:08<26:25, 4.02s/it] 24%|██▍ | 127/520 [08:12<25:45, 3.93s/it] {'loss': 1.2556, 'grad_norm': 0.0014664875550561886, 'learning_rate': 0.1770036235239263, 'epoch': 0.24} + 24%|██▍ | 127/520 [08:12<25:45, 3.93s/it] 25%|██▍ | 128/520 [08:15<25:15, 3.87s/it] {'loss': 1.3014, 'grad_norm': 0.0013508487763252827, 'learning_rate': 0.1766044443118978, 'epoch': 0.25} + 25%|██▍ | 128/520 [08:15<25:15, 3.87s/it] 25%|██▍ | 129/520 [08:19<24:46, 3.80s/it] {'loss': 1.2595, 'grad_norm': 0.001205840269721774, 'learning_rate': 0.17620228870030108, 'epoch': 0.25} + 25%|██▍ | 129/520 [08:19<24:46, 3.80s/it] 25%|██▌ | 130/520 [08:23<24:27, 3.76s/it] {'loss': 1.2918, 'grad_norm': 0.0012141054196543627, 'learning_rate': 0.1757971723145453, 'epoch': 0.25} + 25%|██▌ | 130/520 [08:23<24:27, 3.76s/it] 25%|██▌ | 131/520 [08:26<24:12, 3.73s/it] {'loss': 1.242, 'grad_norm': 0.0011554028497591292, 'learning_rate': 0.175389110895078, 'epoch': 0.25} + 25%|██▌ | 131/520 [08:26<24:12, 3.73s/it] 25%|██▌ | 132/520 [08:30<24:09, 3.74s/it] {'loss': 1.3345, 'grad_norm': 0.0014061748072721886, 'learning_rate': 0.17497812029677343, 'epoch': 0.25} + 25%|██▌ | 132/520 [08:30<24:09, 3.74s/it] 26%|██▌ | 133/520 [08:34<24:25, 3.79s/it] {'loss': 1.2562, 'grad_norm': 0.0014431250921017717, 'learning_rate': 0.17456421648831655, 'epoch': 0.26} + 26%|██▌ | 133/520 [08:34<24:25, 3.79s/it] 26%|██▌ | 134/520 [08:38<24:34, 3.82s/it] {'loss': 1.3294, 'grad_norm': 0.0012898745205024164, 'learning_rate': 0.17414741555158267, 'epoch': 0.26} + 26%|██▌ | 134/520 [08:38<24:34, 3.82s/it] 26%|██▌ | 135/520 [08:42<24:44, 3.86s/it] {'loss': 1.3858, 'grad_norm': 0.0013648812244398728, 'learning_rate': 0.1737277336810124, 'epoch': 0.26} + 26%|██▌ | 135/520 [08:42<24:44, 3.86s/it] 26%|██▌ | 136/520 [08:46<24:47, 3.87s/it] {'loss': 1.3296, 'grad_norm': 0.0013602764057663696, 'learning_rate': 0.17330518718298263, 'epoch': 0.26} + 26%|██▌ | 136/520 [08:46<24:47, 3.87s/it] 26%|██▋ | 137/520 [08:50<24:46, 3.88s/it] {'loss': 1.246, 'grad_norm': 0.001521278826451884, 'learning_rate': 0.17287979247517285, 'epoch': 0.26} + 26%|██▋ | 137/520 [08:50<24:46, 3.88s/it] 27%|██▋ | 138/520 [08:53<24:45, 3.89s/it] {'loss': 1.2624, 'grad_norm': 0.0011988090149939266, 'learning_rate': 0.17245156608592727, 'epoch': 0.27} + 27%|██▋ | 138/520 [08:53<24:45, 3.89s/it] 27%|██▋ | 139/520 [08:57<24:48, 3.91s/it] {'loss': 1.1487, 'grad_norm': 0.0011569039772965544, 'learning_rate': 0.17202052465361267, 'epoch': 0.27} + 27%|██▋ | 139/520 [08:57<24:48, 3.91s/it] 27%|██▋ | 140/520 [09:01<24:48, 3.92s/it] {'loss': 1.2879, 'grad_norm': 0.0011285238808301104, 'learning_rate': 0.17158668492597184, 'epoch': 0.27} + 27%|██▋ | 140/520 [09:01<24:48, 3.92s/it] 27%|██▋ | 141/520 [09:05<24:48, 3.93s/it] {'loss': 1.3601, 'grad_norm': 0.0011632226632954147, 'learning_rate': 0.17115006375947303, 'epoch': 0.27} + 27%|██▋ | 141/520 [09:05<24:48, 3.93s/it] 27%|██▋ | 142/520 [09:09<24:45, 3.93s/it] {'loss': 1.2995, 'grad_norm': 0.0011376957080457741, 'learning_rate': 0.17071067811865476, 'epoch': 0.27} + 27%|██▋ | 142/520 [09:09<24:45, 3.93s/it] 28%|██▊ | 143/520 [09:13<24:40, 3.93s/it] {'loss': 1.2884, 'grad_norm': 0.0012815015594485886, 'learning_rate': 0.17026854507546693, 'epoch': 0.28} + 28%|██▊ | 143/520 [09:13<24:40, 3.93s/it] 28%|██▊ | 144/520 [09:17<24:36, 3.93s/it] {'loss': 1.2531, 'grad_norm': 0.001331564663380466, 'learning_rate': 0.1698236818086073, 'epoch': 0.28} + 28%|██▊ | 144/520 [09:17<24:36, 3.93s/it] 28%|██▊ | 145/520 [09:21<24:34, 3.93s/it] {'loss': 1.1863, 'grad_norm': 0.001135653180402688, 'learning_rate': 0.16937610560285418, 'epoch': 0.28} + 28%|██▊ | 145/520 [09:21<24:34, 3.93s/it] 28%|██▊ | 146/520 [09:25<24:27, 3.92s/it] {'loss': 1.3619, 'grad_norm': 0.0012715235718492584, 'learning_rate': 0.1689258338483947, 'epoch': 0.28} + 28%|██▊ | 146/520 [09:25<24:27, 3.92s/it] 28%|██▊ | 147/520 [09:29<24:22, 3.92s/it] {'loss': 1.2327, 'grad_norm': 0.0012456211918417527, 'learning_rate': 0.16847288404014937, 'epoch': 0.28} + 28%|██▊ | 147/520 [09:29<24:22, 3.92s/it] 28%|██▊ | 148/520 [09:33<24:16, 3.91s/it] {'loss': 1.2535, 'grad_norm': 0.001167174987747351, 'learning_rate': 0.16801727377709194, 'epoch': 0.28} + 28%|██▊ | 148/520 [09:33<24:16, 3.91s/it] 29%|██▊ | 149/520 [09:37<24:12, 3.92s/it] {'loss': 1.2001, 'grad_norm': 0.0012411184712057817, 'learning_rate': 0.16755902076156604, 'epoch': 0.29} + 29%|██▊ | 149/520 [09:37<24:12, 3.92s/it] 29%|██▉ | 150/520 [09:40<23:46, 3.86s/it] {'loss': 1.424, 'grad_norm': 0.0012123807582453277, 'learning_rate': 0.16709814279859703, 'epoch': 0.29} + 29%|██▉ | 150/520 [09:40<23:46, 3.86s/it] 29%|██▉ | 151/520 [09:44<23:23, 3.80s/it] {'loss': 1.2465, 'grad_norm': 0.0012470525671983111, 'learning_rate': 0.1666346577952004, 'epoch': 0.29} + 29%|██▉ | 151/520 [09:44<23:23, 3.80s/it] 29%|██▉ | 152/520 [09:48<23:02, 3.76s/it] {'loss': 1.2192, 'grad_norm': 0.0012876211381693124, 'learning_rate': 0.16616858375968596, 'epoch': 0.29} + 29%|██▉ | 152/520 [09:48<23:02, 3.76s/it] 29%|██▉ | 153/520 [09:51<22:47, 3.73s/it] {'loss': 1.2489, 'grad_norm': 0.001167928751213836, 'learning_rate': 0.16569993880095807, 'epoch': 0.29} + 29%|██▉ | 153/520 [09:51<22:47, 3.73s/it] 30%|██▉ | 154/520 [09:55<22:36, 3.71s/it] {'loss': 1.3357, 'grad_norm': 0.0011382839257638839, 'learning_rate': 0.16522874112781213, 'epoch': 0.3} + 30%|██▉ | 154/520 [09:55<22:36, 3.71s/it] 30%|██▉ | 155/520 [09:59<22:35, 3.71s/it] {'loss': 1.2534, 'grad_norm': 0.0012416998894325803, 'learning_rate': 0.16475500904822704, 'epoch': 0.3} + 30%|██▉ | 155/520 [09:59<22:35, 3.71s/it] 30%|███ | 156/520 [10:03<22:36, 3.73s/it] {'loss': 1.2782, 'grad_norm': 0.0013540220275166971, 'learning_rate': 0.16427876096865393, 'epoch': 0.3} + 30%|███ | 156/520 [10:03<22:36, 3.73s/it] 30%|███ | 157/520 [10:06<22:37, 3.74s/it] {'loss': 1.3467, 'grad_norm': 0.0011606810494358788, 'learning_rate': 0.16380001539330089, 'epoch': 0.3} + 30%|███ | 157/520 [10:06<22:37, 3.74s/it] 30%|███ | 158/520 [10:10<22:25, 3.72s/it] {'loss': 1.246, 'grad_norm': 0.0011984665324840161, 'learning_rate': 0.163318790923414, 'epoch': 0.3} + 30%|███ | 158/520 [10:10<22:25, 3.72s/it] 31%|███ | 159/520 [10:14<22:20, 3.71s/it] {'loss': 1.2925, 'grad_norm': 0.0011812382398161026, 'learning_rate': 0.16283510625655473, 'epoch': 0.31} + 31%|███ | 159/520 [10:14<22:20, 3.71s/it] 31%|███ | 160/520 [10:17<22:21, 3.73s/it] {'loss': 1.303, 'grad_norm': 0.0012472635485908624, 'learning_rate': 0.16234898018587338, 'epoch': 0.31} + 31%|███ | 160/520 [10:17<22:21, 3.73s/it] 31%|███ | 161/520 [10:21<22:11, 3.71s/it] {'loss': 1.2829, 'grad_norm': 0.0012003679022687197, 'learning_rate': 0.16186043159937882, 'epoch': 0.31} + 31%|███ | 161/520 [10:21<22:11, 3.71s/it] 31%|███ | 162/520 [10:25<22:04, 3.70s/it] {'loss': 1.2905, 'grad_norm': 0.0011670027513802051, 'learning_rate': 0.16136947947920477, 'epoch': 0.31} + 31%|███ | 162/520 [10:25<22:04, 3.70s/it] 31%|███▏ | 163/520 [10:29<22:06, 3.72s/it] {'loss': 1.1758, 'grad_norm': 0.0013131494754337464, 'learning_rate': 0.16087614290087207, 'epoch': 0.31} + 31%|███▏ | 163/520 [10:29<22:06, 3.72s/it] 32%|███▏ | 164/520 [10:32<22:14, 3.75s/it] {'loss': 1.1392, 'grad_norm': 0.001102667930604668, 'learning_rate': 0.16038044103254775, 'epoch': 0.32} + 32%|███▏ | 164/520 [10:32<22:14, 3.75s/it] 32%|███▏ | 165/520 [10:36<22:18, 3.77s/it] {'loss': 1.2823, 'grad_norm': 0.0011086640207349752, 'learning_rate': 0.15988239313430005, 'epoch': 0.32} + 32%|███▏ | 165/520 [10:36<22:18, 3.77s/it] 32%|███▏ | 166/520 [10:40<22:19, 3.79s/it] {'loss': 1.257, 'grad_norm': 0.0012493723049839122, 'learning_rate': 0.15938201855735015, 'epoch': 0.32} + 32%|███▏ | 166/520 [10:40<22:19, 3.79s/it] 32%|███▏ | 167/520 [10:44<22:25, 3.81s/it] {'loss': 1.2543, 'grad_norm': 0.0011369262167018914, 'learning_rate': 0.15887933674332047, 'epoch': 0.32} + 32%|███▏ | 167/520 [10:44<22:25, 3.81s/it] 32%|███▏ | 168/520 [10:48<22:21, 3.81s/it] {'loss': 1.1918, 'grad_norm': 0.0011229092989981351, 'learning_rate': 0.158374367223479, 'epoch': 0.32} + 32%|███▏ | 168/520 [10:48<22:21, 3.81s/it] 32%|███▎ | 169/520 [10:51<22:14, 3.80s/it] {'loss': 1.2703, 'grad_norm': 0.0011338544045625281, 'learning_rate': 0.1578671296179806, 'epoch': 0.33} + 32%|███▎ | 169/520 [10:51<22:14, 3.80s/it] 33%|███▎ | 170/520 [10:55<22:01, 3.77s/it] {'loss': 1.2357, 'grad_norm': 0.0010148849813224423, 'learning_rate': 0.15735764363510463, 'epoch': 0.33} + 33%|███▎ | 170/520 [10:55<22:01, 3.77s/it] 33%|███▎ | 171/520 [10:59<21:49, 3.75s/it] {'loss': 1.2054, 'grad_norm': 0.0012063251516556232, 'learning_rate': 0.15684592907048925, 'epoch': 0.33} + 33%|███▎ | 171/520 [10:59<21:49, 3.75s/it] 33%|███▎ | 172/520 [11:02<21:34, 3.72s/it] {'loss': 1.2818, 'grad_norm': 0.0011977758059364993, 'learning_rate': 0.1563320058063622, 'epoch': 0.33} + 33%|███▎ | 172/520 [11:02<21:34, 3.72s/it] 33%|███▎ | 173/520 [11:06<21:25, 3.70s/it] {'loss': 1.2148, 'grad_norm': 0.001122073871702228, 'learning_rate': 0.15581589381076844, 'epoch': 0.33} + 33%|███▎ | 173/520 [11:06<21:25, 3.70s/it] 33%|███▎ | 174/520 [11:10<21:18, 3.70s/it] {'loss': 1.2689, 'grad_norm': 0.001153090413809078, 'learning_rate': 0.15529761313679394, 'epoch': 0.33} + 33%|███▎ | 174/520 [11:10<21:18, 3.70s/it] 34%|███▎ | 175/520 [11:14<21:13, 3.69s/it] {'loss': 1.1896, 'grad_norm': 0.0011119810509938946, 'learning_rate': 0.15477718392178716, 'epoch': 0.34} + 34%|███▎ | 175/520 [11:14<21:13, 3.69s/it] 34%|███▍ | 176/520 [11:17<21:08, 3.69s/it] {'loss': 1.3172, 'grad_norm': 0.00118256707157614, 'learning_rate': 0.15425462638657594, 'epoch': 0.34} + 34%|███▍ | 176/520 [11:17<21:08, 3.69s/it] 34%|███▍ | 177/520 [11:21<21:02, 3.68s/it] {'loss': 1.1855, 'grad_norm': 0.001122190778934444, 'learning_rate': 0.1537299608346824, 'epoch': 0.34} + 34%|███▍ | 177/520 [11:21<21:02, 3.68s/it] 34%|███▍ | 178/520 [11:24<20:54, 3.67s/it] {'loss': 1.2527, 'grad_norm': 0.00119587263774624, 'learning_rate': 0.15320320765153367, 'epoch': 0.34} + 34%|███▍ | 178/520 [11:24<20:54, 3.67s/it] 34%|███▍ | 179/520 [11:28<20:51, 3.67s/it] {'loss': 1.3371, 'grad_norm': 0.001125546454876899, 'learning_rate': 0.15267438730367008, 'epoch': 0.34} + 34%|███▍ | 179/520 [11:28<20:51, 3.67s/it] 35%|███▍ | 180/520 [11:32<20:49, 3.67s/it] {'loss': 1.2487, 'grad_norm': 0.0011771073861991407, 'learning_rate': 0.1521435203379498, 'epoch': 0.35} + 35%|███▍ | 180/520 [11:32<20:49, 3.67s/it] 35%|███▍ | 181/520 [11:36<20:43, 3.67s/it] {'loss': 1.2293, 'grad_norm': 0.0010011087921680962, 'learning_rate': 0.15161062738075068, 'epoch': 0.35} + 35%|███▍ | 181/520 [11:36<20:43, 3.67s/it] 35%|███▌ | 182/520 [11:39<20:40, 3.67s/it] {'loss': 1.2392, 'grad_norm': 0.0011762006167511114, 'learning_rate': 0.1510757291371686, 'epoch': 0.35} + 35%|███▌ | 182/520 [11:39<20:40, 3.67s/it] 35%|███▌ | 183/520 [11:43<20:33, 3.66s/it] {'loss': 1.2649, 'grad_norm': 0.0011345267921900506, 'learning_rate': 0.1505388463902131, 'epoch': 0.35} + 35%|███▌ | 183/520 [11:43<20:33, 3.66s/it] 35%|███▌ | 184/520 [11:46<20:31, 3.66s/it] {'loss': 1.1882, 'grad_norm': 0.0011465298246595619, 'learning_rate': 0.15000000000000002, 'epoch': 0.35} + 35%|███▌ | 184/520 [11:46<20:31, 3.66s/it] 36%|███▌ | 185/520 [11:50<20:26, 3.66s/it] {'loss': 1.3387, 'grad_norm': 0.0011610137202976705, 'learning_rate': 0.14945921090294076, 'epoch': 0.36} + 36%|███▌ | 185/520 [11:50<20:26, 3.66s/it] 36%|███▌ | 186/520 [11:54<20:27, 3.67s/it] {'loss': 1.2128, 'grad_norm': 0.0012139973277319241, 'learning_rate': 0.14891650011092894, 'epoch': 0.36} + 36%|███▌ | 186/520 [11:54<20:27, 3.67s/it] 36%|███▌ | 187/520 [11:58<20:21, 3.67s/it] {'loss': 1.2139, 'grad_norm': 0.0012264067692319093, 'learning_rate': 0.14837188871052398, 'epoch': 0.36} + 36%|███▌ | 187/520 [11:58<20:21, 3.67s/it] 36%|███▌ | 188/520 [12:01<20:14, 3.66s/it] {'loss': 1.2969, 'grad_norm': 0.0011929298299462052, 'learning_rate': 0.14782539786213184, 'epoch': 0.36} + 36%|███▌ | 188/520 [12:01<20:14, 3.66s/it] 36%|███▋ | 189/520 [12:05<20:21, 3.69s/it] {'loss': 1.3014, 'grad_norm': 0.0010442760571392818, 'learning_rate': 0.1472770487991827, 'epoch': 0.36} + 36%|███▋ | 189/520 [12:05<20:21, 3.69s/it] 37%|███▋ | 190/520 [12:09<20:25, 3.71s/it] {'loss': 1.2276, 'grad_norm': 0.001194025071720382, 'learning_rate': 0.1467268628273062, 'epoch': 0.37} + 37%|███▋ | 190/520 [12:09<20:25, 3.71s/it] 37%|███▋ | 191/520 [12:12<20:17, 3.70s/it] {'loss': 1.182, 'grad_norm': 0.0010465831265157716, 'learning_rate': 0.1461748613235034, 'epoch': 0.37} + 37%|███▋ | 191/520 [12:12<20:17, 3.70s/it] 37%|███▋ | 192/520 [12:16<20:12, 3.70s/it] {'loss': 1.2585, 'grad_norm': 0.0010344690082189467, 'learning_rate': 0.1456210657353163, 'epoch': 0.37} + 37%|███▋ | 192/520 [12:16<20:12, 3.70s/it] 37%|███▋ | 193/520 [12:20<20:04, 3.68s/it] {'loss': 1.2392, 'grad_norm': 0.0012120021060262084, 'learning_rate': 0.14506549757999454, 'epoch': 0.37} + 37%|███▋ | 193/520 [12:20<20:04, 3.68s/it] 37%|███▋ | 194/520 [12:23<20:00, 3.68s/it] {'loss': 1.1313, 'grad_norm': 0.0010365598889512267, 'learning_rate': 0.14450817844365924, 'epoch': 0.37} + 37%|███▋ | 194/520 [12:23<20:00, 3.68s/it] 38%|███▊ | 195/520 [12:27<20:13, 3.73s/it] {'loss': 1.2826, 'grad_norm': 0.0011179749869516485, 'learning_rate': 0.1439491299804645, 'epoch': 0.38} + 38%|███▊ | 195/520 [12:27<20:13, 3.73s/it] 38%|███▊ | 196/520 [12:31<20:09, 3.73s/it] {'loss': 1.2519, 'grad_norm': 0.001240277926790677, 'learning_rate': 0.14338837391175582, 'epoch': 0.38} + 38%|███▊ | 196/520 [12:31<20:09, 3.73s/it] 38%|███▊ | 197/520 [12:35<20:02, 3.72s/it] {'loss': 1.2025, 'grad_norm': 0.0011435613162722586, 'learning_rate': 0.14282593202522628, 'epoch': 0.38} + 38%|███▊ | 197/520 [12:35<20:02, 3.72s/it] 38%|███▊ | 198/520 [12:38<20:03, 3.74s/it] {'loss': 1.2782, 'grad_norm': 0.0012064202570115346, 'learning_rate': 0.14226182617406996, 'epoch': 0.38} + 38%|███▊ | 198/520 [12:38<20:03, 3.74s/it] 38%|███▊ | 199/520 [12:42<20:15, 3.79s/it] {'loss': 1.1926, 'grad_norm': 0.001119354099690946, 'learning_rate': 0.14169607827613281, 'epoch': 0.38} + 38%|███▊ | 199/520 [12:42<20:15, 3.79s/it] 38%|███▊ | 200/520 [12:46<20:09, 3.78s/it] {'loss': 1.1905, 'grad_norm': 0.0011196387288295083, 'learning_rate': 0.14112871031306118, 'epoch': 0.38} + 38%|███▊ | 200/520 [12:46<20:09, 3.78s/it] 39%|███▊ | 201/520 [12:50<19:57, 3.75s/it] {'loss': 1.2122, 'grad_norm': 0.0009980832499247444, 'learning_rate': 0.1405597443294475, 'epoch': 0.39} + 39%|███▊ | 201/520 [12:50<19:57, 3.75s/it] 39%|███▉ | 202/520 [12:53<19:48, 3.74s/it] {'loss': 1.1988, 'grad_norm': 0.001131212127009666, 'learning_rate': 0.13998920243197407, 'epoch': 0.39} + 39%|███▉ | 202/520 [12:53<19:48, 3.74s/it] 39%|███▉ | 203/520 [12:57<19:40, 3.72s/it] {'loss': 1.2479, 'grad_norm': 0.0011624506299903784, 'learning_rate': 0.13941710678855396, 'epoch': 0.39} + 39%|███▉ | 203/520 [12:57<19:40, 3.72s/it] 39%|███▉ | 204/520 [13:01<19:35, 3.72s/it] {'loss': 1.2639, 'grad_norm': 0.0011291770571377612, 'learning_rate': 0.13884347962746948, 'epoch': 0.39} + 39%|███▉ | 204/520 [13:01<19:35, 3.72s/it] 39%|███▉ | 205/520 [13:05<19:30, 3.72s/it] {'loss': 1.2093, 'grad_norm': 0.0010974734554053214, 'learning_rate': 0.138268343236509, 'epoch': 0.39} + 39%|███▉ | 205/520 [13:05<19:30, 3.72s/it] 40%|███▉ | 206/520 [13:08<19:23, 3.71s/it] {'loss': 1.3025, 'grad_norm': 0.0010999802079036613, 'learning_rate': 0.13769171996210053, 'epoch': 0.4} + 40%|███▉ | 206/520 [13:08<19:23, 3.71s/it] 40%|███▉ | 207/520 [13:12<19:18, 3.70s/it] {'loss': 1.1853, 'grad_norm': 0.0010253224760570728, 'learning_rate': 0.1371136322084438, 'epoch': 0.4} + 40%|███▉ | 207/520 [13:12<19:18, 3.70s/it] 40%|████ | 208/520 [13:16<19:13, 3.70s/it] {'loss': 1.2878, 'grad_norm': 0.0012272548211114144, 'learning_rate': 0.13653410243663952, 'epoch': 0.4} + 40%|████ | 208/520 [13:16<19:13, 3.70s/it] 40%|████ | 209/520 [13:19<19:06, 3.69s/it] {'loss': 1.2028, 'grad_norm': 0.0010575895353714527, 'learning_rate': 0.13595315316381676, 'epoch': 0.4} + 40%|████ | 209/520 [13:19<19:06, 3.69s/it] 40%|████ | 210/520 [13:23<19:00, 3.68s/it] {'loss': 1.2693, 'grad_norm': 0.001122162421321105, 'learning_rate': 0.13537080696225814, 'epoch': 0.4} + 40%|████ | 210/520 [13:23<19:00, 3.68s/it] 41%|████ | 211/520 [13:27<18:57, 3.68s/it] {'loss': 1.2753, 'grad_norm': 0.0010453421067043768, 'learning_rate': 0.13478708645852272, 'epoch': 0.41} + 41%|████ | 211/520 [13:27<18:57, 3.68s/it] 41%|████ | 212/520 [13:30<18:51, 3.67s/it] {'loss': 1.2697, 'grad_norm': 0.0011102317018174977, 'learning_rate': 0.1342020143325669, 'epoch': 0.41} + 41%|████ | 212/520 [13:30<18:51, 3.67s/it] 41%|████ | 213/520 [13:34<18:53, 3.69s/it] {'loss': 1.2179, 'grad_norm': 0.0013670602155042275, 'learning_rate': 0.13361561331686309, 'epoch': 0.41} + 41%|████ | 213/520 [13:34<18:53, 3.69s/it] 41%|████ | 214/520 [13:38<18:51, 3.70s/it] {'loss': 1.2187, 'grad_norm': 0.0011660191694592922, 'learning_rate': 0.13302790619551672, 'epoch': 0.41} + 41%|████ | 214/520 [13:38<18:51, 3.70s/it] 41%|████▏ | 215/520 [13:41<18:51, 3.71s/it] {'loss': 1.1446, 'grad_norm': 0.0010491558845190509, 'learning_rate': 0.1324389158033807, 'epoch': 0.41} + 41%|████▏ | 215/520 [13:41<18:51, 3.71s/it] 42%|████▏ | 216/520 [13:45<19:01, 3.75s/it] {'loss': 1.1336, 'grad_norm': 0.0011275283014999912, 'learning_rate': 0.13184866502516845, 'epoch': 0.42} + 42%|████▏ | 216/520 [13:45<19:01, 3.75s/it] 42%|████▏ | 217/520 [13:49<19:09, 3.79s/it] {'loss': 1.2668, 'grad_norm': 0.001169685860175375, 'learning_rate': 0.13125717679456447, 'epoch': 0.42} + 42%|████▏ | 217/520 [13:49<19:09, 3.79s/it] 42%|████▏ | 218/520 [13:53<19:09, 3.81s/it] {'loss': 1.2444, 'grad_norm': 0.0012068723122087672, 'learning_rate': 0.13066447409333345, 'epoch': 0.42} + 42%|████▏ | 218/520 [13:53<19:09, 3.81s/it] 42%|████▏ | 219/520 [13:57<19:09, 3.82s/it] {'loss': 1.2452, 'grad_norm': 0.00100856470698655, 'learning_rate': 0.1300705799504273, 'epoch': 0.42} + 42%|████▏ | 219/520 [13:57<19:09, 3.82s/it] 42%|████▏ | 220/520 [14:01<19:04, 3.82s/it] {'loss': 1.1925, 'grad_norm': 0.001066507802932065, 'learning_rate': 0.12947551744109043, 'epoch': 0.42} + 42%|████▏ | 220/520 [14:01<19:04, 3.82s/it] 42%|████▎ | 221/520 [14:04<18:46, 3.77s/it] {'loss': 1.2452, 'grad_norm': 0.0010944275060142488, 'learning_rate': 0.128879309685963, 'epoch': 0.42} + 42%|████▎ | 221/520 [14:04<18:46, 3.77s/it] 43%|████▎ | 222/520 [14:08<18:32, 3.73s/it] {'loss': 1.188, 'grad_norm': 0.0010616545011702919, 'learning_rate': 0.12828197985018275, 'epoch': 0.43} + 43%|████▎ | 222/520 [14:08<18:32, 3.73s/it] 43%|████▎ | 223/520 [14:12<18:21, 3.71s/it] {'loss': 1.1786, 'grad_norm': 0.001037842347052185, 'learning_rate': 0.12768355114248495, 'epoch': 0.43} + 43%|████▎ | 223/520 [14:12<18:21, 3.71s/it] 43%|████▎ | 224/520 [14:15<18:13, 3.69s/it] {'loss': 1.2714, 'grad_norm': 0.0010588189497740582, 'learning_rate': 0.12708404681430052, 'epoch': 0.43} + 43%|████▎ | 224/520 [14:15<18:13, 3.69s/it] 43%|████▎ | 225/520 [14:19<18:06, 3.68s/it] {'loss': 1.1815, 'grad_norm': 0.0010870614045613983, 'learning_rate': 0.1264834901588527, 'epoch': 0.43} + 43%|████▎ | 225/520 [14:19<18:06, 3.68s/it] 43%|████▎ | 226/520 [14:23<17:58, 3.67s/it] {'loss': 1.2814, 'grad_norm': 0.001075478550908176, 'learning_rate': 0.12588190451025208, 'epoch': 0.43} + 43%|████▎ | 226/520 [14:23<17:58, 3.67s/it] 44%|████▎ | 227/520 [14:26<17:52, 3.66s/it] {'loss': 1.2708, 'grad_norm': 0.0010588312297888651, 'learning_rate': 0.12527931324258976, 'epoch': 0.44} + 44%|████▎ | 227/520 [14:26<17:52, 3.66s/it] 44%|████▍ | 228/520 [14:30<17:50, 3.67s/it] {'loss': 1.3013, 'grad_norm': 0.0011487032372810453, 'learning_rate': 0.12467573976902935, 'epoch': 0.44} + 44%|████▍ | 228/520 [14:30<17:50, 3.67s/it] 44%|████▍ | 229/520 [14:34<17:45, 3.66s/it] {'loss': 1.2421, 'grad_norm': 0.001003324904787233, 'learning_rate': 0.12407120754089732, 'epoch': 0.44} + 44%|████▍ | 229/520 [14:34<17:45, 3.66s/it] 44%|████▍ | 230/520 [14:37<17:43, 3.67s/it] {'loss': 1.1396, 'grad_norm': 0.001078963186005138, 'learning_rate': 0.12346574004677154, 'epoch': 0.44} + 44%|████▍ | 230/520 [14:37<17:43, 3.67s/it] 44%|████▍ | 231/520 [14:41<17:37, 3.66s/it] {'loss': 1.1977, 'grad_norm': 0.0009986064533620616, 'learning_rate': 0.12285936081156897, 'epoch': 0.44} + 44%|████▍ | 231/520 [14:41<17:37, 3.66s/it] 45%|████▍ | 232/520 [14:45<17:36, 3.67s/it] {'loss': 1.3406, 'grad_norm': 0.0011306929903330525, 'learning_rate': 0.12225209339563144, 'epoch': 0.45} + 45%|████▍ | 232/520 [14:45<17:36, 3.67s/it] 45%|████▍ | 233/520 [14:48<17:35, 3.68s/it] {'loss': 1.2267, 'grad_norm': 0.0011425912871629487, 'learning_rate': 0.12164396139381028, 'epoch': 0.45} + 45%|████▍ | 233/520 [14:48<17:35, 3.68s/it] 45%|████▌ | 234/520 [14:52<17:28, 3.67s/it] {'loss': 1.1487, 'grad_norm': 0.0011151094826325348, 'learning_rate': 0.12103498843454959, 'epoch': 0.45} + 45%|████▌ | 234/520 [14:52<17:28, 3.67s/it] 45%|████▌ | 235/520 [14:56<17:24, 3.66s/it] {'loss': 1.2028, 'grad_norm': 0.0010854655177638495, 'learning_rate': 0.12042519817896805, 'epoch': 0.45} + 45%|████▌ | 235/520 [14:56<17:24, 3.66s/it] 45%|████▌ | 236/520 [14:59<17:18, 3.66s/it] {'loss': 1.2851, 'grad_norm': 0.0010296475059085248, 'learning_rate': 0.11981461431993977, 'epoch': 0.45} + 45%|████▌ | 236/520 [14:59<17:18, 3.66s/it] 46%|████▌ | 237/520 [15:03<17:12, 3.65s/it] {'loss': 1.272, 'grad_norm': 0.001050845508823717, 'learning_rate': 0.11920326058117364, 'epoch': 0.46} + 46%|████▌ | 237/520 [15:03<17:12, 3.65s/it] 46%|████▌ | 238/520 [15:07<17:12, 3.66s/it] {'loss': 1.2113, 'grad_norm': 0.0010930037121908859, 'learning_rate': 0.11859116071629149, 'epoch': 0.46} + 46%|████▌ | 238/520 [15:07<17:12, 3.66s/it] 46%|████▌ | 239/520 [15:10<17:06, 3.65s/it] {'loss': 1.2876, 'grad_norm': 0.0011294599070943905, 'learning_rate': 0.11797833850790528, 'epoch': 0.46} + 46%|████▌ | 239/520 [15:10<17:06, 3.65s/it] 46%|████▌ | 240/520 [15:14<17:03, 3.66s/it] {'loss': 1.0968, 'grad_norm': 0.0009700435920070099, 'learning_rate': 0.11736481776669305, 'epoch': 0.46} + 46%|████▌ | 240/520 [15:14<17:03, 3.66s/it] 46%|████▋ | 241/520 [15:18<17:02, 3.67s/it] {'loss': 1.1769, 'grad_norm': 0.0010321276010095682, 'learning_rate': 0.11675062233047365, 'epoch': 0.46} + 46%|████▋ | 241/520 [15:18<17:02, 3.67s/it] 47%|████▋ | 242/520 [15:21<16:59, 3.67s/it] {'loss': 1.1928, 'grad_norm': 0.001019805242195925, 'learning_rate': 0.11613577606328068, 'epoch': 0.47} + 47%|████▋ | 242/520 [15:21<16:59, 3.67s/it] 47%|████▋ | 243/520 [15:25<17:04, 3.70s/it] {'loss': 1.1909, 'grad_norm': 0.0010917719771280328, 'learning_rate': 0.11552030285443515, 'epoch': 0.47} + 47%|████▋ | 243/520 [15:25<17:04, 3.70s/it] 47%|████▋ | 244/520 [15:29<17:09, 3.73s/it] {'loss': 1.3025, 'grad_norm': 0.0010811418203701753, 'learning_rate': 0.11490422661761744, 'epoch': 0.47} + 47%|████▋ | 244/520 [15:29<17:09, 3.73s/it] 47%|████▋ | 245/520 [15:33<17:11, 3.75s/it] {'loss': 1.1651, 'grad_norm': 0.0010188390522165385, 'learning_rate': 0.11428757128993801, 'epoch': 0.47} + 47%|████▋ | 245/520 [15:33<17:11, 3.75s/it] 47%|████▋ | 246/520 [15:36<17:12, 3.77s/it] {'loss': 1.322, 'grad_norm': 0.0011236672774095776, 'learning_rate': 0.11367036083100734, 'epoch': 0.47} + 47%|████▋ | 246/520 [15:36<17:12, 3.77s/it] 48%|████▊ | 247/520 [15:40<17:12, 3.78s/it] {'loss': 1.3408, 'grad_norm': 0.0011081306255991733, 'learning_rate': 0.11305261922200519, 'epoch': 0.47} + 48%|████▊ | 247/520 [15:40<17:12, 3.78s/it] 48%|████▊ | 248/520 [15:44<17:09, 3.78s/it] {'loss': 1.1699, 'grad_norm': 0.0011354647881871391, 'learning_rate': 0.11243437046474854, 'epoch': 0.48} + 48%|████▊ | 248/520 [15:44<17:09, 3.78s/it] 48%|████▊ | 249/520 [15:48<16:59, 3.76s/it] {'loss': 1.261, 'grad_norm': 0.001052545480001972, 'learning_rate': 0.1118156385807593, 'epoch': 0.48} + 48%|████▊ | 249/520 [15:48<16:59, 3.76s/it] 48%|████▊ | 250/520 [15:51<16:50, 3.74s/it] {'loss': 1.1968, 'grad_norm': 0.0011056453118728454, 'learning_rate': 0.11119644761033079, 'epoch': 0.48} + 48%|████▊ | 250/520 [15:51<16:50, 3.74s/it] 48%|████▊ | 251/520 [15:55<16:44, 3.74s/it] {'loss': 1.2662, 'grad_norm': 0.0009987775150843258, 'learning_rate': 0.1105768216115938, 'epoch': 0.48} + 48%|████▊ | 251/520 [15:55<16:44, 3.74s/it] 48%|████▊ | 252/520 [15:59<16:40, 3.73s/it] {'loss': 1.2221, 'grad_norm': 0.0010243289266753351, 'learning_rate': 0.10995678465958168, 'epoch': 0.48} + 48%|████▊ | 252/520 [15:59<16:40, 3.73s/it] 49%|████▊ | 253/520 [16:03<16:33, 3.72s/it] {'loss': 1.26, 'grad_norm': 0.001153705811405822, 'learning_rate': 0.10933636084529506, 'epoch': 0.49} + 49%|████▊ | 253/520 [16:03<16:33, 3.72s/it] 49%|████▉ | 254/520 [16:06<16:38, 3.75s/it] {'loss': 1.1972, 'grad_norm': 0.001040564746951141, 'learning_rate': 0.10871557427476584, 'epoch': 0.49} + 49%|████▉ | 254/520 [16:06<16:38, 3.75s/it] 49%|████▉ | 255/520 [16:10<16:39, 3.77s/it] {'loss': 1.1999, 'grad_norm': 0.0011244375866314067, 'learning_rate': 0.10809444906812034, 'epoch': 0.49} + 49%|████▉ | 255/520 [16:10<16:39, 3.77s/it] 49%|████▉ | 256/520 [16:14<16:38, 3.78s/it] {'loss': 1.2562, 'grad_norm': 0.0011161650133955502, 'learning_rate': 0.10747300935864244, 'epoch': 0.49} + 49%|████▉ | 256/520 [16:14<16:38, 3.78s/it] 49%|████▉ | 257/520 [16:18<16:37, 3.79s/it] {'loss': 1.2396, 'grad_norm': 0.001131535178802793, 'learning_rate': 0.10685127929183567, 'epoch': 0.49} + 49%|████▉ | 257/520 [16:18<16:37, 3.79s/it] 50%|████▉ | 258/520 [16:22<16:36, 3.80s/it] {'loss': 1.2456, 'grad_norm': 0.0009724714397576124, 'learning_rate': 0.10622928302448523, 'epoch': 0.5} + 50%|████▉ | 258/520 [16:22<16:36, 3.80s/it] 50%|████▉ | 259/520 [16:25<16:24, 3.77s/it] {'loss': 1.3193, 'grad_norm': 0.0011954435781286938, 'learning_rate': 0.10560704472371918, 'epoch': 0.5} + 50%|████▉ | 259/520 [16:25<16:24, 3.77s/it] 50%|█████ | 260/520 [16:29<16:35, 3.83s/it] {'loss': 1.2715, 'grad_norm': 0.0009148554658119536, 'learning_rate': 0.10498458856606972, 'epoch': 0.5} + 50%|█████ | 260/520 [16:29<16:35, 3.83s/it] 50%|█████ | 261/520 [16:33<16:40, 3.86s/it] {'loss': 1.2119, 'grad_norm': 0.001031215708304372, 'learning_rate': 0.10436193873653361, 'epoch': 0.5} + 50%|█████ | 261/520 [16:33<16:40, 3.86s/it] 50%|█████ | 262/520 [16:37<16:40, 3.88s/it] {'loss': 1.1716, 'grad_norm': 0.001055402037698522, 'learning_rate': 0.10373911942763259, 'epoch': 0.5} + 50%|█████ | 262/520 [16:37<16:40, 3.88s/it] 51%|█████ | 263/520 [16:41<16:38, 3.88s/it] {'loss': 1.2336, 'grad_norm': 0.0010676596100922867, 'learning_rate': 0.10311615483847332, 'epoch': 0.51} + 51%|█████ | 263/520 [16:41<16:38, 3.88s/it] 51%|█████ | 264/520 [16:45<16:39, 3.90s/it] {'loss': 1.2668, 'grad_norm': 0.0010228394660938645, 'learning_rate': 0.1024930691738073, 'epoch': 0.51} + 51%|█████ | 264/520 [16:45<16:39, 3.90s/it] 51%|█████ | 265/520 [16:49<16:39, 3.92s/it] {'loss': 1.1818, 'grad_norm': 0.0011778212234898893, 'learning_rate': 0.10186988664309023, 'epoch': 0.51} + 51%|█████ | 265/520 [16:49<16:39, 3.92s/it] 51%|█████ | 266/520 [16:53<16:35, 3.92s/it] {'loss': 1.056, 'grad_norm': 0.0008979765600305425, 'learning_rate': 0.10124663145954152, 'epoch': 0.51} + 51%|█████ | 266/520 [16:53<16:35, 3.92s/it] 51%|█████▏ | 267/520 [16:57<16:34, 3.93s/it] {'loss': 1.1851, 'grad_norm': 0.0010250343902547263, 'learning_rate': 0.10062332783920336, 'epoch': 0.51} + 51%|█████▏ | 267/520 [16:57<16:34, 3.93s/it] 52%|█████▏ | 268/520 [17:01<16:33, 3.94s/it] {'loss': 1.3366, 'grad_norm': 0.0010981537855734992, 'learning_rate': 0.1, 'epoch': 0.52} + 52%|█████▏ | 268/520 [17:01<16:33, 3.94s/it] 52%|█████▏ | 269/520 [17:05<16:25, 3.93s/it] {'loss': 1.2864, 'grad_norm': 0.0010920182729976281, 'learning_rate': 0.09937667216079665, 'epoch': 0.52} + 52%|█████▏ | 269/520 [17:05<16:25, 3.93s/it] 52%|█████▏ | 270/520 [17:09<16:25, 3.94s/it] {'loss': 1.1633, 'grad_norm': 0.0010506889113737056, 'learning_rate': 0.0987533685404585, 'epoch': 0.52} + 52%|█████▏ | 270/520 [17:09<16:25, 3.94s/it] 52%|█████▏ | 271/520 [17:13<16:19, 3.93s/it] {'loss': 1.267, 'grad_norm': 0.001069317957345299, 'learning_rate': 0.0981301133569098, 'epoch': 0.52} + 52%|█████▏ | 271/520 [17:13<16:19, 3.93s/it] 52%|█████▏ | 272/520 [17:17<16:15, 3.94s/it] {'loss': 1.183, 'grad_norm': 0.0010505783255831862, 'learning_rate': 0.09750693082619273, 'epoch': 0.52} + 52%|█████▏ | 272/520 [17:17<16:15, 3.94s/it] 52%|█████▎ | 273/520 [17:20<16:10, 3.93s/it] {'loss': 1.2901, 'grad_norm': 0.0010316599750085898, 'learning_rate': 0.0968838451615267, 'epoch': 0.53} + 52%|█████▎ | 273/520 [17:20<16:10, 3.93s/it] 53%|█████▎ | 274/520 [17:24<16:04, 3.92s/it] {'loss': 1.2465, 'grad_norm': 0.0011031084379233634, 'learning_rate': 0.09626088057236745, 'epoch': 0.53} + 53%|█████▎ | 274/520 [17:24<16:04, 3.92s/it] 53%|█████▎ | 275/520 [17:28<16:00, 3.92s/it] {'loss': 1.1909, 'grad_norm': 0.001026947882107057, 'learning_rate': 0.09563806126346641, 'epoch': 0.53} + 53%|█████▎ | 275/520 [17:28<16:00, 3.92s/it] 53%|█████▎ | 276/520 [17:32<15:55, 3.92s/it] {'loss': 1.2504, 'grad_norm': 0.0010952618285787147, 'learning_rate': 0.09501541143393027, 'epoch': 0.53} + 53%|█████▎ | 276/520 [17:32<15:55, 3.92s/it] 53%|█████▎ | 277/520 [17:36<15:53, 3.92s/it] {'loss': 1.3041, 'grad_norm': 0.0010038483371936214, 'learning_rate': 0.09439295527628082, 'epoch': 0.53} + 53%|█████▎ | 277/520 [17:36<15:53, 3.92s/it] 53%|█████▎ | 278/520 [17:40<15:50, 3.93s/it] {'loss': 1.1472, 'grad_norm': 0.0009677759251536606, 'learning_rate': 0.0937707169755148, 'epoch': 0.53} + 53%|█████▎ | 278/520 [17:40<15:50, 3.93s/it] 54%|█████▎ | 279/520 [17:44<15:37, 3.89s/it] {'loss': 1.179, 'grad_norm': 0.0011047707990749281, 'learning_rate': 0.09314872070816434, 'epoch': 0.54} + 54%|█████▎ | 279/520 [17:44<15:37, 3.89s/it] 54%|█████▍ | 280/520 [17:48<15:20, 3.84s/it] {'loss': 1.1863, 'grad_norm': 0.0012417646080622987, 'learning_rate': 0.09252699064135758, 'epoch': 0.54} + 54%|█████▍ | 280/520 [17:48<15:20, 3.84s/it] 54%|█████▍ | 281/520 [17:51<15:03, 3.78s/it] {'loss': 1.29, 'grad_norm': 0.0011104638364854973, 'learning_rate': 0.09190555093187967, 'epoch': 0.54} + 54%|█████▍ | 281/520 [17:51<15:03, 3.78s/it] 54%|█████▍ | 282/520 [17:55<14:54, 3.76s/it] {'loss': 1.165, 'grad_norm': 0.0010051742341933094, 'learning_rate': 0.09128442572523418, 'epoch': 0.54} + 54%|█████▍ | 282/520 [17:55<14:54, 3.76s/it] 54%|█████▍ | 283/520 [17:59<14:45, 3.74s/it] {'loss': 1.3072, 'grad_norm': 0.0011592368693722275, 'learning_rate': 0.09066363915470495, 'epoch': 0.54} + 54%|█████▍ | 283/520 [17:59<14:45, 3.74s/it] 55%|█████▍ | 284/520 [18:02<14:37, 3.72s/it] {'loss': 1.1789, 'grad_norm': 0.0010798492191431324, 'learning_rate': 0.09004321534041836, 'epoch': 0.55} + 55%|█████▍ | 284/520 [18:02<14:37, 3.72s/it] 55%|█████▍ | 285/520 [18:06<14:30, 3.71s/it] {'loss': 1.1851, 'grad_norm': 0.0010599958486798174, 'learning_rate': 0.08942317838840624, 'epoch': 0.55} + 55%|█████▍ | 285/520 [18:06<14:30, 3.71s/it] 55%|█████▌ | 286/520 [18:10<14:27, 3.71s/it] {'loss': 1.0654, 'grad_norm': 0.0010256238349407544, 'learning_rate': 0.08880355238966922, 'epoch': 0.55} + 55%|█████▌ | 286/520 [18:10<14:27, 3.71s/it] 55%|█████▌ | 287/520 [18:13<14:22, 3.70s/it] {'loss': 1.2882, 'grad_norm': 0.0010374950690304773, 'learning_rate': 0.08818436141924073, 'epoch': 0.55} + 55%|█████▌ | 287/520 [18:13<14:22, 3.70s/it] 55%|█████▌ | 288/520 [18:17<14:17, 3.70s/it] {'loss': 1.3208, 'grad_norm': 0.0010124473666185167, 'learning_rate': 0.08756562953525152, 'epoch': 0.55} + 55%|█████▌ | 288/520 [18:17<14:17, 3.70s/it] 56%|█████▌ | 289/520 [18:21<14:14, 3.70s/it] {'loss': 1.1966, 'grad_norm': 0.000985451109781226, 'learning_rate': 0.08694738077799487, 'epoch': 0.56} + 56%|█████▌ | 289/520 [18:21<14:14, 3.70s/it] 56%|█████▌ | 290/520 [18:24<14:07, 3.69s/it] {'loss': 1.1208, 'grad_norm': 0.000963819788961858, 'learning_rate': 0.08632963916899268, 'epoch': 0.56} + 56%|█████▌ | 290/520 [18:24<14:07, 3.69s/it] 56%|█████▌ | 291/520 [18:28<14:05, 3.69s/it] {'loss': 1.1725, 'grad_norm': 0.0010286754683639138, 'learning_rate': 0.08571242871006202, 'epoch': 0.56} + 56%|█████▌ | 291/520 [18:28<14:05, 3.69s/it] 56%|█████▌ | 292/520 [18:32<13:59, 3.68s/it] {'loss': 1.2188, 'grad_norm': 0.0010338263829175035, 'learning_rate': 0.08509577338238256, 'epoch': 0.56} + 56%|█████▌ | 292/520 [18:32<13:59, 3.68s/it] 56%|█████▋ | 293/520 [18:35<13:53, 3.67s/it] {'loss': 1.1711, 'grad_norm': 0.0011106363770439253, 'learning_rate': 0.08447969714556484, 'epoch': 0.56} + 56%|█████▋ | 293/520 [18:35<13:53, 3.67s/it] 57%|█████▋ | 294/520 [18:39<13:50, 3.67s/it] {'loss': 1.1848, 'grad_norm': 0.0010710344907268008, 'learning_rate': 0.08386422393671933, 'epoch': 0.57} + 57%|█████▋ | 294/520 [18:39<13:50, 3.67s/it] 57%|█████▋ | 295/520 [18:43<13:46, 3.67s/it] {'loss': 1.2218, 'grad_norm': 0.0010364827822883927, 'learning_rate': 0.08324937766952638, 'epoch': 0.57} + 57%|█████▋ | 295/520 [18:43<13:46, 3.67s/it] 57%|█████▋ | 296/520 [18:46<13:40, 3.66s/it] {'loss': 1.1427, 'grad_norm': 0.0010645732934233022, 'learning_rate': 0.08263518223330697, 'epoch': 0.57} + 57%|█████▋ | 296/520 [18:46<13:40, 3.66s/it] 57%|█████▋ | 297/520 [18:50<13:36, 3.66s/it] {'loss': 1.2708, 'grad_norm': 0.001177736183061831, 'learning_rate': 0.08202166149209474, 'epoch': 0.57} + 57%|█████▋ | 297/520 [18:50<13:36, 3.66s/it] 57%|█████▋ | 298/520 [18:54<13:31, 3.66s/it] {'loss': 1.2356, 'grad_norm': 0.0009842202946525663, 'learning_rate': 0.08140883928370855, 'epoch': 0.57} + 57%|█████▋ | 298/520 [18:54<13:31, 3.66s/it] 57%|█████▊ | 299/520 [18:57<13:30, 3.67s/it] {'loss': 1.2574, 'grad_norm': 0.0009851001475346497, 'learning_rate': 0.0807967394188264, 'epoch': 0.57} + 57%|█████▊ | 299/520 [18:57<13:30, 3.67s/it] 58%|█████▊ | 300/520 [19:01<13:29, 3.68s/it] {'loss': 1.2809, 'grad_norm': 0.0010479761805175613, 'learning_rate': 0.08018538568006027, 'epoch': 0.58} + 58%|█████▊ | 300/520 [19:01<13:29, 3.68s/it] 58%|█████▊ | 301/520 [19:05<13:27, 3.69s/it] {'loss': 1.2625, 'grad_norm': 0.001038257594847717, 'learning_rate': 0.07957480182103199, 'epoch': 0.58} + 58%|█████▊ | 301/520 [19:05<13:27, 3.69s/it] 58%|█████▊ | 302/520 [19:09<13:23, 3.69s/it] {'loss': 1.2708, 'grad_norm': 0.0010233147396271596, 'learning_rate': 0.07896501156545044, 'epoch': 0.58} + 58%|█████▊ | 302/520 [19:09<13:23, 3.69s/it] 58%|█████▊ | 303/520 [19:12<13:18, 3.68s/it] {'loss': 1.1839, 'grad_norm': 0.0011432418860942985, 'learning_rate': 0.07835603860618973, 'epoch': 0.58} + 58%|█████▊ | 303/520 [19:12<13:18, 3.68s/it] 58%|█████▊ | 304/520 [19:16<13:19, 3.70s/it] {'loss': 1.1768, 'grad_norm': 0.0010901217639764998, 'learning_rate': 0.07774790660436857, 'epoch': 0.58} + 58%|█████▊ | 304/520 [19:16<13:19, 3.70s/it] 59%|█████▊ | 305/520 [19:20<13:13, 3.69s/it] {'loss': 1.2975, 'grad_norm': 0.0012154702581704093, 'learning_rate': 0.07714063918843106, 'epoch': 0.59} + 59%|█████▊ | 305/520 [19:20<13:13, 3.69s/it] 59%|█████▉ | 306/520 [19:23<13:07, 3.68s/it] {'loss': 1.2441, 'grad_norm': 0.001079224445859883, 'learning_rate': 0.0765342599532285, 'epoch': 0.59} + 59%|█████▉ | 306/520 [19:23<13:07, 3.68s/it] 59%|█████▉ | 307/520 [19:27<13:00, 3.67s/it] {'loss': 1.1769, 'grad_norm': 0.000990812133431772, 'learning_rate': 0.07592879245910272, 'epoch': 0.59} + 59%|█████▉ | 307/520 [19:27<13:00, 3.67s/it] 59%|█████▉ | 308/520 [19:31<13:06, 3.71s/it] {'loss': 1.2925, 'grad_norm': 0.0010097938374514052, 'learning_rate': 0.07532426023097064, 'epoch': 0.59} + 59%|█████▉ | 308/520 [19:31<13:06, 3.71s/it] 59%|█████▉ | 309/520 [19:35<13:32, 3.85s/it] {'loss': 1.1792, 'grad_norm': 0.0009826679875321103, 'learning_rate': 0.07472068675741024, 'epoch': 0.59} + 59%|█████▉ | 309/520 [19:35<13:32, 3.85s/it] 60%|█████▉ | 310/520 [19:39<13:17, 3.80s/it] {'loss': 1.1537, 'grad_norm': 0.001047679937873588, 'learning_rate': 0.07411809548974792, 'epoch': 0.6} + 60%|█████▉ | 310/520 [19:39<13:17, 3.80s/it] 60%|█████▉ | 311/520 [19:42<13:04, 3.76s/it] {'loss': 1.1357, 'grad_norm': 0.0010280136793863882, 'learning_rate': 0.07351650984114727, 'epoch': 0.6} + 60%|█████▉ | 311/520 [19:42<13:04, 3.76s/it] 60%|██████ | 312/520 [19:46<12:57, 3.74s/it] {'loss': 1.1326, 'grad_norm': 0.0010755759275900714, 'learning_rate': 0.0729159531856995, 'epoch': 0.6} + 60%|██████ | 312/520 [19:46<12:57, 3.74s/it] 60%|██████ | 313/520 [19:50<12:47, 3.71s/it] {'loss': 1.1145, 'grad_norm': 0.0009322021872082693, 'learning_rate': 0.07231644885751508, 'epoch': 0.6} + 60%|██████ | 313/520 [19:50<12:47, 3.71s/it] 60%|██████ | 314/520 [19:54<13:11, 3.84s/it] {'loss': 1.1501, 'grad_norm': 0.000982674086079517, 'learning_rate': 0.07171802014981725, 'epoch': 0.6} + 60%|██████ | 314/520 [19:54<13:11, 3.84s/it] 61%|██████ | 315/520 [19:57<12:58, 3.80s/it] {'loss': 1.2267, 'grad_norm': 0.0011333720860271029, 'learning_rate': 0.07112069031403703, 'epoch': 0.61} + 61%|██████ | 315/520 [19:57<12:58, 3.80s/it] 61%|██████ | 316/520 [20:01<13:12, 3.89s/it] {'loss': 1.125, 'grad_norm': 0.0010482032026221772, 'learning_rate': 0.07052448255890957, 'epoch': 0.61} + 61%|██████ | 316/520 [20:01<13:12, 3.89s/it] 61%|██████ | 317/520 [20:05<12:57, 3.83s/it] {'loss': 1.1427, 'grad_norm': 0.0009511609210682086, 'learning_rate': 0.0699294200495727, 'epoch': 0.61} + 61%|██████ | 317/520 [20:05<12:57, 3.83s/it] 61%|██████ | 318/520 [20:09<12:44, 3.78s/it] {'loss': 1.2525, 'grad_norm': 0.0011042210115628562, 'learning_rate': 0.06933552590666658, 'epoch': 0.61} + 61%|██████ | 318/520 [20:09<12:44, 3.78s/it] 61%|██████▏ | 319/520 [20:13<12:57, 3.87s/it] {'loss': 1.132, 'grad_norm': 0.0009775740444587936, 'learning_rate': 0.06874282320543557, 'epoch': 0.61} + 61%|██████▏ | 319/520 [20:13<12:57, 3.87s/it] 62%|██████▏ | 320/520 [20:17<12:41, 3.81s/it] {'loss': 1.0833, 'grad_norm': 0.0010021786783135696, 'learning_rate': 0.06815133497483157, 'epoch': 0.62} + 62%|██████▏ | 320/520 [20:17<12:41, 3.81s/it] 62%|██████▏ | 321/520 [20:20<12:29, 3.76s/it] {'loss': 1.274, 'grad_norm': 0.0010169414552449337, 'learning_rate': 0.06756108419661931, 'epoch': 0.62} + 62%|██████▏ | 321/520 [20:20<12:29, 3.76s/it] 62%|██████▏ | 322/520 [20:24<12:18, 3.73s/it] {'loss': 1.1239, 'grad_norm': 0.0009989068417784732, 'learning_rate': 0.06697209380448332, 'epoch': 0.62} + 62%|██████▏ | 322/520 [20:24<12:18, 3.73s/it] 62%|██████▏ | 323/520 [20:28<12:09, 3.70s/it] {'loss': 1.1853, 'grad_norm': 0.0010142186336248545, 'learning_rate': 0.06638438668313694, 'epoch': 0.62} + 62%|██████▏ | 323/520 [20:28<12:09, 3.70s/it] 62%|██████▏ | 324/520 [20:31<12:02, 3.69s/it] {'loss': 1.2153, 'grad_norm': 0.0010051315805766948, 'learning_rate': 0.06579798566743314, 'epoch': 0.62} + 62%|██████▏ | 324/520 [20:31<12:02, 3.69s/it] 62%|██████▎ | 325/520 [20:35<11:56, 3.68s/it] {'loss': 1.2202, 'grad_norm': 0.0011000519054201864, 'learning_rate': 0.06521291354147728, 'epoch': 0.62} + 62%|██████▎ | 325/520 [20:35<11:56, 3.68s/it] 63%|██████▎ | 326/520 [20:39<11:54, 3.68s/it] {'loss': 1.2074, 'grad_norm': 0.0010974231506277858, 'learning_rate': 0.06462919303774187, 'epoch': 0.63} + 63%|██████▎ | 326/520 [20:39<11:54, 3.68s/it] 63%|██████▎ | 327/520 [20:42<11:48, 3.67s/it] {'loss': 1.2377, 'grad_norm': 0.001071040211196713, 'learning_rate': 0.06404684683618325, 'epoch': 0.63} + 63%|██████▎ | 327/520 [20:42<11:48, 3.67s/it] 63%|██████▎ | 328/520 [20:46<11:50, 3.70s/it] {'loss': 1.2603, 'grad_norm': 0.0010920170966743842, 'learning_rate': 0.0634658975633605, 'epoch': 0.63} + 63%|██████▎ | 328/520 [20:46<11:50, 3.70s/it] 63%|██████▎ | 329/520 [20:50<11:46, 3.70s/it] {'loss': 1.1372, 'grad_norm': 0.0008907801164559475, 'learning_rate': 0.06288636779155621, 'epoch': 0.63} + 63%|██████▎ | 329/520 [20:50<11:46, 3.70s/it] 63%|██████▎ | 330/520 [20:53<11:40, 3.68s/it] {'loss': 1.2095, 'grad_norm': 0.0009502215597224936, 'learning_rate': 0.06230828003789948, 'epoch': 0.63} + 63%|██████▎ | 330/520 [20:53<11:40, 3.68s/it] 64%|██████▎ | 331/520 [20:57<11:37, 3.69s/it] {'loss': 1.1676, 'grad_norm': 0.0010485007205664992, 'learning_rate': 0.06173165676349103, 'epoch': 0.64} + 64%|██████▎ | 331/520 [20:57<11:37, 3.69s/it] 64%|██████▍ | 332/520 [21:01<11:32, 3.68s/it] {'loss': 1.2554, 'grad_norm': 0.0009910475355934254, 'learning_rate': 0.06115652037253053, 'epoch': 0.64} + 64%|██████▍ | 332/520 [21:01<11:32, 3.68s/it] 64%|██████▍ | 333/520 [21:04<11:29, 3.69s/it] {'loss': 1.3057, 'grad_norm': 0.001075185094479071, 'learning_rate': 0.06058289321144608, 'epoch': 0.64} + 64%|██████▍ | 333/520 [21:04<11:29, 3.69s/it] 64%|██████▍ | 334/520 [21:08<11:25, 3.69s/it] {'loss': 1.2167, 'grad_norm': 0.0010780505208710343, 'learning_rate': 0.06001079756802592, 'epoch': 0.64} + 64%|██████▍ | 334/520 [21:08<11:25, 3.69s/it] 64%|██████▍ | 335/520 [21:12<11:19, 3.67s/it] {'loss': 1.2183, 'grad_norm': 0.0009755008167540266, 'learning_rate': 0.059440255670552514, 'epoch': 0.64} + 64%|██████▍ | 335/520 [21:12<11:19, 3.67s/it] 65%|██████▍ | 336/520 [21:15<11:15, 3.67s/it] {'loss': 1.1143, 'grad_norm': 0.001101429449365093, 'learning_rate': 0.05887128968693887, 'epoch': 0.65} + 65%|██████▍ | 336/520 [21:15<11:15, 3.67s/it] 65%|██████▍ | 337/520 [21:19<11:11, 3.67s/it] {'loss': 1.1074, 'grad_norm': 0.0010043811339039979, 'learning_rate': 0.058303921723867225, 'epoch': 0.65} + 65%|██████▍ | 337/520 [21:19<11:11, 3.67s/it] 65%|██████▌ | 338/520 [21:23<11:07, 3.67s/it] {'loss': 1.2215, 'grad_norm': 0.0010268896943298712, 'learning_rate': 0.05773817382593008, 'epoch': 0.65} + 65%|██████▌ | 338/520 [21:23<11:07, 3.67s/it] 65%|██████▌ | 339/520 [21:26<11:05, 3.68s/it] {'loss': 1.1717, 'grad_norm': 0.0010360796275539852, 'learning_rate': 0.057174067974773715, 'epoch': 0.65} + 65%|██████▌ | 339/520 [21:26<11:05, 3.68s/it] 65%|██████▌ | 340/520 [21:30<11:02, 3.68s/it] {'loss': 1.1577, 'grad_norm': 0.0010036558854803947, 'learning_rate': 0.056611626088244195, 'epoch': 0.65} + 65%|██████▌ | 340/520 [21:30<11:02, 3.68s/it] 66%|██████▌ | 341/520 [21:34<10:57, 3.68s/it] {'loss': 1.18, 'grad_norm': 0.001068718524558516, 'learning_rate': 0.056050870019535494, 'epoch': 0.66} + 66%|██████▌ | 341/520 [21:34<10:57, 3.68s/it] 66%|██████▌ | 342/520 [21:37<10:53, 3.67s/it] {'loss': 1.2289, 'grad_norm': 0.0011856995768500475, 'learning_rate': 0.05549182155634076, 'epoch': 0.66} + 66%|██████▌ | 342/520 [21:37<10:53, 3.67s/it] 66%|██████▌ | 343/520 [21:41<10:50, 3.67s/it] {'loss': 1.186, 'grad_norm': 0.0008904514053079337, 'learning_rate': 0.054934502420005464, 'epoch': 0.66} + 66%|██████▌ | 343/520 [21:41<10:50, 3.67s/it] 66%|██████▌ | 344/520 [21:45<10:46, 3.67s/it] {'loss': 1.1322, 'grad_norm': 0.0009294322118819447, 'learning_rate': 0.0543789342646837, 'epoch': 0.66} + 66%|██████▌ | 344/520 [21:45<10:46, 3.67s/it] 66%|██████▋ | 345/520 [21:48<10:42, 3.67s/it] {'loss': 1.2449, 'grad_norm': 0.0010937352931883416, 'learning_rate': 0.05382513867649663, 'epoch': 0.66} + 66%|██████▋ | 345/520 [21:48<10:42, 3.67s/it] 67%|██████▋ | 346/520 [21:52<10:37, 3.67s/it] {'loss': 1.1979, 'grad_norm': 0.0010349352736782418, 'learning_rate': 0.0532731371726938, 'epoch': 0.67} + 67%|██████▋ | 346/520 [21:52<10:37, 3.67s/it] 67%|██████▋ | 347/520 [21:56<10:35, 3.67s/it] {'loss': 1.1516, 'grad_norm': 0.0009367544616119211, 'learning_rate': 0.05272295120081732, 'epoch': 0.67} + 67%|██████▋ | 347/520 [21:56<10:35, 3.67s/it]Token indices sequence length is longer than the specified maximum sequence length for this model (2778 > 2048). Running this sequence through the model will result in indexing errors + 67%|██████▋ | 348/520 [21:59<10:30, 3.67s/it] {'loss': 1.1164, 'grad_norm': 0.001163666641143426, 'learning_rate': 0.05217460213786821, 'epoch': 0.67} + 67%|██████▋ | 348/520 [21:59<10:30, 3.67s/it] 67%|██████▋ | 349/520 [22:03<10:25, 3.66s/it] {'loss': 1.1462, 'grad_norm': 0.0010057418952873537, 'learning_rate': 0.051628111289476024, 'epoch': 0.67} + 67%|██████▋ | 349/520 [22:03<10:25, 3.66s/it] 67%|██████▋ | 350/520 [22:07<10:21, 3.65s/it] {'loss': 1.1906, 'grad_norm': 0.0010388503688436826, 'learning_rate': 0.051083499889071106, 'epoch': 0.67} + 67%|██████▋ | 350/520 [22:07<10:21, 3.65s/it] 68%|██████▊ | 351/520 [22:10<10:17, 3.66s/it] {'loss': 1.1053, 'grad_norm': 0.0009605143019067341, 'learning_rate': 0.05054078909705926, 'epoch': 0.68} + 68%|██████▊ | 351/520 [22:10<10:17, 3.66s/it] 68%|██████▊ | 352/520 [22:14<10:14, 3.66s/it] {'loss': 1.2191, 'grad_norm': 0.000968924874384306, 'learning_rate': 0.050000000000000024, 'epoch': 0.68} + 68%|██████▊ | 352/520 [22:14<10:14, 3.66s/it] 68%|██████▊ | 353/520 [22:18<10:11, 3.66s/it] {'loss': 1.16, 'grad_norm': 0.0008698058437885715, 'learning_rate': 0.04946115360978696, 'epoch': 0.68} + 68%|██████▊ | 353/520 [22:18<10:11, 3.66s/it] 68%|██████▊ | 354/520 [22:21<10:06, 3.65s/it] {'loss': 1.273, 'grad_norm': 0.000953898521007405, 'learning_rate': 0.048924270862831465, 'epoch': 0.68} + 68%|██████▊ | 354/520 [22:21<10:06, 3.65s/it] 68%|██████▊ | 355/520 [22:25<10:03, 3.66s/it] {'loss': 1.1625, 'grad_norm': 0.000985103367373824, 'learning_rate': 0.04838937261924933, 'epoch': 0.68} + 68%|██████▊ | 355/520 [22:25<10:03, 3.66s/it] 68%|██████▊ | 356/520 [22:29<10:00, 3.66s/it] {'loss': 1.1619, 'grad_norm': 0.001041810698763071, 'learning_rate': 0.0478564796620502, 'epoch': 0.68} + 68%|██████▊ | 356/520 [22:29<10:00, 3.66s/it] 69%|██████▊ | 357/520 [22:32<09:54, 3.65s/it] {'loss': 1.1947, 'grad_norm': 0.0009703523431770904, 'learning_rate': 0.04732561269632993, 'epoch': 0.69} + 69%|██████▊ | 357/520 [22:32<09:54, 3.65s/it] 69%|██████▉ | 358/520 [22:36<09:51, 3.65s/it] {'loss': 1.1272, 'grad_norm': 0.000991726484232695, 'learning_rate': 0.04679679234846636, 'epoch': 0.69} + 69%|██████▉ | 358/520 [22:36<09:51, 3.65s/it] 69%|██████▉ | 359/520 [22:40<09:49, 3.66s/it] {'loss': 1.2016, 'grad_norm': 0.0010579032774989943, 'learning_rate': 0.046270039165317606, 'epoch': 0.69} + 69%|██████▉ | 359/520 [22:40<09:49, 3.66s/it] 69%|██████▉ | 360/520 [22:43<09:45, 3.66s/it] {'loss': 1.2203, 'grad_norm': 0.0010685459813001854, 'learning_rate': 0.04574537361342407, 'epoch': 0.69} + 69%|██████▉ | 360/520 [22:43<09:45, 3.66s/it] 69%|██████▉ | 361/520 [22:47<09:45, 3.68s/it] {'loss': 1.2231, 'grad_norm': 0.0009322044757482808, 'learning_rate': 0.04522281607821288, 'epoch': 0.69} + 69%|██████▉ | 361/520 [22:47<09:45, 3.68s/it] 70%|██████▉ | 362/520 [22:51<09:39, 3.67s/it] {'loss': 1.1781, 'grad_norm': 0.0010779918489727787, 'learning_rate': 0.04470238686320606, 'epoch': 0.7} + 70%|██████▉ | 362/520 [22:51<09:39, 3.67s/it] 70%|██████▉ | 363/520 [22:54<09:35, 3.66s/it] {'loss': 1.2094, 'grad_norm': 0.0010220634832078489, 'learning_rate': 0.044184106189231624, 'epoch': 0.7} + 70%|██████▉ | 363/520 [22:54<09:35, 3.66s/it] 70%|███████ | 364/520 [22:58<09:34, 3.68s/it] {'loss': 1.2462, 'grad_norm': 0.0010554369615372694, 'learning_rate': 0.043667994193637795, 'epoch': 0.7} + 70%|███████ | 364/520 [22:58<09:34, 3.68s/it] 70%|███████ | 365/520 [23:02<09:37, 3.72s/it] {'loss': 1.2566, 'grad_norm': 0.001026866381473444, 'learning_rate': 0.043154070929510784, 'epoch': 0.7} + 70%|███████ | 365/520 [23:02<09:37, 3.72s/it] 70%|███████ | 366/520 [23:06<09:40, 3.77s/it] {'loss': 1.2193, 'grad_norm': 0.0009766441243634575, 'learning_rate': 0.04264235636489542, 'epoch': 0.7} + 70%|███████ | 366/520 [23:06<09:40, 3.77s/it] 71%|███████ | 367/520 [23:09<09:36, 3.77s/it] {'loss': 1.2197, 'grad_norm': 0.001042982093943529, 'learning_rate': 0.04213287038201943, 'epoch': 0.71} + 71%|███████ | 367/520 [23:09<09:36, 3.77s/it] 71%|███████ | 368/520 [23:13<09:28, 3.74s/it] {'loss': 1.0731, 'grad_norm': 0.0010303290933443797, 'learning_rate': 0.04162563277652104, 'epoch': 0.71} + 71%|███████ | 368/520 [23:13<09:28, 3.74s/it] 71%|███████ | 369/520 [23:17<09:21, 3.72s/it] {'loss': 1.2001, 'grad_norm': 0.0009230490640617261, 'learning_rate': 0.04112066325667954, 'epoch': 0.71} + 71%|███████ | 369/520 [23:17<09:21, 3.72s/it] 71%|███████ | 370/520 [23:20<09:14, 3.69s/it] {'loss': 1.1352, 'grad_norm': 0.0009669737959447692, 'learning_rate': 0.04061798144264986, 'epoch': 0.71} + 71%|███████ | 370/520 [23:20<09:14, 3.69s/it] 71%|███████▏ | 371/520 [23:24<09:07, 3.67s/it] {'loss': 1.134, 'grad_norm': 0.0010702300114250893, 'learning_rate': 0.04011760686569998, 'epoch': 0.71} + 71%|███████▏ | 371/520 [23:24<09:07, 3.67s/it] 72%|███████▏ | 372/520 [23:28<09:02, 3.67s/it] {'loss': 1.2789, 'grad_norm': 0.0009400083544229035, 'learning_rate': 0.03961955896745224, 'epoch': 0.72} + 72%|███████▏ | 372/520 [23:28<09:02, 3.67s/it] 72%|███████▏ | 373/520 [23:31<09:00, 3.67s/it] {'loss': 1.1613, 'grad_norm': 0.001120302940315765, 'learning_rate': 0.03912385709912794, 'epoch': 0.72} + 72%|███████▏ | 373/520 [23:31<09:00, 3.67s/it] 72%|███████▏ | 374/520 [23:35<09:02, 3.71s/it] {'loss': 1.2223, 'grad_norm': 0.0010255306450976055, 'learning_rate': 0.038630520520795276, 'epoch': 0.72} + 72%|███████▏ | 374/520 [23:35<09:02, 3.71s/it] 72%|███████▏ | 375/520 [23:39<09:01, 3.73s/it] {'loss': 1.1299, 'grad_norm': 0.0009792777658616855, 'learning_rate': 0.03813956840062119, 'epoch': 0.72} + 72%|███████▏ | 375/520 [23:39<09:01, 3.73s/it] 72%|███████▏ | 376/520 [23:43<08:58, 3.74s/it] {'loss': 1.2399, 'grad_norm': 0.0009507704556800704, 'learning_rate': 0.037651019814126656, 'epoch': 0.72} + 72%|███████▏ | 376/520 [23:43<08:58, 3.74s/it] 72%|███████▎ | 377/520 [23:47<08:56, 3.75s/it] {'loss': 1.1837, 'grad_norm': 0.0011927642186971685, 'learning_rate': 0.037164893743445275, 'epoch': 0.72} + 72%|███████▎ | 377/520 [23:47<08:56, 3.75s/it] 73%|███████▎ | 378/520 [23:50<08:52, 3.75s/it] {'loss': 1.2317, 'grad_norm': 0.0009608473473754067, 'learning_rate': 0.03668120907658603, 'epoch': 0.73} + 73%|███████▎ | 378/520 [23:50<08:52, 3.75s/it] 73%|███████▎ | 379/520 [23:54<08:49, 3.75s/it] {'loss': 1.2146, 'grad_norm': 0.0009818153693474283, 'learning_rate': 0.036199984606699154, 'epoch': 0.73} + 73%|███████▎ | 379/520 [23:54<08:49, 3.75s/it] 73%|███████▎ | 380/520 [23:58<08:41, 3.72s/it] {'loss': 1.2527, 'grad_norm': 0.0010304964609639518, 'learning_rate': 0.035721239031346066, 'epoch': 0.73} + 73%|███████▎ | 380/520 [23:58<08:41, 3.72s/it] 73%|███████▎ | 381/520 [24:01<08:34, 3.70s/it] {'loss': 1.2168, 'grad_norm': 0.0009853585246232814, 'learning_rate': 0.03524499095177297, 'epoch': 0.73} + 73%|███████▎ | 381/520 [24:01<08:34, 3.70s/it] 73%|███████▎ | 382/520 [24:05<08:30, 3.70s/it] {'loss': 1.2159, 'grad_norm': 0.0009761782983782391, 'learning_rate': 0.03477125887218792, 'epoch': 0.73} + 73%|███████▎ | 382/520 [24:05<08:30, 3.70s/it] 74%|███████▎ | 383/520 [24:09<08:26, 3.70s/it] {'loss': 1.0633, 'grad_norm': 0.0011318348924010807, 'learning_rate': 0.03430006119904196, 'epoch': 0.74} + 74%|███████▎ | 383/520 [24:09<08:26, 3.70s/it] 74%|███████▍ | 384/520 [24:12<08:21, 3.69s/it] {'loss': 1.2675, 'grad_norm': 0.0009282259947433767, 'learning_rate': 0.033831416240314084, 'epoch': 0.74} + 74%|███████▍ | 384/520 [24:12<08:21, 3.69s/it] 74%|███████▍ | 385/520 [24:16<08:17, 3.69s/it] {'loss': 1.1946, 'grad_norm': 0.000922880568169277, 'learning_rate': 0.03336534220479961, 'epoch': 0.74} + 74%|███████▍ | 385/520 [24:16<08:17, 3.69s/it] 74%|███████▍ | 386/520 [24:20<08:16, 3.71s/it] {'loss': 1.1529, 'grad_norm': 0.0008925499209404142, 'learning_rate': 0.032901857201403005, 'epoch': 0.74} + 74%|███████▍ | 386/520 [24:20<08:16, 3.71s/it] 74%|███████▍ | 387/520 [24:24<08:18, 3.75s/it] {'loss': 1.278, 'grad_norm': 0.0010091663148886973, 'learning_rate': 0.032440979238433976, 'epoch': 0.74} + 74%|███████▍ | 387/520 [24:24<08:18, 3.75s/it] 75%|███████▍ | 388/520 [24:27<08:12, 3.73s/it] {'loss': 1.1018, 'grad_norm': 0.0009304529571650742, 'learning_rate': 0.03198272622290804, 'epoch': 0.75} + 75%|███████▍ | 388/520 [24:27<08:12, 3.73s/it] 75%|███████▍ | 389/520 [24:31<08:07, 3.72s/it] {'loss': 1.1508, 'grad_norm': 0.0011473570177098745, 'learning_rate': 0.03152711595985065, 'epoch': 0.75} + 75%|███████▍ | 389/520 [24:31<08:07, 3.72s/it] 75%|███████▌ | 390/520 [24:35<08:01, 3.71s/it] {'loss': 1.2163, 'grad_norm': 0.0010257300059895325, 'learning_rate': 0.031074166151605298, 'epoch': 0.75} + 75%|███████▌ | 390/520 [24:35<08:01, 3.71s/it] 75%|███████▌ | 391/520 [24:39<08:00, 3.73s/it] {'loss': 1.289, 'grad_norm': 0.0010336049633587162, 'learning_rate': 0.030623894397145836, 'epoch': 0.75} + 75%|███████▌ | 391/520 [24:39<08:00, 3.73s/it] 75%|███████▌ | 392/520 [24:43<08:06, 3.80s/it] {'loss': 1.113, 'grad_norm': 0.0009637481719523064, 'learning_rate': 0.03017631819139273, 'epoch': 0.75} + 75%|███████▌ | 392/520 [24:43<08:06, 3.80s/it] 76%|███████▌ | 393/520 [24:46<08:03, 3.81s/it] {'loss': 1.1185, 'grad_norm': 0.0008722993994227826, 'learning_rate': 0.029731454924533086, 'epoch': 0.76} + 76%|███████▌ | 393/520 [24:46<08:03, 3.81s/it] 76%|███████▌ | 394/520 [24:50<07:59, 3.81s/it] {'loss': 1.174, 'grad_norm': 0.001054630998791634, 'learning_rate': 0.029289321881345254, 'epoch': 0.76} + 76%|███████▌ | 394/520 [24:50<07:59, 3.81s/it] 76%|███████▌ | 395/520 [24:54<07:56, 3.82s/it] {'loss': 1.1394, 'grad_norm': 0.0010522549687479756, 'learning_rate': 0.028849936240527008, 'epoch': 0.76} + 76%|███████▌ | 395/520 [24:54<07:56, 3.82s/it] 76%|███████▌ | 396/520 [24:58<07:53, 3.82s/it] {'loss': 1.2297, 'grad_norm': 0.0010840979554588026, 'learning_rate': 0.028413315074028157, 'epoch': 0.76} + 76%|███████▌ | 396/520 [24:58<07:53, 3.82s/it] 76%|███████▋ | 397/520 [25:02<07:48, 3.81s/it] {'loss': 1.1984, 'grad_norm': 0.0009710186041574596, 'learning_rate': 0.027979475346387363, 'epoch': 0.76} + 76%|███████▋ | 397/520 [25:02<07:48, 3.81s/it] 77%|███████▋ | 398/520 [25:05<07:44, 3.80s/it] {'loss': 1.1962, 'grad_norm': 0.0010463786481082957, 'learning_rate': 0.027548433914072735, 'epoch': 0.77} + 77%|███████▋ | 398/520 [25:05<07:44, 3.80s/it] 77%|███████▋ | 399/520 [25:09<07:41, 3.81s/it] {'loss': 1.1622, 'grad_norm': 0.0009769915201144378, 'learning_rate': 0.027120207524827168, 'epoch': 0.77} + 77%|███████▋ | 399/520 [25:09<07:41, 3.81s/it] 77%|███████▋ | 400/520 [25:13<07:36, 3.80s/it] {'loss': 1.1927, 'grad_norm': 0.0009192063076503197, 'learning_rate': 0.02669481281701739, 'epoch': 0.77} + 77%|███████▋ | 400/520 [25:13<07:36, 3.80s/it] 77%|███████▋ | 401/520 [25:17<07:33, 3.81s/it] {'loss': 1.0265, 'grad_norm': 0.0010931059683525301, 'learning_rate': 0.026272266318987603, 'epoch': 0.77} + 77%|███████▋ | 401/520 [25:17<07:33, 3.81s/it] 77%|███████▋ | 402/520 [25:21<07:28, 3.80s/it] {'loss': 1.1575, 'grad_norm': 0.0010266879891921634, 'learning_rate': 0.02585258444841733, 'epoch': 0.77} + 77%|███████▋ | 402/520 [25:21<07:28, 3.80s/it] 78%|███████▊ | 403/520 [25:24<07:24, 3.80s/it] {'loss': 1.1836, 'grad_norm': 0.0010942487963900888, 'learning_rate': 0.025435783511683442, 'epoch': 0.78} + 78%|███████▊ | 403/520 [25:24<07:24, 3.80s/it] 78%|███████▊ | 404/520 [25:28<07:20, 3.79s/it] {'loss': 1.0914, 'grad_norm': 0.0011182865540411736, 'learning_rate': 0.02502187970322657, 'epoch': 0.78} + 78%|███████▊ | 404/520 [25:28<07:20, 3.79s/it] 78%|███████▊ | 405/520 [25:32<07:15, 3.78s/it] {'loss': 1.1734, 'grad_norm': 0.0009687294225231329, 'learning_rate': 0.02461088910492202, 'epoch': 0.78} + 78%|███████▊ | 405/520 [25:32<07:15, 3.78s/it] 78%|███████▊ | 406/520 [25:36<07:08, 3.76s/it] {'loss': 1.0962, 'grad_norm': 0.001175719358983435, 'learning_rate': 0.02420282768545469, 'epoch': 0.78} + 78%|███████▊ | 406/520 [25:36<07:08, 3.76s/it] 78%|███████▊ | 407/520 [25:39<07:00, 3.72s/it] {'loss': 1.2667, 'grad_norm': 0.001029906253720282, 'learning_rate': 0.02379771129969892, 'epoch': 0.78} + 78%|███████▊ | 407/520 [25:39<07:00, 3.72s/it] 78%|███████▊ | 408/520 [25:43<06:53, 3.69s/it] {'loss': 1.1681, 'grad_norm': 0.0010880420933110193, 'learning_rate': 0.023395555688102213, 'epoch': 0.78} + 78%|███████▊ | 408/520 [25:43<06:53, 3.69s/it] 79%|███████▊ | 409/520 [25:47<06:49, 3.69s/it] {'loss': 1.2879, 'grad_norm': 0.0010678020055341628, 'learning_rate': 0.02299637647607372, 'epoch': 0.79} + 79%|███████▊ | 409/520 [25:47<06:49, 3.69s/it] 79%|███████▉ | 410/520 [25:50<06:45, 3.69s/it] {'loss': 1.0253, 'grad_norm': 0.001037506311691759, 'learning_rate': 0.022600189173377264, 'epoch': 0.79} + 79%|███████▉ | 410/520 [25:50<06:45, 3.69s/it] 79%|███████▉ | 411/520 [25:54<06:40, 3.67s/it] {'loss': 1.2671, 'grad_norm': 0.0010819056816006782, 'learning_rate': 0.022207009173528525, 'epoch': 0.79} + 79%|███████▉ | 411/520 [25:54<06:40, 3.67s/it] 79%|███████▉ | 412/520 [25:58<06:36, 3.67s/it] {'loss': 1.1782, 'grad_norm': 0.0010458115919827753, 'learning_rate': 0.02181685175319702, 'epoch': 0.79} + 79%|███████▉ | 412/520 [25:58<06:36, 3.67s/it] 79%|███████▉ | 413/520 [26:01<06:31, 3.66s/it] {'loss': 1.1884, 'grad_norm': 0.0010063255855467096, 'learning_rate': 0.021429732071612653, 'epoch': 0.79} + 79%|███████▉ | 413/520 [26:01<06:31, 3.66s/it] 80%|███████▉ | 414/520 [26:05<06:28, 3.67s/it] {'loss': 0.9978, 'grad_norm': 0.0009187017352712183, 'learning_rate': 0.02104566516997647, 'epoch': 0.8} + 80%|███████▉ | 414/520 [26:05<06:28, 3.67s/it] 80%|███████▉ | 415/520 [26:09<06:25, 3.67s/it] {'loss': 1.1603, 'grad_norm': 0.0009556104133310733, 'learning_rate': 0.020664665970876496, 'epoch': 0.8} + 80%|███████▉ | 415/520 [26:09<06:25, 3.67s/it] 80%|████████ | 416/520 [26:12<06:21, 3.67s/it] {'loss': 1.0839, 'grad_norm': 0.001095794098877447, 'learning_rate': 0.020286749277707784, 'epoch': 0.8} + 80%|████████ | 416/520 [26:12<06:21, 3.67s/it] 80%|████████ | 417/520 [26:16<06:17, 3.67s/it] {'loss': 1.2329, 'grad_norm': 0.0010090218713627282, 'learning_rate': 0.019911929774097215, 'epoch': 0.8} + 80%|████████ | 417/520 [26:16<06:17, 3.67s/it] 80%|████████ | 418/520 [26:20<06:13, 3.67s/it] {'loss': 1.2258, 'grad_norm': 0.0009577748442433951, 'learning_rate': 0.019540222023333165, 'epoch': 0.8} + 80%|████████ | 418/520 [26:20<06:13, 3.67s/it] 81%|████████ | 419/520 [26:23<06:09, 3.66s/it] {'loss': 1.2113, 'grad_norm': 0.001110848468612451, 'learning_rate': 0.01917164046779948, 'epoch': 0.81} + 81%|████████ | 419/520 [26:23<06:09, 3.66s/it] 81%|████████ | 420/520 [26:27<06:06, 3.67s/it] {'loss': 1.1039, 'grad_norm': 0.0010584954665321777, 'learning_rate': 0.018806199428414352, 'epoch': 0.81} + 81%|████████ | 420/520 [26:27<06:06, 3.67s/it] 81%|████████ | 421/520 [26:31<06:03, 3.67s/it] {'loss': 1.0437, 'grad_norm': 0.001069458968058384, 'learning_rate': 0.018443913104073985, 'epoch': 0.81} + 81%|████████ | 421/520 [26:31<06:03, 3.67s/it] 81%|████████ | 422/520 [26:34<05:58, 3.66s/it] {'loss': 1.1604, 'grad_norm': 0.0010712950361667166, 'learning_rate': 0.01808479557110081, 'epoch': 0.81} + 81%|████████ | 422/520 [26:34<05:58, 3.66s/it] 81%|████████▏ | 423/520 [26:38<05:55, 3.67s/it] {'loss': 1.1371, 'grad_norm': 0.0010998887519769142, 'learning_rate': 0.017728860782696667, 'epoch': 0.81} + 81%|████████▏ | 423/520 [26:38<05:55, 3.67s/it] 82%|████████▏ | 424/520 [26:42<05:53, 3.68s/it] {'loss': 1.2741, 'grad_norm': 0.000985815450128948, 'learning_rate': 0.017376122568400532, 'epoch': 0.82} + 82%|████████▏ | 424/520 [26:42<05:53, 3.68s/it] 82%|████████▏ | 425/520 [26:45<05:51, 3.70s/it] {'loss': 1.1546, 'grad_norm': 0.0010396644689990217, 'learning_rate': 0.017026594633551252, 'epoch': 0.82} + 82%|████████▏ | 425/520 [26:45<05:51, 3.70s/it] 82%|████████▏ | 426/520 [26:49<05:46, 3.69s/it] {'loss': 1.185, 'grad_norm': 0.0012916415393997081, 'learning_rate': 0.01668029055875512, 'epoch': 0.82} + 82%|████████▏ | 426/520 [26:49<05:46, 3.69s/it] 82%|████████▏ | 427/520 [26:53<05:42, 3.68s/it] {'loss': 1.0914, 'grad_norm': 0.000981809872315531, 'learning_rate': 0.016337223799358026, 'epoch': 0.82} + 82%|████████▏ | 427/520 [26:53<05:42, 3.68s/it] 82%|████████▏ | 428/520 [26:56<05:37, 3.67s/it] {'loss': 1.0731, 'grad_norm': 0.0010831107789565167, 'learning_rate': 0.01599740768492286, 'epoch': 0.82} + 82%|████████▏ | 428/520 [26:56<05:37, 3.67s/it] 82%|████████▎ | 429/520 [27:00<05:34, 3.68s/it] {'loss': 1.169, 'grad_norm': 0.001021272723451405, 'learning_rate': 0.015660855418711452, 'epoch': 0.82} + 82%|████████▎ | 429/520 [27:00<05:34, 3.68s/it]Token indices sequence length is longer than the specified maximum sequence length for this model (2076 > 2048). Running this sequence through the model will result in indexing errors + 83%|████████▎ | 430/520 [27:04<05:29, 3.66s/it] {'loss': 1.1688, 'grad_norm': 0.0009518907080681958, 'learning_rate': 0.015327580077171589, 'epoch': 0.83} + 83%|████████▎ | 430/520 [27:04<05:29, 3.66s/it] 83%|████████▎ | 431/520 [27:07<05:25, 3.66s/it] {'loss': 1.1602, 'grad_norm': 0.0010476567708265212, 'learning_rate': 0.014997594609429088, 'epoch': 0.83} + 83%|████████▎ | 431/520 [27:07<05:25, 3.66s/it] 83%|████████▎ | 432/520 [27:11<05:21, 3.66s/it] {'loss': 1.0795, 'grad_norm': 0.0010234593533020827, 'learning_rate': 0.01467091183678444, 'epoch': 0.83} + 83%|████████▎ | 432/520 [27:11<05:21, 3.66s/it] 83%|████████▎ | 433/520 [27:15<05:17, 3.65s/it] {'loss': 1.2108, 'grad_norm': 0.0009870133330663953, 'learning_rate': 0.014347544452214867, 'epoch': 0.83} + 83%|████████▎ | 433/520 [27:15<05:17, 3.65s/it] 83%|████████▎ | 434/520 [27:18<05:14, 3.66s/it] {'loss': 0.9543, 'grad_norm': 0.001018819443845783, 'learning_rate': 0.014027505019880971, 'epoch': 0.83} + 83%|████████▎ | 434/520 [27:18<05:14, 3.66s/it] 84%|████████▎ | 435/520 [27:22<05:12, 3.68s/it] {'loss': 1.2486, 'grad_norm': 0.0011750579000592229, 'learning_rate': 0.013710805974638696, 'epoch': 0.84} + 84%|████████▎ | 435/520 [27:22<05:12, 3.68s/it] 84%|████████▍ | 436/520 [27:26<05:08, 3.67s/it] {'loss': 1.0501, 'grad_norm': 0.001060250121337156, 'learning_rate': 0.01339745962155613, 'epoch': 0.84} + 84%|████████▍ | 436/520 [27:26<05:08, 3.67s/it] 84%|████████▍ | 437/520 [27:29<05:05, 3.68s/it] {'loss': 1.2702, 'grad_norm': 0.0010491783120745643, 'learning_rate': 0.01308747813543536, 'epoch': 0.84} + 84%|████████▍ | 437/520 [27:29<05:05, 3.68s/it] 84%|████████▍ | 438/520 [27:33<05:02, 3.69s/it] {'loss': 1.0864, 'grad_norm': 0.0010450607971780681, 'learning_rate': 0.012780873560339467, 'epoch': 0.84} + 84%|████████▍ | 438/520 [27:33<05:02, 3.69s/it] 84%|████████▍ | 439/520 [27:37<04:59, 3.70s/it] {'loss': 1.1431, 'grad_norm': 0.0008684243863504343, 'learning_rate': 0.012477657809124632, 'epoch': 0.84} + 84%|████████▍ | 439/520 [27:37<04:59, 3.70s/it] 85%|████████▍ | 440/520 [27:40<04:54, 3.68s/it] {'loss': 1.1256, 'grad_norm': 0.0010488451432756154, 'learning_rate': 0.012177842662977134, 'epoch': 0.85} + 85%|████████▍ | 440/520 [27:40<04:54, 3.68s/it] 85%|████████▍ | 441/520 [27:44<04:50, 3.68s/it] {'loss': 1.156, 'grad_norm': 0.0009745061749362523, 'learning_rate': 0.01188143977095576, 'epoch': 0.85} + 85%|████████▍ | 441/520 [27:44<04:50, 3.68s/it] 85%|████████▌ | 442/520 [27:48<04:46, 3.67s/it] {'loss': 1.1865, 'grad_norm': 0.0011131032586972952, 'learning_rate': 0.011588460649539035, 'epoch': 0.85} + 85%|████████▌ | 442/520 [27:48<04:46, 3.67s/it] 85%|████████▌ | 443/520 [27:52<04:45, 3.71s/it] {'loss': 1.2027, 'grad_norm': 0.0010058038940847886, 'learning_rate': 0.011298916682177829, 'epoch': 0.85} + 85%|████████▌ | 443/520 [27:52<04:45, 3.71s/it] 85%|████████▌ | 444/520 [27:55<04:41, 3.70s/it] {'loss': 1.1651, 'grad_norm': 0.0009187533613916089, 'learning_rate': 0.011012819118853146, 'epoch': 0.85} + 85%|████████▌ | 444/520 [27:55<04:41, 3.70s/it] 86%|████████▌ | 445/520 [27:59<04:40, 3.74s/it] {'loss': 1.0965, 'grad_norm': 0.0009875180278054378, 'learning_rate': 0.01073017907563887, 'epoch': 0.86} + 86%|████████▌ | 445/520 [27:59<04:40, 3.74s/it] 86%|████████▌ | 446/520 [28:03<04:35, 3.72s/it] {'loss': 1.2352, 'grad_norm': 0.00093996225996516, 'learning_rate': 0.010451007534269908, 'epoch': 0.86} + 86%|████████▌ | 446/520 [28:03<04:35, 3.72s/it] 86%|████████▌ | 447/520 [28:06<04:31, 3.72s/it] {'loss': 1.1738, 'grad_norm': 0.001014327012958264, 'learning_rate': 0.010175315341715598, 'epoch': 0.86} + 86%|████████▌ | 447/520 [28:06<04:31, 3.72s/it] 86%|████████▌ | 448/520 [28:10<04:27, 3.71s/it] {'loss': 1.1607, 'grad_norm': 0.0011810011682095133, 'learning_rate': 0.009903113209758098, 'epoch': 0.86} + 86%|████████▌ | 448/520 [28:10<04:27, 3.71s/it] 86%|████████▋ | 449/520 [28:14<04:23, 3.71s/it] {'loss': 1.1883, 'grad_norm': 0.0010380370678473288, 'learning_rate': 0.009634411714576352, 'epoch': 0.86} + 86%|████████▋ | 449/520 [28:14<04:23, 3.71s/it] 87%|████████▋ | 450/520 [28:17<04:19, 3.70s/it] {'loss': 1.1962, 'grad_norm': 0.001046218672844649, 'learning_rate': 0.009369221296335007, 'epoch': 0.87} + 87%|████████▋ | 450/520 [28:18<04:19, 3.70s/it] 87%|████████▋ | 451/520 [28:21<04:15, 3.71s/it] {'loss': 1.1933, 'grad_norm': 0.0010336791834376566, 'learning_rate': 0.009107552258778906, 'epoch': 0.87} + 87%|████████▋ | 451/520 [28:21<04:15, 3.71s/it] 87%|████████▋ | 452/520 [28:25<04:11, 3.70s/it] {'loss': 1.2364, 'grad_norm': 0.0009501637993968304, 'learning_rate': 0.008849414768832687, 'epoch': 0.87} + 87%|████████▋ | 452/520 [28:25<04:11, 3.70s/it] 87%|████████▋ | 453/520 [28:29<04:09, 3.72s/it] {'loss': 1.2107, 'grad_norm': 0.0009642675666293609, 'learning_rate': 0.008594818856205699, 'epoch': 0.87} + 87%|████████▋ | 453/520 [28:29<04:09, 3.72s/it] 87%|████████▋ | 454/520 [28:32<04:06, 3.73s/it] {'loss': 1.1022, 'grad_norm': 0.0010743823040171982, 'learning_rate': 0.00834377441300238, 'epoch': 0.87} + 87%|████████▋ | 454/520 [28:32<04:06, 3.73s/it] 88%|████████▊ | 455/520 [28:36<04:03, 3.75s/it] {'loss': 1.2388, 'grad_norm': 0.0010137022670156752, 'learning_rate': 0.008096291193337934, 'epoch': 0.88} + 88%|████████▊ | 455/520 [28:36<04:03, 3.75s/it] 88%|████████▊ | 456/520 [28:40<04:01, 3.77s/it] {'loss': 1.1705, 'grad_norm': 0.0010278821839031344, 'learning_rate': 0.007852378812959226, 'epoch': 0.88} + 88%|████████▊ | 456/520 [28:40<04:01, 3.77s/it] 88%|████████▊ | 457/520 [28:44<03:58, 3.79s/it] {'loss': 1.1212, 'grad_norm': 0.0008912251258715418, 'learning_rate': 0.007612046748871327, 'epoch': 0.88} + 88%|████████▊ | 457/520 [28:44<03:58, 3.79s/it] 88%|████████▊ | 458/520 [28:48<03:54, 3.78s/it] {'loss': 1.293, 'grad_norm': 0.0011203203774243855, 'learning_rate': 0.007375304338969136, 'epoch': 0.88} + 88%|████████▊ | 458/520 [28:48<03:54, 3.78s/it] 88%|████████▊ | 459/520 [28:51<03:50, 3.78s/it] {'loss': 1.2186, 'grad_norm': 0.000990247020300417, 'learning_rate': 0.007142160781674645, 'epoch': 0.88} + 88%|████████▊ | 459/520 [28:51<03:50, 3.78s/it] 88%|████████▊ | 460/520 [28:55<03:45, 3.75s/it] {'loss': 1.1116, 'grad_norm': 0.0009814822094575439, 'learning_rate': 0.006912625135579587, 'epoch': 0.88} + 88%|████████▊ | 460/520 [28:55<03:45, 3.75s/it] 89%|████████▊ | 461/520 [28:59<03:42, 3.77s/it] {'loss': 1.2103, 'grad_norm': 0.0008172018863829436, 'learning_rate': 0.0066867063190933496, 'epoch': 0.89} + 89%|████████▊ | 461/520 [28:59<03:42, 3.77s/it] 89%|████████▉ | 462/520 [29:03<03:37, 3.76s/it] {'loss': 1.2809, 'grad_norm': 0.0009990542959858594, 'learning_rate': 0.006464413110096601, 'epoch': 0.89} + 89%|████████▉ | 462/520 [29:03<03:37, 3.76s/it] 89%|████████▉ | 463/520 [29:06<03:33, 3.75s/it] {'loss': 1.0751, 'grad_norm': 0.0010443863901121344, 'learning_rate': 0.006245754145600091, 'epoch': 0.89} + 89%|████████▉ | 463/520 [29:06<03:33, 3.75s/it] 89%|████████▉ | 464/520 [29:10<03:29, 3.74s/it] {'loss': 1.2083, 'grad_norm': 0.0010401208909634645, 'learning_rate': 0.006030737921409169, 'epoch': 0.89} + 89%|████████▉ | 464/520 [29:10<03:29, 3.74s/it] 89%|████████▉ | 465/520 [29:14<03:24, 3.72s/it] {'loss': 1.3155, 'grad_norm': 0.0010599530423036525, 'learning_rate': 0.005819372791793654, 'epoch': 0.89} + 89%|████████▉ | 465/520 [29:14<03:24, 3.72s/it] 90%|████████▉ | 466/520 [29:17<03:20, 3.71s/it] {'loss': 1.2024, 'grad_norm': 0.0009170933911406784, 'learning_rate': 0.005611666969163243, 'epoch': 0.9} + 90%|████████▉ | 466/520 [29:17<03:20, 3.71s/it] 90%|████████▉ | 467/520 [29:21<03:16, 3.72s/it] {'loss': 1.172, 'grad_norm': 0.0009469656747915778, 'learning_rate': 0.005407628523748398, 'epoch': 0.9} + 90%|████████▉ | 467/520 [29:21<03:16, 3.72s/it] 90%|█████████ | 468/520 [29:25<03:13, 3.71s/it] {'loss': 1.1768, 'grad_norm': 0.0011685716077953553, 'learning_rate': 0.00520726538328683, 'epoch': 0.9} + 90%|█████████ | 468/520 [29:25<03:13, 3.71s/it] 90%|█████████ | 469/520 [29:29<03:09, 3.72s/it] {'loss': 1.2369, 'grad_norm': 0.0010676106606575687, 'learning_rate': 0.005010585332715401, 'epoch': 0.9} + 90%|█████████ | 469/520 [29:29<03:09, 3.72s/it] 90%|█████████ | 470/520 [29:32<03:05, 3.72s/it] {'loss': 1.1136, 'grad_norm': 0.000926908361943266, 'learning_rate': 0.004817596013867765, 'epoch': 0.9} + 90%|█████████ | 470/520 [29:32<03:05, 3.72s/it] 91%|█████████ | 471/520 [29:36<03:02, 3.73s/it] {'loss': 1.1366, 'grad_norm': 0.0010737488649097979, 'learning_rate': 0.004628304925177318, 'epoch': 0.91} + 91%|█████████ | 471/520 [29:36<03:02, 3.73s/it] 91%|█████████ | 472/520 [29:40<02:58, 3.73s/it] {'loss': 1.1052, 'grad_norm': 0.0009590143017722953, 'learning_rate': 0.004442719421385921, 'epoch': 0.91} + 91%|█████████ | 472/520 [29:40<02:58, 3.73s/it] 91%|█████████ | 473/520 [29:44<02:55, 3.72s/it] {'loss': 1.1712, 'grad_norm': 0.0010554217431173805, 'learning_rate': 0.004260846713258193, 'epoch': 0.91} + 91%|█████████ | 473/520 [29:44<02:55, 3.72s/it] 91%|█████████ | 474/520 [29:47<02:50, 3.71s/it] {'loss': 1.2082, 'grad_norm': 0.00097059888240832, 'learning_rate': 0.004082693867301224, 'epoch': 0.91} + 91%|█████████ | 474/520 [29:47<02:50, 3.71s/it] 91%|█████████▏| 475/520 [29:51<02:46, 3.71s/it] {'loss': 1.1261, 'grad_norm': 0.000967022662747355, 'learning_rate': 0.003908267805490051, 'epoch': 0.91} + 91%|█████████▏| 475/520 [29:51<02:46, 3.71s/it] 92%|█████████▏| 476/520 [29:55<02:42, 3.69s/it] {'loss': 1.1618, 'grad_norm': 0.0010765463361416718, 'learning_rate': 0.003737575304998797, 'epoch': 0.92} + 92%|█████████▏| 476/520 [29:55<02:42, 3.69s/it] 92%|█████████▏| 477/520 [29:58<02:40, 3.74s/it] {'loss': 1.152, 'grad_norm': 0.0010509533882927691, 'learning_rate': 0.003570622997937234, 'epoch': 0.92} + 92%|█████████▏| 477/520 [29:58<02:40, 3.74s/it] 92%|█████████▏| 478/520 [30:02<02:36, 3.73s/it] {'loss': 1.101, 'grad_norm': 0.0009950605561280055, 'learning_rate': 0.00340741737109318, 'epoch': 0.92} + 92%|█████████▏| 478/520 [30:02<02:36, 3.73s/it] 92%|█████████▏| 479/520 [30:06<02:32, 3.72s/it] {'loss': 1.1705, 'grad_norm': 0.0010647612661538084, 'learning_rate': 0.003247964765680389, 'epoch': 0.92} + 92%|█████████▏| 479/520 [30:06<02:32, 3.72s/it] 92%|█████████▏| 480/520 [30:10<02:28, 3.71s/it] {'loss': 1.1922, 'grad_norm': 0.0009467547647997204, 'learning_rate': 0.0030922713770922153, 'epoch': 0.92} + 92%|█████████▏| 480/520 [30:10<02:28, 3.71s/it] 92%|█████████▎| 481/520 [30:13<02:24, 3.70s/it] {'loss': 1.1811, 'grad_norm': 0.00091917887585295, 'learning_rate': 0.0029403432546609046, 'epoch': 0.93} + 92%|█████████▎| 481/520 [30:13<02:24, 3.70s/it] 93%|█████████▎| 482/520 [30:17<02:20, 3.71s/it] {'loss': 1.2021, 'grad_norm': 0.0009446674537735768, 'learning_rate': 0.0027921863014225504, 'epoch': 0.93} + 93%|█████████▎| 482/520 [30:17<02:20, 3.71s/it] 93%|█████████▎| 483/520 [30:21<02:17, 3.71s/it] {'loss': 1.1712, 'grad_norm': 0.0009996390857155426, 'learning_rate': 0.002647806273887665, 'epoch': 0.93} + 93%|█████████▎| 483/520 [30:21<02:17, 3.71s/it] 93%|█████████▎| 484/520 [30:24<02:13, 3.71s/it] {'loss': 1.1719, 'grad_norm': 0.0010316333756730787, 'learning_rate': 0.0025072087818176383, 'epoch': 0.93} + 93%|█████████▎| 484/520 [30:24<02:13, 3.71s/it] 93%|█████████▎| 485/520 [30:28<02:09, 3.71s/it] {'loss': 1.1301, 'grad_norm': 0.000993376627184422, 'learning_rate': 0.002370399288006664, 'epoch': 0.93} + 93%|█████████▎| 485/520 [30:28<02:09, 3.71s/it] 93%|█████████▎| 486/520 [30:32<02:05, 3.70s/it] {'loss': 1.2515, 'grad_norm': 0.001049545795545984, 'learning_rate': 0.0022373831080695463, 'epoch': 0.93} + 93%|█████████▎| 486/520 [30:32<02:05, 3.70s/it] 94%|█████████▎| 487/520 [30:35<02:02, 3.70s/it] {'loss': 1.104, 'grad_norm': 0.0009795808953260755, 'learning_rate': 0.0021081654102351635, 'epoch': 0.94} + 94%|█████████▎| 487/520 [30:35<02:02, 3.70s/it] 94%|█████████▍| 488/520 [30:39<01:58, 3.71s/it] {'loss': 1.0546, 'grad_norm': 0.0010219154335858254, 'learning_rate': 0.0019827512151456175, 'epoch': 0.94} + 94%|█████████▍| 488/520 [30:39<01:58, 3.71s/it] 94%|█████████▍| 489/520 [30:43<01:55, 3.71s/it] {'loss': 1.2037, 'grad_norm': 0.000874464007752385, 'learning_rate': 0.0018611453956612345, 'epoch': 0.94} + 94%|█████████▍| 489/520 [30:43<01:55, 3.71s/it] 94%|█████████▍| 490/520 [30:47<01:51, 3.71s/it] {'loss': 1.1744, 'grad_norm': 0.0010409365252051046, 'learning_rate': 0.0017433526766711727, 'epoch': 0.94} + 94%|█████████▍| 490/520 [30:47<01:51, 3.71s/it] 94%|█████████▍| 491/520 [30:50<01:47, 3.71s/it] {'loss': 1.1374, 'grad_norm': 0.001041509248018266, 'learning_rate': 0.0016293776349098677, 'epoch': 0.94} + 94%|█████████▍| 491/520 [30:50<01:47, 3.71s/it] 95%|█████████▍| 492/520 [30:54<01:43, 3.71s/it] {'loss': 1.2461, 'grad_norm': 0.0010582665916096985, 'learning_rate': 0.0015192246987791981, 'epoch': 0.95} + 95%|█████████▍| 492/520 [30:54<01:43, 3.71s/it] 95%|█████████▍| 493/520 [30:58<01:40, 3.71s/it] {'loss': 1.2072, 'grad_norm': 0.0010180844296685723, 'learning_rate': 0.0014128981481764114, 'epoch': 0.95} + 95%|█████████▍| 493/520 [30:58<01:40, 3.71s/it] 95%|█████████▌| 494/520 [31:01<01:36, 3.72s/it] {'loss': 1.1848, 'grad_norm': 0.0009551603297550791, 'learning_rate': 0.0013104021143278911, 'epoch': 0.95} + 95%|█████████▌| 494/520 [31:01<01:36, 3.72s/it] 95%|█████████▌| 495/520 [31:05<01:33, 3.73s/it] {'loss': 1.1561, 'grad_norm': 0.001009584882918316, 'learning_rate': 0.0012117405796285285, 'epoch': 0.95} + 95%|█████████▌| 495/520 [31:05<01:33, 3.73s/it] 95%|█████████▌| 496/520 [31:09<01:29, 3.74s/it] {'loss': 1.0734, 'grad_norm': 0.0010231061764634454, 'learning_rate': 0.0011169173774871477, 'epoch': 0.95} + 95%|█████████▌| 496/520 [31:09<01:29, 3.74s/it] 96%|█████████▌| 497/520 [31:13<01:25, 3.73s/it] {'loss': 1.1325, 'grad_norm': 0.0008962950209467639, 'learning_rate': 0.0010259361921774012, 'epoch': 0.96} + 96%|█████████▌| 497/520 [31:13<01:25, 3.73s/it] 96%|█████████▌| 498/520 [31:16<01:21, 3.72s/it] {'loss': 1.1471, 'grad_norm': 0.000999322795716895, 'learning_rate': 0.000938800558694719, 'epoch': 0.96} + 96%|█████████▌| 498/520 [31:16<01:21, 3.72s/it] 96%|█████████▌| 499/520 [31:20<01:18, 3.73s/it] {'loss': 1.2714, 'grad_norm': 0.0009525749304024124, 'learning_rate': 0.0008555138626189618, 'epoch': 0.96} + 96%|█████████▌| 499/520 [31:20<01:18, 3.73s/it] 96%|█████████▌| 500/520 [31:24<01:14, 3.71s/it] {'loss': 1.2666, 'grad_norm': 0.0011475380244201656, 'learning_rate': 0.0007760793399827937, 'epoch': 0.96} + 96%|█████████▌| 500/520 [31:24<01:14, 3.71s/it] 96%|█████████▋| 501/520 [31:27<01:10, 3.71s/it] {'loss': 1.1778, 'grad_norm': 0.0010792199697102073, 'learning_rate': 0.000700500077146038, 'epoch': 0.96} + 96%|█████████▋| 501/520 [31:27<01:10, 3.71s/it] 97%|█████████▋| 502/520 [31:31<01:06, 3.70s/it] {'loss': 1.1868, 'grad_norm': 0.0009816920728617656, 'learning_rate': 0.0006287790106757397, 'epoch': 0.97} + 97%|█████████▋| 502/520 [31:31<01:06, 3.70s/it] 97%|█████████▋| 503/520 [31:35<01:02, 3.70s/it] {'loss': 1.1649, 'grad_norm': 0.0010362599021628174, 'learning_rate': 0.0005609189272320237, 'epoch': 0.97} + 97%|█████████▋| 503/520 [31:35<01:02, 3.70s/it] 97%|█████████▋| 504/520 [31:39<00:59, 3.70s/it] {'loss': 1.1819, 'grad_norm': 0.0011877816594212806, 'learning_rate': 0.000496922463459859, 'epoch': 0.97} + 97%|█████████▋| 504/520 [31:39<00:59, 3.70s/it] 97%|█████████▋| 505/520 [31:42<00:55, 3.69s/it] {'loss': 1.2142, 'grad_norm': 0.0010109515781496384, 'learning_rate': 0.0004367921058866187, 'epoch': 0.97} + 97%|█████████▋| 505/520 [31:42<00:55, 3.69s/it] 97%|█████████▋| 506/520 [31:46<00:51, 3.68s/it] {'loss': 1.1421, 'grad_norm': 0.0010394479218179431, 'learning_rate': 0.0003805301908254455, 'epoch': 0.97} + 97%|█████████▋| 506/520 [31:46<00:51, 3.68s/it] 98%|█████████▊| 507/520 [31:50<00:47, 3.69s/it] {'loss': 1.3146, 'grad_norm': 0.0009739992691940968, 'learning_rate': 0.0003281389042844918, 'epoch': 0.97} + 98%|█████████▊| 507/520 [31:50<00:47, 3.69s/it] 98%|█████████▊| 508/520 [31:53<00:44, 3.69s/it] {'loss': 1.2597, 'grad_norm': 0.0010371151692124111, 'learning_rate': 0.00027962028188198705, 'epoch': 0.98} + 98%|█████████▊| 508/520 [31:53<00:44, 3.69s/it] 98%|█████████▊| 509/520 [31:57<00:40, 3.70s/it] {'loss': 1.2321, 'grad_norm': 0.0009916292322614383, 'learning_rate': 0.00023497620876711256, 'epoch': 0.98} + 98%|█████████▊| 509/520 [31:57<00:40, 3.70s/it] 98%|█████████▊| 510/520 [32:01<00:36, 3.68s/it] {'loss': 1.1781, 'grad_norm': 0.0010249047316303447, 'learning_rate': 0.00019420841954681523, 'epoch': 0.98} + 98%|█████████▊| 510/520 [32:01<00:36, 3.68s/it] 98%|█████████▊| 511/520 [32:04<00:33, 3.67s/it] {'loss': 1.148, 'grad_norm': 0.0009980133915307985, 'learning_rate': 0.00015731849821833956, 'epoch': 0.98} + 98%|█████████▊| 511/520 [32:04<00:33, 3.67s/it] 98%|█████████▊| 512/520 [32:08<00:29, 3.70s/it] {'loss': 1.0395, 'grad_norm': 0.0010068539732754306, 'learning_rate': 0.00012430787810776555, 'epoch': 0.98} + 98%|█████████▊| 512/520 [32:08<00:29, 3.70s/it] 99%|█████████▊| 513/520 [32:12<00:26, 3.71s/it] {'loss': 1.2414, 'grad_norm': 0.0012055174255335744, 'learning_rate': 9.517784181422018e-05, 'epoch': 0.99} + 99%|█████████▊| 513/520 [32:12<00:26, 3.71s/it] 99%|█████████▉| 514/520 [32:16<00:22, 3.72s/it] {'loss': 1.2022, 'grad_norm': 0.0009347623786238357, 'learning_rate': 6.992952116013917e-05, 'epoch': 0.99} + 99%|█████████▉| 514/520 [32:16<00:22, 3.72s/it] 99%|█████████▉| 515/520 [32:19<00:18, 3.71s/it] {'loss': 1.2517, 'grad_norm': 0.0011988394281002585, 'learning_rate': 4.856389714723575e-05, 'epoch': 0.99} + 99%|█████████▉| 515/520 [32:19<00:18, 3.71s/it] 99%|█████████▉| 516/520 [32:23<00:14, 3.72s/it] {'loss': 1.1603, 'grad_norm': 0.0009882462162350194, 'learning_rate': 3.108179991837545e-05, 'epoch': 0.99} + 99%|█████████▉| 516/520 [32:23<00:14, 3.72s/it] 99%|█████████▉| 517/520 [32:27<00:11, 3.71s/it] {'loss': 1.2108, 'grad_norm': 0.0009822925666062422, 'learning_rate': 1.7483908725357544e-05, 'epoch': 0.99} + 99%|█████████▉| 517/520 [32:27<00:11, 3.71s/it] 100%|█████████▉| 518/520 [32:30<00:07, 3.69s/it] {'loss': 1.1733, 'grad_norm': 0.0011708592921859519, 'learning_rate': 7.770751902513862e-06, 'epoch': 1.0} + 100%|█████████▉| 518/520 [32:30<00:07, 3.69s/it] 100%|█████████▉| 519/520 [32:34<00:03, 3.69s/it] {'loss': 1.176, 'grad_norm': 0.0010082788857408967, 'learning_rate': 1.9427068461808086e-06, 'epoch': 1.0} + 100%|█████████▉| 519/520 [32:34<00:03, 3.69s/it] 100%|██████████| 520/520 [32:39<00:00, 3.94s/it] {'loss': 1.1789, 'grad_norm': 0.0009405886026133902, 'learning_rate': 0.0, 'epoch': 1.0} + 100%|██████████| 520/520 [32:39<00:00, 3.94s/it] {'train_runtime': 1959.0677, 'train_samples_per_second': 33.96, 'train_steps_per_second': 0.265, 'train_loss': 1.2405493433658894, 'epoch': 1.0} + 100%|██████████| 520/520 [32:39<00:00, 3.94s/it] 100%|██████████| 520/520 [32:39<00:00, 3.77s/it] +[2025-10-12 14:41:03,034] [INFO] [launch.py:348:main] Process 379804 exits successfully. +[2025-10-12 14:41:03,034] [INFO] [launch.py:348:main] Process 379803 exits successfully. +[2025-10-12 14:41:04,036] [INFO] [launch.py:348:main] Process 379805 exits successfully. +[2025-10-12 14:41:04,036] [INFO] [launch.py:348:main] Process 379802 exits successfully. +[2025-10-12 14:41:04,037] [INFO] [launch.py:348:main] Process 379799 exits successfully. +[2025-10-12 14:41:04,037] [INFO] [launch.py:348:main] Process 379800 exits successfully. +[2025-10-12 14:41:05,038] [INFO] [launch.py:348:main] Process 379801 exits successfully. +[2025-10-12 14:41:08,042] [INFO] [launch.py:348:main] Process 379798 exits successfully. +==== EXPERIMENT COMPLETED: qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.7_2e-1_connector-3.0_0.7_2e-1_ablation ==== +Log File: qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.7_2e-1_connector-3.0_0.7_2e-1_ablation_20251012_133702.log +Timestamp: 2025-10-12 14:41:10 +===================================== diff --git a/logs_oct11/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.9_2e-1_connector-3.0_0.9_2e-1_ablation_20251012_144110.log b/logs_oct11/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.9_2e-1_connector-3.0_0.9_2e-1_ablation_20251012_144110.log new file mode 100644 index 0000000000000000000000000000000000000000..33bdc7bb53e703730f29bfb310472b9a8213cc65 --- /dev/null +++ b/logs_oct11/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.9_2e-1_connector-3.0_0.9_2e-1_ablation_20251012_144110.log @@ -0,0 +1,2312 @@ +==== STARTING EXPERIMENT: qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.9_2e-1_connector-3.0_0.9_2e-1_ablation ==== +Log File: qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.9_2e-1_connector-3.0_0.9_2e-1_ablation_20251012_144110.log +Timestamp: 2025-10-12 14:41:10 +===================================== +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 14:41:13,167] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 14:41:15,848] [WARNING] [runner.py:202:fetch_hostfile] Unable to find hostfile, will proceed with training with local resources only. +[2025-10-12 14:41:15,849] [INFO] [runner.py:568:main] cmd = /opt/conda/envs/tinyllava/bin/python3.10 -u -m deepspeed.launcher.launch --world_info=eyJsb2NhbGhvc3QiOiBbMCwgMSwgMiwgMywgNCwgNSwgNiwgN119 --master_addr=127.0.0.1 --master_port=29501 --enable_each_rank_log=None tinyllava/train/train.py --deepspeed ./scripts/zero3.json --data_path /s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json --image_folder /s3-code/ywang29/datasets/tinyllava --is_multimodal True --conv_version qwen2_base --model_name_or_path Qwen/Qwen2.5-0.5B --vision_tower google/siglip-so400m-patch14-384 --vision_tower2 --connector_type mlp2x_gelu --mm_vision_select_layer -2 --image_aspect_ratio square --attn_implementation flash_attention_2 --bf16 True --training_recipe common --tune_type_llm full --tune_type_vision_tower frozen --tune_vision_tower_from_layer 0 --tune_type_connector full --group_by_modality_length True --pretrained_model_path /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain --output_dir /nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.9_2e-1_connector-3.0_0.9_2e-1_ablation --num_train_epochs 1 --per_device_train_batch_size 4 --per_device_eval_batch_size 4 --gradient_accumulation_steps 4 --evaluation_strategy no --learning_rate 2e-1 --weight_decay 0. --warmup_ratio 0.03 --lr_scheduler_type cosine --logging_steps 1 --tf32 False --model_max_length 2048 --gradient_checkpointing True --dataloader_num_workers 8 --lazy_preprocess True --report_to tensorboard --tokenizer_use_fast False --run_name tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune --subnet_mode_text both --subnet_type_text None --mask_type_text soft --init_mean_text 3.0 --temperature_attn_text 0.9 --temperature_mlp_text 0.9 --backward_type_text normal --masked_layers_text all --subnet_mode_vision both --subnet_type_vision None --mask_type_vision soft --init_mean_vision 3.0 --temperature_attn_vision 0.9 --temperature_mlp_vision 0.9 --backward_type_vision normal --masked_layers_vision all --subnet_type_connector global --mask_type_connector soft --init_mean_connector 3.0 --temperature_connector 0.9 --backward_type_connector normal --mm_projector_lr 2e-1 --seed 42 --mask_model llm-connector --save_strategy steps --save_steps 50000 --save_total_limit 1 --train_data_ratio 0.1 +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 14:41:18,419] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 14:41:19,512] [INFO] [launch.py:138:main] 0 NCCL_VERSION=2.21.5 +[2025-10-12 14:41:19,512] [INFO] [launch.py:138:main] 0 NCCL_SOCKET_IFNAME=eth +[2025-10-12 14:41:19,512] [INFO] [launch.py:145:main] WORLD INFO DICT: {'localhost': [0, 1, 2, 3, 4, 5, 6, 7]} +[2025-10-12 14:41:19,512] [INFO] [launch.py:151:main] nnodes=1, num_local_procs=8, node_rank=0 +[2025-10-12 14:41:19,512] [INFO] [launch.py:162:main] global_rank_mapping=defaultdict(, {'localhost': [0, 1, 2, 3, 4, 5, 6, 7]}) +[2025-10-12 14:41:19,512] [INFO] [launch.py:163:main] dist_world_size=8 +[2025-10-12 14:41:19,512] [INFO] [launch.py:165:main] Setting CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 +[2025-10-12 14:41:19,515] [INFO] [launch.py:253:main] process 405736 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=0', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.9_2e-1_connector-3.0_0.9_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.9', '--temperature_mlp_text', '0.9', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.9', '--temperature_mlp_vision', '0.9', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.9', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 14:41:19,517] [INFO] [launch.py:253:main] process 405737 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=1', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.9_2e-1_connector-3.0_0.9_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.9', '--temperature_mlp_text', '0.9', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.9', '--temperature_mlp_vision', '0.9', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.9', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 14:41:19,519] [INFO] [launch.py:253:main] process 405738 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=2', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.9_2e-1_connector-3.0_0.9_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.9', '--temperature_mlp_text', '0.9', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.9', '--temperature_mlp_vision', '0.9', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.9', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 14:41:19,521] [INFO] [launch.py:253:main] process 405739 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=3', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.9_2e-1_connector-3.0_0.9_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.9', '--temperature_mlp_text', '0.9', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.9', '--temperature_mlp_vision', '0.9', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.9', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 14:41:19,523] [INFO] [launch.py:253:main] process 405740 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=4', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.9_2e-1_connector-3.0_0.9_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.9', '--temperature_mlp_text', '0.9', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.9', '--temperature_mlp_vision', '0.9', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.9', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 14:41:19,525] [INFO] [launch.py:253:main] process 405741 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=5', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.9_2e-1_connector-3.0_0.9_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.9', '--temperature_mlp_text', '0.9', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.9', '--temperature_mlp_vision', '0.9', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.9', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 14:41:19,528] [INFO] [launch.py:253:main] process 405742 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=6', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.9_2e-1_connector-3.0_0.9_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.9', '--temperature_mlp_text', '0.9', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.9', '--temperature_mlp_vision', '0.9', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.9', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 14:41:19,530] [INFO] [launch.py:253:main] process 405743 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=7', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.9_2e-1_connector-3.0_0.9_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '0.9', '--temperature_mlp_text', '0.9', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '0.9', '--temperature_mlp_vision', '0.9', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '0.9', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 14:41:26,265] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 14:41:26,379] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 14:41:26,386] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 14:41:26,430] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 14:41:26,442] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 14:41:26,446] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 14:41:26,446] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 14:41:26,451] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 14:41:26,672] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 14:41:26,787] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 14:41:26,787] [INFO] [comm.py:668:init_distributed] Initializing TorchBackend in DeepSpeed with backend nccl +[2025-10-12 14:41:26,789] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 14:41:26,831] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 14:41:26,845] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 14:41:26,846] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 14:41:26,846] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 14:41:26,861] [INFO] [comm.py:637:init_distributed] cdb=None +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +{'llm': {'model_name_or_path': 'Qwen/Qwen2.5-0.5B', 'cache_dir': None, 'attn_implementation': 'flash_attention_2', 'subnet_mode': 'both', 'subnet_type': 'None', 'sparsity_attn': None, 'sparsity_mlp': None, 'threshold_attn': None, 'threshold_mlp': None, 'temperature_attn': 0.9, 'temperature_mlp': 0.9, 'masked_layers': 'all', 'mask_type': 'soft', 'backward_type': 'normal'}, 'vision_tower': {'model_name_or_path': 'google/siglip-so400m-patch14-384'}, 'connector': {'connector_type': 'mlp2x_gelu', 'subnet_type': 'global', 'threshold': None, 'sparsity': None, 'temperature': 0.9, 'mask_type': 'soft', 'backward_type': 'normal'}} +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +TinyLlavaConfig { + "backward_type_connector": "normal", + "cache_dir": null, + "connector_type": "mlp2x_gelu", + "hidden_size": 896, + "ignore_index": -100, + "image_aspect_ratio": "square", + "image_token_index": -200, + "llm_model_name_or_path": "Qwen/Qwen2.5-0.5B", + "mask_model": [ + "llm", + "connector" + ], + "mask_type_connector": "soft", + "model_type": "tinyllava", + "num_queries": 128, + "num_resampler_layers": 3, + "pad_token": null, + "resampler_hidden_size": 768, + "sparsity_connector": null, + "subnet_type_connector": "global", + "temperature_connector": 0.9, + "text_config": { + "_name_or_path": "Qwen/Qwen2.5-0.5B", + "architectures": [ + "Qwen2ForCausalLM" + ], + "backward_type": "normal", + "bos_token_id": 151643, + "eos_token_id": 151643, + "hidden_size": 896, + "intermediate_size": 4864, + "mask_type": "soft", + "masked_layers": "all", + "max_position_embeddings": 32768, + "max_window_layers": 24, + "model_type": "qwen2", + "num_attention_heads": 14, + "num_hidden_layers": 24, + "num_key_value_heads": 2, + "rope_theta": 1000000.0, + "sliding_window": 32768, + "subnet_mode": "both", + "subnet_type": "None", + "temperature_attn": 0.9, + "temperature_mlp": 0.9, + "tie_word_embeddings": true, + "torch_dtype": "bfloat16", + "use_mrope": false, + "use_sliding_window": false, + "vocab_size": 151936 + }, + "threshold_connector": null, + "tokenizer_model_max_length": 2048, + "tokenizer_name_or_path": "Qwen/Qwen2.5-0.5B", + "tokenizer_padding_side": "right", + "tokenizer_use_fast": false, + "transformers_version": "4.40.1", + "tune_type_connector": "frozen", + "tune_type_llm": "frozen", + "tune_type_vision_tower": "frozen", + "tune_vision_tower_from_layer": -1, + "use_cache": false, + "vision_config": { + "hidden_act": "gelu_pytorch_tanh", + "hidden_size": 1152, + "image_size": 384, + "intermediate_size": 4304, + "layer_norm_eps": 1e-06, + "model_name_or_path": "google/siglip-so400m-patch14-384", + "model_name_or_path2": "", + "model_type": "siglip_vision_model", + "num_attention_heads": 16, + "num_hidden_layers": 27, + "patch_size": 14 + }, + "vision_feature_layer": -2, + "vision_feature_select_strategy": "patch", + "vision_hidden_size": 1152, + "vision_model_name_or_path": "google/siglip-so400m-patch14-384", + "vision_model_name_or_path2": "", + "vocab_size": 151936 +} + +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test1-worker-0:405736:405736 [0] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:405736:405736 [0] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:405736:405736 [0] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:405736:405736 [0] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:405736:405736 [0] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test1-worker-0:405736:405736 [0] NCCL INFO cudaDriverVersion 12040 +NCCL version 2.21.5+cuda12.1 +ywang29-vrdb-test1-worker-0:405737:405737 [1] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test1-worker-0:405737:405737 [1] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:405737:405737 [1] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:405737:405737 [1] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:405737:405737 [1] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:405737:405737 [1] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test1-worker-0:405736:407304 [0] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:405736:407304 [0] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:405736:407304 [0] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:405736:407304 [0] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:405736:407304 [0] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:405736:407304 [0] NCCL INFO Using network Socket +ywang29-vrdb-test1-worker-0:405737:407305 [1] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:405737:407305 [1] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:405737:407305 [1] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:405737:407305 [1] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:405737:407305 [1] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:405737:407305 [1] NCCL INFO Using network Socket +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test1-worker-0:405742:405742 [6] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test1-worker-0:405742:405742 [6] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:405742:405742 [6] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:405742:405742 [6] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:405742:405742 [6] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:405742:405742 [6] NCCL INFO NET/Plugin: Using internal network plugin. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test1-worker-0:405738:405738 [2] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test1-worker-0:405738:405738 [2] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:405738:405738 [2] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:405738:405738 [2] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:405738:405738 [2] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:405738:405738 [2] NCCL INFO NET/Plugin: Using internal network plugin. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test1-worker-0:405741:405741 [5] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test1-worker-0:405741:405741 [5] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:405741:405741 [5] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:405741:405741 [5] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:405741:405741 [5] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:405741:405741 [5] NCCL INFO NET/Plugin: Using internal network plugin. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test1-worker-0:405739:405739 [3] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test1-worker-0:405739:405739 [3] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:405739:405739 [3] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:405739:405739 [3] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:405739:405739 [3] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:405739:405739 [3] NCCL INFO NET/Plugin: Using internal network plugin. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test1-worker-0:405742:407306 [6] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:405742:407306 [6] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:405742:407306 [6] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:405742:407306 [6] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:405742:407306 [6] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:405742:407306 [6] NCCL INFO Using network Socket +ywang29-vrdb-test1-worker-0:405740:405740 [4] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test1-worker-0:405740:405740 [4] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:405740:405740 [4] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:405740:405740 [4] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:405740:405740 [4] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:405740:405740 [4] NCCL INFO NET/Plugin: Using internal network plugin. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test1-worker-0:405738:407307 [2] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:405738:407307 [2] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:405738:407307 [2] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:405738:407307 [2] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:405738:407307 [2] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:405738:407307 [2] NCCL INFO Using network Socket +ywang29-vrdb-test1-worker-0:405743:405743 [7] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test1-worker-0:405743:405743 [7] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:405743:405743 [7] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:405743:405743 [7] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:405743:405743 [7] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:405743:405743 [7] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test1-worker-0:405741:407308 [5] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:405741:407308 [5] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:405741:407308 [5] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:405741:407308 [5] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:405741:407308 [5] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:405741:407308 [5] NCCL INFO Using network Socket +ywang29-vrdb-test1-worker-0:405739:407309 [3] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:405739:407309 [3] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:405739:407309 [3] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:405739:407309 [3] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:405739:407309 [3] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:405739:407309 [3] NCCL INFO Using network Socket +ywang29-vrdb-test1-worker-0:405740:407310 [4] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:405740:407310 [4] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:405740:407310 [4] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:405740:407310 [4] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:405740:407310 [4] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:405740:407310 [4] NCCL INFO Using network Socket +ywang29-vrdb-test1-worker-0:405743:407319 [7] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:405743:407319 [7] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:405743:407319 [7] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:405743:407319 [7] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:405743:407319 [7] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:405743:407319 [7] NCCL INFO Using network Socket +ywang29-vrdb-test1-worker-0:405743:407319 [7] NCCL INFO ncclCommInitRank comm 0x555f2adfd7e0 rank 7 nranks 8 cudaDev 7 nvmlDev 7 busId a01d0 commId 0x393efc2ff08be14b - Init START +ywang29-vrdb-test1-worker-0:405737:407305 [1] NCCL INFO ncclCommInitRank comm 0x55bf1466c580 rank 1 nranks 8 cudaDev 1 nvmlDev 1 busId 101d0 commId 0x393efc2ff08be14b - Init START +ywang29-vrdb-test1-worker-0:405738:407307 [2] NCCL INFO ncclCommInitRank comm 0x564ea713b740 rank 2 nranks 8 cudaDev 2 nvmlDev 2 busId 201c0 commId 0x393efc2ff08be14b - Init START +ywang29-vrdb-test1-worker-0:405742:407306 [6] NCCL INFO ncclCommInitRank comm 0x5623bb2d3d50 rank 6 nranks 8 cudaDev 6 nvmlDev 6 busId a01c0 commId 0x393efc2ff08be14b - Init START +ywang29-vrdb-test1-worker-0:405741:407308 [5] NCCL INFO ncclCommInitRank comm 0x55da54f5b280 rank 5 nranks 8 cudaDev 5 nvmlDev 5 busId 901d0 commId 0x393efc2ff08be14b - Init START +ywang29-vrdb-test1-worker-0:405736:407304 [0] NCCL INFO ncclCommInitRank comm 0x55bd09a355d0 rank 0 nranks 8 cudaDev 0 nvmlDev 0 busId 101c0 commId 0x393efc2ff08be14b - Init START +ywang29-vrdb-test1-worker-0:405740:407310 [4] NCCL INFO ncclCommInitRank comm 0x5606a94dd090 rank 4 nranks 8 cudaDev 4 nvmlDev 4 busId 901c0 commId 0x393efc2ff08be14b - Init START +ywang29-vrdb-test1-worker-0:405739:407309 [3] NCCL INFO ncclCommInitRank comm 0x559f23476c50 rank 3 nranks 8 cudaDev 3 nvmlDev 3 busId 201d0 commId 0x393efc2ff08be14b - Init START +ywang29-vrdb-test1-worker-0:405736:407304 [0] NCCL INFO Setting affinity for GPU 0 to ff,ffff0000,00ffffff +ywang29-vrdb-test1-worker-0:405736:407304 [0] NCCL INFO NVLS multicast support is not available on dev 0 +ywang29-vrdb-test1-worker-0:405742:407306 [6] NCCL INFO Setting affinity for GPU 6 to ffffff00,0000ffff,ff000000 +ywang29-vrdb-test1-worker-0:405742:407306 [6] NCCL INFO NVLS multicast support is not available on dev 6 +ywang29-vrdb-test1-worker-0:405739:407309 [3] NCCL INFO Setting affinity for GPU 3 to ff,ffff0000,00ffffff +ywang29-vrdb-test1-worker-0:405741:407308 [5] NCCL INFO Setting affinity for GPU 5 to ffffff00,0000ffff,ff000000 +ywang29-vrdb-test1-worker-0:405739:407309 [3] NCCL INFO NVLS multicast support is not available on dev 3 +ywang29-vrdb-test1-worker-0:405741:407308 [5] NCCL INFO NVLS multicast support is not available on dev 5 +ywang29-vrdb-test1-worker-0:405743:407319 [7] NCCL INFO Setting affinity for GPU 7 to ffffff00,0000ffff,ff000000 +ywang29-vrdb-test1-worker-0:405743:407319 [7] NCCL INFO NVLS multicast support is not available on dev 7 +ywang29-vrdb-test1-worker-0:405740:407310 [4] NCCL INFO Setting affinity for GPU 4 to ffffff00,0000ffff,ff000000 +ywang29-vrdb-test1-worker-0:405740:407310 [4] NCCL INFO NVLS multicast support is not available on dev 4 +ywang29-vrdb-test1-worker-0:405738:407307 [2] NCCL INFO Setting affinity for GPU 2 to ff,ffff0000,00ffffff +ywang29-vrdb-test1-worker-0:405738:407307 [2] NCCL INFO NVLS multicast support is not available on dev 2 +ywang29-vrdb-test1-worker-0:405737:407305 [1] NCCL INFO Setting affinity for GPU 1 to ff,ffff0000,00ffffff +ywang29-vrdb-test1-worker-0:405737:407305 [1] NCCL INFO NVLS multicast support is not available on dev 1 +ywang29-vrdb-test1-worker-0:405741:407308 [5] NCCL INFO comm 0x55da54f5b280 rank 5 nRanks 8 nNodes 1 localRanks 8 localRank 5 MNNVL 0 +ywang29-vrdb-test1-worker-0:405740:407310 [4] NCCL INFO comm 0x5606a94dd090 rank 4 nRanks 8 nNodes 1 localRanks 8 localRank 4 MNNVL 0 +ywang29-vrdb-test1-worker-0:405738:407307 [2] NCCL INFO comm 0x564ea713b740 rank 2 nRanks 8 nNodes 1 localRanks 8 localRank 2 MNNVL 0 +ywang29-vrdb-test1-worker-0:405739:407309 [3] NCCL INFO comm 0x559f23476c50 rank 3 nRanks 8 nNodes 1 localRanks 8 localRank 3 MNNVL 0 +ywang29-vrdb-test1-worker-0:405737:407305 [1] NCCL INFO comm 0x55bf1466c580 rank 1 nRanks 8 nNodes 1 localRanks 8 localRank 1 MNNVL 0 +ywang29-vrdb-test1-worker-0:405743:407319 [7] NCCL INFO comm 0x555f2adfd7e0 rank 7 nRanks 8 nNodes 1 localRanks 8 localRank 7 MNNVL 0 +ywang29-vrdb-test1-worker-0:405742:407306 [6] NCCL INFO comm 0x5623bb2d3d50 rank 6 nRanks 8 nNodes 1 localRanks 8 localRank 6 MNNVL 0 +ywang29-vrdb-test1-worker-0:405736:407304 [0] NCCL INFO comm 0x55bd09a355d0 rank 0 nRanks 8 nNodes 1 localRanks 8 localRank 0 MNNVL 0 +ywang29-vrdb-test1-worker-0:405736:407304 [0] NCCL INFO Channel 00/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:405740:407310 [4] NCCL INFO Trees [0] 5/-1/-1->4->3 [1] 5/-1/-1->4->3 [2] 5/-1/-1->4->3 [3] 5/-1/-1->4->3 [4] 5/-1/-1->4->3 [5] 5/-1/-1->4->3 [6] 5/-1/-1->4->3 [7] 5/-1/-1->4->3 [8] 5/-1/-1->4->3 [9] 5/-1/-1->4->3 [10] 5/-1/-1->4->3 [11] 5/-1/-1->4->3 [12] 5/-1/-1->4->3 [13] 5/-1/-1->4->3 [14] 5/-1/-1->4->3 [15] 5/-1/-1->4->3 [16] 5/-1/-1->4->3 [17] 5/-1/-1->4->3 [18] 5/-1/-1->4->3 [19] 5/-1/-1->4->3 [20] 5/-1/-1->4->3 [21] 5/-1/-1->4->3 [22] 5/-1/-1->4->3 [23] 5/-1/-1->4->3 +ywang29-vrdb-test1-worker-0:405738:407307 [2] NCCL INFO Trees [0] 3/-1/-1->2->1 [1] 3/-1/-1->2->1 [2] 3/-1/-1->2->1 [3] 3/-1/-1->2->1 [4] 3/-1/-1->2->1 [5] 3/-1/-1->2->1 [6] 3/-1/-1->2->1 [7] 3/-1/-1->2->1 [8] 3/-1/-1->2->1 [9] 3/-1/-1->2->1 [10] 3/-1/-1->2->1 [11] 3/-1/-1->2->1 [12] 3/-1/-1->2->1 [13] 3/-1/-1->2->1 [14] 3/-1/-1->2->1 [15] 3/-1/-1->2->1 [16] 3/-1/-1->2->1 [17] 3/-1/-1->2->1 [18] 3/-1/-1->2->1 [19] 3/-1/-1->2->1 [20] 3/-1/-1->2->1 [21] 3/-1/-1->2->1 [22] 3/-1/-1->2->1 [23] 3/-1/-1->2->1 +ywang29-vrdb-test1-worker-0:405739:407309 [3] NCCL INFO Trees [0] 4/-1/-1->3->2 [1] 4/-1/-1->3->2 [2] 4/-1/-1->3->2 [3] 4/-1/-1->3->2 [4] 4/-1/-1->3->2 [5] 4/-1/-1->3->2 [6] 4/-1/-1->3->2 [7] 4/-1/-1->3->2 [8] 4/-1/-1->3->2 [9] 4/-1/-1->3->2 [10] 4/-1/-1->3->2 [11] 4/-1/-1->3->2 [12] 4/-1/-1->3->2 [13] 4/-1/-1->3->2 [14] 4/-1/-1->3->2 [15] 4/-1/-1->3->2 [16] 4/-1/-1->3->2 [17] 4/-1/-1->3->2 [18] 4/-1/-1->3->2 [19] 4/-1/-1->3->2 [20] 4/-1/-1->3->2 [21] 4/-1/-1->3->2 [22] 4/-1/-1->3->2 [23] 4/-1/-1->3->2 +ywang29-vrdb-test1-worker-0:405741:407308 [5] NCCL INFO Trees [0] 6/-1/-1->5->4 [1] 6/-1/-1->5->4 [2] 6/-1/-1->5->4 [3] 6/-1/-1->5->4 [4] 6/-1/-1->5->4 [5] 6/-1/-1->5->4 [6] 6/-1/-1->5->4 [7] 6/-1/-1->5->4 [8] 6/-1/-1->5->4 [9] 6/-1/-1->5->4 [10] 6/-1/-1->5->4 [11] 6/-1/-1->5->4 [12] 6/-1/-1->5->4 [13] 6/-1/-1->5->4 [14] 6/-1/-1->5->4 [15] 6/-1/-1->5->4 [16] 6/-1/-1->5->4 [17] 6/-1/-1->5->4 [18] 6/-1/-1->5->4 [19] 6/-1/-1->5->4 [20] 6/-1/-1->5->4 [21] 6/-1/-1->5->4 [22] 6/-1/-1->5->4 [23] 6/-1/-1->5->4 +ywang29-vrdb-test1-worker-0:405736:407304 [0] NCCL INFO Channel 01/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:405738:407307 [2] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:405739:407309 [3] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:405740:407310 [4] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:405736:407304 [0] NCCL INFO Channel 02/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:405741:407308 [5] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:405736:407304 [0] NCCL INFO Channel 03/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:405736:407304 [0] NCCL INFO Channel 04/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:405737:407305 [1] NCCL INFO Trees [0] 2/-1/-1->1->0 [1] 2/-1/-1->1->0 [2] 2/-1/-1->1->0 [3] 2/-1/-1->1->0 [4] 2/-1/-1->1->0 [5] 2/-1/-1->1->0 [6] 2/-1/-1->1->0 [7] 2/-1/-1->1->0 [8] 2/-1/-1->1->0 [9] 2/-1/-1->1->0 [10] 2/-1/-1->1->0 [11] 2/-1/-1->1->0 [12] 2/-1/-1->1->0 [13] 2/-1/-1->1->0 [14] 2/-1/-1->1->0 [15] 2/-1/-1->1->0 [16] 2/-1/-1->1->0 [17] 2/-1/-1->1->0 [18] 2/-1/-1->1->0 [19] 2/-1/-1->1->0 [20] 2/-1/-1->1->0 [21] 2/-1/-1->1->0 [22] 2/-1/-1->1->0 [23] 2/-1/-1->1->0 +ywang29-vrdb-test1-worker-0:405736:407304 [0] NCCL INFO Channel 05/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:405743:407319 [7] NCCL INFO Trees [0] -1/-1/-1->7->6 [1] -1/-1/-1->7->6 [2] -1/-1/-1->7->6 [3] -1/-1/-1->7->6 [4] -1/-1/-1->7->6 [5] -1/-1/-1->7->6 [6] -1/-1/-1->7->6 [7] -1/-1/-1->7->6 [8] -1/-1/-1->7->6 [9] -1/-1/-1->7->6 [10] -1/-1/-1->7->6 [11] -1/-1/-1->7->6 [12] -1/-1/-1->7->6 [13] -1/-1/-1->7->6 [14] -1/-1/-1->7->6 [15] -1/-1/-1->7->6 [16] -1/-1/-1->7->6 [17] -1/-1/-1->7->6 [18] -1/-1/-1->7->6 [19] -1/-1/-1->7->6 [20] -1/-1/-1->7->6 [21] -1/-1/-1->7->6 [22] -1/-1/-1->7->6 [23] -1/-1/-1->7->6 +ywang29-vrdb-test1-worker-0:405736:407304 [0] NCCL INFO Channel 06/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:405737:407305 [1] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:405736:407304 [0] NCCL INFO Channel 07/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:405743:407319 [7] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:405736:407304 [0] NCCL INFO Channel 08/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:405742:407306 [6] NCCL INFO Trees [0] 7/-1/-1->6->5 [1] 7/-1/-1->6->5 [2] 7/-1/-1->6->5 [3] 7/-1/-1->6->5 [4] 7/-1/-1->6->5 [5] 7/-1/-1->6->5 [6] 7/-1/-1->6->5 [7] 7/-1/-1->6->5 [8] 7/-1/-1->6->5 [9] 7/-1/-1->6->5 [10] 7/-1/-1->6->5 [11] 7/-1/-1->6->5 [12] 7/-1/-1->6->5 [13] 7/-1/-1->6->5 [14] 7/-1/-1->6->5 [15] 7/-1/-1->6->5 [16] 7/-1/-1->6->5 [17] 7/-1/-1->6->5 [18] 7/-1/-1->6->5 [19] 7/-1/-1->6->5 [20] 7/-1/-1->6->5 [21] 7/-1/-1->6->5 [22] 7/-1/-1->6->5 [23] 7/-1/-1->6->5 +ywang29-vrdb-test1-worker-0:405736:407304 [0] NCCL INFO Channel 09/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:405736:407304 [0] NCCL INFO Channel 10/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:405736:407304 [0] NCCL INFO Channel 11/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:405742:407306 [6] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:405736:407304 [0] NCCL INFO Channel 12/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:405736:407304 [0] NCCL INFO Channel 13/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:405736:407304 [0] NCCL INFO Channel 14/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:405736:407304 [0] NCCL INFO Channel 15/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:405736:407304 [0] NCCL INFO Channel 16/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:405736:407304 [0] NCCL INFO Channel 17/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:405736:407304 [0] NCCL INFO Channel 18/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:405736:407304 [0] NCCL INFO Channel 19/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:405736:407304 [0] NCCL INFO Channel 20/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:405736:407304 [0] NCCL INFO Channel 21/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:405736:407304 [0] NCCL INFO Channel 22/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:405736:407304 [0] NCCL INFO Channel 23/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:405736:407304 [0] NCCL INFO Trees [0] 1/-1/-1->0->-1 [1] 1/-1/-1->0->-1 [2] 1/-1/-1->0->-1 [3] 1/-1/-1->0->-1 [4] 1/-1/-1->0->-1 [5] 1/-1/-1->0->-1 [6] 1/-1/-1->0->-1 [7] 1/-1/-1->0->-1 [8] 1/-1/-1->0->-1 [9] 1/-1/-1->0->-1 [10] 1/-1/-1->0->-1 [11] 1/-1/-1->0->-1 [12] 1/-1/-1->0->-1 [13] 1/-1/-1->0->-1 [14] 1/-1/-1->0->-1 [15] 1/-1/-1->0->-1 [16] 1/-1/-1->0->-1 [17] 1/-1/-1->0->-1 [18] 1/-1/-1->0->-1 [19] 1/-1/-1->0->-1 [20] 1/-1/-1->0->-1 [21] 1/-1/-1->0->-1 [22] 1/-1/-1->0->-1 [23] 1/-1/-1->0->-1 +ywang29-vrdb-test1-worker-0:405736:407304 [0] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:405740:407310 [4] NCCL INFO Channel 00/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:407309 [3] NCCL INFO Channel 00/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:407308 [5] NCCL INFO Channel 00/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:407310 [4] NCCL INFO Channel 01/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:407307 [2] NCCL INFO Channel 00/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:407309 [3] NCCL INFO Channel 01/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:407308 [5] NCCL INFO Channel 01/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:407310 [4] NCCL INFO Channel 02/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:407307 [2] NCCL INFO Channel 01/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:407309 [3] NCCL INFO Channel 02/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:407308 [5] NCCL INFO Channel 02/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:407310 [4] NCCL INFO Channel 03/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:407307 [2] NCCL INFO Channel 02/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405736:407304 [0] NCCL INFO Channel 00/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:407309 [3] NCCL INFO Channel 03/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:407319 [7] NCCL INFO Channel 00/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:407308 [5] NCCL INFO Channel 03/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:407310 [4] NCCL INFO Channel 04/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:407306 [6] NCCL INFO Channel 00/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:407307 [2] NCCL INFO Channel 03/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405736:407304 [0] NCCL INFO Channel 01/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:407309 [3] NCCL INFO Channel 04/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:407319 [7] NCCL INFO Channel 01/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:407308 [5] NCCL INFO Channel 04/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:407310 [4] NCCL INFO Channel 05/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:407306 [6] NCCL INFO Channel 01/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:407307 [2] NCCL INFO Channel 04/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405736:407304 [0] NCCL INFO Channel 02/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:407309 [3] NCCL INFO Channel 05/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:407319 [7] NCCL INFO Channel 02/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:407308 [5] NCCL INFO Channel 05/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:407310 [4] NCCL INFO Channel 06/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:407306 [6] NCCL INFO Channel 02/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:407307 [2] NCCL INFO Channel 05/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405736:407304 [0] NCCL INFO Channel 03/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:407309 [3] NCCL INFO Channel 06/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:407319 [7] NCCL INFO Channel 03/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:407308 [5] NCCL INFO Channel 06/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:407310 [4] NCCL INFO Channel 07/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:407306 [6] NCCL INFO Channel 03/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405736:407304 [0] NCCL INFO Channel 04/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:407307 [2] NCCL INFO Channel 06/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:407309 [3] NCCL INFO Channel 07/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:407308 [5] NCCL INFO Channel 07/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:407319 [7] NCCL INFO Channel 04/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:407310 [4] NCCL INFO Channel 08/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:407306 [6] NCCL INFO Channel 04/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:407307 [2] NCCL INFO Channel 07/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405736:407304 [0] NCCL INFO Channel 05/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:407309 [3] NCCL INFO Channel 08/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:407308 [5] NCCL INFO Channel 08/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:407310 [4] NCCL INFO Channel 09/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:407319 [7] NCCL INFO Channel 05/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:407306 [6] NCCL INFO Channel 05/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:407307 [2] NCCL INFO Channel 08/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:407309 [3] NCCL INFO Channel 09/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405736:407304 [0] NCCL INFO Channel 06/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:407308 [5] NCCL INFO Channel 09/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:407310 [4] NCCL INFO Channel 10/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:407319 [7] NCCL INFO Channel 06/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:407306 [6] NCCL INFO Channel 06/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:407307 [2] NCCL INFO Channel 09/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:407309 [3] NCCL INFO Channel 10/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405736:407304 [0] NCCL INFO Channel 07/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:407308 [5] NCCL INFO Channel 10/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:407310 [4] NCCL INFO Channel 11/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:407319 [7] NCCL INFO Channel 07/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:407306 [6] NCCL INFO Channel 07/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:407307 [2] NCCL INFO Channel 10/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:407309 [3] NCCL INFO Channel 11/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405736:407304 [0] NCCL INFO Channel 08/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:407308 [5] NCCL INFO Channel 11/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:407310 [4] NCCL INFO Channel 12/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:407306 [6] NCCL INFO Channel 08/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:407319 [7] NCCL INFO Channel 08/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:407307 [2] NCCL INFO Channel 11/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:407309 [3] NCCL INFO Channel 12/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405736:407304 [0] NCCL INFO Channel 09/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:407308 [5] NCCL INFO Channel 12/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:407310 [4] NCCL INFO Channel 13/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:407306 [6] NCCL INFO Channel 09/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:407319 [7] NCCL INFO Channel 09/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:407307 [2] NCCL INFO Channel 12/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:407309 [3] NCCL INFO Channel 13/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405736:407304 [0] NCCL INFO Channel 10/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:407308 [5] NCCL INFO Channel 13/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:407310 [4] NCCL INFO Channel 14/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:407306 [6] NCCL INFO Channel 10/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:407319 [7] NCCL INFO Channel 10/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:407307 [2] NCCL INFO Channel 13/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:407309 [3] NCCL INFO Channel 14/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:407310 [4] NCCL INFO Channel 15/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:407308 [5] NCCL INFO Channel 14/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405736:407304 [0] NCCL INFO Channel 11/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:407306 [6] NCCL INFO Channel 11/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:407307 [2] NCCL INFO Channel 14/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:407319 [7] NCCL INFO Channel 11/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:407309 [3] NCCL INFO Channel 15/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:407310 [4] NCCL INFO Channel 16/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:407308 [5] NCCL INFO Channel 15/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405736:407304 [0] NCCL INFO Channel 12/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:407306 [6] NCCL INFO Channel 12/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:407307 [2] NCCL INFO Channel 15/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:407319 [7] NCCL INFO Channel 12/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:407309 [3] NCCL INFO Channel 16/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:407310 [4] NCCL INFO Channel 17/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:407308 [5] NCCL INFO Channel 16/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405736:407304 [0] NCCL INFO Channel 13/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:407306 [6] NCCL INFO Channel 13/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:407307 [2] NCCL INFO Channel 16/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:407319 [7] NCCL INFO Channel 13/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:407309 [3] NCCL INFO Channel 17/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:407310 [4] NCCL INFO Channel 18/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:407308 [5] NCCL INFO Channel 17/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405736:407304 [0] NCCL INFO Channel 14/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:407307 [2] NCCL INFO Channel 17/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:407306 [6] NCCL INFO Channel 14/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:407319 [7] NCCL INFO Channel 14/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:407309 [3] NCCL INFO Channel 18/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:407310 [4] NCCL INFO Channel 19/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:407308 [5] NCCL INFO Channel 18/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405736:407304 [0] NCCL INFO Channel 15/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:407307 [2] NCCL INFO Channel 18/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:407306 [6] NCCL INFO Channel 15/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:407319 [7] NCCL INFO Channel 15/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:407309 [3] NCCL INFO Channel 19/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:407310 [4] NCCL INFO Channel 20/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:407308 [5] NCCL INFO Channel 19/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405736:407304 [0] NCCL INFO Channel 16/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:407307 [2] NCCL INFO Channel 19/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:407309 [3] NCCL INFO Channel 20/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:407306 [6] NCCL INFO Channel 16/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:407319 [7] NCCL INFO Channel 16/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:407305 [1] NCCL INFO Channel 00/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:407310 [4] NCCL INFO Channel 21/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:407308 [5] NCCL INFO Channel 20/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405736:407304 [0] NCCL INFO Channel 17/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:407307 [2] NCCL INFO Channel 20/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:407309 [3] NCCL INFO Channel 21/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:407306 [6] NCCL INFO Channel 17/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:407305 [1] NCCL INFO Channel 01/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:407319 [7] NCCL INFO Channel 17/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:407310 [4] NCCL INFO Channel 22/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:407308 [5] NCCL INFO Channel 21/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405736:407304 [0] NCCL INFO Channel 18/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:407307 [2] NCCL INFO Channel 21/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:407309 [3] NCCL INFO Channel 22/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:407306 [6] NCCL INFO Channel 18/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:407305 [1] NCCL INFO Channel 02/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:407319 [7] NCCL INFO Channel 18/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:407310 [4] NCCL INFO Channel 23/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:407308 [5] NCCL INFO Channel 22/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:407307 [2] NCCL INFO Channel 22/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405736:407304 [0] NCCL INFO Channel 19/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:407309 [3] NCCL INFO Channel 23/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:407306 [6] NCCL INFO Channel 19/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:407305 [1] NCCL INFO Channel 03/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:407319 [7] NCCL INFO Channel 19/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:407308 [5] NCCL INFO Channel 23/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:407307 [2] NCCL INFO Channel 23/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405736:407304 [0] NCCL INFO Channel 20/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:407306 [6] NCCL INFO Channel 20/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:407305 [1] NCCL INFO Channel 04/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:407319 [7] NCCL INFO Channel 20/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405736:407304 [0] NCCL INFO Channel 21/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:407306 [6] NCCL INFO Channel 21/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:407305 [1] NCCL INFO Channel 05/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405736:407304 [0] NCCL INFO Channel 22/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:407319 [7] NCCL INFO Channel 21/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:407306 [6] NCCL INFO Channel 22/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405736:407304 [0] NCCL INFO Channel 23/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:407319 [7] NCCL INFO Channel 22/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:407305 [1] NCCL INFO Channel 06/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:407319 [7] NCCL INFO Channel 23/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:407305 [1] NCCL INFO Channel 07/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:407306 [6] NCCL INFO Channel 23/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:407305 [1] NCCL INFO Channel 08/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:407305 [1] NCCL INFO Channel 09/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:407305 [1] NCCL INFO Channel 10/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:407305 [1] NCCL INFO Channel 11/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:407305 [1] NCCL INFO Channel 12/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:407305 [1] NCCL INFO Channel 13/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:407305 [1] NCCL INFO Channel 14/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:407305 [1] NCCL INFO Channel 15/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:407305 [1] NCCL INFO Channel 16/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:407305 [1] NCCL INFO Channel 17/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:407305 [1] NCCL INFO Channel 18/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:407305 [1] NCCL INFO Channel 19/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:407305 [1] NCCL INFO Channel 20/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:407305 [1] NCCL INFO Channel 21/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:407305 [1] NCCL INFO Channel 22/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:407305 [1] NCCL INFO Channel 23/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:407309 [3] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:405740:407310 [4] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:405738:407307 [2] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:405737:407305 [1] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:405736:407304 [0] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:405741:407308 [5] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:405742:407306 [6] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:405739:407309 [3] NCCL INFO Channel 00/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:407309 [3] NCCL INFO Channel 01/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:407309 [3] NCCL INFO Channel 02/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:407309 [3] NCCL INFO Channel 03/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:407309 [3] NCCL INFO Channel 04/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:407319 [7] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:405743:407319 [7] NCCL INFO Channel 00/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:407309 [3] NCCL INFO Channel 05/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:407319 [7] NCCL INFO Channel 01/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:407309 [3] NCCL INFO Channel 06/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:407319 [7] NCCL INFO Channel 02/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:407309 [3] NCCL INFO Channel 07/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:407319 [7] NCCL INFO Channel 03/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:407309 [3] NCCL INFO Channel 08/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:407319 [7] NCCL INFO Channel 04/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:407309 [3] NCCL INFO Channel 09/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:407319 [7] NCCL INFO Channel 05/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:407309 [3] NCCL INFO Channel 10/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:407319 [7] NCCL INFO Channel 06/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:407307 [2] NCCL INFO Channel 00/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:407309 [3] NCCL INFO Channel 11/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:407319 [7] NCCL INFO Channel 07/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:407307 [2] NCCL INFO Channel 01/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:407309 [3] NCCL INFO Channel 12/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:407319 [7] NCCL INFO Channel 08/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:407305 [1] NCCL INFO Channel 00/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:407307 [2] NCCL INFO Channel 02/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:407309 [3] NCCL INFO Channel 13/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:407319 [7] NCCL INFO Channel 09/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:407305 [1] NCCL INFO Channel 01/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:407307 [2] NCCL INFO Channel 03/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:407309 [3] NCCL INFO Channel 14/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:407319 [7] NCCL INFO Channel 10/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:407305 [1] NCCL INFO Channel 02/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:407307 [2] NCCL INFO Channel 04/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:407309 [3] NCCL INFO Channel 15/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:407319 [7] NCCL INFO Channel 11/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:407305 [1] NCCL INFO Channel 03/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:407307 [2] NCCL INFO Channel 05/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:407309 [3] NCCL INFO Channel 16/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:407310 [4] NCCL INFO Channel 00/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:407319 [7] NCCL INFO Channel 12/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:407305 [1] NCCL INFO Channel 04/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:407307 [2] NCCL INFO Channel 06/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:407309 [3] NCCL INFO Channel 17/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:407310 [4] NCCL INFO Channel 01/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:407319 [7] NCCL INFO Channel 13/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:407305 [1] NCCL INFO Channel 05/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:407307 [2] NCCL INFO Channel 07/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:407309 [3] NCCL INFO Channel 18/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:407310 [4] NCCL INFO Channel 02/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:407319 [7] NCCL INFO Channel 14/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:407305 [1] NCCL INFO Channel 06/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:407307 [2] NCCL INFO Channel 08/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:407309 [3] NCCL INFO Channel 19/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:407310 [4] NCCL INFO Channel 03/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:407319 [7] NCCL INFO Channel 15/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:407305 [1] NCCL INFO Channel 07/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:407307 [2] NCCL INFO Channel 09/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:407309 [3] NCCL INFO Channel 20/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:407310 [4] NCCL INFO Channel 04/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:407319 [7] NCCL INFO Channel 16/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:407305 [1] NCCL INFO Channel 08/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:407307 [2] NCCL INFO Channel 10/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:407309 [3] NCCL INFO Channel 21/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:407310 [4] NCCL INFO Channel 05/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:407319 [7] NCCL INFO Channel 17/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:407305 [1] NCCL INFO Channel 09/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:407307 [2] NCCL INFO Channel 11/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:407309 [3] NCCL INFO Channel 22/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:407310 [4] NCCL INFO Channel 06/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:407319 [7] NCCL INFO Channel 18/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:407305 [1] NCCL INFO Channel 10/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:407307 [2] NCCL INFO Channel 12/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:407309 [3] NCCL INFO Channel 23/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:407310 [4] NCCL INFO Channel 07/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:407308 [5] NCCL INFO Channel 00/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:407306 [6] NCCL INFO Channel 00/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:407319 [7] NCCL INFO Channel 19/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:407305 [1] NCCL INFO Channel 11/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:407307 [2] NCCL INFO Channel 13/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:407310 [4] NCCL INFO Channel 08/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:407308 [5] NCCL INFO Channel 01/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:407306 [6] NCCL INFO Channel 01/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:407319 [7] NCCL INFO Channel 20/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:407305 [1] NCCL INFO Channel 12/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:407307 [2] NCCL INFO Channel 14/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:407310 [4] NCCL INFO Channel 09/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:407306 [6] NCCL INFO Channel 02/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:407319 [7] NCCL INFO Channel 21/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:407305 [1] NCCL INFO Channel 13/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:407307 [2] NCCL INFO Channel 15/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:407310 [4] NCCL INFO Channel 10/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:407306 [6] NCCL INFO Channel 03/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:407319 [7] NCCL INFO Channel 22/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:407305 [1] NCCL INFO Channel 14/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:407307 [2] NCCL INFO Channel 16/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:407310 [4] NCCL INFO Channel 11/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:407306 [6] NCCL INFO Channel 04/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:407305 [1] NCCL INFO Channel 15/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:407319 [7] NCCL INFO Channel 23/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:407307 [2] NCCL INFO Channel 17/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:407310 [4] NCCL INFO Channel 12/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:407306 [6] NCCL INFO Channel 05/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:407305 [1] NCCL INFO Channel 16/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:407307 [2] NCCL INFO Channel 18/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:407310 [4] NCCL INFO Channel 13/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:407306 [6] NCCL INFO Channel 06/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:407305 [1] NCCL INFO Channel 17/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:407307 [2] NCCL INFO Channel 19/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:407310 [4] NCCL INFO Channel 14/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:407306 [6] NCCL INFO Channel 07/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:407305 [1] NCCL INFO Channel 18/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:407307 [2] NCCL INFO Channel 20/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:407310 [4] NCCL INFO Channel 15/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:407306 [6] NCCL INFO Channel 08/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:407305 [1] NCCL INFO Channel 19/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:407307 [2] NCCL INFO Channel 21/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:407310 [4] NCCL INFO Channel 16/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:407306 [6] NCCL INFO Channel 09/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:407305 [1] NCCL INFO Channel 20/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:407307 [2] NCCL INFO Channel 22/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:407310 [4] NCCL INFO Channel 17/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:407306 [6] NCCL INFO Channel 10/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:407305 [1] NCCL INFO Channel 21/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:407307 [2] NCCL INFO Channel 23/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:407310 [4] NCCL INFO Channel 18/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:407306 [6] NCCL INFO Channel 11/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:407305 [1] NCCL INFO Channel 22/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:407310 [4] NCCL INFO Channel 19/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:407306 [6] NCCL INFO Channel 12/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:407305 [1] NCCL INFO Channel 23/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:407310 [4] NCCL INFO Channel 20/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:407306 [6] NCCL INFO Channel 13/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:407310 [4] NCCL INFO Channel 21/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:407306 [6] NCCL INFO Channel 14/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:407310 [4] NCCL INFO Channel 22/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:407306 [6] NCCL INFO Channel 15/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:407310 [4] NCCL INFO Channel 23/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:407306 [6] NCCL INFO Channel 16/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:407306 [6] NCCL INFO Channel 17/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:407306 [6] NCCL INFO Channel 18/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:407306 [6] NCCL INFO Channel 19/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:407306 [6] NCCL INFO Channel 20/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:407306 [6] NCCL INFO Channel 21/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:407308 [5] NCCL INFO Channel 02/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:407306 [6] NCCL INFO Channel 22/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:407308 [5] NCCL INFO Channel 03/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:407306 [6] NCCL INFO Channel 23/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:407308 [5] NCCL INFO Channel 04/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:407308 [5] NCCL INFO Channel 05/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:407308 [5] NCCL INFO Channel 06/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:407308 [5] NCCL INFO Channel 07/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:407308 [5] NCCL INFO Channel 08/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:407308 [5] NCCL INFO Channel 09/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:407308 [5] NCCL INFO Channel 10/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:407308 [5] NCCL INFO Channel 11/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:407308 [5] NCCL INFO Channel 12/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:407308 [5] NCCL INFO Channel 13/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:407308 [5] NCCL INFO Channel 14/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:407308 [5] NCCL INFO Channel 15/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:407308 [5] NCCL INFO Channel 16/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:407308 [5] NCCL INFO Channel 17/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:407308 [5] NCCL INFO Channel 18/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:407308 [5] NCCL INFO Channel 19/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:407308 [5] NCCL INFO Channel 20/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:407308 [5] NCCL INFO Channel 21/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:407308 [5] NCCL INFO Channel 22/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:407308 [5] NCCL INFO Channel 23/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405736:407304 [0] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:405736:407304 [0] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:405736:407304 [0] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:405737:407305 [1] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:405737:407305 [1] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:405737:407305 [1] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:405738:407307 [2] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:405738:407307 [2] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:405738:407307 [2] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:405743:407319 [7] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:405743:407319 [7] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:405743:407319 [7] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:405742:407306 [6] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:405742:407306 [6] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:405742:407306 [6] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:405739:407309 [3] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:405739:407309 [3] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:405739:407309 [3] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:405741:407308 [5] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:405741:407308 [5] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:405741:407308 [5] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:405740:407310 [4] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:405740:407310 [4] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:405740:407310 [4] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:405737:407305 [1] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:405739:407309 [3] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:405739:407309 [3] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:405739:407309 [3] NCCL INFO ncclCommInitRank comm 0x559f23476c50 rank 3 nranks 8 cudaDev 3 nvmlDev 3 busId 201d0 commId 0x393efc2ff08be14b - Init COMPLETE +ywang29-vrdb-test1-worker-0:405741:407308 [5] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:405743:407319 [7] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:405737:407305 [1] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:405741:407308 [5] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:405743:407319 [7] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:405737:407305 [1] NCCL INFO ncclCommInitRank comm 0x55bf1466c580 rank 1 nranks 8 cudaDev 1 nvmlDev 1 busId 101d0 commId 0x393efc2ff08be14b - Init COMPLETE +ywang29-vrdb-test1-worker-0:405741:407308 [5] NCCL INFO ncclCommInitRank comm 0x55da54f5b280 rank 5 nranks 8 cudaDev 5 nvmlDev 5 busId 901d0 commId 0x393efc2ff08be14b - Init COMPLETE +ywang29-vrdb-test1-worker-0:405743:407319 [7] NCCL INFO ncclCommInitRank comm 0x555f2adfd7e0 rank 7 nranks 8 cudaDev 7 nvmlDev 7 busId a01d0 commId 0x393efc2ff08be14b - Init COMPLETE +ywang29-vrdb-test1-worker-0:405738:407307 [2] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:405738:407307 [2] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:405738:407307 [2] NCCL INFO ncclCommInitRank comm 0x564ea713b740 rank 2 nranks 8 cudaDev 2 nvmlDev 2 busId 201c0 commId 0x393efc2ff08be14b - Init COMPLETE +ywang29-vrdb-test1-worker-0:405742:407306 [6] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:405740:407310 [4] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:405740:407310 [4] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:405740:407310 [4] NCCL INFO ncclCommInitRank comm 0x5606a94dd090 rank 4 nranks 8 cudaDev 4 nvmlDev 4 busId 901c0 commId 0x393efc2ff08be14b - Init COMPLETE +ywang29-vrdb-test1-worker-0:405742:407306 [6] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:405736:407304 [0] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:405742:407306 [6] NCCL INFO ncclCommInitRank comm 0x5623bb2d3d50 rank 6 nranks 8 cudaDev 6 nvmlDev 6 busId a01c0 commId 0x393efc2ff08be14b - Init COMPLETE +ywang29-vrdb-test1-worker-0:405736:407304 [0] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:405736:407304 [0] NCCL INFO ncclCommInitRank comm 0x55bd09a355d0 rank 0 nranks 8 cudaDev 0 nvmlDev 0 busId 101c0 commId 0x393efc2ff08be14b - Init COMPLETE +[2025-10-12 14:42:14,141] [INFO] [partition_parameters.py:343:__exit__] finished initializing model - num_params = 459, num_elems = 0.99B +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +[2025-10-12 14:42:15,872] [INFO] [partition_parameters.py:343:__exit__] finished initializing model - num_params = 907, num_elems = 1.42B +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +TinyLlavaForConditionalGeneration( + (language_model): Qwen2ForCausalLM( + (model): Qwen2Model( + (embed_tokens): Embedding(151936, 896) + (layers): ModuleList( + (0-23): 24 x Qwen2DecoderLayer( + (self_attn): Qwen2FlashAttention2( + (q_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=True) + (k_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=128, bias=True) + (v_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=128, bias=True) + (o_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=False) + (rotary_emb): Qwen2RotaryEmbedding() + ) + (mlp): Qwen2MLP( + (gate_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=4864, bias=False) + (up_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=4864, bias=False) + (down_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=4864, out_features=896, bias=False) + (act_fn): SiLU() + ) + (input_layernorm): Qwen2RMSNorm() + (post_attention_layernorm): Qwen2RMSNorm() + ) + ) + (norm): Qwen2RMSNorm() + ) + (lm_head): Linear(in_features=896, out_features=151936, bias=False) + ) + (vision_tower): SIGLIPVisionTower( + (_vision_tower): SiglipVisionModel( + (vision_model): SiglipVisionTransformer( + (embeddings): SiglipVisionEmbeddings( + (patch_embedding): Conv2d(3, 1152, kernel_size=(14, 14), stride=(14, 14), padding=valid) + (position_embedding): Embedding(729, 1152) + ) + (encoder): SiglipEncoder( + (layers): ModuleList( + (0-26): 27 x SiglipEncoderLayer( + (self_attn): SiglipAttention( + (k_proj): Linear(in_features=1152, out_features=1152, bias=True) + (v_proj): Linear(in_features=1152, out_features=1152, bias=True) + (q_proj): Linear(in_features=1152, out_features=1152, bias=True) + (out_proj): Linear(in_features=1152, out_features=1152, bias=True) + ) + (layer_norm1): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + (layer_norm2): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + ) + ) + ) + (post_layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (head): SiglipMultiheadAttentionPoolingHead( + (attention): MultiheadAttention( + (out_proj): NonDynamicallyQuantizableLinear(in_features=1152, out_features=1152, bias=True) + ) + (layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + ) + ) + ) + ) + (connector): MLPConnector( + (_connector): Sequential( + (0): SupermaskLinearSparsity_SoftForward_Normal(in_features=1152, out_features=896, bias=True) + (1): GELU(approximate='none') + (2): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=True) + ) + ) +) +Pre-training init language_model.model.layers.0.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.0.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.0.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.0.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.0.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.0.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.0.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.mlp.down_proj.scores: Mean=3.000000 +Pre-training init connector._connector.0.scores: Mean=3.000005 +Pre-training init connector._connector.2.scores: Mean=2.999970 +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +2025-10-12 14:42:34,109 | INFO: Total Parameters: 1283756736, Total Trainable Parameters: 359661568 +2025-10-12 14:42:34,117 | INFO: Trainable Parameters: +language_model.model.layers.0.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.0.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.0.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.0.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.0.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.0.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.0.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.1.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.1.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.1.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.1.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.1.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.1.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.1.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.2.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.2.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.2.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.2.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.2.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.2.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.2.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.3.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.3.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.3.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.3.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.3.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.3.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.3.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.4.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.4.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.4.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.4.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.4.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.4.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.4.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.5.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.5.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.5.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.5.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.5.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.5.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.5.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.6.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.6.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.6.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.6.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.6.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.6.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.6.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.7.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.7.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.7.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.7.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.7.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.7.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.7.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.8.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.8.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.8.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.8.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.8.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.8.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.8.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.9.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.9.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.9.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.9.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.9.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.9.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.9.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.10.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.10.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.10.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.10.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.10.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.10.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.10.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.11.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.11.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.11.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.11.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.11.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.11.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.11.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.12.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.12.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.12.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.12.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.12.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.12.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.12.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.13.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.13.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.13.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.13.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.13.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.13.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.13.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.14.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.14.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.14.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.14.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.14.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.14.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.14.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.15.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.15.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.15.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.15.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.15.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.15.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.15.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.16.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.16.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.16.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.16.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.16.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.16.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.16.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.17.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.17.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.17.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.17.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.17.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.17.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.17.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.18.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.18.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.18.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.18.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.18.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.18.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.18.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.19.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.19.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.19.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.19.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.19.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.19.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.19.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.20.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.20.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.20.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.20.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.20.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.20.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.20.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.21.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.21.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.21.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.21.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.21.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.21.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.21.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.22.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.22.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.22.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.22.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.22.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.22.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.22.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.23.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.23.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.23.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.23.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.23.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.23.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.23.mlp.down_proj.scores: 4358144 parameters +connector._connector.0.scores: 1032192 parameters +connector._connector.2.scores: 802816 parameters +Parameter Offload: Total persistent parameters: 486464 in 403 params + 0%| | 0/520 [00:002->1 [1] 3/-1/-1->2->1 [2] 3/-1/-1->2->1 [3] 3/-1/-1->2->1 [4] 3/-1/-1->2->1 [5] 3/-1/-1->2->1 [6] 3/-1/-1->2->1 [7] 3/-1/-1->2->1 [8] 3/-1/-1->2->1 [9] 3/-1/-1->2->1 [10] 3/-1/-1->2->1 [11] 3/-1/-1->2->1 [12] 3/-1/-1->2->1 [13] 3/-1/-1->2->1 [14] 3/-1/-1->2->1 [15] 3/-1/-1->2->1 [16] 3/-1/-1->2->1 [17] 3/-1/-1->2->1 [18] 3/-1/-1->2->1 [19] 3/-1/-1->2->1 [20] 3/-1/-1->2->1 [21] 3/-1/-1->2->1 [22] 3/-1/-1->2->1 [23] 3/-1/-1->2->1 +ywang29-vrdb-test1-worker-0:405736:412239 [0] NCCL INFO Channel 02/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:405739:412244 [3] NCCL INFO Trees [0] 4/-1/-1->3->2 [1] 4/-1/-1->3->2 [2] 4/-1/-1->3->2 [3] 4/-1/-1->3->2 [4] 4/-1/-1->3->2 [5] 4/-1/-1->3->2 [6] 4/-1/-1->3->2 [7] 4/-1/-1->3->2 [8] 4/-1/-1->3->2 [9] 4/-1/-1->3->2 [10] 4/-1/-1->3->2 [11] 4/-1/-1->3->2 [12] 4/-1/-1->3->2 [13] 4/-1/-1->3->2 [14] 4/-1/-1->3->2 [15] 4/-1/-1->3->2 [16] 4/-1/-1->3->2 [17] 4/-1/-1->3->2 [18] 4/-1/-1->3->2 [19] 4/-1/-1->3->2 [20] 4/-1/-1->3->2 [21] 4/-1/-1->3->2 [22] 4/-1/-1->3->2 [23] 4/-1/-1->3->2 +ywang29-vrdb-test1-worker-0:405738:412246 [2] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:405739:412244 [3] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:405736:412239 [0] NCCL INFO Channel 03/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:405736:412239 [0] NCCL INFO Channel 04/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:405736:412239 [0] NCCL INFO Channel 05/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:405736:412239 [0] NCCL INFO Channel 06/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:405737:412243 [1] NCCL INFO Trees [0] 2/-1/-1->1->0 [1] 2/-1/-1->1->0 [2] 2/-1/-1->1->0 [3] 2/-1/-1->1->0 [4] 2/-1/-1->1->0 [5] 2/-1/-1->1->0 [6] 2/-1/-1->1->0 [7] 2/-1/-1->1->0 [8] 2/-1/-1->1->0 [9] 2/-1/-1->1->0 [10] 2/-1/-1->1->0 [11] 2/-1/-1->1->0 [12] 2/-1/-1->1->0 [13] 2/-1/-1->1->0 [14] 2/-1/-1->1->0 [15] 2/-1/-1->1->0 [16] 2/-1/-1->1->0 [17] 2/-1/-1->1->0 [18] 2/-1/-1->1->0 [19] 2/-1/-1->1->0 [20] 2/-1/-1->1->0 [21] 2/-1/-1->1->0 [22] 2/-1/-1->1->0 [23] 2/-1/-1->1->0 +ywang29-vrdb-test1-worker-0:405737:412243 [1] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:405736:412239 [0] NCCL INFO Channel 07/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:405736:412239 [0] NCCL INFO Channel 08/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:405736:412239 [0] NCCL INFO Channel 09/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:405736:412239 [0] NCCL INFO Channel 10/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:405736:412239 [0] NCCL INFO Channel 11/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:405740:412242 [4] NCCL INFO Trees [0] 5/-1/-1->4->3 [1] 5/-1/-1->4->3 [2] 5/-1/-1->4->3 [3] 5/-1/-1->4->3 [4] 5/-1/-1->4->3 [5] 5/-1/-1->4->3 [6] 5/-1/-1->4->3 [7] 5/-1/-1->4->3 [8] 5/-1/-1->4->3 [9] 5/-1/-1->4->3 [10] 5/-1/-1->4->3 [11] 5/-1/-1->4->3 [12] 5/-1/-1->4->3 [13] 5/-1/-1->4->3 [14] 5/-1/-1->4->3 [15] 5/-1/-1->4->3 [16] 5/-1/-1->4->3 [17] 5/-1/-1->4->3 [18] 5/-1/-1->4->3 [19] 5/-1/-1->4->3 [20] 5/-1/-1->4->3 [21] 5/-1/-1->4->3 [22] 5/-1/-1->4->3 [23] 5/-1/-1->4->3 +ywang29-vrdb-test1-worker-0:405736:412239 [0] NCCL INFO Channel 12/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:405736:412239 [0] NCCL INFO Channel 13/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:405740:412242 [4] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:405741:412241 [5] NCCL INFO Trees [0] 6/-1/-1->5->4 [1] 6/-1/-1->5->4 [2] 6/-1/-1->5->4 [3] 6/-1/-1->5->4 [4] 6/-1/-1->5->4 [5] 6/-1/-1->5->4 [6] 6/-1/-1->5->4 [7] 6/-1/-1->5->4 [8] 6/-1/-1->5->4 [9] 6/-1/-1->5->4 [10] 6/-1/-1->5->4 [11] 6/-1/-1->5->4 [12] 6/-1/-1->5->4 [13] 6/-1/-1->5->4 [14] 6/-1/-1->5->4 [15] 6/-1/-1->5->4 [16] 6/-1/-1->5->4 [17] 6/-1/-1->5->4 [18] 6/-1/-1->5->4 [19] 6/-1/-1->5->4 [20] 6/-1/-1->5->4 [21] 6/-1/-1->5->4 [22] 6/-1/-1->5->4 [23] 6/-1/-1->5->4 +ywang29-vrdb-test1-worker-0:405742:412240 [6] NCCL INFO Trees [0] 7/-1/-1->6->5 [1] 7/-1/-1->6->5 [2] 7/-1/-1->6->5 [3] 7/-1/-1->6->5 [4] 7/-1/-1->6->5 [5] 7/-1/-1->6->5 [6] 7/-1/-1->6->5 [7] 7/-1/-1->6->5 [8] 7/-1/-1->6->5 [9] 7/-1/-1->6->5 [10] 7/-1/-1->6->5 [11] 7/-1/-1->6->5 [12] 7/-1/-1->6->5 [13] 7/-1/-1->6->5 [14] 7/-1/-1->6->5 [15] 7/-1/-1->6->5 [16] 7/-1/-1->6->5 [17] 7/-1/-1->6->5 [18] 7/-1/-1->6->5 [19] 7/-1/-1->6->5 [20] 7/-1/-1->6->5 [21] 7/-1/-1->6->5 [22] 7/-1/-1->6->5 [23] 7/-1/-1->6->5 +ywang29-vrdb-test1-worker-0:405736:412239 [0] NCCL INFO Channel 14/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:405736:412239 [0] NCCL INFO Channel 15/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:405741:412241 [5] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:405742:412240 [6] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:405736:412239 [0] NCCL INFO Channel 16/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:405743:412245 [7] NCCL INFO Trees [0] -1/-1/-1->7->6 [1] -1/-1/-1->7->6 [2] -1/-1/-1->7->6 [3] -1/-1/-1->7->6 [4] -1/-1/-1->7->6 [5] -1/-1/-1->7->6 [6] -1/-1/-1->7->6 [7] -1/-1/-1->7->6 [8] -1/-1/-1->7->6 [9] -1/-1/-1->7->6 [10] -1/-1/-1->7->6 [11] -1/-1/-1->7->6 [12] -1/-1/-1->7->6 [13] -1/-1/-1->7->6 [14] -1/-1/-1->7->6 [15] -1/-1/-1->7->6 [16] -1/-1/-1->7->6 [17] -1/-1/-1->7->6 [18] -1/-1/-1->7->6 [19] -1/-1/-1->7->6 [20] -1/-1/-1->7->6 [21] -1/-1/-1->7->6 [22] -1/-1/-1->7->6 [23] -1/-1/-1->7->6 +ywang29-vrdb-test1-worker-0:405736:412239 [0] NCCL INFO Channel 17/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:405736:412239 [0] NCCL INFO Channel 18/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:405743:412245 [7] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:405736:412239 [0] NCCL INFO Channel 19/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:405736:412239 [0] NCCL INFO Channel 20/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:405736:412239 [0] NCCL INFO Channel 21/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:405736:412239 [0] NCCL INFO Channel 22/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:405736:412239 [0] NCCL INFO Channel 23/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:405736:412239 [0] NCCL INFO Trees [0] 1/-1/-1->0->-1 [1] 1/-1/-1->0->-1 [2] 1/-1/-1->0->-1 [3] 1/-1/-1->0->-1 [4] 1/-1/-1->0->-1 [5] 1/-1/-1->0->-1 [6] 1/-1/-1->0->-1 [7] 1/-1/-1->0->-1 [8] 1/-1/-1->0->-1 [9] 1/-1/-1->0->-1 [10] 1/-1/-1->0->-1 [11] 1/-1/-1->0->-1 [12] 1/-1/-1->0->-1 [13] 1/-1/-1->0->-1 [14] 1/-1/-1->0->-1 [15] 1/-1/-1->0->-1 [16] 1/-1/-1->0->-1 [17] 1/-1/-1->0->-1 [18] 1/-1/-1->0->-1 [19] 1/-1/-1->0->-1 [20] 1/-1/-1->0->-1 [21] 1/-1/-1->0->-1 [22] 1/-1/-1->0->-1 [23] 1/-1/-1->0->-1 +ywang29-vrdb-test1-worker-0:405736:412239 [0] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:405736:412239 [0] NCCL INFO Channel 00/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:412246 [2] NCCL INFO Channel 00/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405736:412239 [0] NCCL INFO Channel 01/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:412243 [1] NCCL INFO Channel 00/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:412246 [2] NCCL INFO Channel 01/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405736:412239 [0] NCCL INFO Channel 02/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:412243 [1] NCCL INFO Channel 01/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:412246 [2] NCCL INFO Channel 02/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405736:412239 [0] NCCL INFO Channel 03/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:412243 [1] NCCL INFO Channel 02/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:412246 [2] NCCL INFO Channel 03/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405736:412239 [0] NCCL INFO Channel 04/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:412245 [7] NCCL INFO Channel 00/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:412242 [4] NCCL INFO Channel 00/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:412243 [1] NCCL INFO Channel 03/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:412246 [2] NCCL INFO Channel 04/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405736:412239 [0] NCCL INFO Channel 05/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:412240 [6] NCCL INFO Channel 00/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:412242 [4] NCCL INFO Channel 01/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:412241 [5] NCCL INFO Channel 00/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:412245 [7] NCCL INFO Channel 01/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:412243 [1] NCCL INFO Channel 04/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:412246 [2] NCCL INFO Channel 05/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405736:412239 [0] NCCL INFO Channel 06/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:412240 [6] NCCL INFO Channel 01/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:412242 [4] NCCL INFO Channel 02/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:412243 [1] NCCL INFO Channel 05/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:412241 [5] NCCL INFO Channel 01/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:412245 [7] NCCL INFO Channel 02/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:412246 [2] NCCL INFO Channel 06/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405736:412239 [0] NCCL INFO Channel 07/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:412240 [6] NCCL INFO Channel 02/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:412243 [1] NCCL INFO Channel 06/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:412242 [4] NCCL INFO Channel 03/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:412241 [5] NCCL INFO Channel 02/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:412245 [7] NCCL INFO Channel 03/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:412246 [2] NCCL INFO Channel 07/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405736:412239 [0] NCCL INFO Channel 08/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:412244 [3] NCCL INFO Channel 00/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:412243 [1] NCCL INFO Channel 07/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:412242 [4] NCCL INFO Channel 04/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:412240 [6] NCCL INFO Channel 03/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:412245 [7] NCCL INFO Channel 04/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:412241 [5] NCCL INFO Channel 03/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:412246 [2] NCCL INFO Channel 08/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405736:412239 [0] NCCL INFO Channel 09/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:412244 [3] NCCL INFO Channel 01/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:412243 [1] NCCL INFO Channel 08/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:412242 [4] NCCL INFO Channel 05/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:412240 [6] NCCL INFO Channel 04/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:412245 [7] NCCL INFO Channel 05/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:412241 [5] NCCL INFO Channel 04/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:412246 [2] NCCL INFO Channel 09/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405736:412239 [0] NCCL INFO Channel 10/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:412244 [3] NCCL INFO Channel 02/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:412243 [1] NCCL INFO Channel 09/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:412242 [4] NCCL INFO Channel 06/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:412240 [6] NCCL INFO Channel 05/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:412245 [7] NCCL INFO Channel 06/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:412246 [2] NCCL INFO Channel 10/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:412241 [5] NCCL INFO Channel 05/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405736:412239 [0] NCCL INFO Channel 11/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:412244 [3] NCCL INFO Channel 03/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:412243 [1] NCCL INFO Channel 10/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:412242 [4] NCCL INFO Channel 07/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:412245 [7] NCCL INFO Channel 07/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:412240 [6] NCCL INFO Channel 06/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:412246 [2] NCCL INFO Channel 11/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:412241 [5] NCCL INFO Channel 06/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405736:412239 [0] NCCL INFO Channel 12/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:412244 [3] NCCL INFO Channel 04/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:412243 [1] NCCL INFO Channel 11/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:412242 [4] NCCL INFO Channel 08/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:412245 [7] NCCL INFO Channel 08/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:412240 [6] NCCL INFO Channel 07/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405736:412239 [0] NCCL INFO Channel 13/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:412241 [5] NCCL INFO Channel 07/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:412244 [3] NCCL INFO Channel 05/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:412243 [1] NCCL INFO Channel 12/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:412245 [7] NCCL INFO Channel 09/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:412242 [4] NCCL INFO Channel 09/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:412240 [6] NCCL INFO Channel 08/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405736:412239 [0] NCCL INFO Channel 14/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:412241 [5] NCCL INFO Channel 08/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:412244 [3] NCCL INFO Channel 06/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:412243 [1] NCCL INFO Channel 13/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:412245 [7] NCCL INFO Channel 10/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:412242 [4] NCCL INFO Channel 10/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:412240 [6] NCCL INFO Channel 09/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405736:412239 [0] NCCL INFO Channel 15/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:412244 [3] NCCL INFO Channel 07/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:412241 [5] NCCL INFO Channel 09/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:412243 [1] NCCL INFO Channel 14/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:412245 [7] NCCL INFO Channel 11/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:412242 [4] NCCL INFO Channel 11/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:412240 [6] NCCL INFO Channel 10/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405736:412239 [0] NCCL INFO Channel 16/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:412244 [3] NCCL INFO Channel 08/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:412241 [5] NCCL INFO Channel 10/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:412243 [1] NCCL INFO Channel 15/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:412245 [7] NCCL INFO Channel 12/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:412242 [4] NCCL INFO Channel 12/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:412240 [6] NCCL INFO Channel 11/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405736:412239 [0] NCCL INFO Channel 17/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:412244 [3] NCCL INFO Channel 09/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:412241 [5] NCCL INFO Channel 11/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:412243 [1] NCCL INFO Channel 16/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:412245 [7] NCCL INFO Channel 13/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:412242 [4] NCCL INFO Channel 13/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:412240 [6] NCCL INFO Channel 12/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405736:412239 [0] NCCL INFO Channel 18/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:412244 [3] NCCL INFO Channel 10/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:412243 [1] NCCL INFO Channel 17/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:412241 [5] NCCL INFO Channel 12/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:412245 [7] NCCL INFO Channel 14/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:412242 [4] NCCL INFO Channel 14/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:412240 [6] NCCL INFO Channel 13/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405736:412239 [0] NCCL INFO Channel 19/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:412244 [3] NCCL INFO Channel 11/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:412243 [1] NCCL INFO Channel 18/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:412241 [5] NCCL INFO Channel 13/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:412245 [7] NCCL INFO Channel 15/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:412242 [4] NCCL INFO Channel 15/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:412240 [6] NCCL INFO Channel 14/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405736:412239 [0] NCCL INFO Channel 20/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:412244 [3] NCCL INFO Channel 12/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:412243 [1] NCCL INFO Channel 19/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:412245 [7] NCCL INFO Channel 16/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:412241 [5] NCCL INFO Channel 14/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:412242 [4] NCCL INFO Channel 16/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405736:412239 [0] NCCL INFO Channel 21/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:412240 [6] NCCL INFO Channel 15/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:412244 [3] NCCL INFO Channel 13/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:412243 [1] NCCL INFO Channel 20/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:412245 [7] NCCL INFO Channel 17/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:412241 [5] NCCL INFO Channel 15/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:412242 [4] NCCL INFO Channel 17/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405736:412239 [0] NCCL INFO Channel 22/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:412244 [3] NCCL INFO Channel 14/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:412243 [1] NCCL INFO Channel 21/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:412240 [6] NCCL INFO Channel 16/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:412245 [7] NCCL INFO Channel 18/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:412241 [5] NCCL INFO Channel 16/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:412242 [4] NCCL INFO Channel 18/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405736:412239 [0] NCCL INFO Channel 23/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:412244 [3] NCCL INFO Channel 15/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:412243 [1] NCCL INFO Channel 22/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:412245 [7] NCCL INFO Channel 19/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:412240 [6] NCCL INFO Channel 17/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:412241 [5] NCCL INFO Channel 17/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:412242 [4] NCCL INFO Channel 19/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:412244 [3] NCCL INFO Channel 16/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:412243 [1] NCCL INFO Channel 23/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:412245 [7] NCCL INFO Channel 20/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:412240 [6] NCCL INFO Channel 18/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:412242 [4] NCCL INFO Channel 20/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:412241 [5] NCCL INFO Channel 18/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:412244 [3] NCCL INFO Channel 17/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:412245 [7] NCCL INFO Channel 21/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:412240 [6] NCCL INFO Channel 19/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:412242 [4] NCCL INFO Channel 21/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:412241 [5] NCCL INFO Channel 19/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:412244 [3] NCCL INFO Channel 18/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:412246 [2] NCCL INFO Channel 12/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:412245 [7] NCCL INFO Channel 22/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:412242 [4] NCCL INFO Channel 22/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:412240 [6] NCCL INFO Channel 20/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:412244 [3] NCCL INFO Channel 19/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:412246 [2] NCCL INFO Channel 13/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:412245 [7] NCCL INFO Channel 23/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:412242 [4] NCCL INFO Channel 23/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:412240 [6] NCCL INFO Channel 21/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:412244 [3] NCCL INFO Channel 20/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:412246 [2] NCCL INFO Channel 14/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:412240 [6] NCCL INFO Channel 22/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:412244 [3] NCCL INFO Channel 21/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:412246 [2] NCCL INFO Channel 15/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:412240 [6] NCCL INFO Channel 23/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:412244 [3] NCCL INFO Channel 22/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:412246 [2] NCCL INFO Channel 16/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:412244 [3] NCCL INFO Channel 23/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:412246 [2] NCCL INFO Channel 17/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:412246 [2] NCCL INFO Channel 18/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:412246 [2] NCCL INFO Channel 19/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:412246 [2] NCCL INFO Channel 20/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:412246 [2] NCCL INFO Channel 21/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:412246 [2] NCCL INFO Channel 22/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:412246 [2] NCCL INFO Channel 23/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:412241 [5] NCCL INFO Channel 20/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:412241 [5] NCCL INFO Channel 21/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:412241 [5] NCCL INFO Channel 22/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:412241 [5] NCCL INFO Channel 23/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:412243 [1] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:405736:412239 [0] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:405738:412246 [2] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:405739:412244 [3] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:405740:412242 [4] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:405737:412243 [1] NCCL INFO Channel 00/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:412243 [1] NCCL INFO Channel 01/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:412243 [1] NCCL INFO Channel 02/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:412243 [1] NCCL INFO Channel 03/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:412243 [1] NCCL INFO Channel 04/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:412243 [1] NCCL INFO Channel 05/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:412246 [2] NCCL INFO Channel 00/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:412243 [1] NCCL INFO Channel 06/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:412246 [2] NCCL INFO Channel 01/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:412243 [1] NCCL INFO Channel 07/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:412246 [2] NCCL INFO Channel 02/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:412243 [1] NCCL INFO Channel 08/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:412243 [1] NCCL INFO Channel 09/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:412243 [1] NCCL INFO Channel 10/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:412246 [2] NCCL INFO Channel 03/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:412243 [1] NCCL INFO Channel 11/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:412246 [2] NCCL INFO Channel 04/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:412243 [1] NCCL INFO Channel 12/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:412246 [2] NCCL INFO Channel 05/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:412243 [1] NCCL INFO Channel 13/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:412246 [2] NCCL INFO Channel 06/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:412243 [1] NCCL INFO Channel 14/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:412243 [1] NCCL INFO Channel 15/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:412246 [2] NCCL INFO Channel 07/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:412243 [1] NCCL INFO Channel 16/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:412246 [2] NCCL INFO Channel 08/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:412243 [1] NCCL INFO Channel 17/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:412246 [2] NCCL INFO Channel 09/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:412243 [1] NCCL INFO Channel 18/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:412246 [2] NCCL INFO Channel 10/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:412245 [7] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:405743:412245 [7] NCCL INFO Channel 00/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:412241 [5] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:405742:412240 [6] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:405738:412246 [2] NCCL INFO Channel 11/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:412243 [1] NCCL INFO Channel 19/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:412245 [7] NCCL INFO Channel 01/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:412246 [2] NCCL INFO Channel 12/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:412243 [1] NCCL INFO Channel 20/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:412246 [2] NCCL INFO Channel 13/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:412245 [7] NCCL INFO Channel 02/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:412243 [1] NCCL INFO Channel 21/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:412244 [3] NCCL INFO Channel 00/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:412246 [2] NCCL INFO Channel 14/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:412245 [7] NCCL INFO Channel 03/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:412243 [1] NCCL INFO Channel 22/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:412244 [3] NCCL INFO Channel 01/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:412246 [2] NCCL INFO Channel 15/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:412245 [7] NCCL INFO Channel 04/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405737:412243 [1] NCCL INFO Channel 23/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:412244 [3] NCCL INFO Channel 02/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:412246 [2] NCCL INFO Channel 16/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:412245 [7] NCCL INFO Channel 05/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:412244 [3] NCCL INFO Channel 03/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:412246 [2] NCCL INFO Channel 17/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:412245 [7] NCCL INFO Channel 06/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:412244 [3] NCCL INFO Channel 04/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:412246 [2] NCCL INFO Channel 18/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:412245 [7] NCCL INFO Channel 07/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:412242 [4] NCCL INFO Channel 00/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:412244 [3] NCCL INFO Channel 05/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:412246 [2] NCCL INFO Channel 19/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:412246 [2] NCCL INFO Channel 20/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:412244 [3] NCCL INFO Channel 06/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:412244 [3] NCCL INFO Channel 07/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:412242 [4] NCCL INFO Channel 01/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:412244 [3] NCCL INFO Channel 08/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:412242 [4] NCCL INFO Channel 02/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:412245 [7] NCCL INFO Channel 08/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:412244 [3] NCCL INFO Channel 09/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:412242 [4] NCCL INFO Channel 03/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:412245 [7] NCCL INFO Channel 09/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:412242 [4] NCCL INFO Channel 04/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:412246 [2] NCCL INFO Channel 21/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:412245 [7] NCCL INFO Channel 10/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:412246 [2] NCCL INFO Channel 22/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:412244 [3] NCCL INFO Channel 10/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405738:412246 [2] NCCL INFO Channel 23/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:412244 [3] NCCL INFO Channel 11/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:412245 [7] NCCL INFO Channel 11/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:412244 [3] NCCL INFO Channel 12/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:412245 [7] NCCL INFO Channel 12/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:412244 [3] NCCL INFO Channel 13/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:412245 [7] NCCL INFO Channel 13/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:412244 [3] NCCL INFO Channel 14/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:412244 [3] NCCL INFO Channel 15/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:412245 [7] NCCL INFO Channel 14/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:412244 [3] NCCL INFO Channel 16/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:412242 [4] NCCL INFO Channel 05/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:412244 [3] NCCL INFO Channel 17/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:412242 [4] NCCL INFO Channel 06/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:412244 [3] NCCL INFO Channel 18/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:412242 [4] NCCL INFO Channel 07/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:412244 [3] NCCL INFO Channel 19/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:412242 [4] NCCL INFO Channel 08/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:412244 [3] NCCL INFO Channel 20/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:412242 [4] NCCL INFO Channel 09/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:412241 [5] NCCL INFO Channel 00/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:412245 [7] NCCL INFO Channel 15/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:412244 [3] NCCL INFO Channel 21/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:412241 [5] NCCL INFO Channel 01/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:412245 [7] NCCL INFO Channel 16/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:412244 [3] NCCL INFO Channel 22/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:412241 [5] NCCL INFO Channel 02/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:412242 [4] NCCL INFO Channel 10/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405739:412244 [3] NCCL INFO Channel 23/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:412241 [5] NCCL INFO Channel 03/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:412242 [4] NCCL INFO Channel 11/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:412241 [5] NCCL INFO Channel 04/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:412240 [6] NCCL INFO Channel 00/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:412242 [4] NCCL INFO Channel 12/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:412245 [7] NCCL INFO Channel 17/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:412240 [6] NCCL INFO Channel 01/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:412242 [4] NCCL INFO Channel 13/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:412245 [7] NCCL INFO Channel 18/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:412240 [6] NCCL INFO Channel 02/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:412242 [4] NCCL INFO Channel 14/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:412245 [7] NCCL INFO Channel 19/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:412240 [6] NCCL INFO Channel 03/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:412242 [4] NCCL INFO Channel 15/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:412245 [7] NCCL INFO Channel 20/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:412240 [6] NCCL INFO Channel 04/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:412242 [4] NCCL INFO Channel 16/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:412245 [7] NCCL INFO Channel 21/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:412241 [5] NCCL INFO Channel 05/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:412240 [6] NCCL INFO Channel 05/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:412245 [7] NCCL INFO Channel 22/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:412242 [4] NCCL INFO Channel 17/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:412241 [5] NCCL INFO Channel 06/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:412241 [5] NCCL INFO Channel 07/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405743:412245 [7] NCCL INFO Channel 23/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:412240 [6] NCCL INFO Channel 06/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:412242 [4] NCCL INFO Channel 18/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:412241 [5] NCCL INFO Channel 08/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:412242 [4] NCCL INFO Channel 19/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:412242 [4] NCCL INFO Channel 20/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:412241 [5] NCCL INFO Channel 09/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:412242 [4] NCCL INFO Channel 21/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:412240 [6] NCCL INFO Channel 07/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:412242 [4] NCCL INFO Channel 22/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405740:412242 [4] NCCL INFO Channel 23/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:412241 [5] NCCL INFO Channel 10/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:412240 [6] NCCL INFO Channel 08/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:412241 [5] NCCL INFO Channel 11/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:412240 [6] NCCL INFO Channel 09/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:412241 [5] NCCL INFO Channel 12/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:412240 [6] NCCL INFO Channel 10/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:412241 [5] NCCL INFO Channel 13/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:412240 [6] NCCL INFO Channel 11/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:412241 [5] NCCL INFO Channel 14/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:412240 [6] NCCL INFO Channel 12/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:412241 [5] NCCL INFO Channel 15/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:412240 [6] NCCL INFO Channel 13/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:412241 [5] NCCL INFO Channel 16/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:412241 [5] NCCL INFO Channel 17/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:412240 [6] NCCL INFO Channel 14/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:412241 [5] NCCL INFO Channel 18/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:412241 [5] NCCL INFO Channel 19/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:412241 [5] NCCL INFO Channel 20/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:412240 [6] NCCL INFO Channel 15/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:412240 [6] NCCL INFO Channel 16/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:412241 [5] NCCL INFO Channel 21/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:412240 [6] NCCL INFO Channel 17/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:412241 [5] NCCL INFO Channel 22/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:412240 [6] NCCL INFO Channel 18/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405741:412241 [5] NCCL INFO Channel 23/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:412240 [6] NCCL INFO Channel 19/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:412240 [6] NCCL INFO Channel 20/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:412240 [6] NCCL INFO Channel 21/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:412240 [6] NCCL INFO Channel 22/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405742:412240 [6] NCCL INFO Channel 23/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:405736:412239 [0] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:405736:412239 [0] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:405736:412239 [0] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:405737:412243 [1] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:405737:412243 [1] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:405737:412243 [1] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:405738:412246 [2] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:405738:412246 [2] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:405738:412246 [2] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:405740:412242 [4] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:405740:412242 [4] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:405740:412242 [4] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:405739:412244 [3] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:405739:412244 [3] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:405739:412244 [3] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:405741:412241 [5] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:405741:412241 [5] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:405741:412241 [5] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:405742:412240 [6] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:405742:412240 [6] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:405742:412240 [6] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:405743:412245 [7] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:405743:412245 [7] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:405743:412245 [7] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:405739:412244 [3] NCCL INFO ncclCommInitRank comm 0x7f88d406b0b0 rank 3 nranks 8 cudaDev 3 nvmlDev 3 busId 201d0 commId 0x80c5b17ae6ec6fe5 - Init COMPLETE +ywang29-vrdb-test1-worker-0:405743:412245 [7] NCCL INFO ncclCommInitRank comm 0x7f3dd806ab70 rank 7 nranks 8 cudaDev 7 nvmlDev 7 busId a01d0 commId 0x80c5b17ae6ec6fe5 - Init COMPLETE +ywang29-vrdb-test1-worker-0:405740:412242 [4] NCCL INFO ncclCommInitRank comm 0x7fb33806b0b0 rank 4 nranks 8 cudaDev 4 nvmlDev 4 busId 901c0 commId 0x80c5b17ae6ec6fe5 - Init COMPLETE +ywang29-vrdb-test1-worker-0:405736:412239 [0] NCCL INFO ncclCommInitRank comm 0x7fef9806b610 rank 0 nranks 8 cudaDev 0 nvmlDev 0 busId 101c0 commId 0x80c5b17ae6ec6fe5 - Init COMPLETE +ywang29-vrdb-test1-worker-0:405741:412241 [5] NCCL INFO ncclCommInitRank comm 0x7f1dcc06b400 rank 5 nranks 8 cudaDev 5 nvmlDev 5 busId 901d0 commId 0x80c5b17ae6ec6fe5 - Init COMPLETE +ywang29-vrdb-test1-worker-0:405737:412243 [1] NCCL INFO ncclCommInitRank comm 0x7f918c06b370 rank 1 nranks 8 cudaDev 1 nvmlDev 1 busId 101d0 commId 0x80c5b17ae6ec6fe5 - Init COMPLETE +ywang29-vrdb-test1-worker-0:405738:412246 [2] NCCL INFO ncclCommInitRank comm 0x7fdc5806abb0 rank 2 nranks 8 cudaDev 2 nvmlDev 2 busId 201c0 commId 0x80c5b17ae6ec6fe5 - Init COMPLETE +ywang29-vrdb-test1-worker-0:405742:412240 [6] NCCL INFO ncclCommInitRank comm 0x7f1a0406b870 rank 6 nranks 8 cudaDev 6 nvmlDev 6 busId a01c0 commId 0x80c5b17ae6ec6fe5 - Init COMPLETE + 0%| | 1/520 [00:21<3:04:22, 21.31s/it] {'loss': 2.0555, 'grad_norm': 0.022258482719725345, 'learning_rate': 0.0125, 'epoch': 0.0} + 0%| | 1/520 [00:21<3:04:22, 21.31s/it] 0%| | 2/520 [00:25<1:35:30, 11.06s/it] {'loss': 2.0597, 'grad_norm': 0.024061563744108373, 'learning_rate': 0.025, 'epoch': 0.0} + 0%| | 2/520 [00:25<1:35:30, 11.06s/it] 1%| | 3/520 [00:29<1:06:58, 7.77s/it] {'loss': 1.6842, 'grad_norm': 0.012833364384109647, 'learning_rate': 0.037500000000000006, 'epoch': 0.01} + 1%| | 3/520 [00:29<1:06:58, 7.77s/it] 1%| | 4/520 [00:32<53:30, 6.22s/it] {'loss': 1.5841, 'grad_norm': 0.004586789336104231, 'learning_rate': 0.05, 'epoch': 0.01} + 1%| | 4/520 [00:32<53:30, 6.22s/it] 1%| | 5/520 [00:36<46:11, 5.38s/it] {'loss': 1.5957, 'grad_norm': 0.0035757557819500675, 'learning_rate': 0.0625, 'epoch': 0.01} + 1%| | 5/520 [00:36<46:11, 5.38s/it] 1%| | 6/520 [00:40<41:41, 4.87s/it] {'loss': 1.3743, 'grad_norm': 0.0029776354282375333, 'learning_rate': 0.07500000000000001, 'epoch': 0.01} + 1%| | 6/520 [00:40<41:41, 4.87s/it] 1%|▏ | 7/520 [00:44<38:44, 4.53s/it] {'loss': 1.4023, 'grad_norm': 0.002862011407481585, 'learning_rate': 0.08750000000000001, 'epoch': 0.01} + 1%|▏ | 7/520 [00:44<38:44, 4.53s/it] 2%|▏ | 8/520 [00:48<38:30, 4.51s/it] {'loss': 1.4491, 'grad_norm': 0.002575324730416944, 'learning_rate': 0.1, 'epoch': 0.02} + 2%|▏ | 8/520 [00:48<38:30, 4.51s/it] 2%|▏ | 9/520 [00:53<37:53, 4.45s/it] {'loss': 1.5128, 'grad_norm': 0.0023342344688286267, 'learning_rate': 0.1125, 'epoch': 0.02} + 2%|▏ | 9/520 [00:53<37:53, 4.45s/it] 2%|▏ | 10/520 [00:57<36:10, 4.26s/it] {'loss': 1.3394, 'grad_norm': 0.0024826563848327813, 'learning_rate': 0.125, 'epoch': 0.02} + 2%|▏ | 10/520 [00:57<36:10, 4.26s/it] 2%|▏ | 11/520 [01:00<34:49, 4.11s/it] {'loss': 1.4264, 'grad_norm': 0.0028048126920766432, 'learning_rate': 0.1375, 'epoch': 0.02} + 2%|▏ | 11/520 [01:00<34:49, 4.11s/it] 2%|▏ | 12/520 [01:04<33:41, 3.98s/it] {'loss': 1.3273, 'grad_norm': 0.002256395423290961, 'learning_rate': 0.15000000000000002, 'epoch': 0.02} + 2%|▏ | 12/520 [01:04<33:41, 3.98s/it][2025-10-12 14:43:47,966] [WARNING] [stage3.py:2069:step] 1 pytorch allocator cache flushes since last step. this happens when there is high memory pressure and is detrimental to performance. if this is happening frequently consider adjusting settings to reduce memory consumption. If you are unable to make the cache flushes go away consider adding get_accelerator().empty_cache() calls in your training loop to ensure that all ranks flush their caches at the same time + 2%|▎ | 13/520 [01:08<34:35, 4.09s/it] {'loss': 1.3679, 'grad_norm': 0.0024510787603127682, 'learning_rate': 0.1625, 'epoch': 0.03} + 2%|▎ | 13/520 [01:08<34:35, 4.09s/it] 3%|▎ | 14/520 [01:12<33:22, 3.96s/it] {'loss': 1.4223, 'grad_norm': 0.0024127631876388676, 'learning_rate': 0.17500000000000002, 'epoch': 0.03} + 3%|▎ | 14/520 [01:12<33:22, 3.96s/it] 3%|▎ | 15/520 [01:16<32:31, 3.87s/it] {'loss': 1.3843, 'grad_norm': 0.001680870312346185, 'learning_rate': 0.1875, 'epoch': 0.03} + 3%|▎ | 15/520 [01:16<32:31, 3.87s/it] 3%|▎ | 16/520 [01:19<31:50, 3.79s/it] {'loss': 1.3659, 'grad_norm': 0.0025110190511046146, 'learning_rate': 0.2, 'epoch': 0.03} + 3%|▎ | 16/520 [01:19<31:50, 3.79s/it] 3%|▎ | 17/520 [01:23<31:24, 3.75s/it] {'loss': 1.4577, 'grad_norm': 0.00223648813739522, 'learning_rate': 0.1999980572931538, 'epoch': 0.03} + 3%|▎ | 17/520 [01:23<31:24, 3.75s/it] 3%|▎ | 18/520 [01:27<31:08, 3.72s/it] {'loss': 1.3227, 'grad_norm': 0.002040889627124711, 'learning_rate': 0.19999222924809748, 'epoch': 0.03} + 3%|▎ | 18/520 [01:27<31:08, 3.72s/it] 4%|▎ | 19/520 [01:30<30:54, 3.70s/it] {'loss': 1.349, 'grad_norm': 0.0018646003822792264, 'learning_rate': 0.19998251609127465, 'epoch': 0.04} + 4%|▎ | 19/520 [01:30<30:54, 3.70s/it] 4%|▍ | 20/520 [01:34<30:42, 3.69s/it] {'loss': 1.3138, 'grad_norm': 0.002352100188638858, 'learning_rate': 0.19996891820008164, 'epoch': 0.04} + 4%|▍ | 20/520 [01:34<30:42, 3.69s/it] 4%|▍ | 21/520 [01:38<30:38, 3.68s/it] {'loss': 1.3333, 'grad_norm': 0.001948756318240298, 'learning_rate': 0.19995143610285276, 'epoch': 0.04} + 4%|▍ | 21/520 [01:38<30:38, 3.68s/it] 4%|▍ | 22/520 [01:41<30:30, 3.67s/it] {'loss': 1.4566, 'grad_norm': 0.0017821363785514733, 'learning_rate': 0.19993007047883987, 'epoch': 0.04} + 4%|▍ | 22/520 [01:41<30:30, 3.67s/it] 4%|▍ | 23/520 [01:45<30:18, 3.66s/it] {'loss': 1.4046, 'grad_norm': 0.0020063804599633217, 'learning_rate': 0.1999048221581858, 'epoch': 0.04} + 4%|▍ | 23/520 [01:45<30:18, 3.66s/it] 5%|▍ | 24/520 [01:49<30:09, 3.65s/it] {'loss': 1.338, 'grad_norm': 0.0017049640988065788, 'learning_rate': 0.19987569212189224, 'epoch': 0.05} + 5%|▍ | 24/520 [01:49<30:09, 3.65s/it] 5%|▍ | 25/520 [01:52<30:00, 3.64s/it] {'loss': 1.3961, 'grad_norm': 0.0019050155981425545, 'learning_rate': 0.19984268150178167, 'epoch': 0.05} + 5%|▍ | 25/520 [01:52<30:00, 3.64s/it] 5%|▌ | 26/520 [01:56<29:59, 3.64s/it] {'loss': 1.3786, 'grad_norm': 0.001759812405229477, 'learning_rate': 0.1998057915804532, 'epoch': 0.05} + 5%|▌ | 26/520 [01:56<29:59, 3.64s/it] 5%|▌ | 27/520 [01:59<29:49, 3.63s/it] {'loss': 1.2989, 'grad_norm': 0.0018643339346227767, 'learning_rate': 0.1997650237912329, 'epoch': 0.05} + 5%|▌ | 27/520 [01:59<29:49, 3.63s/it] 5%|▌ | 28/520 [02:03<29:43, 3.62s/it] {'loss': 1.3188, 'grad_norm': 0.0018980138276566887, 'learning_rate': 0.199720379718118, 'epoch': 0.05} + 5%|▌ | 28/520 [02:03<29:43, 3.62s/it] 6%|▌ | 29/520 [02:07<29:37, 3.62s/it] {'loss': 1.3358, 'grad_norm': 0.0019056812725426484, 'learning_rate': 0.19967186109571552, 'epoch': 0.06} + 6%|▌ | 29/520 [02:07<29:37, 3.62s/it] 6%|▌ | 30/520 [02:10<29:31, 3.61s/it] {'loss': 1.4169, 'grad_norm': 0.0017215421841542849, 'learning_rate': 0.19961946980917455, 'epoch': 0.06} + 6%|▌ | 30/520 [02:10<29:31, 3.61s/it] 6%|▌ | 31/520 [02:14<29:26, 3.61s/it] {'loss': 1.3176, 'grad_norm': 0.0015886000384247474, 'learning_rate': 0.1995632078941134, 'epoch': 0.06} + 6%|▌ | 31/520 [02:14<29:26, 3.61s/it] 6%|▌ | 32/520 [02:17<29:30, 3.63s/it] {'loss': 1.2683, 'grad_norm': 0.001950606394117378, 'learning_rate': 0.19950307753654017, 'epoch': 0.06} + 6%|▌ | 32/520 [02:17<29:30, 3.63s/it] 6%|▋ | 33/520 [02:21<29:29, 3.63s/it] {'loss': 1.3267, 'grad_norm': 0.0018999434031757307, 'learning_rate': 0.19943908107276798, 'epoch': 0.06} + 6%|▋ | 33/520 [02:21<29:29, 3.63s/it] 7%|▋ | 34/520 [02:25<29:31, 3.65s/it] {'loss': 1.3191, 'grad_norm': 0.002009114292028153, 'learning_rate': 0.19937122098932428, 'epoch': 0.07} + 7%|▋ | 34/520 [02:25<29:31, 3.65s/it] 7%|▋ | 35/520 [02:28<29:28, 3.65s/it] {'loss': 1.3222, 'grad_norm': 0.0020185458642257266, 'learning_rate': 0.19929949992285395, 'epoch': 0.07} + 7%|▋ | 35/520 [02:28<29:28, 3.65s/it] 7%|▋ | 36/520 [02:32<29:30, 3.66s/it] {'loss': 1.4295, 'grad_norm': 0.001914736608980602, 'learning_rate': 0.19922392066001723, 'epoch': 0.07} + 7%|▋ | 36/520 [02:32<29:30, 3.66s/it] 7%|▋ | 37/520 [02:36<29:21, 3.65s/it] {'loss': 1.408, 'grad_norm': 0.0018113710872781467, 'learning_rate': 0.19914448613738106, 'epoch': 0.07} + 7%|▋ | 37/520 [02:36<29:21, 3.65s/it] 7%|▋ | 38/520 [02:40<29:54, 3.72s/it] {'loss': 1.5037, 'grad_norm': 0.0018000778917680901, 'learning_rate': 0.1990611994413053, 'epoch': 0.07} + 7%|▋ | 38/520 [02:40<29:54, 3.72s/it] 8%|▊ | 39/520 [02:44<30:16, 3.78s/it] {'loss': 1.3614, 'grad_norm': 0.002344787200094458, 'learning_rate': 0.19897406380782262, 'epoch': 0.07} + 8%|▊ | 39/520 [02:44<30:16, 3.78s/it] 8%|▊ | 40/520 [02:47<30:19, 3.79s/it] {'loss': 1.3905, 'grad_norm': 0.0017566201775907346, 'learning_rate': 0.19888308262251286, 'epoch': 0.08} + 8%|▊ | 40/520 [02:47<30:19, 3.79s/it] 8%|▊ | 41/520 [02:51<30:26, 3.81s/it] {'loss': 1.3634, 'grad_norm': 0.0017967948835718164, 'learning_rate': 0.19878825942037148, 'epoch': 0.08} + 8%|▊ | 41/520 [02:51<30:26, 3.81s/it] 8%|▊ | 42/520 [02:55<30:35, 3.84s/it] {'loss': 1.3917, 'grad_norm': 0.002535091005542377, 'learning_rate': 0.19868959788567211, 'epoch': 0.08} + 8%|▊ | 42/520 [02:55<30:35, 3.84s/it] 8%|▊ | 43/520 [02:59<30:38, 3.85s/it] {'loss': 1.3248, 'grad_norm': 0.0025078265115845927, 'learning_rate': 0.1985871018518236, 'epoch': 0.08} + 8%|▊ | 43/520 [02:59<30:38, 3.85s/it] 8%|▊ | 44/520 [03:03<30:36, 3.86s/it] {'loss': 1.4175, 'grad_norm': 0.0019491456157533327, 'learning_rate': 0.19848077530122082, 'epoch': 0.08} + 8%|▊ | 44/520 [03:03<30:36, 3.86s/it] 9%|▊ | 45/520 [03:07<30:40, 3.87s/it] {'loss': 1.3905, 'grad_norm': 0.0020180430749293737, 'learning_rate': 0.19837062236509015, 'epoch': 0.09} + 9%|▊ | 45/520 [03:07<30:40, 3.87s/it] 9%|▉ | 46/520 [03:11<30:32, 3.87s/it] {'loss': 1.4737, 'grad_norm': 0.0018517841439232164, 'learning_rate': 0.19825664732332884, 'epoch': 0.09} + 9%|▉ | 46/520 [03:11<30:32, 3.87s/it] 9%|▉ | 47/520 [03:15<30:24, 3.86s/it] {'loss': 1.3767, 'grad_norm': 0.0017800458565566965, 'learning_rate': 0.19813885460433878, 'epoch': 0.09} + 9%|▉ | 47/520 [03:15<30:24, 3.86s/it] 9%|▉ | 48/520 [03:18<30:21, 3.86s/it] {'loss': 1.3535, 'grad_norm': 0.0021336817550622884, 'learning_rate': 0.19801724878485438, 'epoch': 0.09} + 9%|▉ | 48/520 [03:18<30:21, 3.86s/it] 9%|▉ | 49/520 [03:22<30:16, 3.86s/it] {'loss': 1.4069, 'grad_norm': 0.001970717182590638, 'learning_rate': 0.19789183458976486, 'epoch': 0.09} + 9%|▉ | 49/520 [03:22<30:16, 3.86s/it] 10%|▉ | 50/520 [03:26<30:10, 3.85s/it] {'loss': 1.3936, 'grad_norm': 0.0018270685685970454, 'learning_rate': 0.19776261689193048, 'epoch': 0.1} + 10%|▉ | 50/520 [03:26<30:10, 3.85s/it] 10%|▉ | 51/520 [03:30<30:07, 3.85s/it] {'loss': 1.3242, 'grad_norm': 0.0021962198267656005, 'learning_rate': 0.19762960071199334, 'epoch': 0.1} + 10%|▉ | 51/520 [03:30<30:07, 3.85s/it] 10%|█ | 52/520 [03:34<29:40, 3.80s/it] {'loss': 1.4587, 'grad_norm': 0.0022247763034993645, 'learning_rate': 0.19749279121818236, 'epoch': 0.1} + 10%|█ | 52/520 [03:34<29:40, 3.80s/it] 10%|█ | 53/520 [03:37<29:15, 3.76s/it] {'loss': 1.4348, 'grad_norm': 0.0018313625666137076, 'learning_rate': 0.19735219372611235, 'epoch': 0.1} + 10%|█ | 53/520 [03:37<29:15, 3.76s/it] 10%|█ | 54/520 [03:41<28:49, 3.71s/it] {'loss': 1.3503, 'grad_norm': 0.0018890557236561991, 'learning_rate': 0.19720781369857746, 'epoch': 0.1} + 10%|█ | 54/520 [03:41<28:49, 3.71s/it] 11%|█ | 55/520 [03:45<28:39, 3.70s/it] {'loss': 1.3314, 'grad_norm': 0.0021151654986927932, 'learning_rate': 0.1970596567453391, 'epoch': 0.11} + 11%|█ | 55/520 [03:45<28:39, 3.70s/it] 11%|█ | 56/520 [03:48<28:33, 3.69s/it] {'loss': 1.465, 'grad_norm': 0.002017507996588744, 'learning_rate': 0.1969077286229078, 'epoch': 0.11} + 11%|█ | 56/520 [03:48<28:33, 3.69s/it] 11%|█ | 57/520 [03:52<28:20, 3.67s/it] {'loss': 1.3274, 'grad_norm': 0.0023445006845813064, 'learning_rate': 0.19675203523431964, 'epoch': 0.11} + 11%|█ | 57/520 [03:52<28:20, 3.67s/it] 11%|█ | 58/520 [03:56<28:22, 3.68s/it] {'loss': 1.4872, 'grad_norm': 0.0018730603312284177, 'learning_rate': 0.19659258262890683, 'epoch': 0.11} + 11%|█ | 58/520 [03:56<28:22, 3.68s/it] 11%|█▏ | 59/520 [03:59<28:22, 3.69s/it] {'loss': 1.3073, 'grad_norm': 0.0018255639635573779, 'learning_rate': 0.19642937700206278, 'epoch': 0.11} + 11%|█▏ | 59/520 [03:59<28:22, 3.69s/it] 12%|█▏ | 60/520 [04:03<28:10, 3.67s/it] {'loss': 1.4026, 'grad_norm': 0.0019659415158854066, 'learning_rate': 0.19626242469500121, 'epoch': 0.12} + 12%|█▏ | 60/520 [04:03<28:10, 3.67s/it] 12%|█▏ | 61/520 [04:07<28:05, 3.67s/it] {'loss': 1.4059, 'grad_norm': 0.0035114734070699153, 'learning_rate': 0.19609173219450998, 'epoch': 0.12} + 12%|█▏ | 61/520 [04:07<28:05, 3.67s/it] 12%|█▏ | 62/520 [04:10<28:09, 3.69s/it] {'loss': 1.383, 'grad_norm': 0.0019206811129658743, 'learning_rate': 0.19591730613269878, 'epoch': 0.12} + 12%|█▏ | 62/520 [04:10<28:09, 3.69s/it] 12%|█▏ | 63/520 [04:14<28:24, 3.73s/it] {'loss': 1.3621, 'grad_norm': 0.0017704350479838104, 'learning_rate': 0.19573915328674182, 'epoch': 0.12} + 12%|█▏ | 63/520 [04:14<28:24, 3.73s/it] 12%|█▏ | 64/520 [04:18<28:31, 3.75s/it] {'loss': 1.4015, 'grad_norm': 0.0020323771408274124, 'learning_rate': 0.1955572805786141, 'epoch': 0.12} + 12%|█▏ | 64/520 [04:18<28:31, 3.75s/it] 12%|█▎ | 65/520 [04:22<28:20, 3.74s/it] {'loss': 1.4009, 'grad_norm': 0.0023398099574200785, 'learning_rate': 0.1953716950748227, 'epoch': 0.12} + 12%|█▎ | 65/520 [04:22<28:20, 3.74s/it] 13%|█▎ | 66/520 [04:25<28:05, 3.71s/it] {'loss': 1.3567, 'grad_norm': 0.001761147847625208, 'learning_rate': 0.19518240398613226, 'epoch': 0.13} + 13%|█▎ | 66/520 [04:25<28:05, 3.71s/it] 13%|█▎ | 67/520 [04:29<27:50, 3.69s/it] {'loss': 1.2591, 'grad_norm': 0.0018894086308648493, 'learning_rate': 0.1949894146672846, 'epoch': 0.13} + 13%|█▎ | 67/520 [04:29<27:50, 3.69s/it] 13%|█▎ | 68/520 [04:33<27:59, 3.72s/it] {'loss': 1.32, 'grad_norm': 0.0021188818292089046, 'learning_rate': 0.1947927346167132, 'epoch': 0.13} + 13%|█▎ | 68/520 [04:33<27:59, 3.72s/it] 13%|█▎ | 69/520 [04:37<28:10, 3.75s/it] {'loss': 1.298, 'grad_norm': 0.0018092073743544697, 'learning_rate': 0.1945923714762516, 'epoch': 0.13} + 13%|█▎ | 69/520 [04:37<28:10, 3.75s/it] 13%|█▎ | 70/520 [04:40<28:16, 3.77s/it] {'loss': 1.3392, 'grad_norm': 0.0019593239475290563, 'learning_rate': 0.19438833303083677, 'epoch': 0.13} + 13%|█▎ | 70/520 [04:40<28:16, 3.77s/it] 14%|█▎ | 71/520 [04:44<28:18, 3.78s/it] {'loss': 1.275, 'grad_norm': 0.001774817122016827, 'learning_rate': 0.19418062720820636, 'epoch': 0.14} + 14%|█▎ | 71/520 [04:44<28:18, 3.78s/it] 14%|█▍ | 72/520 [04:48<28:24, 3.80s/it] {'loss': 1.4254, 'grad_norm': 0.0021273924128218384, 'learning_rate': 0.19396926207859086, 'epoch': 0.14} + 14%|█▍ | 72/520 [04:48<28:24, 3.80s/it] 14%|█▍ | 73/520 [04:52<28:25, 3.82s/it] {'loss': 1.2531, 'grad_norm': 0.0017568810555284196, 'learning_rate': 0.19375424585439993, 'epoch': 0.14} + 14%|█▍ | 73/520 [04:52<28:25, 3.82s/it] 14%|█▍ | 74/520 [04:56<28:24, 3.82s/it] {'loss': 1.369, 'grad_norm': 0.0020233169508287363, 'learning_rate': 0.1935355868899034, 'epoch': 0.14} + 14%|█▍ | 74/520 [04:56<28:24, 3.82s/it] 14%|█▍ | 75/520 [04:59<28:18, 3.82s/it] {'loss': 1.2741, 'grad_norm': 0.001688208975687015, 'learning_rate': 0.19331329368090666, 'epoch': 0.14} + 14%|█▍ | 75/520 [04:59<28:18, 3.82s/it] 15%|█▍ | 76/520 [05:03<28:19, 3.83s/it] {'loss': 1.4834, 'grad_norm': 0.00410215203668756, 'learning_rate': 0.19308737486442043, 'epoch': 0.15} + 15%|█▍ | 76/520 [05:03<28:19, 3.83s/it] 15%|█▍ | 77/520 [05:07<28:16, 3.83s/it] {'loss': 1.2017, 'grad_norm': 0.0019277934939380497, 'learning_rate': 0.19285783921832536, 'epoch': 0.15} + 15%|█▍ | 77/520 [05:07<28:16, 3.83s/it] 15%|█▌ | 78/520 [05:11<28:36, 3.88s/it] {'loss': 1.3222, 'grad_norm': 0.0018688962971227438, 'learning_rate': 0.19262469566103088, 'epoch': 0.15} + 15%|█▌ | 78/520 [05:11<28:36, 3.88s/it] 15%|█▌ | 79/520 [05:15<29:00, 3.95s/it] {'loss': 1.3089, 'grad_norm': 0.0019239785504654647, 'learning_rate': 0.19238795325112867, 'epoch': 0.15} + 15%|█▌ | 79/520 [05:15<29:00, 3.95s/it] 15%|█▌ | 80/520 [05:19<29:14, 3.99s/it] {'loss': 1.4744, 'grad_norm': 0.0029802117774753176, 'learning_rate': 0.19214762118704076, 'epoch': 0.15} + 15%|█▌ | 80/520 [05:19<29:14, 3.99s/it] 16%|█▌ | 81/520 [05:23<29:24, 4.02s/it] {'loss': 1.4576, 'grad_norm': 0.0022987628238765817, 'learning_rate': 0.19190370880666208, 'epoch': 0.16} + 16%|█▌ | 81/520 [05:23<29:24, 4.02s/it] 16%|█▌ | 82/520 [05:28<29:32, 4.05s/it] {'loss': 1.3864, 'grad_norm': 0.001860648466581515, 'learning_rate': 0.19165622558699763, 'epoch': 0.16} + 16%|█▌ | 82/520 [05:28<29:32, 4.05s/it] 16%|█▌ | 83/520 [05:32<29:32, 4.06s/it] {'loss': 1.394, 'grad_norm': 0.0019500748322763641, 'learning_rate': 0.19140518114379435, 'epoch': 0.16} + 16%|█▌ | 83/520 [05:32<29:32, 4.06s/it] 16%|█▌ | 84/520 [05:36<29:18, 4.03s/it] {'loss': 1.4086, 'grad_norm': 0.0018838446116297095, 'learning_rate': 0.19115058523116735, 'epoch': 0.16} + 16%|█▌ | 84/520 [05:36<29:18, 4.03s/it] 16%|█▋ | 85/520 [05:39<28:46, 3.97s/it] {'loss': 1.4401, 'grad_norm': 0.0017296649441544117, 'learning_rate': 0.1908924477412211, 'epoch': 0.16} + 16%|█▋ | 85/520 [05:39<28:46, 3.97s/it] 17%|█▋ | 86/520 [05:43<28:23, 3.93s/it] {'loss': 1.44, 'grad_norm': 0.0018479334300556525, 'learning_rate': 0.19063077870366502, 'epoch': 0.17} + 17%|█▋ | 86/520 [05:43<28:23, 3.93s/it] 17%|█▋ | 87/520 [05:47<27:51, 3.86s/it] {'loss': 1.3979, 'grad_norm': 0.0018178232837688943, 'learning_rate': 0.1903655882854237, 'epoch': 0.17} + 17%|█▋ | 87/520 [05:47<27:51, 3.86s/it] 17%|█▋ | 88/520 [05:51<27:21, 3.80s/it] {'loss': 1.3726, 'grad_norm': 0.0020504015003726776, 'learning_rate': 0.19009688679024192, 'epoch': 0.17} + 17%|█▋ | 88/520 [05:51<27:21, 3.80s/it] 17%|█▋ | 89/520 [05:54<26:56, 3.75s/it] {'loss': 1.3935, 'grad_norm': 0.001894363145714689, 'learning_rate': 0.18982468465828442, 'epoch': 0.17} + 17%|█▋ | 89/520 [05:54<26:56, 3.75s/it] 17%|█▋ | 90/520 [05:58<26:41, 3.72s/it] {'loss': 1.3296, 'grad_norm': 0.0019581024308436296, 'learning_rate': 0.1895489924657301, 'epoch': 0.17} + 17%|█▋ | 90/520 [05:58<26:41, 3.72s/it] 18%|█▊ | 91/520 [06:02<26:24, 3.69s/it] {'loss': 1.4006, 'grad_norm': 0.0017010944874500515, 'learning_rate': 0.18926982092436118, 'epoch': 0.17} + 18%|█▊ | 91/520 [06:02<26:24, 3.69s/it] 18%|█▊ | 92/520 [06:05<26:19, 3.69s/it] {'loss': 1.3257, 'grad_norm': 0.0017697986282858616, 'learning_rate': 0.18898718088114688, 'epoch': 0.18} + 18%|█▊ | 92/520 [06:05<26:19, 3.69s/it] 18%|█▊ | 93/520 [06:09<26:09, 3.68s/it] {'loss': 1.3488, 'grad_norm': 0.0017894429276792206, 'learning_rate': 0.18870108331782218, 'epoch': 0.18} + 18%|█▊ | 93/520 [06:09<26:09, 3.68s/it] 18%|█▊ | 94/520 [06:12<25:57, 3.66s/it] {'loss': 1.4327, 'grad_norm': 0.0018492719735284133, 'learning_rate': 0.18841153935046098, 'epoch': 0.18} + 18%|█▊ | 94/520 [06:12<25:57, 3.66s/it] 18%|█▊ | 95/520 [06:16<25:49, 3.65s/it] {'loss': 1.3235, 'grad_norm': 0.001936124333376416, 'learning_rate': 0.18811856022904425, 'epoch': 0.18} + 18%|█▊ | 95/520 [06:16<25:49, 3.65s/it] 18%|█▊ | 96/520 [06:20<25:48, 3.65s/it] {'loss': 1.3364, 'grad_norm': 0.0015352160128268828, 'learning_rate': 0.18782215733702287, 'epoch': 0.18} + 18%|█▊ | 96/520 [06:20<25:48, 3.65s/it] 19%|█▊ | 97/520 [06:23<25:44, 3.65s/it] {'loss': 1.3009, 'grad_norm': 0.002042265454339007, 'learning_rate': 0.18752234219087538, 'epoch': 0.19} + 19%|█▊ | 97/520 [06:23<25:44, 3.65s/it] 19%|█▉ | 98/520 [06:27<25:38, 3.65s/it] {'loss': 1.304, 'grad_norm': 0.0015319157678965372, 'learning_rate': 0.18721912643966054, 'epoch': 0.19} + 19%|█▉ | 98/520 [06:27<25:38, 3.65s/it] 19%|█▉ | 99/520 [06:31<25:32, 3.64s/it] {'loss': 1.3159, 'grad_norm': 0.0017083561060768804, 'learning_rate': 0.18691252186456464, 'epoch': 0.19} + 19%|█▉ | 99/520 [06:31<25:32, 3.64s/it] 19%|█▉ | 100/520 [06:34<25:29, 3.64s/it] {'loss': 1.3323, 'grad_norm': 0.0018688123471502182, 'learning_rate': 0.1866025403784439, 'epoch': 0.19} + 19%|█▉ | 100/520 [06:34<25:29, 3.64s/it] 19%|█▉ | 101/520 [06:38<25:32, 3.66s/it] {'loss': 1.3167, 'grad_norm': 0.0017476359941507732, 'learning_rate': 0.18628919402536132, 'epoch': 0.19} + 19%|█▉ | 101/520 [06:38<25:32, 3.66s/it] 20%|█▉ | 102/520 [06:42<25:29, 3.66s/it] {'loss': 1.3281, 'grad_norm': 0.0017203137598773499, 'learning_rate': 0.18597249498011903, 'epoch': 0.2} + 20%|█▉ | 102/520 [06:42<25:29, 3.66s/it] 20%|█▉ | 103/520 [06:45<25:27, 3.66s/it] {'loss': 1.2535, 'grad_norm': 0.0015649010030676699, 'learning_rate': 0.18565245554778517, 'epoch': 0.2} + 20%|█▉ | 103/520 [06:45<25:27, 3.66s/it] 20%|██ | 104/520 [06:49<25:27, 3.67s/it] {'loss': 1.3329, 'grad_norm': 0.001649178598991316, 'learning_rate': 0.18532908816321558, 'epoch': 0.2} + 20%|██ | 104/520 [06:49<25:27, 3.67s/it] 20%|██ | 105/520 [06:53<25:27, 3.68s/it] {'loss': 1.3229, 'grad_norm': 0.0015108377574577028, 'learning_rate': 0.18500240539057092, 'epoch': 0.2} + 20%|██ | 105/520 [06:53<25:27, 3.68s/it] 20%|██ | 106/520 [06:56<25:24, 3.68s/it] {'loss': 1.3484, 'grad_norm': 0.001627518735487245, 'learning_rate': 0.18467241992282843, 'epoch': 0.2} + 20%|██ | 106/520 [06:56<25:24, 3.68s/it] 21%|██ | 107/520 [07:00<25:21, 3.68s/it] {'loss': 1.3219, 'grad_norm': 0.0016543337621240189, 'learning_rate': 0.18433914458128858, 'epoch': 0.21} + 21%|██ | 107/520 [07:00<25:21, 3.68s/it] 21%|██ | 108/520 [07:04<25:14, 3.68s/it] {'loss': 1.2825, 'grad_norm': 0.0017177265417466332, 'learning_rate': 0.18400259231507718, 'epoch': 0.21} + 21%|██ | 108/520 [07:04<25:14, 3.68s/it] 21%|██ | 109/520 [07:07<25:09, 3.67s/it] {'loss': 1.3067, 'grad_norm': 0.0015043813708827366, 'learning_rate': 0.18366277620064198, 'epoch': 0.21} + 21%|██ | 109/520 [07:07<25:09, 3.67s/it] 21%|██ | 110/520 [07:11<25:00, 3.66s/it] {'loss': 1.4693, 'grad_norm': 0.0016940750998395103, 'learning_rate': 0.1833197094412449, 'epoch': 0.21} + 21%|██ | 110/520 [07:11<25:00, 3.66s/it] 21%|██▏ | 111/520 [07:15<25:13, 3.70s/it] {'loss': 1.4684, 'grad_norm': 0.0017854304725589412, 'learning_rate': 0.18297340536644877, 'epoch': 0.21} + 21%|██▏ | 111/520 [07:15<25:13, 3.70s/it] 22%|██▏ | 112/520 [07:19<25:28, 3.75s/it] {'loss': 1.3504, 'grad_norm': 0.0015772177216845467, 'learning_rate': 0.1826238774315995, 'epoch': 0.22} + 22%|██▏ | 112/520 [07:19<25:28, 3.75s/it] 22%|██▏ | 113/520 [07:23<25:38, 3.78s/it] {'loss': 1.2362, 'grad_norm': 0.0015927355899037735, 'learning_rate': 0.18227113921730334, 'epoch': 0.22} + 22%|██▏ | 113/520 [07:23<25:38, 3.78s/it] 22%|██▏ | 114/520 [07:26<25:44, 3.80s/it] {'loss': 1.334, 'grad_norm': 0.001535022265055132, 'learning_rate': 0.1819152044288992, 'epoch': 0.22} + 22%|██▏ | 114/520 [07:26<25:44, 3.80s/it] 22%|██▏ | 115/520 [07:30<25:49, 3.83s/it] {'loss': 1.4445, 'grad_norm': 0.0015670522059212833, 'learning_rate': 0.18155608689592603, 'epoch': 0.22} + 22%|██▏ | 115/520 [07:30<25:49, 3.83s/it] 22%|██▏ | 116/520 [07:34<25:51, 3.84s/it] {'loss': 1.4524, 'grad_norm': 0.0015567096416337289, 'learning_rate': 0.18119380057158568, 'epoch': 0.22} + 22%|██▏ | 116/520 [07:34<25:51, 3.84s/it] 22%|██▎ | 117/520 [07:38<25:50, 3.85s/it] {'loss': 1.4196, 'grad_norm': 0.0016820355570207695, 'learning_rate': 0.18082835953220056, 'epoch': 0.23} + 22%|██▎ | 117/520 [07:38<25:50, 3.85s/it] 23%|██▎ | 118/520 [07:42<25:43, 3.84s/it] {'loss': 1.3129, 'grad_norm': 0.001446611302658097, 'learning_rate': 0.18045977797666685, 'epoch': 0.23} + 23%|██▎ | 118/520 [07:42<25:43, 3.84s/it] 23%|██▎ | 119/520 [07:46<25:39, 3.84s/it] {'loss': 1.2693, 'grad_norm': 0.0016061424892427655, 'learning_rate': 0.1800880702259028, 'epoch': 0.23} + 23%|██▎ | 119/520 [07:46<25:39, 3.84s/it] 23%|██▎ | 120/520 [07:50<25:35, 3.84s/it] {'loss': 1.2933, 'grad_norm': 0.0018026988659038251, 'learning_rate': 0.17971325072229227, 'epoch': 0.23} + 23%|██▎ | 120/520 [07:50<25:35, 3.84s/it] 23%|██▎ | 121/520 [07:53<25:33, 3.84s/it] {'loss': 1.3491, 'grad_norm': 0.0017300242295934637, 'learning_rate': 0.17933533402912352, 'epoch': 0.23} + 23%|██▎ | 121/520 [07:53<25:33, 3.84s/it] 23%|██▎ | 122/520 [07:57<25:04, 3.78s/it] {'loss': 1.2468, 'grad_norm': 0.0014592984126265188, 'learning_rate': 0.17895433483002354, 'epoch': 0.23} + 23%|██▎ | 122/520 [07:57<25:04, 3.78s/it] 24%|██▎ | 123/520 [08:01<24:41, 3.73s/it] {'loss': 1.3935, 'grad_norm': 0.001598388402024712, 'learning_rate': 0.17857026792838737, 'epoch': 0.24} + 24%|██▎ | 123/520 [08:01<24:41, 3.73s/it] 24%|██▍ | 124/520 [08:04<24:31, 3.72s/it] {'loss': 1.3133, 'grad_norm': 0.0016885269406771996, 'learning_rate': 0.178183148246803, 'epoch': 0.24} + 24%|██▍ | 124/520 [08:04<24:31, 3.72s/it] 24%|██▍ | 125/520 [08:08<24:19, 3.70s/it] {'loss': 1.307, 'grad_norm': 0.0015740992045274727, 'learning_rate': 0.1777929908264715, 'epoch': 0.24} + 24%|██▍ | 125/520 [08:08<24:19, 3.70s/it] 24%|██▍ | 126/520 [08:12<25:24, 3.87s/it] {'loss': 1.3114, 'grad_norm': 0.001369897415827549, 'learning_rate': 0.17739981082662276, 'epoch': 0.24} + 24%|██▍ | 126/520 [08:12<25:24, 3.87s/it] 24%|██▍ | 127/520 [08:16<24:55, 3.81s/it] {'loss': 1.2925, 'grad_norm': 0.001792641273869612, 'learning_rate': 0.1770036235239263, 'epoch': 0.24} + 24%|██▍ | 127/520 [08:16<24:55, 3.81s/it] 25%|██▍ | 128/520 [08:20<24:29, 3.75s/it] {'loss': 1.339, 'grad_norm': 0.0016203057271198362, 'learning_rate': 0.1766044443118978, 'epoch': 0.25} + 25%|██▍ | 128/520 [08:20<24:29, 3.75s/it] 25%|██▍ | 129/520 [08:23<24:11, 3.71s/it] {'loss': 1.2866, 'grad_norm': 0.0014381819102395174, 'learning_rate': 0.17620228870030108, 'epoch': 0.25} + 25%|██▍ | 129/520 [08:23<24:11, 3.71s/it] 25%|██▌ | 130/520 [08:27<23:57, 3.69s/it] {'loss': 1.3249, 'grad_norm': 0.0014301030568672646, 'learning_rate': 0.1757971723145453, 'epoch': 0.25} + 25%|██▌ | 130/520 [08:27<23:57, 3.69s/it] 25%|██▌ | 131/520 [08:30<23:47, 3.67s/it] {'loss': 1.288, 'grad_norm': 0.0014341102107319106, 'learning_rate': 0.175389110895078, 'epoch': 0.25} + 25%|██▌ | 131/520 [08:30<23:47, 3.67s/it] 25%|██▌ | 132/520 [08:34<23:39, 3.66s/it] {'loss': 1.3729, 'grad_norm': 0.0016905837248988509, 'learning_rate': 0.17497812029677343, 'epoch': 0.25} + 25%|██▌ | 132/520 [08:34<23:39, 3.66s/it] 26%|██▌ | 133/520 [08:38<23:38, 3.67s/it] {'loss': 1.2855, 'grad_norm': 0.001741721935787191, 'learning_rate': 0.17456421648831655, 'epoch': 0.26} + 26%|██▌ | 133/520 [08:38<23:38, 3.67s/it] 26%|██▌ | 134/520 [08:41<23:30, 3.65s/it] {'loss': 1.3645, 'grad_norm': 0.0015637318589052497, 'learning_rate': 0.17414741555158267, 'epoch': 0.26} + 26%|██▌ | 134/520 [08:41<23:30, 3.65s/it] 26%|██▌ | 135/520 [08:45<23:25, 3.65s/it] {'loss': 1.4201, 'grad_norm': 0.001609909535058473, 'learning_rate': 0.1737277336810124, 'epoch': 0.26} + 26%|██▌ | 135/520 [08:45<23:25, 3.65s/it] 26%|██▌ | 136/520 [08:49<23:19, 3.64s/it] {'loss': 1.3545, 'grad_norm': 0.001610725146619169, 'learning_rate': 0.17330518718298263, 'epoch': 0.26} + 26%|██▌ | 136/520 [08:49<23:19, 3.64s/it] 26%|██▋ | 137/520 [08:52<23:10, 3.63s/it] {'loss': 1.2744, 'grad_norm': 0.0017516696279460222, 'learning_rate': 0.17287979247517285, 'epoch': 0.26} + 26%|██▋ | 137/520 [08:52<23:10, 3.63s/it] 27%|██▋ | 138/520 [08:56<23:08, 3.64s/it] {'loss': 1.2882, 'grad_norm': 0.0014801379308906843, 'learning_rate': 0.17245156608592727, 'epoch': 0.27} + 27%|██▋ | 138/520 [08:56<23:08, 3.64s/it] 27%|██▋ | 139/520 [09:00<23:06, 3.64s/it] {'loss': 1.1906, 'grad_norm': 0.0015477642229186116, 'learning_rate': 0.17202052465361267, 'epoch': 0.27} + 27%|██▋ | 139/520 [09:00<23:06, 3.64s/it] 27%|██▋ | 140/520 [09:03<23:07, 3.65s/it] {'loss': 1.332, 'grad_norm': 0.0014590968016335395, 'learning_rate': 0.17158668492597184, 'epoch': 0.27} + 27%|██▋ | 140/520 [09:03<23:07, 3.65s/it] 27%|██▋ | 141/520 [09:07<23:03, 3.65s/it] {'loss': 1.3955, 'grad_norm': 0.00158710603309713, 'learning_rate': 0.17115006375947303, 'epoch': 0.27} + 27%|██▋ | 141/520 [09:07<23:03, 3.65s/it] 27%|██▋ | 142/520 [09:10<22:57, 3.64s/it] {'loss': 1.3469, 'grad_norm': 0.0014499186423657662, 'learning_rate': 0.17071067811865476, 'epoch': 0.27} + 27%|██▋ | 142/520 [09:10<22:57, 3.64s/it] 28%|██▊ | 143/520 [09:14<22:55, 3.65s/it] {'loss': 1.3172, 'grad_norm': 0.001710968185197135, 'learning_rate': 0.17026854507546693, 'epoch': 0.28} + 28%|██▊ | 143/520 [09:14<22:55, 3.65s/it] 28%|██▊ | 144/520 [09:18<22:48, 3.64s/it] {'loss': 1.2767, 'grad_norm': 0.0016279604104813428, 'learning_rate': 0.1698236818086073, 'epoch': 0.28} + 28%|██▊ | 144/520 [09:18<22:48, 3.64s/it] 28%|██▊ | 145/520 [09:21<22:47, 3.65s/it] {'loss': 1.2088, 'grad_norm': 0.0013982283420346578, 'learning_rate': 0.16937610560285418, 'epoch': 0.28} + 28%|██▊ | 145/520 [09:21<22:47, 3.65s/it] 28%|██▊ | 146/520 [09:25<22:41, 3.64s/it] {'loss': 1.4053, 'grad_norm': 0.0016054638373152847, 'learning_rate': 0.1689258338483947, 'epoch': 0.28} + 28%|██▊ | 146/520 [09:25<22:41, 3.64s/it] 28%|██▊ | 147/520 [09:29<22:37, 3.64s/it] {'loss': 1.2651, 'grad_norm': 0.0015914923677445487, 'learning_rate': 0.16847288404014937, 'epoch': 0.28} + 28%|██▊ | 147/520 [09:29<22:37, 3.64s/it] 28%|██▊ | 148/520 [09:32<22:32, 3.64s/it] {'loss': 1.2852, 'grad_norm': 0.0014634282092856686, 'learning_rate': 0.16801727377709194, 'epoch': 0.28} + 28%|██▊ | 148/520 [09:32<22:32, 3.64s/it] 29%|██▊ | 149/520 [09:36<22:32, 3.65s/it] {'loss': 1.2313, 'grad_norm': 0.0015391250380710035, 'learning_rate': 0.16755902076156604, 'epoch': 0.29} + 29%|██▊ | 149/520 [09:36<22:32, 3.65s/it] 29%|██▉ | 150/520 [09:40<22:26, 3.64s/it] {'loss': 1.4619, 'grad_norm': 0.0015884844423072902, 'learning_rate': 0.16709814279859703, 'epoch': 0.29} + 29%|██▉ | 150/520 [09:40<22:26, 3.64s/it] 29%|██▉ | 151/520 [09:43<22:21, 3.64s/it] {'loss': 1.2794, 'grad_norm': 0.0014725205298785743, 'learning_rate': 0.1666346577952004, 'epoch': 0.29} + 29%|██▉ | 151/520 [09:43<22:21, 3.64s/it] 29%|██▉ | 152/520 [09:47<22:16, 3.63s/it] {'loss': 1.2515, 'grad_norm': 0.0015731858675081306, 'learning_rate': 0.16616858375968596, 'epoch': 0.29} + 29%|██▉ | 152/520 [09:47<22:16, 3.63s/it] 29%|██▉ | 153/520 [09:50<22:11, 3.63s/it] {'loss': 1.2746, 'grad_norm': 0.0014911202131381546, 'learning_rate': 0.16569993880095807, 'epoch': 0.29} + 29%|██▉ | 153/520 [09:50<22:11, 3.63s/it] 30%|██▉ | 154/520 [09:54<22:10, 3.64s/it] {'loss': 1.3691, 'grad_norm': 0.001435044964898139, 'learning_rate': 0.16522874112781213, 'epoch': 0.3} + 30%|██▉ | 154/520 [09:54<22:10, 3.64s/it] 30%|██▉ | 155/520 [09:58<22:06, 3.63s/it] {'loss': 1.2795, 'grad_norm': 0.001502674760123467, 'learning_rate': 0.16475500904822704, 'epoch': 0.3} + 30%|██▉ | 155/520 [09:58<22:06, 3.63s/it] 30%|███ | 156/520 [10:01<22:10, 3.65s/it] {'loss': 1.3063, 'grad_norm': 0.0016835734134760106, 'learning_rate': 0.16427876096865393, 'epoch': 0.3} + 30%|███ | 156/520 [10:01<22:10, 3.65s/it] 30%|███ | 157/520 [10:05<22:03, 3.64s/it] {'loss': 1.3945, 'grad_norm': 0.001500670650416316, 'learning_rate': 0.16380001539330089, 'epoch': 0.3} + 30%|███ | 157/520 [10:05<22:03, 3.64s/it] 30%|███ | 158/520 [10:09<21:54, 3.63s/it] {'loss': 1.2858, 'grad_norm': 0.0015882780198306938, 'learning_rate': 0.163318790923414, 'epoch': 0.3} + 30%|███ | 158/520 [10:09<21:54, 3.63s/it] 31%|███ | 159/520 [10:12<21:45, 3.62s/it] {'loss': 1.32, 'grad_norm': 0.0014729187164322303, 'learning_rate': 0.16283510625655473, 'epoch': 0.31} + 31%|███ | 159/520 [10:12<21:45, 3.62s/it] 31%|███ | 160/520 [10:16<21:44, 3.62s/it] {'loss': 1.3387, 'grad_norm': 0.0015636554204741721, 'learning_rate': 0.16234898018587338, 'epoch': 0.31} + 31%|███ | 160/520 [10:16<21:44, 3.62s/it] 31%|███ | 161/520 [10:20<22:03, 3.69s/it] {'loss': 1.3128, 'grad_norm': 0.001457100349689279, 'learning_rate': 0.16186043159937882, 'epoch': 0.31} + 31%|███ | 161/520 [10:20<22:03, 3.69s/it] 31%|███ | 162/520 [10:24<22:18, 3.74s/it] {'loss': 1.3423, 'grad_norm': 0.0015096601745162009, 'learning_rate': 0.16136947947920477, 'epoch': 0.31} + 31%|███ | 162/520 [10:24<22:18, 3.74s/it] 31%|███▏ | 163/520 [10:27<22:01, 3.70s/it] {'loss': 1.1981, 'grad_norm': 0.0017026832431445455, 'learning_rate': 0.16087614290087207, 'epoch': 0.31} + 31%|███▏ | 163/520 [10:27<22:01, 3.70s/it] 32%|███▏ | 164/520 [10:31<21:51, 3.68s/it] {'loss': 1.1719, 'grad_norm': 0.0014573247827889392, 'learning_rate': 0.16038044103254775, 'epoch': 0.32} + 32%|███▏ | 164/520 [10:31<21:51, 3.68s/it] 32%|███▏ | 165/520 [10:35<21:43, 3.67s/it] {'loss': 1.3112, 'grad_norm': 0.001331726807630174, 'learning_rate': 0.15988239313430005, 'epoch': 0.32} + 32%|███▏ | 165/520 [10:35<21:43, 3.67s/it] 32%|███▏ | 166/520 [10:38<21:35, 3.66s/it] {'loss': 1.2978, 'grad_norm': 0.001638603024550063, 'learning_rate': 0.15938201855735015, 'epoch': 0.32} + 32%|███▏ | 166/520 [10:38<21:35, 3.66s/it] 32%|███▏ | 167/520 [10:42<21:27, 3.65s/it] {'loss': 1.2922, 'grad_norm': 0.001459426439882226, 'learning_rate': 0.15887933674332047, 'epoch': 0.32} + 32%|███▏ | 167/520 [10:42<21:27, 3.65s/it] 32%|███▏ | 168/520 [10:45<21:26, 3.65s/it] {'loss': 1.2207, 'grad_norm': 0.0014611467791701893, 'learning_rate': 0.158374367223479, 'epoch': 0.32} + 32%|███▏ | 168/520 [10:45<21:26, 3.65s/it] 32%|███▎ | 169/520 [10:49<21:21, 3.65s/it] {'loss': 1.3101, 'grad_norm': 0.0013757150563029213, 'learning_rate': 0.1578671296179806, 'epoch': 0.33} + 32%|███▎ | 169/520 [10:49<21:21, 3.65s/it] 33%|███▎ | 170/520 [10:53<21:18, 3.65s/it] {'loss': 1.2708, 'grad_norm': 0.0012677354033716812, 'learning_rate': 0.15735764363510463, 'epoch': 0.33} + 33%|███▎ | 170/520 [10:53<21:18, 3.65s/it] 33%|███▎ | 171/520 [10:56<21:18, 3.66s/it] {'loss': 1.2417, 'grad_norm': 0.0014916938288539886, 'learning_rate': 0.15684592907048925, 'epoch': 0.33} + 33%|███▎ | 171/520 [10:56<21:18, 3.66s/it] 33%|███▎ | 172/520 [11:00<21:12, 3.66s/it] {'loss': 1.3084, 'grad_norm': 0.001418159297456361, 'learning_rate': 0.1563320058063622, 'epoch': 0.33} + 33%|███▎ | 172/520 [11:00<21:12, 3.66s/it] 33%|███▎ | 173/520 [11:04<21:08, 3.66s/it] {'loss': 1.2483, 'grad_norm': 0.00142929128537375, 'learning_rate': 0.15581589381076844, 'epoch': 0.33} + 33%|███▎ | 173/520 [11:04<21:08, 3.66s/it] 33%|███▎ | 174/520 [11:07<21:07, 3.66s/it] {'loss': 1.3025, 'grad_norm': 0.00155875880576651, 'learning_rate': 0.15529761313679394, 'epoch': 0.33} + 33%|███▎ | 174/520 [11:07<21:07, 3.66s/it] 34%|███▎ | 175/520 [11:11<21:07, 3.67s/it] {'loss': 1.218, 'grad_norm': 0.001426701873872309, 'learning_rate': 0.15477718392178716, 'epoch': 0.34} + 34%|███▎ | 175/520 [11:11<21:07, 3.67s/it] 34%|███▍ | 176/520 [11:15<21:02, 3.67s/it] {'loss': 1.3633, 'grad_norm': 0.00145127041409917, 'learning_rate': 0.15425462638657594, 'epoch': 0.34} + 34%|███▍ | 176/520 [11:15<21:02, 3.67s/it] 34%|███▍ | 177/520 [11:18<21:00, 3.68s/it] {'loss': 1.2205, 'grad_norm': 0.0014047890466567395, 'learning_rate': 0.1537299608346824, 'epoch': 0.34} + 34%|███▍ | 177/520 [11:18<21:00, 3.68s/it] 34%|███▍ | 178/520 [11:22<20:52, 3.66s/it] {'loss': 1.2843, 'grad_norm': 0.0015192899835394873, 'learning_rate': 0.15320320765153367, 'epoch': 0.34} + 34%|███▍ | 178/520 [11:22<20:52, 3.66s/it] 34%|███▍ | 179/520 [11:26<20:48, 3.66s/it] {'loss': 1.3676, 'grad_norm': 0.0014140391166190683, 'learning_rate': 0.15267438730367008, 'epoch': 0.34} + 34%|███▍ | 179/520 [11:26<20:48, 3.66s/it] 35%|███▍ | 180/520 [11:29<20:43, 3.66s/it] {'loss': 1.2829, 'grad_norm': 0.0015147349185867227, 'learning_rate': 0.1521435203379498, 'epoch': 0.35} + 35%|███▍ | 180/520 [11:29<20:43, 3.66s/it] 35%|███▍ | 181/520 [11:33<20:38, 3.65s/it] {'loss': 1.2555, 'grad_norm': 0.0012344294847445917, 'learning_rate': 0.15161062738075068, 'epoch': 0.35} + 35%|███▍ | 181/520 [11:33<20:38, 3.65s/it] 35%|███▌ | 182/520 [11:37<20:34, 3.65s/it] {'loss': 1.2696, 'grad_norm': 0.001470622457789686, 'learning_rate': 0.1510757291371686, 'epoch': 0.35} + 35%|███▌ | 182/520 [11:37<20:34, 3.65s/it] 35%|███▌ | 183/520 [11:40<20:29, 3.65s/it] {'loss': 1.2954, 'grad_norm': 0.0014076143474277237, 'learning_rate': 0.1505388463902131, 'epoch': 0.35} + 35%|███▌ | 183/520 [11:40<20:29, 3.65s/it] 35%|███▌ | 184/520 [11:44<20:30, 3.66s/it] {'loss': 1.2199, 'grad_norm': 0.0015826973128003634, 'learning_rate': 0.15000000000000002, 'epoch': 0.35} + 35%|███▌ | 184/520 [11:44<20:30, 3.66s/it] 36%|███▌ | 185/520 [11:48<20:29, 3.67s/it] {'loss': 1.3751, 'grad_norm': 0.0014976281326783793, 'learning_rate': 0.14945921090294076, 'epoch': 0.36} + 36%|███▌ | 185/520 [11:48<20:29, 3.67s/it] 36%|███▌ | 186/520 [11:51<20:23, 3.66s/it] {'loss': 1.2463, 'grad_norm': 0.0014477144223401025, 'learning_rate': 0.14891650011092894, 'epoch': 0.36} + 36%|███▌ | 186/520 [11:51<20:23, 3.66s/it] 36%|███▌ | 187/520 [11:55<20:19, 3.66s/it] {'loss': 1.2444, 'grad_norm': 0.00177441998120183, 'learning_rate': 0.14837188871052398, 'epoch': 0.36} + 36%|███▌ | 187/520 [11:55<20:19, 3.66s/it] 36%|███▌ | 188/520 [11:59<20:14, 3.66s/it] {'loss': 1.3198, 'grad_norm': 0.0014637427048048795, 'learning_rate': 0.14782539786213184, 'epoch': 0.36} + 36%|███▌ | 188/520 [11:59<20:14, 3.66s/it] 36%|███▋ | 189/520 [12:02<20:09, 3.65s/it] {'loss': 1.3337, 'grad_norm': 0.0013288505815333808, 'learning_rate': 0.1472770487991827, 'epoch': 0.36} + 36%|███▋ | 189/520 [12:02<20:09, 3.65s/it] 37%|███▋ | 190/520 [12:06<20:05, 3.65s/it] {'loss': 1.2526, 'grad_norm': 0.0014875143397374033, 'learning_rate': 0.1467268628273062, 'epoch': 0.37} + 37%|███▋ | 190/520 [12:06<20:05, 3.65s/it] 37%|███▋ | 191/520 [12:10<20:01, 3.65s/it] {'loss': 1.205, 'grad_norm': 0.0012891364266170843, 'learning_rate': 0.1461748613235034, 'epoch': 0.37} + 37%|███▋ | 191/520 [12:10<20:01, 3.65s/it] 37%|███▋ | 192/520 [12:13<19:56, 3.65s/it] {'loss': 1.2903, 'grad_norm': 0.0014013624416305062, 'learning_rate': 0.1456210657353163, 'epoch': 0.37} + 37%|███▋ | 192/520 [12:13<19:56, 3.65s/it] 37%|███▋ | 193/520 [12:17<19:51, 3.64s/it] {'loss': 1.2785, 'grad_norm': 0.0015312865252239218, 'learning_rate': 0.14506549757999454, 'epoch': 0.37} + 37%|███▋ | 193/520 [12:17<19:51, 3.64s/it] 37%|███▋ | 194/520 [12:20<19:45, 3.64s/it] {'loss': 1.1639, 'grad_norm': 0.001336533515495972, 'learning_rate': 0.14450817844365924, 'epoch': 0.37} + 37%|███▋ | 194/520 [12:20<19:45, 3.64s/it] 38%|███▊ | 195/520 [12:24<19:43, 3.64s/it] {'loss': 1.3057, 'grad_norm': 0.001403102624417084, 'learning_rate': 0.1439491299804645, 'epoch': 0.38} + 38%|███▊ | 195/520 [12:24<19:43, 3.64s/it] 38%|███▊ | 196/520 [12:28<19:39, 3.64s/it] {'loss': 1.2747, 'grad_norm': 0.00150742304206603, 'learning_rate': 0.14338837391175582, 'epoch': 0.38} + 38%|███▊ | 196/520 [12:28<19:39, 3.64s/it] 38%|███▊ | 197/520 [12:31<19:38, 3.65s/it] {'loss': 1.2349, 'grad_norm': 0.0014065993577229822, 'learning_rate': 0.14282593202522628, 'epoch': 0.38} + 38%|███▊ | 197/520 [12:31<19:38, 3.65s/it] 38%|███▊ | 198/520 [12:35<19:35, 3.65s/it] {'loss': 1.309, 'grad_norm': 0.0015197100940245744, 'learning_rate': 0.14226182617406996, 'epoch': 0.38} + 38%|███▊ | 198/520 [12:35<19:35, 3.65s/it] 38%|███▊ | 199/520 [12:39<19:34, 3.66s/it] {'loss': 1.2187, 'grad_norm': 0.001414328788579178, 'learning_rate': 0.14169607827613281, 'epoch': 0.38} + 38%|███▊ | 199/520 [12:39<19:34, 3.66s/it] 38%|███▊ | 200/520 [12:42<19:27, 3.65s/it] {'loss': 1.2213, 'grad_norm': 0.0014707844072585326, 'learning_rate': 0.14112871031306118, 'epoch': 0.38} + 38%|███▊ | 200/520 [12:42<19:27, 3.65s/it] 39%|███▊ | 201/520 [12:46<19:23, 3.65s/it] {'loss': 1.2448, 'grad_norm': 0.0012486991071186562, 'learning_rate': 0.1405597443294475, 'epoch': 0.39} + 39%|███▊ | 201/520 [12:46<19:23, 3.65s/it] 39%|███▉ | 202/520 [12:50<19:22, 3.65s/it] {'loss': 1.223, 'grad_norm': 0.0014544755507226717, 'learning_rate': 0.13998920243197407, 'epoch': 0.39} + 39%|███▉ | 202/520 [12:50<19:22, 3.65s/it] 39%|███▉ | 203/520 [12:53<19:14, 3.64s/it] {'loss': 1.2801, 'grad_norm': 0.0014854725992460912, 'learning_rate': 0.13941710678855396, 'epoch': 0.39} + 39%|███▉ | 203/520 [12:53<19:14, 3.64s/it] 39%|███▉ | 204/520 [12:57<19:14, 3.65s/it] {'loss': 1.2975, 'grad_norm': 0.0014178401189589634, 'learning_rate': 0.13884347962746948, 'epoch': 0.39} + 39%|███▉ | 204/520 [12:57<19:14, 3.65s/it] 39%|███▉ | 205/520 [13:01<19:11, 3.65s/it] {'loss': 1.2481, 'grad_norm': 0.001375749675235421, 'learning_rate': 0.138268343236509, 'epoch': 0.39} + 39%|███▉ | 205/520 [13:01<19:11, 3.65s/it] 40%|███▉ | 206/520 [13:04<19:06, 3.65s/it] {'loss': 1.3373, 'grad_norm': 0.0014403449274058273, 'learning_rate': 0.13769171996210053, 'epoch': 0.4} + 40%|███▉ | 206/520 [13:04<19:06, 3.65s/it] 40%|███▉ | 207/520 [13:08<18:57, 3.63s/it] {'loss': 1.2273, 'grad_norm': 0.0012741545915324543, 'learning_rate': 0.1371136322084438, 'epoch': 0.4} + 40%|███▉ | 207/520 [13:08<18:57, 3.63s/it] 40%|████ | 208/520 [13:12<18:53, 3.63s/it] {'loss': 1.3166, 'grad_norm': 0.0015276095032046568, 'learning_rate': 0.13653410243663952, 'epoch': 0.4} + 40%|████ | 208/520 [13:12<18:53, 3.63s/it] 40%|████ | 209/520 [13:15<18:49, 3.63s/it] {'loss': 1.2277, 'grad_norm': 0.0013128441698091623, 'learning_rate': 0.13595315316381676, 'epoch': 0.4} + 40%|████ | 209/520 [13:15<18:49, 3.63s/it] 40%|████ | 210/520 [13:19<18:45, 3.63s/it] {'loss': 1.302, 'grad_norm': 0.0014309453407662504, 'learning_rate': 0.13537080696225814, 'epoch': 0.4} + 40%|████ | 210/520 [13:19<18:45, 3.63s/it] 41%|████ | 211/520 [13:22<18:42, 3.63s/it] {'loss': 1.3122, 'grad_norm': 0.0013181106551085533, 'learning_rate': 0.13478708645852272, 'epoch': 0.41} + 41%|████ | 211/520 [13:22<18:42, 3.63s/it] 41%|████ | 212/520 [13:26<18:38, 3.63s/it] {'loss': 1.2975, 'grad_norm': 0.001410660259999712, 'learning_rate': 0.1342020143325669, 'epoch': 0.41} + 41%|████ | 212/520 [13:26<18:38, 3.63s/it] 41%|████ | 213/520 [13:30<18:38, 3.64s/it] {'loss': 1.2441, 'grad_norm': 0.0016420382875104527, 'learning_rate': 0.13361561331686309, 'epoch': 0.41} + 41%|████ | 213/520 [13:30<18:38, 3.64s/it] 41%|████ | 214/520 [13:33<18:33, 3.64s/it] {'loss': 1.239, 'grad_norm': 0.0013926474981198833, 'learning_rate': 0.13302790619551672, 'epoch': 0.41} + 41%|████ | 214/520 [13:33<18:33, 3.64s/it] 41%|████▏ | 215/520 [13:37<18:32, 3.65s/it] {'loss': 1.176, 'grad_norm': 0.0013385534926246363, 'learning_rate': 0.1324389158033807, 'epoch': 0.41} + 41%|████▏ | 215/520 [13:37<18:32, 3.65s/it] 42%|████▏ | 216/520 [13:41<18:27, 3.64s/it] {'loss': 1.1572, 'grad_norm': 0.0013609252687871818, 'learning_rate': 0.13184866502516845, 'epoch': 0.42} + 42%|████▏ | 216/520 [13:41<18:27, 3.64s/it] 42%|████▏ | 217/520 [13:44<18:22, 3.64s/it] {'loss': 1.2835, 'grad_norm': 0.0013979893749468564, 'learning_rate': 0.13125717679456447, 'epoch': 0.42} + 42%|████▏ | 217/520 [13:44<18:22, 3.64s/it] 42%|████▏ | 218/520 [13:48<18:16, 3.63s/it] {'loss': 1.2672, 'grad_norm': 0.001495147730548393, 'learning_rate': 0.13066447409333345, 'epoch': 0.42} + 42%|████▏ | 218/520 [13:48<18:16, 3.63s/it] 42%|████▏ | 219/520 [13:52<18:11, 3.63s/it] {'loss': 1.264, 'grad_norm': 0.001255515013483906, 'learning_rate': 0.1300705799504273, 'epoch': 0.42} + 42%|████▏ | 219/520 [13:52<18:11, 3.63s/it] 42%|████▏ | 220/520 [13:55<18:10, 3.63s/it] {'loss': 1.2222, 'grad_norm': 0.0013387466255039308, 'learning_rate': 0.12947551744109043, 'epoch': 0.42} + 42%|████▏ | 220/520 [13:55<18:10, 3.63s/it] 42%|████▎ | 221/520 [13:59<18:08, 3.64s/it] {'loss': 1.2766, 'grad_norm': 0.0013830809672187767, 'learning_rate': 0.128879309685963, 'epoch': 0.42} + 42%|████▎ | 221/520 [13:59<18:08, 3.64s/it] 43%|████▎ | 222/520 [14:03<18:09, 3.66s/it] {'loss': 1.2098, 'grad_norm': 0.001338750278832306, 'learning_rate': 0.12828197985018275, 'epoch': 0.43} + 43%|████▎ | 222/520 [14:03<18:09, 3.66s/it] 43%|████▎ | 223/520 [14:06<18:03, 3.65s/it] {'loss': 1.2005, 'grad_norm': 0.001361157741521117, 'learning_rate': 0.12768355114248495, 'epoch': 0.43} + 43%|████▎ | 223/520 [14:06<18:03, 3.65s/it] 43%|████▎ | 224/520 [14:10<17:58, 3.64s/it] {'loss': 1.3172, 'grad_norm': 0.0013400409389070895, 'learning_rate': 0.12708404681430052, 'epoch': 0.43} + 43%|████▎ | 224/520 [14:10<17:58, 3.64s/it] 43%|████▎ | 225/520 [14:13<17:51, 3.63s/it] {'loss': 1.2109, 'grad_norm': 0.0013596418388368322, 'learning_rate': 0.1264834901588527, 'epoch': 0.43} + 43%|████▎ | 225/520 [14:13<17:51, 3.63s/it] 43%|████▎ | 226/520 [14:17<17:46, 3.63s/it] {'loss': 1.3104, 'grad_norm': 0.0013730671086912578, 'learning_rate': 0.12588190451025208, 'epoch': 0.43} + 43%|████▎ | 226/520 [14:17<17:46, 3.63s/it] 44%|████▎ | 227/520 [14:21<17:44, 3.63s/it] {'loss': 1.2952, 'grad_norm': 0.0012742364265154744, 'learning_rate': 0.12527931324258976, 'epoch': 0.44} + 44%|████▎ | 227/520 [14:21<17:44, 3.63s/it] 44%|████▍ | 228/520 [14:24<17:43, 3.64s/it] {'loss': 1.3445, 'grad_norm': 0.0014039473047742883, 'learning_rate': 0.12467573976902935, 'epoch': 0.44} + 44%|████▍ | 228/520 [14:24<17:43, 3.64s/it] 44%|████▍ | 229/520 [14:28<17:39, 3.64s/it] {'loss': 1.2683, 'grad_norm': 0.001235318310073495, 'learning_rate': 0.12407120754089732, 'epoch': 0.44} + 44%|████▍ | 229/520 [14:28<17:39, 3.64s/it] 44%|████▍ | 230/520 [14:32<17:37, 3.65s/it] {'loss': 1.1528, 'grad_norm': 0.0013086054951266788, 'learning_rate': 0.12346574004677154, 'epoch': 0.44} + 44%|████▍ | 230/520 [14:32<17:37, 3.65s/it] 44%|████▍ | 231/520 [14:35<17:35, 3.65s/it] {'loss': 1.2168, 'grad_norm': 0.00125728515605982, 'learning_rate': 0.12285936081156897, 'epoch': 0.44} + 44%|████▍ | 231/520 [14:35<17:35, 3.65s/it] 45%|████▍ | 232/520 [14:39<17:34, 3.66s/it] {'loss': 1.3771, 'grad_norm': 0.001440554723445132, 'learning_rate': 0.12225209339563144, 'epoch': 0.45} + 45%|████▍ | 232/520 [14:39<17:34, 3.66s/it] 45%|████▍ | 233/520 [14:43<17:30, 3.66s/it] {'loss': 1.2651, 'grad_norm': 0.0015015280176622144, 'learning_rate': 0.12164396139381028, 'epoch': 0.45} + 45%|████▍ | 233/520 [14:43<17:30, 3.66s/it] 45%|████▌ | 234/520 [14:46<17:28, 3.67s/it] {'loss': 1.1697, 'grad_norm': 0.0014311563614175734, 'learning_rate': 0.12103498843454959, 'epoch': 0.45} + 45%|████▌ | 234/520 [14:46<17:28, 3.67s/it] 45%|████▌ | 235/520 [14:50<17:23, 3.66s/it] {'loss': 1.2245, 'grad_norm': 0.0013832505415629205, 'learning_rate': 0.12042519817896805, 'epoch': 0.45} + 45%|████▌ | 235/520 [14:50<17:23, 3.66s/it] 45%|████▌ | 236/520 [14:54<17:30, 3.70s/it] {'loss': 1.3093, 'grad_norm': 0.0013021531940274479, 'learning_rate': 0.11981461431993977, 'epoch': 0.45} + 45%|████▌ | 236/520 [14:54<17:30, 3.70s/it] 46%|████▌ | 237/520 [14:58<17:34, 3.73s/it] {'loss': 1.3001, 'grad_norm': 0.0013115632682051937, 'learning_rate': 0.11920326058117364, 'epoch': 0.46} + 46%|████▌ | 237/520 [14:58<17:34, 3.73s/it] 46%|████▌ | 238/520 [15:01<17:35, 3.74s/it] {'loss': 1.232, 'grad_norm': 0.0013863434568208341, 'learning_rate': 0.11859116071629149, 'epoch': 0.46} + 46%|████▌ | 238/520 [15:01<17:35, 3.74s/it] 46%|████▌ | 239/520 [15:05<17:38, 3.77s/it] {'loss': 1.3168, 'grad_norm': 0.0014502345625499056, 'learning_rate': 0.11797833850790528, 'epoch': 0.46} + 46%|████▌ | 239/520 [15:05<17:38, 3.77s/it] 46%|████▌ | 240/520 [15:09<17:37, 3.78s/it] {'loss': 1.1086, 'grad_norm': 0.001259983390997156, 'learning_rate': 0.11736481776669305, 'epoch': 0.46} + 46%|████▌ | 240/520 [15:09<17:37, 3.78s/it] 46%|████▋ | 241/520 [15:13<17:35, 3.78s/it] {'loss': 1.1983, 'grad_norm': 0.0012968298976403024, 'learning_rate': 0.11675062233047365, 'epoch': 0.46} + 46%|████▋ | 241/520 [15:13<17:35, 3.78s/it] 47%|████▋ | 242/520 [15:17<17:32, 3.79s/it] {'loss': 1.214, 'grad_norm': 0.00126322147349843, 'learning_rate': 0.11613577606328068, 'epoch': 0.47} + 47%|████▋ | 242/520 [15:17<17:32, 3.79s/it] 47%|████▋ | 243/520 [15:20<17:31, 3.79s/it] {'loss': 1.2106, 'grad_norm': 0.0013675030437134302, 'learning_rate': 0.11552030285443515, 'epoch': 0.47} + 47%|████▋ | 243/520 [15:20<17:31, 3.79s/it] 47%|████▋ | 244/520 [15:24<17:27, 3.80s/it] {'loss': 1.3313, 'grad_norm': 0.0013857632083820594, 'learning_rate': 0.11490422661761744, 'epoch': 0.47} + 47%|████▋ | 244/520 [15:24<17:27, 3.80s/it] 47%|████▋ | 245/520 [15:28<17:25, 3.80s/it] {'loss': 1.186, 'grad_norm': 0.0013589891267231156, 'learning_rate': 0.11428757128993801, 'epoch': 0.47} + 47%|████▋ | 245/520 [15:28<17:25, 3.80s/it] 47%|████▋ | 246/520 [15:32<17:20, 3.80s/it] {'loss': 1.3549, 'grad_norm': 0.0014424625943554503, 'learning_rate': 0.11367036083100734, 'epoch': 0.47} + 47%|████▋ | 246/520 [15:32<17:20, 3.80s/it] 48%|████▊ | 247/520 [15:36<17:17, 3.80s/it] {'loss': 1.3631, 'grad_norm': 0.0013756297755245185, 'learning_rate': 0.11305261922200519, 'epoch': 0.47} + 48%|████▊ | 247/520 [15:36<17:17, 3.80s/it] 48%|████▊ | 248/520 [15:39<17:12, 3.80s/it] {'loss': 1.1915, 'grad_norm': 0.0013985394173174912, 'learning_rate': 0.11243437046474854, 'epoch': 0.48} + 48%|████▊ | 248/520 [15:39<17:12, 3.80s/it] 48%|████▊ | 249/520 [15:43<17:09, 3.80s/it] {'loss': 1.2873, 'grad_norm': 0.0013617474240016057, 'learning_rate': 0.1118156385807593, 'epoch': 0.48} + 48%|████▊ | 249/520 [15:43<17:09, 3.80s/it] 48%|████▊ | 250/520 [15:47<17:05, 3.80s/it] {'loss': 1.2199, 'grad_norm': 0.0014734458030415392, 'learning_rate': 0.11119644761033079, 'epoch': 0.48} + 48%|████▊ | 250/520 [15:47<17:05, 3.80s/it] 48%|████▊ | 251/520 [15:51<17:01, 3.80s/it] {'loss': 1.288, 'grad_norm': 0.0013166785727250741, 'learning_rate': 0.1105768216115938, 'epoch': 0.48} + 48%|████▊ | 251/520 [15:51<17:01, 3.80s/it] 48%|████▊ | 252/520 [15:55<16:56, 3.79s/it] {'loss': 1.2538, 'grad_norm': 0.0013538708677515612, 'learning_rate': 0.10995678465958168, 'epoch': 0.48} + 48%|████▊ | 252/520 [15:55<16:56, 3.79s/it] 49%|████▊ | 253/520 [15:58<16:52, 3.79s/it] {'loss': 1.2939, 'grad_norm': 0.0014296770196404979, 'learning_rate': 0.10933636084529506, 'epoch': 0.49} + 49%|████▊ | 253/520 [15:58<16:52, 3.79s/it] 49%|████▉ | 254/520 [16:02<16:48, 3.79s/it] {'loss': 1.2114, 'grad_norm': 0.0013034984458290533, 'learning_rate': 0.10871557427476584, 'epoch': 0.49} + 49%|████▉ | 254/520 [16:02<16:48, 3.79s/it] 49%|████▉ | 255/520 [16:06<16:43, 3.79s/it] {'loss': 1.2243, 'grad_norm': 0.0014203428518939857, 'learning_rate': 0.10809444906812034, 'epoch': 0.49} + 49%|████▉ | 255/520 [16:06<16:43, 3.79s/it] 49%|████▉ | 256/520 [16:10<16:38, 3.78s/it] {'loss': 1.2739, 'grad_norm': 0.0014093259506644305, 'learning_rate': 0.10747300935864244, 'epoch': 0.49} + 49%|████▉ | 256/520 [16:10<16:38, 3.78s/it] 49%|████▉ | 257/520 [16:13<16:34, 3.78s/it] {'loss': 1.2558, 'grad_norm': 0.0013973591848123346, 'learning_rate': 0.10685127929183567, 'epoch': 0.49} + 49%|████▉ | 257/520 [16:13<16:34, 3.78s/it] 50%|████▉ | 258/520 [16:17<16:36, 3.80s/it] {'loss': 1.267, 'grad_norm': 0.0012259185043141188, 'learning_rate': 0.10622928302448523, 'epoch': 0.5} + 50%|████▉ | 258/520 [16:17<16:36, 3.80s/it] 50%|████▉ | 259/520 [16:21<16:30, 3.80s/it] {'loss': 1.3476, 'grad_norm': 0.001580772702938286, 'learning_rate': 0.10560704472371918, 'epoch': 0.5} + 50%|████▉ | 259/520 [16:21<16:30, 3.80s/it] 50%|█████ | 260/520 [16:25<16:27, 3.80s/it] {'loss': 1.3035, 'grad_norm': 0.00120165421587445, 'learning_rate': 0.10498458856606972, 'epoch': 0.5} + 50%|█████ | 260/520 [16:25<16:27, 3.80s/it] 50%|█████ | 261/520 [16:29<16:20, 3.79s/it] {'loss': 1.2423, 'grad_norm': 0.001353362091183934, 'learning_rate': 0.10436193873653361, 'epoch': 0.5} + 50%|█████ | 261/520 [16:29<16:20, 3.79s/it] 50%|█████ | 262/520 [16:32<16:17, 3.79s/it] {'loss': 1.195, 'grad_norm': 0.0013507638791492374, 'learning_rate': 0.10373911942763259, 'epoch': 0.5} + 50%|█████ | 262/520 [16:32<16:17, 3.79s/it] 51%|█████ | 263/520 [16:36<16:12, 3.79s/it] {'loss': 1.2648, 'grad_norm': 0.0013455003122762178, 'learning_rate': 0.10311615483847332, 'epoch': 0.51} + 51%|█████ | 263/520 [16:36<16:12, 3.79s/it] 51%|█████ | 264/520 [16:40<16:09, 3.79s/it] {'loss': 1.2954, 'grad_norm': 0.0012813922328728119, 'learning_rate': 0.1024930691738073, 'epoch': 0.51} + 51%|█████ | 264/520 [16:40<16:09, 3.79s/it] 51%|█████ | 265/520 [16:44<16:03, 3.78s/it] {'loss': 1.1982, 'grad_norm': 0.001491380783571623, 'learning_rate': 0.10186988664309023, 'epoch': 0.51} + 51%|█████ | 265/520 [16:44<16:03, 3.78s/it] 51%|█████ | 266/520 [16:48<16:02, 3.79s/it] {'loss': 1.0701, 'grad_norm': 0.0011728844084613716, 'learning_rate': 0.10124663145954152, 'epoch': 0.51} + 51%|█████ | 266/520 [16:48<16:02, 3.79s/it] 51%|█████▏ | 267/520 [16:51<15:58, 3.79s/it] {'loss': 1.2031, 'grad_norm': 0.001265892381166571, 'learning_rate': 0.10062332783920336, 'epoch': 0.51} + 51%|█████▏ | 267/520 [16:51<15:58, 3.79s/it] 52%|█████▏ | 268/520 [16:55<15:53, 3.79s/it] {'loss': 1.3699, 'grad_norm': 0.001488853226800848, 'learning_rate': 0.1, 'epoch': 0.52} + 52%|█████▏ | 268/520 [16:55<15:53, 3.79s/it] 52%|█████▏ | 269/520 [16:59<15:48, 3.78s/it] {'loss': 1.3024, 'grad_norm': 0.0014091635272721877, 'learning_rate': 0.09937667216079665, 'epoch': 0.52} + 52%|█████▏ | 269/520 [16:59<15:48, 3.78s/it] 52%|█████▏ | 270/520 [17:03<15:46, 3.78s/it] {'loss': 1.1914, 'grad_norm': 0.0013031507147992585, 'learning_rate': 0.0987533685404585, 'epoch': 0.52} + 52%|█████▏ | 270/520 [17:03<15:46, 3.78s/it] 52%|█████▏ | 271/520 [17:07<15:45, 3.80s/it] {'loss': 1.292, 'grad_norm': 0.0013413614525852344, 'learning_rate': 0.0981301133569098, 'epoch': 0.52} + 52%|█████▏ | 271/520 [17:07<15:45, 3.80s/it] 52%|█████▏ | 272/520 [17:10<15:40, 3.79s/it] {'loss': 1.2127, 'grad_norm': 0.001328829177843661, 'learning_rate': 0.09750693082619273, 'epoch': 0.52} + 52%|█████▏ | 272/520 [17:10<15:40, 3.79s/it] 52%|█████▎ | 273/520 [17:14<15:34, 3.78s/it] {'loss': 1.3227, 'grad_norm': 0.0012941923541237134, 'learning_rate': 0.0968838451615267, 'epoch': 0.53} + 52%|█████▎ | 273/520 [17:14<15:34, 3.78s/it] 53%|█████▎ | 274/520 [17:18<15:30, 3.78s/it] {'loss': 1.2644, 'grad_norm': 0.0014232314041934563, 'learning_rate': 0.09626088057236745, 'epoch': 0.53} + 53%|█████▎ | 274/520 [17:18<15:30, 3.78s/it] 53%|█████▎ | 275/520 [17:22<15:27, 3.79s/it] {'loss': 1.2092, 'grad_norm': 0.0013430810971062092, 'learning_rate': 0.09563806126346641, 'epoch': 0.53} + 53%|█████▎ | 275/520 [17:22<15:27, 3.79s/it] 53%|█████▎ | 276/520 [17:25<15:24, 3.79s/it] {'loss': 1.2721, 'grad_norm': 0.0014557043711921536, 'learning_rate': 0.09501541143393027, 'epoch': 0.53} + 53%|█████▎ | 276/520 [17:25<15:24, 3.79s/it] 53%|█████▎ | 277/520 [17:29<15:23, 3.80s/it] {'loss': 1.3276, 'grad_norm': 0.0012793615469035712, 'learning_rate': 0.09439295527628082, 'epoch': 0.53} + 53%|█████▎ | 277/520 [17:29<15:23, 3.80s/it] 53%|█████▎ | 278/520 [17:33<15:20, 3.80s/it] {'loss': 1.1546, 'grad_norm': 0.0012018861131475331, 'learning_rate': 0.0937707169755148, 'epoch': 0.53} + 53%|█████▎ | 278/520 [17:33<15:20, 3.80s/it] 54%|█████▎ | 279/520 [17:37<15:15, 3.80s/it] {'loss': 1.2079, 'grad_norm': 0.0013758718081197776, 'learning_rate': 0.09314872070816434, 'epoch': 0.54} + 54%|█████▎ | 279/520 [17:37<15:15, 3.80s/it] 54%|█████▍ | 280/520 [17:41<15:15, 3.81s/it] {'loss': 1.2068, 'grad_norm': 0.0015327962136808136, 'learning_rate': 0.09252699064135758, 'epoch': 0.54} + 54%|█████▍ | 280/520 [17:41<15:15, 3.81s/it] 54%|█████▍ | 281/520 [17:44<15:10, 3.81s/it] {'loss': 1.303, 'grad_norm': 0.0013829793298141303, 'learning_rate': 0.09190555093187967, 'epoch': 0.54} + 54%|█████▍ | 281/520 [17:44<15:10, 3.81s/it] 54%|█████▍ | 282/520 [17:48<15:05, 3.81s/it] {'loss': 1.1751, 'grad_norm': 0.0012655914744424724, 'learning_rate': 0.09128442572523418, 'epoch': 0.54} + 54%|█████▍ | 282/520 [17:48<15:05, 3.81s/it] 54%|█████▍ | 283/520 [17:52<14:59, 3.80s/it] {'loss': 1.3295, 'grad_norm': 0.0014532379503920756, 'learning_rate': 0.09066363915470495, 'epoch': 0.54} + 54%|█████▍ | 283/520 [17:52<14:59, 3.80s/it] 55%|█████▍ | 284/520 [17:56<14:47, 3.76s/it] {'loss': 1.2035, 'grad_norm': 0.0014009957737909826, 'learning_rate': 0.09004321534041836, 'epoch': 0.55} + 55%|█████▍ | 284/520 [17:56<14:47, 3.76s/it] 55%|█████▍ | 285/520 [17:59<14:31, 3.71s/it] {'loss': 1.1955, 'grad_norm': 0.0013168436115621374, 'learning_rate': 0.08942317838840624, 'epoch': 0.55} + 55%|█████▍ | 285/520 [17:59<14:31, 3.71s/it] 55%|█████▌ | 286/520 [18:03<14:21, 3.68s/it] {'loss': 1.0745, 'grad_norm': 0.0013325117998717667, 'learning_rate': 0.08880355238966922, 'epoch': 0.55} + 55%|█████▌ | 286/520 [18:03<14:21, 3.68s/it] 55%|█████▌ | 287/520 [18:07<14:15, 3.67s/it] {'loss': 1.3088, 'grad_norm': 0.0013023215380890145, 'learning_rate': 0.08818436141924073, 'epoch': 0.55} + 55%|█████▌ | 287/520 [18:07<14:15, 3.67s/it] 55%|█████▌ | 288/520 [18:10<14:08, 3.66s/it] {'loss': 1.3402, 'grad_norm': 0.0012993686254203329, 'learning_rate': 0.08756562953525152, 'epoch': 0.55} + 55%|█████▌ | 288/520 [18:10<14:08, 3.66s/it] 56%|█████▌ | 289/520 [18:14<14:02, 3.65s/it] {'loss': 1.2115, 'grad_norm': 0.0012282342360205505, 'learning_rate': 0.08694738077799487, 'epoch': 0.56} + 56%|█████▌ | 289/520 [18:14<14:02, 3.65s/it] 56%|█████▌ | 290/520 [18:18<13:59, 3.65s/it] {'loss': 1.1356, 'grad_norm': 0.0012149736939465058, 'learning_rate': 0.08632963916899268, 'epoch': 0.56} + 56%|█████▌ | 290/520 [18:18<13:59, 3.65s/it] 56%|█████▌ | 291/520 [18:21<13:53, 3.64s/it] {'loss': 1.1863, 'grad_norm': 0.0012811110152590394, 'learning_rate': 0.08571242871006202, 'epoch': 0.56} + 56%|█████▌ | 291/520 [18:21<13:53, 3.64s/it] 56%|█████▌ | 292/520 [18:25<13:47, 3.63s/it] {'loss': 1.2344, 'grad_norm': 0.0012931305478231128, 'learning_rate': 0.08509577338238256, 'epoch': 0.56} + 56%|█████▌ | 292/520 [18:25<13:47, 3.63s/it] 56%|█████▋ | 293/520 [18:28<13:47, 3.65s/it] {'loss': 1.1811, 'grad_norm': 0.0013977713512836297, 'learning_rate': 0.08447969714556484, 'epoch': 0.56} + 56%|█████▋ | 293/520 [18:28<13:47, 3.65s/it] 57%|█████▋ | 294/520 [18:32<13:43, 3.64s/it] {'loss': 1.195, 'grad_norm': 0.0013283631274092406, 'learning_rate': 0.08386422393671933, 'epoch': 0.57} + 57%|█████▋ | 294/520 [18:32<13:43, 3.64s/it] 57%|█████▋ | 295/520 [18:36<13:39, 3.64s/it] {'loss': 1.2538, 'grad_norm': 0.0013331712034916053, 'learning_rate': 0.08324937766952638, 'epoch': 0.57} + 57%|█████▋ | 295/520 [18:36<13:39, 3.64s/it] 57%|█████▋ | 296/520 [18:39<13:33, 3.63s/it] {'loss': 1.1622, 'grad_norm': 0.00136715591342425, 'learning_rate': 0.08263518223330697, 'epoch': 0.57} + 57%|█████▋ | 296/520 [18:39<13:33, 3.63s/it] 57%|█████▋ | 297/520 [18:43<13:30, 3.63s/it] {'loss': 1.2836, 'grad_norm': 0.0014433831235635193, 'learning_rate': 0.08202166149209474, 'epoch': 0.57} + 57%|█████▋ | 297/520 [18:43<13:30, 3.63s/it] 57%|█████▋ | 298/520 [18:47<13:24, 3.63s/it] {'loss': 1.2524, 'grad_norm': 0.0012631867934841789, 'learning_rate': 0.08140883928370855, 'epoch': 0.57} + 57%|█████▋ | 298/520 [18:47<13:24, 3.63s/it] 57%|█████▊ | 299/520 [18:50<13:20, 3.62s/it] {'loss': 1.2788, 'grad_norm': 0.0012527299845838478, 'learning_rate': 0.0807967394188264, 'epoch': 0.57} + 57%|█████▊ | 299/520 [18:50<13:20, 3.62s/it] 58%|█████▊ | 300/520 [18:54<13:16, 3.62s/it] {'loss': 1.2986, 'grad_norm': 0.001309332347083669, 'learning_rate': 0.08018538568006027, 'epoch': 0.58} + 58%|█████▊ | 300/520 [18:54<13:16, 3.62s/it] 58%|█████▊ | 301/520 [18:58<13:20, 3.66s/it] {'loss': 1.2761, 'grad_norm': 0.0012772533050386975, 'learning_rate': 0.07957480182103199, 'epoch': 0.58} + 58%|█████▊ | 301/520 [18:58<13:20, 3.66s/it] 58%|█████▊ | 302/520 [19:01<13:15, 3.65s/it] {'loss': 1.3017, 'grad_norm': 0.0013237944994808988, 'learning_rate': 0.07896501156545044, 'epoch': 0.58} + 58%|█████▊ | 302/520 [19:01<13:15, 3.65s/it] 58%|█████▊ | 303/520 [19:05<13:14, 3.66s/it] {'loss': 1.201, 'grad_norm': 0.0014461176704191146, 'learning_rate': 0.07835603860618973, 'epoch': 0.58} + 58%|█████▊ | 303/520 [19:05<13:14, 3.66s/it] 58%|█████▊ | 304/520 [19:09<13:11, 3.66s/it] {'loss': 1.1943, 'grad_norm': 0.0013644942727814085, 'learning_rate': 0.07774790660436857, 'epoch': 0.58} + 58%|█████▊ | 304/520 [19:09<13:11, 3.66s/it] 59%|█████▊ | 305/520 [19:12<13:05, 3.65s/it] {'loss': 1.3129, 'grad_norm': 0.0014974094390361758, 'learning_rate': 0.07714063918843106, 'epoch': 0.59} + 59%|█████▊ | 305/520 [19:12<13:05, 3.65s/it] 59%|█████▉ | 306/520 [19:16<13:01, 3.65s/it] {'loss': 1.2581, 'grad_norm': 0.0013507588181570383, 'learning_rate': 0.0765342599532285, 'epoch': 0.59} + 59%|█████▉ | 306/520 [19:16<13:01, 3.65s/it] 59%|█████▉ | 307/520 [19:20<13:18, 3.75s/it] {'loss': 1.1928, 'grad_norm': 0.0012444254194301804, 'learning_rate': 0.07592879245910272, 'epoch': 0.59} + 59%|█████▉ | 307/520 [19:20<13:18, 3.75s/it] 59%|█████▉ | 308/520 [19:23<13:07, 3.71s/it] {'loss': 1.307, 'grad_norm': 0.0012990593609309643, 'learning_rate': 0.07532426023097064, 'epoch': 0.59} + 59%|█████▉ | 308/520 [19:23<13:07, 3.71s/it] 59%|█████▉ | 309/520 [19:27<12:58, 3.69s/it] {'loss': 1.1937, 'grad_norm': 0.0012258719542871418, 'learning_rate': 0.07472068675741024, 'epoch': 0.59} + 59%|█████▉ | 309/520 [19:27<12:58, 3.69s/it] 60%|█████▉ | 310/520 [19:31<12:50, 3.67s/it] {'loss': 1.1728, 'grad_norm': 0.0013560193765132663, 'learning_rate': 0.07411809548974792, 'epoch': 0.6} + 60%|█████▉ | 310/520 [19:31<12:50, 3.67s/it] 60%|█████▉ | 311/520 [19:34<12:45, 3.66s/it] {'loss': 1.1469, 'grad_norm': 0.0013028350094435796, 'learning_rate': 0.07351650984114727, 'epoch': 0.6} + 60%|█████▉ | 311/520 [19:34<12:45, 3.66s/it] 60%|██████ | 312/520 [19:38<12:41, 3.66s/it] {'loss': 1.1353, 'grad_norm': 0.0013288573271635143, 'learning_rate': 0.0729159531856995, 'epoch': 0.6} + 60%|██████ | 312/520 [19:38<12:41, 3.66s/it] 60%|██████ | 313/520 [19:42<12:38, 3.66s/it] {'loss': 1.1271, 'grad_norm': 0.0011798720888060734, 'learning_rate': 0.07231644885751508, 'epoch': 0.6} + 60%|██████ | 313/520 [19:42<12:38, 3.66s/it] 60%|██████ | 314/520 [19:46<13:01, 3.80s/it] {'loss': 1.1648, 'grad_norm': 0.0012436423945144876, 'learning_rate': 0.07171802014981725, 'epoch': 0.6} + 60%|██████ | 314/520 [19:46<13:01, 3.80s/it] 61%|██████ | 315/520 [19:49<12:48, 3.75s/it] {'loss': 1.2456, 'grad_norm': 0.0014168491221285499, 'learning_rate': 0.07112069031403703, 'epoch': 0.61} + 61%|██████ | 315/520 [19:49<12:48, 3.75s/it] 61%|██████ | 316/520 [19:53<13:01, 3.83s/it] {'loss': 1.1348, 'grad_norm': 0.0013309465700813167, 'learning_rate': 0.07052448255890957, 'epoch': 0.61} + 61%|██████ | 316/520 [19:53<13:01, 3.83s/it] 61%|██████ | 317/520 [19:57<12:47, 3.78s/it] {'loss': 1.1576, 'grad_norm': 0.001211591165528304, 'learning_rate': 0.0699294200495727, 'epoch': 0.61} + 61%|██████ | 317/520 [19:57<12:47, 3.78s/it] 61%|██████ | 318/520 [20:01<12:32, 3.73s/it] {'loss': 1.2719, 'grad_norm': 0.0014513353832034796, 'learning_rate': 0.06933552590666658, 'epoch': 0.61} + 61%|██████ | 318/520 [20:01<12:32, 3.73s/it] 61%|██████▏ | 319/520 [20:05<12:44, 3.80s/it] {'loss': 1.1468, 'grad_norm': 0.0012794794077063318, 'learning_rate': 0.06874282320543557, 'epoch': 0.61} + 61%|██████▏ | 319/520 [20:05<12:44, 3.80s/it] 62%|██████▏ | 320/520 [20:08<12:33, 3.77s/it] {'loss': 1.0924, 'grad_norm': 0.0012401199195681094, 'learning_rate': 0.06815133497483157, 'epoch': 0.62} + 62%|██████▏ | 320/520 [20:08<12:33, 3.77s/it] 62%|██████▏ | 321/520 [20:12<12:19, 3.72s/it] {'loss': 1.2918, 'grad_norm': 0.0013335946566557279, 'learning_rate': 0.06756108419661931, 'epoch': 0.62} + 62%|██████▏ | 321/520 [20:12<12:19, 3.72s/it] 62%|██████▏ | 322/520 [20:16<12:11, 3.69s/it] {'loss': 1.1396, 'grad_norm': 0.0012790741486166835, 'learning_rate': 0.06697209380448332, 'epoch': 0.62} + 62%|██████▏ | 322/520 [20:16<12:11, 3.69s/it] 62%|██████▏ | 323/520 [20:19<12:03, 3.67s/it] {'loss': 1.2066, 'grad_norm': 0.0012932248611490533, 'learning_rate': 0.06638438668313694, 'epoch': 0.62} + 62%|██████▏ | 323/520 [20:19<12:03, 3.67s/it] 62%|██████▏ | 324/520 [20:23<11:57, 3.66s/it] {'loss': 1.2185, 'grad_norm': 0.0012597888058857247, 'learning_rate': 0.06579798566743314, 'epoch': 0.62} + 62%|██████▏ | 324/520 [20:23<11:57, 3.66s/it] 62%|██████▎ | 325/520 [20:26<11:52, 3.65s/it] {'loss': 1.2327, 'grad_norm': 0.0013716763108900512, 'learning_rate': 0.06521291354147728, 'epoch': 0.62} + 62%|██████▎ | 325/520 [20:26<11:52, 3.65s/it] 63%|██████▎ | 326/520 [20:30<11:45, 3.64s/it] {'loss': 1.2251, 'grad_norm': 0.0013689149715867854, 'learning_rate': 0.06462919303774187, 'epoch': 0.63} + 63%|██████▎ | 326/520 [20:30<11:45, 3.64s/it] 63%|██████▎ | 327/520 [20:34<11:40, 3.63s/it] {'loss': 1.2631, 'grad_norm': 0.0014127441264882737, 'learning_rate': 0.06404684683618325, 'epoch': 0.63} + 63%|██████▎ | 327/520 [20:34<11:40, 3.63s/it] 63%|██████▎ | 328/520 [20:37<11:36, 3.63s/it] {'loss': 1.2699, 'grad_norm': 0.0013506109590055864, 'learning_rate': 0.0634658975633605, 'epoch': 0.63} + 63%|██████▎ | 328/520 [20:37<11:36, 3.63s/it] 63%|██████▎ | 329/520 [20:41<11:31, 3.62s/it] {'loss': 1.1469, 'grad_norm': 0.0011572333598502363, 'learning_rate': 0.06288636779155621, 'epoch': 0.63} + 63%|██████▎ | 329/520 [20:41<11:31, 3.62s/it] 63%|██████▎ | 330/520 [20:44<11:27, 3.62s/it] {'loss': 1.223, 'grad_norm': 0.0012127144494009157, 'learning_rate': 0.06230828003789948, 'epoch': 0.63} + 63%|██████▎ | 330/520 [20:44<11:27, 3.62s/it] 64%|██████▎ | 331/520 [20:48<11:25, 3.63s/it] {'loss': 1.1783, 'grad_norm': 0.0012983097095589596, 'learning_rate': 0.06173165676349103, 'epoch': 0.64} + 64%|██████▎ | 331/520 [20:48<11:25, 3.63s/it] 64%|██████▍ | 332/520 [20:52<11:23, 3.64s/it] {'loss': 1.2804, 'grad_norm': 0.0012357840405355619, 'learning_rate': 0.06115652037253053, 'epoch': 0.64} + 64%|██████▍ | 332/520 [20:52<11:23, 3.64s/it] 64%|██████▍ | 333/520 [20:55<11:19, 3.64s/it] {'loss': 1.3212, 'grad_norm': 0.0013767706389533076, 'learning_rate': 0.06058289321144608, 'epoch': 0.64} + 64%|██████▍ | 333/520 [20:55<11:19, 3.64s/it] 64%|██████▍ | 334/520 [20:59<11:16, 3.64s/it] {'loss': 1.2255, 'grad_norm': 0.0013538082779371411, 'learning_rate': 0.06001079756802592, 'epoch': 0.64} + 64%|██████▍ | 334/520 [20:59<11:16, 3.64s/it] 64%|██████▍ | 335/520 [21:03<11:12, 3.63s/it] {'loss': 1.224, 'grad_norm': 0.001208284649720952, 'learning_rate': 0.059440255670552514, 'epoch': 0.64} + 64%|██████▍ | 335/520 [21:03<11:12, 3.63s/it] 65%|██████▍ | 336/520 [21:06<11:09, 3.64s/it] {'loss': 1.1206, 'grad_norm': 0.0014261594435286346, 'learning_rate': 0.05887128968693887, 'epoch': 0.65} + 65%|██████▍ | 336/520 [21:06<11:09, 3.64s/it] 65%|██████▍ | 337/520 [21:10<11:04, 3.63s/it] {'loss': 1.1077, 'grad_norm': 0.0012650696754355612, 'learning_rate': 0.058303921723867225, 'epoch': 0.65} + 65%|██████▍ | 337/520 [21:10<11:04, 3.63s/it] 65%|██████▌ | 338/520 [21:14<11:00, 3.63s/it] {'loss': 1.2299, 'grad_norm': 0.0013339590300868462, 'learning_rate': 0.05773817382593008, 'epoch': 0.65} + 65%|██████▌ | 338/520 [21:14<11:00, 3.63s/it] 65%|██████▌ | 339/520 [21:17<10:56, 3.63s/it] {'loss': 1.1778, 'grad_norm': 0.0012825663707811301, 'learning_rate': 0.057174067974773715, 'epoch': 0.65} + 65%|██████▌ | 339/520 [21:17<10:56, 3.63s/it] 65%|██████▌ | 340/520 [21:21<10:54, 3.63s/it] {'loss': 1.1671, 'grad_norm': 0.001299847749278224, 'learning_rate': 0.056611626088244195, 'epoch': 0.65} + 65%|██████▌ | 340/520 [21:21<10:54, 3.63s/it] 66%|██████▌ | 341/520 [21:24<10:49, 3.63s/it] {'loss': 1.188, 'grad_norm': 0.0013642979768104546, 'learning_rate': 0.056050870019535494, 'epoch': 0.66} + 66%|██████▌ | 341/520 [21:24<10:49, 3.63s/it] 66%|██████▌ | 342/520 [21:28<10:46, 3.63s/it] {'loss': 1.2501, 'grad_norm': 0.0014910690895271778, 'learning_rate': 0.05549182155634076, 'epoch': 0.66} + 66%|██████▌ | 342/520 [21:28<10:46, 3.63s/it] 66%|██████▌ | 343/520 [21:32<10:44, 3.64s/it] {'loss': 1.2078, 'grad_norm': 0.001207167566688789, 'learning_rate': 0.054934502420005464, 'epoch': 0.66} + 66%|██████▌ | 343/520 [21:32<10:44, 3.64s/it] 66%|██████▌ | 344/520 [21:35<10:38, 3.63s/it] {'loss': 1.1398, 'grad_norm': 0.0011864317121628568, 'learning_rate': 0.0543789342646837, 'epoch': 0.66} + 66%|██████▌ | 344/520 [21:35<10:38, 3.63s/it] 66%|██████▋ | 345/520 [21:39<10:36, 3.64s/it] {'loss': 1.2558, 'grad_norm': 0.0014070135974571486, 'learning_rate': 0.05382513867649663, 'epoch': 0.66} + 66%|██████▋ | 345/520 [21:39<10:36, 3.64s/it] 67%|██████▋ | 346/520 [21:43<10:33, 3.64s/it] {'loss': 1.2188, 'grad_norm': 0.001278561654587745, 'learning_rate': 0.0532731371726938, 'epoch': 0.67} + 67%|██████▋ | 346/520 [21:43<10:33, 3.64s/it] 67%|██████▋ | 347/520 [21:46<10:28, 3.63s/it] {'loss': 1.156, 'grad_norm': 0.0011788957178230936, 'learning_rate': 0.05272295120081732, 'epoch': 0.67} + 67%|██████▋ | 347/520 [21:46<10:28, 3.63s/it]Token indices sequence length is longer than the specified maximum sequence length for this model (2778 > 2048). Running this sequence through the model will result in indexing errors + 67%|██████▋ | 348/520 [21:50<10:24, 3.63s/it] {'loss': 1.1239, 'grad_norm': 0.0014675480575507556, 'learning_rate': 0.05217460213786821, 'epoch': 0.67} + 67%|██████▋ | 348/520 [21:50<10:24, 3.63s/it] 67%|██████▋ | 349/520 [21:54<10:20, 3.63s/it] {'loss': 1.1575, 'grad_norm': 0.001283279449915364, 'learning_rate': 0.051628111289476024, 'epoch': 0.67} + 67%|██████▋ | 349/520 [21:54<10:20, 3.63s/it] 67%|██████▋ | 350/520 [21:57<10:18, 3.64s/it] {'loss': 1.1994, 'grad_norm': 0.0013313556990487733, 'learning_rate': 0.051083499889071106, 'epoch': 0.67} + 67%|██████▋ | 350/520 [21:57<10:18, 3.64s/it] 68%|██████▊ | 351/520 [22:01<10:16, 3.65s/it] {'loss': 1.1106, 'grad_norm': 0.001198155872792034, 'learning_rate': 0.05054078909705926, 'epoch': 0.68} + 68%|██████▊ | 351/520 [22:01<10:16, 3.65s/it] 68%|██████▊ | 352/520 [22:05<10:12, 3.65s/it] {'loss': 1.2304, 'grad_norm': 0.0012454953439848826, 'learning_rate': 0.050000000000000024, 'epoch': 0.68} + 68%|██████▊ | 352/520 [22:05<10:12, 3.65s/it] 68%|██████▊ | 353/520 [22:08<10:10, 3.66s/it] {'loss': 1.1686, 'grad_norm': 0.0010895307241250202, 'learning_rate': 0.04946115360978696, 'epoch': 0.68} + 68%|██████▊ | 353/520 [22:08<10:10, 3.66s/it] 68%|██████▊ | 354/520 [22:12<10:05, 3.65s/it] {'loss': 1.2951, 'grad_norm': 0.0012136751856391553, 'learning_rate': 0.048924270862831465, 'epoch': 0.68} + 68%|██████▊ | 354/520 [22:12<10:05, 3.65s/it] 68%|██████▊ | 355/520 [22:15<10:02, 3.65s/it] {'loss': 1.1681, 'grad_norm': 0.0012426596741648313, 'learning_rate': 0.04838937261924933, 'epoch': 0.68} + 68%|██████▊ | 355/520 [22:15<10:02, 3.65s/it] 68%|██████▊ | 356/520 [22:19<09:59, 3.66s/it] {'loss': 1.1642, 'grad_norm': 0.0012822268327229238, 'learning_rate': 0.0478564796620502, 'epoch': 0.68} + 68%|██████▊ | 356/520 [22:19<09:59, 3.66s/it] 69%|██████▊ | 357/520 [22:23<09:55, 3.65s/it] {'loss': 1.1969, 'grad_norm': 0.0012041025287972864, 'learning_rate': 0.04732561269632993, 'epoch': 0.69} + 69%|██████▊ | 357/520 [22:23<09:55, 3.65s/it] 69%|██████▉ | 358/520 [22:27<09:54, 3.67s/it] {'loss': 1.1339, 'grad_norm': 0.0012781310545159073, 'learning_rate': 0.04679679234846636, 'epoch': 0.69} + 69%|██████▉ | 358/520 [22:27<09:54, 3.67s/it] 69%|██████▉ | 359/520 [22:30<09:58, 3.72s/it] {'loss': 1.223, 'grad_norm': 0.0012960637238186629, 'learning_rate': 0.046270039165317606, 'epoch': 0.69} + 69%|██████▉ | 359/520 [22:30<09:58, 3.72s/it] 69%|██████▉ | 360/520 [22:34<10:03, 3.77s/it] {'loss': 1.2338, 'grad_norm': 0.001342438935447942, 'learning_rate': 0.04574537361342407, 'epoch': 0.69} + 69%|██████▉ | 360/520 [22:34<10:03, 3.77s/it] 69%|██████▉ | 361/520 [22:38<09:58, 3.76s/it] {'loss': 1.2428, 'grad_norm': 0.0011772238860645091, 'learning_rate': 0.04522281607821288, 'epoch': 0.69} + 69%|██████▉ | 361/520 [22:38<09:58, 3.76s/it] 70%|██████▉ | 362/520 [22:42<09:47, 3.72s/it] {'loss': 1.1897, 'grad_norm': 0.0013873539989264802, 'learning_rate': 0.04470238686320606, 'epoch': 0.7} + 70%|██████▉ | 362/520 [22:42<09:47, 3.72s/it] 70%|██████▉ | 363/520 [22:45<09:39, 3.69s/it] {'loss': 1.2091, 'grad_norm': 0.0012716310281616762, 'learning_rate': 0.044184106189231624, 'epoch': 0.7} + 70%|██████▉ | 363/520 [22:45<09:39, 3.69s/it] 70%|███████ | 364/520 [22:49<09:34, 3.68s/it] {'loss': 1.2636, 'grad_norm': 0.001313510902772349, 'learning_rate': 0.043667994193637795, 'epoch': 0.7} + 70%|███████ | 364/520 [22:49<09:34, 3.68s/it] 70%|███████ | 365/520 [22:53<09:33, 3.70s/it] {'loss': 1.2668, 'grad_norm': 0.0013069828368192555, 'learning_rate': 0.043154070929510784, 'epoch': 0.7} + 70%|███████ | 365/520 [22:53<09:33, 3.70s/it] 70%|███████ | 366/520 [22:56<09:27, 3.69s/it] {'loss': 1.2276, 'grad_norm': 0.00124880202453895, 'learning_rate': 0.04264235636489542, 'epoch': 0.7} + 70%|███████ | 366/520 [22:56<09:27, 3.69s/it] 71%|███████ | 367/520 [23:00<09:22, 3.68s/it] {'loss': 1.2252, 'grad_norm': 0.001286038533665109, 'learning_rate': 0.04213287038201943, 'epoch': 0.71} + 71%|███████ | 367/520 [23:00<09:22, 3.68s/it] 71%|███████ | 368/520 [23:04<09:16, 3.66s/it] {'loss': 1.0759, 'grad_norm': 0.001281072896944521, 'learning_rate': 0.04162563277652104, 'epoch': 0.71} + 71%|███████ | 368/520 [23:04<09:16, 3.66s/it] 71%|███████ | 369/520 [23:07<09:12, 3.66s/it] {'loss': 1.2147, 'grad_norm': 0.0011748428059380946, 'learning_rate': 0.04112066325667954, 'epoch': 0.71} + 71%|███████ | 369/520 [23:07<09:12, 3.66s/it] 71%|███████ | 370/520 [23:11<09:08, 3.66s/it] {'loss': 1.1424, 'grad_norm': 0.0012289525144807386, 'learning_rate': 0.04061798144264986, 'epoch': 0.71} + 71%|███████ | 370/520 [23:11<09:08, 3.66s/it] 71%|███████▏ | 371/520 [23:15<09:11, 3.70s/it] {'loss': 1.1412, 'grad_norm': 0.0013207701538737852, 'learning_rate': 0.04011760686569998, 'epoch': 0.71} + 71%|███████▏ | 371/520 [23:15<09:11, 3.70s/it] 72%|███████▏ | 372/520 [23:19<09:13, 3.74s/it] {'loss': 1.2982, 'grad_norm': 0.00122003564173973, 'learning_rate': 0.03961955896745224, 'epoch': 0.72} + 72%|███████▏ | 372/520 [23:19<09:13, 3.74s/it] 72%|███████▏ | 373/520 [23:22<09:16, 3.78s/it] {'loss': 1.1814, 'grad_norm': 0.0013974074444752585, 'learning_rate': 0.03912385709912794, 'epoch': 0.72} + 72%|███████▏ | 373/520 [23:22<09:16, 3.78s/it] 72%|███████▏ | 374/520 [23:26<09:15, 3.80s/it] {'loss': 1.2268, 'grad_norm': 0.0012919251523566932, 'learning_rate': 0.038630520520795276, 'epoch': 0.72} + 72%|███████▏ | 374/520 [23:26<09:15, 3.80s/it] 72%|███████▏ | 375/520 [23:30<09:12, 3.81s/it] {'loss': 1.1367, 'grad_norm': 0.0012402952437247582, 'learning_rate': 0.03813956840062119, 'epoch': 0.72} + 72%|███████▏ | 375/520 [23:30<09:12, 3.81s/it] 72%|███████▏ | 376/520 [23:34<09:08, 3.81s/it] {'loss': 1.2484, 'grad_norm': 0.00119657484352812, 'learning_rate': 0.037651019814126656, 'epoch': 0.72} + 72%|███████▏ | 376/520 [23:34<09:08, 3.81s/it] 72%|███████▎ | 377/520 [23:38<09:04, 3.81s/it] {'loss': 1.1912, 'grad_norm': 0.001373688567294995, 'learning_rate': 0.037164893743445275, 'epoch': 0.72} + 72%|███████▎ | 377/520 [23:38<09:04, 3.81s/it] 73%|███████▎ | 378/520 [23:41<09:00, 3.81s/it] {'loss': 1.2395, 'grad_norm': 0.0012185579047184556, 'learning_rate': 0.03668120907658603, 'epoch': 0.73} + 73%|███████▎ | 378/520 [23:41<09:00, 3.81s/it] 73%|███████▎ | 379/520 [23:45<08:55, 3.80s/it] {'loss': 1.227, 'grad_norm': 0.0012171133534395575, 'learning_rate': 0.036199984606699154, 'epoch': 0.73} + 73%|███████▎ | 379/520 [23:45<08:55, 3.80s/it] 73%|███████▎ | 380/520 [23:49<08:47, 3.77s/it] {'loss': 1.2699, 'grad_norm': 0.0013082102746337925, 'learning_rate': 0.035721239031346066, 'epoch': 0.73} + 73%|███████▎ | 380/520 [23:49<08:47, 3.77s/it] 73%|███████▎ | 381/520 [23:53<08:38, 3.73s/it] {'loss': 1.2225, 'grad_norm': 0.0012297809592868281, 'learning_rate': 0.03524499095177297, 'epoch': 0.73} + 73%|███████▎ | 381/520 [23:53<08:38, 3.73s/it] 73%|███████▎ | 382/520 [23:56<08:33, 3.72s/it] {'loss': 1.2283, 'grad_norm': 0.0011855130651185497, 'learning_rate': 0.03477125887218792, 'epoch': 0.73} + 73%|███████▎ | 382/520 [23:56<08:33, 3.72s/it] 74%|███████▎ | 383/520 [24:00<08:26, 3.69s/it] {'loss': 1.0644, 'grad_norm': 0.001362456140922569, 'learning_rate': 0.03430006119904196, 'epoch': 0.74} + 74%|███████▎ | 383/520 [24:00<08:26, 3.69s/it] 74%|███████▍ | 384/520 [24:04<08:20, 3.68s/it] {'loss': 1.2917, 'grad_norm': 0.0012026568194702912, 'learning_rate': 0.033831416240314084, 'epoch': 0.74} + 74%|███████▍ | 384/520 [24:04<08:20, 3.68s/it] 74%|███████▍ | 385/520 [24:07<08:15, 3.67s/it] {'loss': 1.2014, 'grad_norm': 0.0011849924361634393, 'learning_rate': 0.03336534220479961, 'epoch': 0.74} + 74%|███████▍ | 385/520 [24:07<08:15, 3.67s/it] 74%|███████▍ | 386/520 [24:11<08:10, 3.66s/it] {'loss': 1.157, 'grad_norm': 0.001090500185791764, 'learning_rate': 0.032901857201403005, 'epoch': 0.74} + 74%|███████▍ | 386/520 [24:11<08:10, 3.66s/it] 74%|███████▍ | 387/520 [24:15<08:06, 3.66s/it] {'loss': 1.2951, 'grad_norm': 0.0012537826597817212, 'learning_rate': 0.032440979238433976, 'epoch': 0.74} + 74%|███████▍ | 387/520 [24:15<08:06, 3.66s/it] 75%|███████▍ | 388/520 [24:18<08:03, 3.66s/it] {'loss': 1.1081, 'grad_norm': 0.001194883838956544, 'learning_rate': 0.03198272622290804, 'epoch': 0.75} + 75%|███████▍ | 388/520 [24:18<08:03, 3.66s/it] 75%|███████▍ | 389/520 [24:22<07:58, 3.66s/it] {'loss': 1.1544, 'grad_norm': 0.0013963264748534034, 'learning_rate': 0.03152711595985065, 'epoch': 0.75} + 75%|███████▍ | 389/520 [24:22<07:58, 3.66s/it] 75%|███████▌ | 390/520 [24:25<07:55, 3.65s/it] {'loss': 1.2192, 'grad_norm': 0.0012458368125980036, 'learning_rate': 0.031074166151605298, 'epoch': 0.75} + 75%|███████▌ | 390/520 [24:25<07:55, 3.65s/it] 75%|███████▌ | 391/520 [24:29<07:51, 3.66s/it] {'loss': 1.3004, 'grad_norm': 0.0013082924276154758, 'learning_rate': 0.030623894397145836, 'epoch': 0.75} + 75%|███████▌ | 391/520 [24:29<07:51, 3.66s/it] 75%|███████▌ | 392/520 [24:33<07:48, 3.66s/it] {'loss': 1.1136, 'grad_norm': 0.001204832656420224, 'learning_rate': 0.03017631819139273, 'epoch': 0.75} + 75%|███████▌ | 392/520 [24:33<07:48, 3.66s/it] 76%|███████▌ | 393/520 [24:37<07:46, 3.68s/it] {'loss': 1.1271, 'grad_norm': 0.0011029620798768613, 'learning_rate': 0.029731454924533086, 'epoch': 0.76} + 76%|███████▌ | 393/520 [24:37<07:46, 3.68s/it] 76%|███████▌ | 394/520 [24:40<07:42, 3.67s/it] {'loss': 1.1771, 'grad_norm': 0.0012928609526183968, 'learning_rate': 0.029289321881345254, 'epoch': 0.76} + 76%|███████▌ | 394/520 [24:40<07:42, 3.67s/it] 76%|███████▌ | 395/520 [24:44<07:37, 3.66s/it] {'loss': 1.1442, 'grad_norm': 0.0013164085287117773, 'learning_rate': 0.028849936240527008, 'epoch': 0.76} + 76%|███████▌ | 395/520 [24:44<07:37, 3.66s/it] 76%|███████▌ | 396/520 [24:47<07:33, 3.66s/it] {'loss': 1.2329, 'grad_norm': 0.0013882653042026195, 'learning_rate': 0.028413315074028157, 'epoch': 0.76} + 76%|███████▌ | 396/520 [24:47<07:33, 3.66s/it] 76%|███████▋ | 397/520 [24:51<07:28, 3.65s/it] {'loss': 1.2045, 'grad_norm': 0.0012056509097587037, 'learning_rate': 0.027979475346387363, 'epoch': 0.76} + 76%|███████▋ | 397/520 [24:51<07:28, 3.65s/it] 77%|███████▋ | 398/520 [24:55<07:26, 3.66s/it] {'loss': 1.2041, 'grad_norm': 0.0013223668302405275, 'learning_rate': 0.027548433914072735, 'epoch': 0.77} + 77%|███████▋ | 398/520 [24:55<07:26, 3.66s/it] 77%|███████▋ | 399/520 [24:58<07:22, 3.66s/it] {'loss': 1.1734, 'grad_norm': 0.0012380460118554557, 'learning_rate': 0.027120207524827168, 'epoch': 0.77} + 77%|███████▋ | 399/520 [24:58<07:22, 3.66s/it] 77%|███████▋ | 400/520 [25:02<07:19, 3.66s/it] {'loss': 1.2024, 'grad_norm': 0.0011928304714615505, 'learning_rate': 0.02669481281701739, 'epoch': 0.77} + 77%|███████▋ | 400/520 [25:02<07:19, 3.66s/it] 77%|███████▋ | 401/520 [25:06<07:15, 3.66s/it] {'loss': 1.0261, 'grad_norm': 0.001340499080490449, 'learning_rate': 0.026272266318987603, 'epoch': 0.77} + 77%|███████▋ | 401/520 [25:06<07:15, 3.66s/it] 77%|███████▋ | 402/520 [25:09<07:11, 3.66s/it] {'loss': 1.1585, 'grad_norm': 0.0012565182873857262, 'learning_rate': 0.02585258444841733, 'epoch': 0.77} + 77%|███████▋ | 402/520 [25:09<07:11, 3.66s/it] 78%|███████▊ | 403/520 [25:13<07:07, 3.65s/it] {'loss': 1.1804, 'grad_norm': 0.0013586960675347677, 'learning_rate': 0.025435783511683442, 'epoch': 0.78} + 78%|███████▊ | 403/520 [25:13<07:07, 3.65s/it] 78%|███████▊ | 404/520 [25:17<07:03, 3.65s/it] {'loss': 1.0892, 'grad_norm': 0.0015122232572725716, 'learning_rate': 0.02502187970322657, 'epoch': 0.78} + 78%|███████▊ | 404/520 [25:17<07:03, 3.65s/it] 78%|███████▊ | 405/520 [25:20<07:00, 3.66s/it] {'loss': 1.181, 'grad_norm': 0.0012465831356325567, 'learning_rate': 0.02461088910492202, 'epoch': 0.78} + 78%|███████▊ | 405/520 [25:20<07:00, 3.66s/it] 78%|███████▊ | 406/520 [25:24<06:57, 3.66s/it] {'loss': 1.1054, 'grad_norm': 0.0014613125295822025, 'learning_rate': 0.02420282768545469, 'epoch': 0.78} + 78%|███████▊ | 406/520 [25:24<06:57, 3.66s/it] 78%|███████▊ | 407/520 [25:28<06:53, 3.66s/it] {'loss': 1.2655, 'grad_norm': 0.001318568491417933, 'learning_rate': 0.02379771129969892, 'epoch': 0.78} + 78%|███████▊ | 407/520 [25:28<06:53, 3.66s/it] 78%|███████▊ | 408/520 [25:31<06:49, 3.65s/it] {'loss': 1.1695, 'grad_norm': 0.0013746307240631174, 'learning_rate': 0.023395555688102213, 'epoch': 0.78} + 78%|███████▊ | 408/520 [25:31<06:49, 3.65s/it] 79%|███████▊ | 409/520 [25:35<06:45, 3.65s/it] {'loss': 1.293, 'grad_norm': 0.001373366780324279, 'learning_rate': 0.02299637647607372, 'epoch': 0.79} + 79%|███████▊ | 409/520 [25:35<06:45, 3.65s/it] 79%|███████▉ | 410/520 [25:39<06:41, 3.65s/it] {'loss': 1.0241, 'grad_norm': 0.0012594279763946835, 'learning_rate': 0.022600189173377264, 'epoch': 0.79} + 79%|███████▉ | 410/520 [25:39<06:41, 3.65s/it] 79%|███████▉ | 411/520 [25:42<06:37, 3.65s/it] {'loss': 1.2752, 'grad_norm': 0.0013706430789508547, 'learning_rate': 0.022207009173528525, 'epoch': 0.79} + 79%|███████▉ | 411/520 [25:42<06:37, 3.65s/it] 79%|███████▉ | 412/520 [25:46<06:33, 3.65s/it] {'loss': 1.1839, 'grad_norm': 0.0013198707996671874, 'learning_rate': 0.02181685175319702, 'epoch': 0.79} + 79%|███████▉ | 412/520 [25:46<06:33, 3.65s/it] 79%|███████▉ | 413/520 [25:50<06:32, 3.67s/it] {'loss': 1.1983, 'grad_norm': 0.0012291494792168916, 'learning_rate': 0.021429732071612653, 'epoch': 0.79} + 79%|███████▉ | 413/520 [25:50<06:32, 3.67s/it] 80%|███████▉ | 414/520 [25:53<06:29, 3.67s/it] {'loss': 1.0014, 'grad_norm': 0.0010436639697315362, 'learning_rate': 0.02104566516997647, 'epoch': 0.8} + 80%|███████▉ | 414/520 [25:53<06:29, 3.67s/it] 80%|███████▉ | 415/520 [25:57<06:26, 3.68s/it] {'loss': 1.1592, 'grad_norm': 0.0011660432988096702, 'learning_rate': 0.020664665970876496, 'epoch': 0.8} + 80%|███████▉ | 415/520 [25:57<06:26, 3.68s/it] 80%|████████ | 416/520 [26:01<06:21, 3.67s/it] {'loss': 1.0833, 'grad_norm': 0.0013833504551383773, 'learning_rate': 0.020286749277707784, 'epoch': 0.8} + 80%|████████ | 416/520 [26:01<06:21, 3.67s/it] 80%|████████ | 417/520 [26:04<06:18, 3.68s/it] {'loss': 1.2405, 'grad_norm': 0.0013049563498367307, 'learning_rate': 0.019911929774097215, 'epoch': 0.8} + 80%|████████ | 417/520 [26:04<06:18, 3.68s/it] 80%|████████ | 418/520 [26:08<06:14, 3.67s/it] {'loss': 1.2237, 'grad_norm': 0.0011764115498162184, 'learning_rate': 0.019540222023333165, 'epoch': 0.8} + 80%|████████ | 418/520 [26:08<06:14, 3.67s/it] 81%|████████ | 419/520 [26:12<06:10, 3.66s/it] {'loss': 1.2103, 'grad_norm': 0.0013942699472926506, 'learning_rate': 0.01917164046779948, 'epoch': 0.81} + 81%|████████ | 419/520 [26:12<06:10, 3.66s/it] 81%|████████ | 420/520 [26:15<06:08, 3.68s/it] {'loss': 1.108, 'grad_norm': 0.001327869404677157, 'learning_rate': 0.018806199428414352, 'epoch': 0.81} + 81%|████████ | 420/520 [26:15<06:08, 3.68s/it] 81%|████████ | 421/520 [26:19<06:04, 3.68s/it] {'loss': 1.0435, 'grad_norm': 0.0013646708173296106, 'learning_rate': 0.018443913104073985, 'epoch': 0.81} + 81%|████████ | 421/520 [26:19<06:04, 3.68s/it] 81%|████████ | 422/520 [26:23<06:00, 3.68s/it] {'loss': 1.1618, 'grad_norm': 0.0013208803638026536, 'learning_rate': 0.01808479557110081, 'epoch': 0.81} + 81%|████████ | 422/520 [26:23<06:00, 3.68s/it] 81%|████████▏ | 423/520 [26:26<05:56, 3.67s/it] {'loss': 1.144, 'grad_norm': 0.0014107303192713856, 'learning_rate': 0.017728860782696667, 'epoch': 0.81} + 81%|████████▏ | 423/520 [26:26<05:56, 3.67s/it] 82%|████████▏ | 424/520 [26:30<05:52, 3.67s/it] {'loss': 1.2817, 'grad_norm': 0.001255094847577685, 'learning_rate': 0.017376122568400532, 'epoch': 0.82} + 82%|████████▏ | 424/520 [26:30<05:52, 3.67s/it] 82%|████████▏ | 425/520 [26:34<05:48, 3.67s/it] {'loss': 1.1562, 'grad_norm': 0.0012678838809253613, 'learning_rate': 0.017026594633551252, 'epoch': 0.82} + 82%|████████▏ | 425/520 [26:34<05:48, 3.67s/it] 82%|████████▏ | 426/520 [26:38<05:48, 3.71s/it] {'loss': 1.1824, 'grad_norm': 0.0016483453421246676, 'learning_rate': 0.01668029055875512, 'epoch': 0.82} + 82%|████████▏ | 426/520 [26:38<05:48, 3.71s/it] 82%|████████▏ | 427/520 [26:41<05:42, 3.68s/it] {'loss': 1.0914, 'grad_norm': 0.0012189172539971598, 'learning_rate': 0.016337223799358026, 'epoch': 0.82} + 82%|████████▏ | 427/520 [26:41<05:42, 3.68s/it] 82%|████████▏ | 428/520 [26:45<05:39, 3.69s/it] {'loss': 1.0716, 'grad_norm': 0.0013042828808868716, 'learning_rate': 0.01599740768492286, 'epoch': 0.82} + 82%|████████▏ | 428/520 [26:45<05:39, 3.69s/it] 82%|████████▎ | 429/520 [26:49<05:42, 3.76s/it] {'loss': 1.1667, 'grad_norm': 0.0012870137612680374, 'learning_rate': 0.015660855418711452, 'epoch': 0.82} + 82%|████████▎ | 429/520 [26:49<05:42, 3.76s/it]Token indices sequence length is longer than the specified maximum sequence length for this model (2076 > 2048). Running this sequence through the model will result in indexing errors + 83%|████████▎ | 430/520 [26:53<05:40, 3.79s/it] {'loss': 1.1687, 'grad_norm': 0.0011897521885977432, 'learning_rate': 0.015327580077171589, 'epoch': 0.83} + 83%|████████▎ | 430/520 [26:53<05:40, 3.79s/it] 83%|████████▎ | 431/520 [26:56<05:38, 3.81s/it] {'loss': 1.1695, 'grad_norm': 0.001278666109030383, 'learning_rate': 0.014997594609429088, 'epoch': 0.83} + 83%|████████▎ | 431/520 [26:57<05:38, 3.81s/it] 83%|████████▎ | 432/520 [27:00<05:36, 3.82s/it] {'loss': 1.0809, 'grad_norm': 0.0012892154423871686, 'learning_rate': 0.01467091183678444, 'epoch': 0.83} + 83%|████████▎ | 432/520 [27:00<05:36, 3.82s/it] 83%|████████▎ | 433/520 [27:04<05:33, 3.83s/it] {'loss': 1.2118, 'grad_norm': 0.0012254935250953338, 'learning_rate': 0.014347544452214867, 'epoch': 0.83} + 83%|████████▎ | 433/520 [27:04<05:33, 3.83s/it] 83%|████████▎ | 434/520 [27:08<05:30, 3.85s/it] {'loss': 0.9549, 'grad_norm': 0.0013176426944668184, 'learning_rate': 0.014027505019880971, 'epoch': 0.83} + 83%|████████▎ | 434/520 [27:08<05:30, 3.85s/it] 84%|████████▎ | 435/520 [27:12<05:26, 3.84s/it] {'loss': 1.2505, 'grad_norm': 0.0014157971861356198, 'learning_rate': 0.013710805974638696, 'epoch': 0.84} + 84%|████████▎ | 435/520 [27:12<05:26, 3.84s/it] 84%|████████▍ | 436/520 [27:16<05:22, 3.84s/it] {'loss': 1.0486, 'grad_norm': 0.0013169990592864162, 'learning_rate': 0.01339745962155613, 'epoch': 0.84} + 84%|████████▍ | 436/520 [27:16<05:22, 3.84s/it] 84%|████████▍ | 437/520 [27:20<05:19, 3.85s/it] {'loss': 1.27, 'grad_norm': 0.0013094096707016306, 'learning_rate': 0.01308747813543536, 'epoch': 0.84} + 84%|████████▍ | 437/520 [27:20<05:19, 3.85s/it] 84%|████████▍ | 438/520 [27:23<05:15, 3.85s/it] {'loss': 1.0848, 'grad_norm': 0.0012777006614721018, 'learning_rate': 0.012780873560339467, 'epoch': 0.84} + 84%|████████▍ | 438/520 [27:23<05:15, 3.85s/it] 84%|████████▍ | 439/520 [27:27<05:12, 3.86s/it] {'loss': 1.1491, 'grad_norm': 0.0010865197910800406, 'learning_rate': 0.012477657809124632, 'epoch': 0.84} + 84%|████████▍ | 439/520 [27:27<05:12, 3.86s/it] 85%|████████▍ | 440/520 [27:31<05:09, 3.87s/it] {'loss': 1.1267, 'grad_norm': 0.0015251479464252532, 'learning_rate': 0.012177842662977134, 'epoch': 0.85} + 85%|████████▍ | 440/520 [27:31<05:09, 3.87s/it] 85%|████████▍ | 441/520 [27:35<05:06, 3.88s/it] {'loss': 1.1645, 'grad_norm': 0.0012419465165783166, 'learning_rate': 0.01188143977095576, 'epoch': 0.85} + 85%|████████▍ | 441/520 [27:35<05:06, 3.88s/it] 85%|████████▌ | 442/520 [27:39<05:03, 3.89s/it] {'loss': 1.1864, 'grad_norm': 0.0013753246703949189, 'learning_rate': 0.011588460649539035, 'epoch': 0.85} + 85%|████████▌ | 442/520 [27:39<05:03, 3.89s/it] 85%|████████▌ | 443/520 [27:43<04:58, 3.88s/it] {'loss': 1.2021, 'grad_norm': 0.0012489313238293017, 'learning_rate': 0.011298916682177829, 'epoch': 0.85} + 85%|████████▌ | 443/520 [27:43<04:58, 3.88s/it] 85%|████████▌ | 444/520 [27:47<04:55, 3.89s/it] {'loss': 1.17, 'grad_norm': 0.001155812500844116, 'learning_rate': 0.011012819118853146, 'epoch': 0.85} + 85%|████████▌ | 444/520 [27:47<04:55, 3.89s/it] 86%|████████▌ | 445/520 [27:51<04:48, 3.85s/it] {'loss': 1.0943, 'grad_norm': 0.0012408482367036345, 'learning_rate': 0.01073017907563887, 'epoch': 0.86} + 86%|████████▌ | 445/520 [27:51<04:48, 3.85s/it] 86%|████████▌ | 446/520 [27:54<04:41, 3.81s/it] {'loss': 1.2482, 'grad_norm': 0.0011980154660036094, 'learning_rate': 0.010451007534269908, 'epoch': 0.86} + 86%|████████▌ | 446/520 [27:54<04:41, 3.81s/it] 86%|████████▌ | 447/520 [27:58<04:35, 3.78s/it] {'loss': 1.1741, 'grad_norm': 0.0012495327900141702, 'learning_rate': 0.010175315341715598, 'epoch': 0.86} + 86%|████████▌ | 447/520 [27:58<04:35, 3.78s/it] 86%|████████▌ | 448/520 [28:02<04:29, 3.75s/it] {'loss': 1.1635, 'grad_norm': 0.0013203927499392597, 'learning_rate': 0.009903113209758098, 'epoch': 0.86} + 86%|████████▌ | 448/520 [28:02<04:29, 3.75s/it] 86%|████████▋ | 449/520 [28:05<04:24, 3.73s/it] {'loss': 1.1964, 'grad_norm': 0.0013055158472783997, 'learning_rate': 0.009634411714576352, 'epoch': 0.86} + 86%|████████▋ | 449/520 [28:05<04:24, 3.73s/it] 87%|████████▋ | 450/520 [28:09<04:20, 3.73s/it] {'loss': 1.1948, 'grad_norm': 0.0012986569625479023, 'learning_rate': 0.009369221296335007, 'epoch': 0.87} + 87%|████████▋ | 450/520 [28:09<04:20, 3.73s/it] 87%|████████▋ | 451/520 [28:13<04:16, 3.72s/it] {'loss': 1.1912, 'grad_norm': 0.0012823388833823803, 'learning_rate': 0.009107552258778906, 'epoch': 0.87} + 87%|████████▋ | 451/520 [28:13<04:16, 3.72s/it] 87%|████████▋ | 452/520 [28:16<04:11, 3.70s/it] {'loss': 1.2416, 'grad_norm': 0.001193915703622801, 'learning_rate': 0.008849414768832687, 'epoch': 0.87} + 87%|████████▋ | 452/520 [28:16<04:11, 3.70s/it] 87%|████████▋ | 453/520 [28:20<04:08, 3.71s/it] {'loss': 1.2188, 'grad_norm': 0.0012423593288343378, 'learning_rate': 0.008594818856205699, 'epoch': 0.87} + 87%|████████▋ | 453/520 [28:20<04:08, 3.71s/it] 87%|████████▋ | 454/520 [28:24<04:03, 3.68s/it] {'loss': 1.1004, 'grad_norm': 0.0012881673620363833, 'learning_rate': 0.00834377441300238, 'epoch': 0.87} + 87%|████████▋ | 454/520 [28:24<04:03, 3.68s/it] 88%|████████▊ | 455/520 [28:27<03:58, 3.67s/it] {'loss': 1.2396, 'grad_norm': 0.0012692928896639137, 'learning_rate': 0.008096291193337934, 'epoch': 0.88} + 88%|████████▊ | 455/520 [28:27<03:58, 3.67s/it] 88%|████████▊ | 456/520 [28:31<03:54, 3.66s/it] {'loss': 1.168, 'grad_norm': 0.0012921073701460868, 'learning_rate': 0.007852378812959226, 'epoch': 0.88} + 88%|████████▊ | 456/520 [28:31<03:54, 3.66s/it] 88%|████████▊ | 457/520 [28:35<03:51, 3.67s/it] {'loss': 1.1374, 'grad_norm': 0.001132281901634167, 'learning_rate': 0.007612046748871327, 'epoch': 0.88} + 88%|████████▊ | 457/520 [28:35<03:51, 3.67s/it] 88%|████████▊ | 458/520 [28:38<03:47, 3.66s/it] {'loss': 1.299, 'grad_norm': 0.0013712003443238437, 'learning_rate': 0.007375304338969136, 'epoch': 0.88} + 88%|████████▊ | 458/520 [28:38<03:47, 3.66s/it] 88%|████████▊ | 459/520 [28:42<03:42, 3.65s/it] {'loss': 1.227, 'grad_norm': 0.0012683439714100387, 'learning_rate': 0.007142160781674645, 'epoch': 0.88} + 88%|████████▊ | 459/520 [28:42<03:42, 3.65s/it] 88%|████████▊ | 460/520 [28:46<03:38, 3.64s/it] {'loss': 1.109, 'grad_norm': 0.0012452533134066494, 'learning_rate': 0.006912625135579587, 'epoch': 0.88} + 88%|████████▊ | 460/520 [28:46<03:38, 3.64s/it] 89%|████████▊ | 461/520 [28:49<03:34, 3.64s/it] {'loss': 1.223, 'grad_norm': 0.0010390129169985454, 'learning_rate': 0.0066867063190933496, 'epoch': 0.89} + 89%|████████▊ | 461/520 [28:49<03:34, 3.64s/it] 89%|████████▉ | 462/520 [28:53<03:30, 3.64s/it] {'loss': 1.2877, 'grad_norm': 0.001254026215234257, 'learning_rate': 0.006464413110096601, 'epoch': 0.89} + 89%|████████▉ | 462/520 [28:53<03:30, 3.64s/it] 89%|████████▉ | 463/520 [28:57<03:27, 3.65s/it] {'loss': 1.0699, 'grad_norm': 0.0013124445971887857, 'learning_rate': 0.006245754145600091, 'epoch': 0.89} + 89%|████████▉ | 463/520 [28:57<03:27, 3.65s/it] 89%|████████▉ | 464/520 [29:00<03:23, 3.64s/it] {'loss': 1.2074, 'grad_norm': 0.0013315528857001833, 'learning_rate': 0.006030737921409169, 'epoch': 0.89} + 89%|████████▉ | 464/520 [29:00<03:23, 3.64s/it] 89%|████████▉ | 465/520 [29:04<03:19, 3.63s/it] {'loss': 1.3175, 'grad_norm': 0.0013384443130654445, 'learning_rate': 0.005819372791793654, 'epoch': 0.89} + 89%|████████▉ | 465/520 [29:04<03:19, 3.63s/it] 90%|████████▉ | 466/520 [29:07<03:15, 3.63s/it] {'loss': 1.198, 'grad_norm': 0.0011478457810530636, 'learning_rate': 0.005611666969163243, 'epoch': 0.9} + 90%|████████▉ | 466/520 [29:07<03:15, 3.63s/it] 90%|████████▉ | 467/520 [29:11<03:11, 3.62s/it] {'loss': 1.1786, 'grad_norm': 0.001182997889838019, 'learning_rate': 0.005407628523748398, 'epoch': 0.9} + 90%|████████▉ | 467/520 [29:11<03:11, 3.62s/it] 90%|█████████ | 468/520 [29:15<03:08, 3.62s/it] {'loss': 1.1764, 'grad_norm': 0.0014785812822121402, 'learning_rate': 0.00520726538328683, 'epoch': 0.9} + 90%|█████████ | 468/520 [29:15<03:08, 3.62s/it] 90%|█████████ | 469/520 [29:18<03:05, 3.63s/it] {'loss': 1.2343, 'grad_norm': 0.0013850938215348217, 'learning_rate': 0.005010585332715401, 'epoch': 0.9} + 90%|█████████ | 469/520 [29:18<03:05, 3.63s/it] 90%|█████████ | 470/520 [29:22<03:01, 3.64s/it] {'loss': 1.1138, 'grad_norm': 0.0011847642195815964, 'learning_rate': 0.004817596013867765, 'epoch': 0.9} + 90%|█████████ | 470/520 [29:22<03:01, 3.64s/it] 91%|█████████ | 471/520 [29:26<02:58, 3.64s/it] {'loss': 1.138, 'grad_norm': 0.0013528916770866276, 'learning_rate': 0.004628304925177318, 'epoch': 0.91} + 91%|█████████ | 471/520 [29:26<02:58, 3.64s/it] 91%|█████████ | 472/520 [29:29<02:55, 3.65s/it] {'loss': 1.1057, 'grad_norm': 0.0012288341451465543, 'learning_rate': 0.004442719421385921, 'epoch': 0.91} + 91%|█████████ | 472/520 [29:29<02:55, 3.65s/it] 91%|█████████ | 473/520 [29:33<02:52, 3.66s/it] {'loss': 1.1699, 'grad_norm': 0.0013607710055584079, 'learning_rate': 0.004260846713258193, 'epoch': 0.91} + 91%|█████████ | 473/520 [29:33<02:52, 3.66s/it] 91%|█████████ | 474/520 [29:37<02:50, 3.71s/it] {'loss': 1.2136, 'grad_norm': 0.0012436272226910574, 'learning_rate': 0.004082693867301224, 'epoch': 0.91} + 91%|█████████ | 474/520 [29:37<02:50, 3.71s/it] 91%|█████████▏| 475/520 [29:41<02:47, 3.73s/it] {'loss': 1.1335, 'grad_norm': 0.0012094385743864858, 'learning_rate': 0.003908267805490051, 'epoch': 0.91} + 91%|█████████▏| 475/520 [29:41<02:47, 3.73s/it] 92%|█████████▏| 476/520 [29:44<02:45, 3.76s/it] {'loss': 1.1585, 'grad_norm': 0.001339062960632655, 'learning_rate': 0.003737575304998797, 'epoch': 0.92} + 92%|█████████▏| 476/520 [29:44<02:45, 3.76s/it] 92%|█████████▏| 477/520 [29:48<02:40, 3.74s/it] {'loss': 1.1493, 'grad_norm': 0.0013415354551626517, 'learning_rate': 0.003570622997937234, 'epoch': 0.92} + 92%|█████████▏| 477/520 [29:48<02:40, 3.74s/it] 92%|█████████▏| 478/520 [29:52<02:38, 3.78s/it] {'loss': 1.0984, 'grad_norm': 0.0012504900886687713, 'learning_rate': 0.00340741737109318, 'epoch': 0.92} + 92%|█████████▏| 478/520 [29:52<02:38, 3.78s/it] 92%|█████████▏| 479/520 [29:56<02:36, 3.82s/it] {'loss': 1.1808, 'grad_norm': 0.0013356130293240723, 'learning_rate': 0.003247964765680389, 'epoch': 0.92} + 92%|█████████▏| 479/520 [29:56<02:36, 3.82s/it] 92%|█████████▏| 480/520 [30:00<02:33, 3.84s/it] {'loss': 1.1999, 'grad_norm': 0.001229874642760878, 'learning_rate': 0.0030922713770922153, 'epoch': 0.92} + 92%|█████████▏| 480/520 [30:00<02:33, 3.84s/it] 92%|█████████▎| 481/520 [30:04<02:30, 3.86s/it] {'loss': 1.1945, 'grad_norm': 0.0011457076421827741, 'learning_rate': 0.0029403432546609046, 'epoch': 0.93} + 92%|█████████▎| 481/520 [30:04<02:30, 3.86s/it] 93%|█████████▎| 482/520 [30:08<02:27, 3.87s/it] {'loss': 1.2063, 'grad_norm': 0.0011843991921187981, 'learning_rate': 0.0027921863014225504, 'epoch': 0.93} + 93%|█████████▎| 482/520 [30:08<02:27, 3.87s/it] 93%|█████████▎| 483/520 [30:11<02:23, 3.88s/it] {'loss': 1.1715, 'grad_norm': 0.0012624250245580493, 'learning_rate': 0.002647806273887665, 'epoch': 0.93} + 93%|█████████▎| 483/520 [30:11<02:23, 3.88s/it] 93%|█████████▎| 484/520 [30:15<02:19, 3.88s/it] {'loss': 1.1734, 'grad_norm': 0.0013256365912178096, 'learning_rate': 0.0025072087818176383, 'epoch': 0.93} + 93%|█████████▎| 484/520 [30:15<02:19, 3.88s/it] 93%|█████████▎| 485/520 [30:19<02:16, 3.89s/it] {'loss': 1.1283, 'grad_norm': 0.0012247982298017397, 'learning_rate': 0.002370399288006664, 'epoch': 0.93} + 93%|█████████▎| 485/520 [30:19<02:16, 3.89s/it] 93%|█████████▎| 486/520 [30:23<02:12, 3.89s/it] {'loss': 1.2507, 'grad_norm': 0.0013069484629941739, 'learning_rate': 0.0022373831080695463, 'epoch': 0.93} + 93%|█████████▎| 486/520 [30:23<02:12, 3.89s/it] 94%|█████████▎| 487/520 [30:27<02:08, 3.90s/it] {'loss': 1.0995, 'grad_norm': 0.0012207899888665881, 'learning_rate': 0.0021081654102351635, 'epoch': 0.94} + 94%|█████████▎| 487/520 [30:27<02:08, 3.90s/it] 94%|█████████▍| 488/520 [30:31<02:04, 3.90s/it] {'loss': 1.0468, 'grad_norm': 0.0012745291514557984, 'learning_rate': 0.0019827512151456175, 'epoch': 0.94} + 94%|█████████▍| 488/520 [30:31<02:04, 3.90s/it] 94%|█████████▍| 489/520 [30:35<02:00, 3.90s/it] {'loss': 1.2096, 'grad_norm': 0.0010975691900800587, 'learning_rate': 0.0018611453956612345, 'epoch': 0.94} + 94%|█████████▍| 489/520 [30:35<02:00, 3.90s/it] 94%|█████████▍| 490/520 [30:39<01:57, 3.90s/it] {'loss': 1.1712, 'grad_norm': 0.0012901593558227674, 'learning_rate': 0.0017433526766711727, 'epoch': 0.94} + 94%|█████████▍| 490/520 [30:39<01:57, 3.90s/it] 94%|█████████▍| 491/520 [30:43<01:52, 3.89s/it] {'loss': 1.1321, 'grad_norm': 0.0013020138119465858, 'learning_rate': 0.0016293776349098677, 'epoch': 0.94} + 94%|█████████▍| 491/520 [30:43<01:52, 3.89s/it] 95%|█████████▍| 492/520 [30:47<01:49, 3.90s/it] {'loss': 1.2451, 'grad_norm': 0.0013174302188561982, 'learning_rate': 0.0015192246987791981, 'epoch': 0.95} + 95%|█████████▍| 492/520 [30:47<01:49, 3.90s/it] 95%|█████████▍| 493/520 [30:50<01:45, 3.90s/it] {'loss': 1.2211, 'grad_norm': 0.001315427489002492, 'learning_rate': 0.0014128981481764114, 'epoch': 0.95} + 95%|█████████▍| 493/520 [30:50<01:45, 3.90s/it] 95%|█████████▌| 494/520 [30:54<01:41, 3.90s/it] {'loss': 1.1872, 'grad_norm': 0.001171194471695787, 'learning_rate': 0.0013104021143278911, 'epoch': 0.95} + 95%|█████████▌| 494/520 [30:54<01:41, 3.90s/it] 95%|█████████▌| 495/520 [30:58<01:37, 3.90s/it] {'loss': 1.1484, 'grad_norm': 0.0012677508678158645, 'learning_rate': 0.0012117405796285285, 'epoch': 0.95} + 95%|█████████▌| 495/520 [30:58<01:37, 3.90s/it] 95%|█████████▌| 496/520 [31:02<01:33, 3.90s/it] {'loss': 1.0685, 'grad_norm': 0.00149278019856293, 'learning_rate': 0.0011169173774871477, 'epoch': 0.95} + 95%|█████████▌| 496/520 [31:02<01:33, 3.90s/it] 96%|█████████▌| 497/520 [31:06<01:29, 3.91s/it] {'loss': 1.1369, 'grad_norm': 0.0011254230160287128, 'learning_rate': 0.0010259361921774012, 'epoch': 0.96} + 96%|█████████▌| 497/520 [31:06<01:29, 3.91s/it] 96%|█████████▌| 498/520 [31:10<01:25, 3.91s/it] {'loss': 1.1426, 'grad_norm': 0.0012607569514867705, 'learning_rate': 0.000938800558694719, 'epoch': 0.96} + 96%|█████████▌| 498/520 [31:10<01:25, 3.91s/it] 96%|█████████▌| 499/520 [31:14<01:22, 3.91s/it] {'loss': 1.2792, 'grad_norm': 0.0012624113383509854, 'learning_rate': 0.0008555138626189618, 'epoch': 0.96} + 96%|█████████▌| 499/520 [31:14<01:22, 3.91s/it] 96%|█████████▌| 500/520 [31:18<01:18, 3.91s/it] {'loss': 1.2644, 'grad_norm': 0.0015181416771668324, 'learning_rate': 0.0007760793399827937, 'epoch': 0.96} + 96%|█████████▌| 500/520 [31:18<01:18, 3.91s/it] 96%|█████████▋| 501/520 [31:22<01:14, 3.90s/it] {'loss': 1.1889, 'grad_norm': 0.001357298726713291, 'learning_rate': 0.000700500077146038, 'epoch': 0.96} + 96%|█████████▋| 501/520 [31:22<01:14, 3.90s/it] 97%|█████████▋| 502/520 [31:26<01:09, 3.87s/it] {'loss': 1.1878, 'grad_norm': 0.0012367377237071243, 'learning_rate': 0.0006287790106757397, 'epoch': 0.97} + 97%|█████████▋| 502/520 [31:26<01:09, 3.87s/it] 97%|█████████▋| 503/520 [31:29<01:05, 3.85s/it] {'loss': 1.1702, 'grad_norm': 0.001269657262502918, 'learning_rate': 0.0005609189272320237, 'epoch': 0.97} + 97%|█████████▋| 503/520 [31:29<01:05, 3.85s/it] 97%|█████████▋| 504/520 [31:33<01:01, 3.83s/it] {'loss': 1.1791, 'grad_norm': 0.0014527317125354823, 'learning_rate': 0.000496922463459859, 'epoch': 0.97} + 97%|█████████▋| 504/520 [31:33<01:01, 3.83s/it] 97%|█████████▋| 505/520 [31:37<00:57, 3.82s/it] {'loss': 1.2162, 'grad_norm': 0.0012913493924389193, 'learning_rate': 0.0004367921058866187, 'epoch': 0.97} + 97%|█████████▋| 505/520 [31:37<00:57, 3.82s/it] 97%|█████████▋| 506/520 [31:41<00:53, 3.80s/it] {'loss': 1.141, 'grad_norm': 0.0013063257753935119, 'learning_rate': 0.0003805301908254455, 'epoch': 0.97} + 97%|█████████▋| 506/520 [31:41<00:53, 3.80s/it] 98%|█████████▊| 507/520 [31:44<00:49, 3.81s/it] {'loss': 1.3298, 'grad_norm': 0.0012503528586212385, 'learning_rate': 0.0003281389042844918, 'epoch': 0.97} + 98%|█████████▊| 507/520 [31:44<00:49, 3.81s/it] 98%|█████████▊| 508/520 [31:48<00:45, 3.81s/it] {'loss': 1.256, 'grad_norm': 0.0012936861406673919, 'learning_rate': 0.00027962028188198705, 'epoch': 0.98} + 98%|█████████▊| 508/520 [31:48<00:45, 3.81s/it] 98%|█████████▊| 509/520 [31:52<00:41, 3.79s/it] {'loss': 1.2265, 'grad_norm': 0.0012106772552326979, 'learning_rate': 0.00023497620876711256, 'epoch': 0.98} + 98%|█████████▊| 509/520 [31:52<00:41, 3.79s/it] 98%|█████████▊| 510/520 [31:56<00:37, 3.74s/it] {'loss': 1.1751, 'grad_norm': 0.0012700541011907175, 'learning_rate': 0.00019420841954681523, 'epoch': 0.98} + 98%|█████████▊| 510/520 [31:56<00:37, 3.74s/it] 98%|█████████▊| 511/520 [31:59<00:33, 3.71s/it] {'loss': 1.1474, 'grad_norm': 0.00121592905738457, 'learning_rate': 0.00015731849821833956, 'epoch': 0.98} + 98%|█████████▊| 511/520 [31:59<00:33, 3.71s/it] 98%|█████████▊| 512/520 [32:03<00:29, 3.67s/it] {'loss': 1.0361, 'grad_norm': 0.001271243666861461, 'learning_rate': 0.00012430787810776555, 'epoch': 0.98} + 98%|█████████▊| 512/520 [32:03<00:29, 3.67s/it] 99%|█████████▊| 513/520 [32:07<00:25, 3.68s/it] {'loss': 1.2398, 'grad_norm': 0.0014164654764151961, 'learning_rate': 9.517784181422018e-05, 'epoch': 0.99} + 99%|█████████▊| 513/520 [32:07<00:25, 3.68s/it] 99%|█████████▉| 514/520 [32:10<00:22, 3.67s/it] {'loss': 1.204, 'grad_norm': 0.0011650455127815175, 'learning_rate': 6.992952116013917e-05, 'epoch': 0.99} + 99%|█████████▉| 514/520 [32:10<00:22, 3.67s/it] 99%|█████████▉| 515/520 [32:14<00:18, 3.66s/it] {'loss': 1.2486, 'grad_norm': 0.001482459217272228, 'learning_rate': 4.856389714723575e-05, 'epoch': 0.99} + 99%|█████████▉| 515/520 [32:14<00:18, 3.66s/it] 99%|█████████▉| 516/520 [32:18<00:14, 3.65s/it] {'loss': 1.1544, 'grad_norm': 0.001232639662486634, 'learning_rate': 3.108179991837545e-05, 'epoch': 0.99} + 99%|█████████▉| 516/520 [32:18<00:14, 3.65s/it] 99%|█████████▉| 517/520 [32:21<00:10, 3.63s/it] {'loss': 1.2165, 'grad_norm': 0.0012258876023356208, 'learning_rate': 1.7483908725357544e-05, 'epoch': 0.99} + 99%|█████████▉| 517/520 [32:21<00:10, 3.63s/it] 100%|█████████▉| 518/520 [32:25<00:07, 3.62s/it] {'loss': 1.1685, 'grad_norm': 0.0014230167086031148, 'learning_rate': 7.770751902513862e-06, 'epoch': 1.0} + 100%|█████████▉| 518/520 [32:25<00:07, 3.62s/it] 100%|█████████▉| 519/520 [32:28<00:03, 3.61s/it] {'loss': 1.183, 'grad_norm': 0.00126296635675092, 'learning_rate': 1.9427068461808086e-06, 'epoch': 1.0} + 100%|█████████▉| 519/520 [32:28<00:03, 3.61s/it] 100%|██████████| 520/520 [32:33<00:00, 3.86s/it] {'loss': 1.1927, 'grad_norm': 0.001254886612343554, 'learning_rate': 0.0, 'epoch': 1.0} + 100%|██████████| 520/520 [32:33<00:00, 3.86s/it] {'train_runtime': 1953.2016, 'train_samples_per_second': 34.062, 'train_steps_per_second': 0.266, 'train_loss': 1.260267140315129, 'epoch': 1.0} + 100%|██████████| 520/520 [32:33<00:00, 3.86s/it] 100%|██████████| 520/520 [32:33<00:00, 3.76s/it] +[2025-10-12 15:15:18,685] [INFO] [launch.py:348:main] Process 405743 exits successfully. +[2025-10-12 15:15:18,685] [INFO] [launch.py:348:main] Process 405740 exits successfully. +[2025-10-12 15:15:18,686] [INFO] [launch.py:348:main] Process 405737 exits successfully. +[2025-10-12 15:15:18,686] [INFO] [launch.py:348:main] Process 405741 exits successfully. +[2025-10-12 15:15:18,686] [INFO] [launch.py:348:main] Process 405738 exits successfully. +[2025-10-12 15:15:19,688] [INFO] [launch.py:348:main] Process 405742 exits successfully. +[2025-10-12 15:15:19,688] [INFO] [launch.py:348:main] Process 405739 exits successfully. +[2025-10-12 15:15:23,693] [INFO] [launch.py:348:main] Process 405736 exits successfully. +==== EXPERIMENT COMPLETED: qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.9_2e-1_connector-3.0_0.9_2e-1_ablation ==== +Log File: qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_0.9_2e-1_connector-3.0_0.9_2e-1_ablation_20251012_144110.log +Timestamp: 2025-10-12 15:15:26 +===================================== diff --git a/logs_oct11/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.1_2e-1_connector-3.0_1.1_2e-1_ablation_20251012_151526.log b/logs_oct11/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.1_2e-1_connector-3.0_1.1_2e-1_ablation_20251012_151526.log new file mode 100644 index 0000000000000000000000000000000000000000..f18840b142cd03341474e8a4f00bd963dcb1ffaf --- /dev/null +++ b/logs_oct11/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.1_2e-1_connector-3.0_1.1_2e-1_ablation_20251012_151526.log @@ -0,0 +1,2312 @@ +==== STARTING EXPERIMENT: qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.1_2e-1_connector-3.0_1.1_2e-1_ablation ==== +Log File: qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.1_2e-1_connector-3.0_1.1_2e-1_ablation_20251012_151526.log +Timestamp: 2025-10-12 15:15:26 +===================================== +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 15:15:28,817] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 15:15:31,715] [WARNING] [runner.py:202:fetch_hostfile] Unable to find hostfile, will proceed with training with local resources only. +[2025-10-12 15:15:31,716] [INFO] [runner.py:568:main] cmd = /opt/conda/envs/tinyllava/bin/python3.10 -u -m deepspeed.launcher.launch --world_info=eyJsb2NhbGhvc3QiOiBbMCwgMSwgMiwgMywgNCwgNSwgNiwgN119 --master_addr=127.0.0.1 --master_port=29501 --enable_each_rank_log=None tinyllava/train/train.py --deepspeed ./scripts/zero3.json --data_path /s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json --image_folder /s3-code/ywang29/datasets/tinyllava --is_multimodal True --conv_version qwen2_base --model_name_or_path Qwen/Qwen2.5-0.5B --vision_tower google/siglip-so400m-patch14-384 --vision_tower2 --connector_type mlp2x_gelu --mm_vision_select_layer -2 --image_aspect_ratio square --attn_implementation flash_attention_2 --bf16 True --training_recipe common --tune_type_llm full --tune_type_vision_tower frozen --tune_vision_tower_from_layer 0 --tune_type_connector full --group_by_modality_length True --pretrained_model_path /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain --output_dir /nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.1_2e-1_connector-3.0_1.1_2e-1_ablation --num_train_epochs 1 --per_device_train_batch_size 4 --per_device_eval_batch_size 4 --gradient_accumulation_steps 4 --evaluation_strategy no --learning_rate 2e-1 --weight_decay 0. --warmup_ratio 0.03 --lr_scheduler_type cosine --logging_steps 1 --tf32 False --model_max_length 2048 --gradient_checkpointing True --dataloader_num_workers 8 --lazy_preprocess True --report_to tensorboard --tokenizer_use_fast False --run_name tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune --subnet_mode_text both --subnet_type_text None --mask_type_text soft --init_mean_text 3.0 --temperature_attn_text 1.1 --temperature_mlp_text 1.1 --backward_type_text normal --masked_layers_text all --subnet_mode_vision both --subnet_type_vision None --mask_type_vision soft --init_mean_vision 3.0 --temperature_attn_vision 1.1 --temperature_mlp_vision 1.1 --backward_type_vision normal --masked_layers_vision all --subnet_type_connector global --mask_type_connector soft --init_mean_connector 3.0 --temperature_connector 1.1 --backward_type_connector normal --mm_projector_lr 2e-1 --seed 42 --mask_model llm-connector --save_strategy steps --save_steps 50000 --save_total_limit 1 --train_data_ratio 0.1 +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 15:15:34,315] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 15:15:35,339] [INFO] [launch.py:138:main] 0 NCCL_VERSION=2.21.5 +[2025-10-12 15:15:35,339] [INFO] [launch.py:138:main] 0 NCCL_SOCKET_IFNAME=eth +[2025-10-12 15:15:35,339] [INFO] [launch.py:145:main] WORLD INFO DICT: {'localhost': [0, 1, 2, 3, 4, 5, 6, 7]} +[2025-10-12 15:15:35,339] [INFO] [launch.py:151:main] nnodes=1, num_local_procs=8, node_rank=0 +[2025-10-12 15:15:35,339] [INFO] [launch.py:162:main] global_rank_mapping=defaultdict(, {'localhost': [0, 1, 2, 3, 4, 5, 6, 7]}) +[2025-10-12 15:15:35,339] [INFO] [launch.py:163:main] dist_world_size=8 +[2025-10-12 15:15:35,339] [INFO] [launch.py:165:main] Setting CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 +[2025-10-12 15:15:35,342] [INFO] [launch.py:253:main] process 425405 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=0', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.1_2e-1_connector-3.0_1.1_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '1.1', '--temperature_mlp_text', '1.1', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '1.1', '--temperature_mlp_vision', '1.1', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '1.1', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 15:15:35,344] [INFO] [launch.py:253:main] process 425406 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=1', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.1_2e-1_connector-3.0_1.1_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '1.1', '--temperature_mlp_text', '1.1', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '1.1', '--temperature_mlp_vision', '1.1', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '1.1', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 15:15:35,346] [INFO] [launch.py:253:main] process 425407 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=2', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.1_2e-1_connector-3.0_1.1_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '1.1', '--temperature_mlp_text', '1.1', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '1.1', '--temperature_mlp_vision', '1.1', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '1.1', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 15:15:35,349] [INFO] [launch.py:253:main] process 425408 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=3', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.1_2e-1_connector-3.0_1.1_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '1.1', '--temperature_mlp_text', '1.1', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '1.1', '--temperature_mlp_vision', '1.1', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '1.1', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 15:15:35,351] [INFO] [launch.py:253:main] process 425409 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=4', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.1_2e-1_connector-3.0_1.1_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '1.1', '--temperature_mlp_text', '1.1', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '1.1', '--temperature_mlp_vision', '1.1', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '1.1', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 15:15:35,353] [INFO] [launch.py:253:main] process 425410 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=5', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.1_2e-1_connector-3.0_1.1_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '1.1', '--temperature_mlp_text', '1.1', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '1.1', '--temperature_mlp_vision', '1.1', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '1.1', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 15:15:35,355] [INFO] [launch.py:253:main] process 425411 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=6', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.1_2e-1_connector-3.0_1.1_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '1.1', '--temperature_mlp_text', '1.1', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '1.1', '--temperature_mlp_vision', '1.1', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '1.1', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 15:15:35,357] [INFO] [launch.py:253:main] process 425412 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=7', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.1_2e-1_connector-3.0_1.1_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '1.1', '--temperature_mlp_text', '1.1', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '1.1', '--temperature_mlp_vision', '1.1', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '1.1', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 15:15:41,972] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 15:15:42,137] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 15:15:42,137] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 15:15:42,144] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 15:15:42,151] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 15:15:42,152] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 15:15:42,156] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 15:15:42,160] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 15:15:42,379] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 15:15:42,541] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 15:15:42,548] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 15:15:42,551] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 15:15:42,554] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 15:15:42,567] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 15:15:42,575] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 15:15:42,590] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 15:15:42,590] [INFO] [comm.py:668:init_distributed] Initializing TorchBackend in DeepSpeed with backend nccl +Apply masks for the following modules: ['llm', 'connector'] +Apply masks for the following modules: ['llm', 'connector'] +{'llm': {'model_name_or_path': 'Qwen/Qwen2.5-0.5B', 'cache_dir': None, 'attn_implementation': 'flash_attention_2', 'subnet_mode': 'both', 'subnet_type': 'None', 'sparsity_attn': None, 'sparsity_mlp': None, 'threshold_attn': None, 'threshold_mlp': None, 'temperature_attn': 1.1, 'temperature_mlp': 1.1, 'masked_layers': 'all', 'mask_type': 'soft', 'backward_type': 'normal'}, 'vision_tower': {'model_name_or_path': 'google/siglip-so400m-patch14-384'}, 'connector': {'connector_type': 'mlp2x_gelu', 'subnet_type': 'global', 'threshold': None, 'sparsity': None, 'temperature': 1.1, 'mask_type': 'soft', 'backward_type': 'normal'}} +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +TinyLlavaConfig { + "backward_type_connector": "normal", + "cache_dir": null, + "connector_type": "mlp2x_gelu", + "hidden_size": 896, + "ignore_index": -100, + "image_aspect_ratio": "square", + "image_token_index": -200, + "llm_model_name_or_path": "Qwen/Qwen2.5-0.5B", + "mask_model": [ + "llm", + "connector" + ], + "mask_type_connector": "soft", + "model_type": "tinyllava", + "num_queries": 128, + "num_resampler_layers": 3, + "pad_token": null, + "resampler_hidden_size": 768, + "sparsity_connector": null, + "subnet_type_connector": "global", + "temperature_connector": 1.1, + "text_config": { + "_name_or_path": "Qwen/Qwen2.5-0.5B", + "architectures": [ + "Qwen2ForCausalLM" + ], + "backward_type": "normal", + "bos_token_id": 151643, + "eos_token_id": 151643, + "hidden_size": 896, + "intermediate_size": 4864, + "mask_type": "soft", + "masked_layers": "all", + "max_position_embeddings": 32768, + "max_window_layers": 24, + "model_type": "qwen2", + "num_attention_heads": 14, + "num_hidden_layers": 24, + "num_key_value_heads": 2, + "rope_theta": 1000000.0, + "sliding_window": 32768, + "subnet_mode": "both", + "subnet_type": "None", + "temperature_attn": 1.1, + "temperature_mlp": 1.1, + "tie_word_embeddings": true, + "torch_dtype": "bfloat16", + "use_mrope": false, + "use_sliding_window": false, + "vocab_size": 151936 + }, + "threshold_connector": null, + "tokenizer_model_max_length": 2048, + "tokenizer_name_or_path": "Qwen/Qwen2.5-0.5B", + "tokenizer_padding_side": "right", + "tokenizer_use_fast": false, + "transformers_version": "4.40.1", + "tune_type_connector": "frozen", + "tune_type_llm": "frozen", + "tune_type_vision_tower": "frozen", + "tune_vision_tower_from_layer": -1, + "use_cache": false, + "vision_config": { + "hidden_act": "gelu_pytorch_tanh", + "hidden_size": 1152, + "image_size": 384, + "intermediate_size": 4304, + "layer_norm_eps": 1e-06, + "model_name_or_path": "google/siglip-so400m-patch14-384", + "model_name_or_path2": "", + "model_type": "siglip_vision_model", + "num_attention_heads": 16, + "num_hidden_layers": 27, + "patch_size": 14 + }, + "vision_feature_layer": -2, + "vision_feature_select_strategy": "patch", + "vision_hidden_size": 1152, + "vision_model_name_or_path": "google/siglip-so400m-patch14-384", + "vision_model_name_or_path2": "", + "vocab_size": 151936 +} + +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +ywang29-vrdb-test1-worker-0:425405:425405 [0] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:425405:425405 [0] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:425405:425405 [0] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:425405:425405 [0] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:425405:425405 [0] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test1-worker-0:425405:425405 [0] NCCL INFO cudaDriverVersion 12040 +NCCL version 2.21.5+cuda12.1 +ywang29-vrdb-test1-worker-0:425411:425411 [6] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test1-worker-0:425411:425411 [6] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:425411:425411 [6] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:425411:425411 [6] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:425411:425411 [6] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:425411:425411 [6] NCCL INFO NET/Plugin: Using internal network plugin. +Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage() + return self.fget.__get__(instance, owner)() +ywang29-vrdb-test1-worker-0:425405:427002 [0] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:425405:427002 [0] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:425405:427002 [0] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:425405:427002 [0] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:425405:427002 [0] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:425405:427002 [0] NCCL INFO Using network Socket +ywang29-vrdb-test1-worker-0:425411:427007 [6] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:425411:427007 [6] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:425411:427007 [6] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:425411:427007 [6] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:425411:427007 [6] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:425411:427007 [6] NCCL INFO Using network Socket +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test1-worker-0:425406:425406 [1] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test1-worker-0:425406:425406 [1] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:425406:425406 [1] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:425406:425406 [1] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:425406:425406 [1] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:425406:425406 [1] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test1-worker-0:425410:425410 [5] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test1-worker-0:425410:425410 [5] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:425410:425410 [5] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:425410:425410 [5] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:425410:425410 [5] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:425410:425410 [5] NCCL INFO NET/Plugin: Using internal network plugin. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test1-worker-0:425412:425412 [7] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test1-worker-0:425412:425412 [7] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:425412:425412 [7] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:425412:425412 [7] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:425412:425412 [7] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:425412:425412 [7] NCCL INFO NET/Plugin: Using internal network plugin. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test1-worker-0:425409:425409 [4] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test1-worker-0:425409:425409 [4] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:425409:425409 [4] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:425409:425409 [4] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:425409:425409 [4] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:425409:425409 [4] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test1-worker-0:425407:425407 [2] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test1-worker-0:425407:425407 [2] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:425407:425407 [2] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:425407:425407 [2] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:425407:425407 [2] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:425407:425407 [2] NCCL INFO NET/Plugin: Using internal network plugin. +You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`. +ywang29-vrdb-test1-worker-0:425408:425408 [3] NCCL INFO cudaDriverVersion 12040 +ywang29-vrdb-test1-worker-0:425408:425408 [3] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:425408:425408 [3] NCCL INFO Bootstrap : Using eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:425408:425408 [3] NCCL INFO NET/Plugin: No plugin found (libnccl-net.so) +ywang29-vrdb-test1-worker-0:425408:425408 [3] NCCL INFO NET/Plugin: Plugin load returned 2 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-net.so +ywang29-vrdb-test1-worker-0:425408:425408 [3] NCCL INFO NET/Plugin: Using internal network plugin. +ywang29-vrdb-test1-worker-0:425406:427020 [1] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:425406:427020 [1] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:425406:427020 [1] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:425406:427020 [1] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:425406:427020 [1] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:425406:427020 [1] NCCL INFO Using network Socket +ywang29-vrdb-test1-worker-0:425410:427021 [5] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:425410:427021 [5] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:425410:427021 [5] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:425410:427021 [5] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:425410:427021 [5] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:425410:427021 [5] NCCL INFO Using network Socket +ywang29-vrdb-test1-worker-0:425412:427022 [7] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:425412:427022 [7] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:425412:427022 [7] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:425412:427022 [7] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:425412:427022 [7] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:425412:427022 [7] NCCL INFO Using network Socket +ywang29-vrdb-test1-worker-0:425409:427023 [4] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:425409:427023 [4] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:425409:427023 [4] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:425409:427023 [4] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:425409:427023 [4] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:425409:427023 [4] NCCL INFO Using network Socket +ywang29-vrdb-test1-worker-0:425407:427024 [2] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:425407:427024 [2] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:425407:427024 [2] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:425407:427024 [2] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:425407:427024 [2] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:425407:427024 [2] NCCL INFO Using network Socket +ywang29-vrdb-test1-worker-0:425408:427025 [3] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:425408:427025 [3] NCCL INFO NET/IB : No device found. +ywang29-vrdb-test1-worker-0:425408:427025 [3] NCCL INFO NCCL_SOCKET_IFNAME set by environment to eth +ywang29-vrdb-test1-worker-0:425408:427025 [3] NCCL INFO NET/Socket : Using [0]eth0:10.200.136.19<0> +ywang29-vrdb-test1-worker-0:425408:427025 [3] NCCL INFO Using non-device net plugin version 0 +ywang29-vrdb-test1-worker-0:425408:427025 [3] NCCL INFO Using network Socket +ywang29-vrdb-test1-worker-0:425408:427025 [3] NCCL INFO ncclCommInitRank comm 0x556a6bee3f60 rank 3 nranks 8 cudaDev 3 nvmlDev 3 busId 201d0 commId 0x42b5c6f4c4b0197a - Init START +ywang29-vrdb-test1-worker-0:425407:427024 [2] NCCL INFO ncclCommInitRank comm 0x55b55b04e0c0 rank 2 nranks 8 cudaDev 2 nvmlDev 2 busId 201c0 commId 0x42b5c6f4c4b0197a - Init START +ywang29-vrdb-test1-worker-0:425406:427020 [1] NCCL INFO ncclCommInitRank comm 0x556b23711fc0 rank 1 nranks 8 cudaDev 1 nvmlDev 1 busId 101d0 commId 0x42b5c6f4c4b0197a - Init START +ywang29-vrdb-test1-worker-0:425410:427021 [5] NCCL INFO ncclCommInitRank comm 0x5648ed843ca0 rank 5 nranks 8 cudaDev 5 nvmlDev 5 busId 901d0 commId 0x42b5c6f4c4b0197a - Init START +ywang29-vrdb-test1-worker-0:425409:427023 [4] NCCL INFO ncclCommInitRank comm 0x55d67f3afe90 rank 4 nranks 8 cudaDev 4 nvmlDev 4 busId 901c0 commId 0x42b5c6f4c4b0197a - Init START +ywang29-vrdb-test1-worker-0:425411:427007 [6] NCCL INFO ncclCommInitRank comm 0x56163f032c00 rank 6 nranks 8 cudaDev 6 nvmlDev 6 busId a01c0 commId 0x42b5c6f4c4b0197a - Init START +ywang29-vrdb-test1-worker-0:425412:427022 [7] NCCL INFO ncclCommInitRank comm 0x561024e243a0 rank 7 nranks 8 cudaDev 7 nvmlDev 7 busId a01d0 commId 0x42b5c6f4c4b0197a - Init START +ywang29-vrdb-test1-worker-0:425405:427002 [0] NCCL INFO ncclCommInitRank comm 0x56302a9f12b0 rank 0 nranks 8 cudaDev 0 nvmlDev 0 busId 101c0 commId 0x42b5c6f4c4b0197a - Init START +ywang29-vrdb-test1-worker-0:425405:427002 [0] NCCL INFO Setting affinity for GPU 0 to ff,ffff0000,00ffffff +ywang29-vrdb-test1-worker-0:425405:427002 [0] NCCL INFO NVLS multicast support is not available on dev 0 +ywang29-vrdb-test1-worker-0:425409:427023 [4] NCCL INFO Setting affinity for GPU 4 to ffffff00,0000ffff,ff000000 +ywang29-vrdb-test1-worker-0:425409:427023 [4] NCCL INFO NVLS multicast support is not available on dev 4 +ywang29-vrdb-test1-worker-0:425408:427025 [3] NCCL INFO Setting affinity for GPU 3 to ff,ffff0000,00ffffff +ywang29-vrdb-test1-worker-0:425408:427025 [3] NCCL INFO NVLS multicast support is not available on dev 3 +ywang29-vrdb-test1-worker-0:425411:427007 [6] NCCL INFO Setting affinity for GPU 6 to ffffff00,0000ffff,ff000000 +ywang29-vrdb-test1-worker-0:425411:427007 [6] NCCL INFO NVLS multicast support is not available on dev 6 +ywang29-vrdb-test1-worker-0:425412:427022 [7] NCCL INFO Setting affinity for GPU 7 to ffffff00,0000ffff,ff000000 +ywang29-vrdb-test1-worker-0:425410:427021 [5] NCCL INFO Setting affinity for GPU 5 to ffffff00,0000ffff,ff000000 +ywang29-vrdb-test1-worker-0:425410:427021 [5] NCCL INFO NVLS multicast support is not available on dev 5 +ywang29-vrdb-test1-worker-0:425412:427022 [7] NCCL INFO NVLS multicast support is not available on dev 7 +ywang29-vrdb-test1-worker-0:425406:427020 [1] NCCL INFO Setting affinity for GPU 1 to ff,ffff0000,00ffffff +ywang29-vrdb-test1-worker-0:425406:427020 [1] NCCL INFO NVLS multicast support is not available on dev 1 +ywang29-vrdb-test1-worker-0:425407:427024 [2] NCCL INFO Setting affinity for GPU 2 to ff,ffff0000,00ffffff +ywang29-vrdb-test1-worker-0:425407:427024 [2] NCCL INFO NVLS multicast support is not available on dev 2 +ywang29-vrdb-test1-worker-0:425407:427024 [2] NCCL INFO comm 0x55b55b04e0c0 rank 2 nRanks 8 nNodes 1 localRanks 8 localRank 2 MNNVL 0 +ywang29-vrdb-test1-worker-0:425406:427020 [1] NCCL INFO comm 0x556b23711fc0 rank 1 nRanks 8 nNodes 1 localRanks 8 localRank 1 MNNVL 0 +ywang29-vrdb-test1-worker-0:425405:427002 [0] NCCL INFO comm 0x56302a9f12b0 rank 0 nRanks 8 nNodes 1 localRanks 8 localRank 0 MNNVL 0 +ywang29-vrdb-test1-worker-0:425412:427022 [7] NCCL INFO comm 0x561024e243a0 rank 7 nRanks 8 nNodes 1 localRanks 8 localRank 7 MNNVL 0 +ywang29-vrdb-test1-worker-0:425411:427007 [6] NCCL INFO comm 0x56163f032c00 rank 6 nRanks 8 nNodes 1 localRanks 8 localRank 6 MNNVL 0 +ywang29-vrdb-test1-worker-0:425410:427021 [5] NCCL INFO comm 0x5648ed843ca0 rank 5 nRanks 8 nNodes 1 localRanks 8 localRank 5 MNNVL 0 +ywang29-vrdb-test1-worker-0:425409:427023 [4] NCCL INFO comm 0x55d67f3afe90 rank 4 nRanks 8 nNodes 1 localRanks 8 localRank 4 MNNVL 0 +ywang29-vrdb-test1-worker-0:425408:427025 [3] NCCL INFO comm 0x556a6bee3f60 rank 3 nRanks 8 nNodes 1 localRanks 8 localRank 3 MNNVL 0 +ywang29-vrdb-test1-worker-0:425407:427024 [2] NCCL INFO Trees [0] 3/-1/-1->2->1 [1] 3/-1/-1->2->1 [2] 3/-1/-1->2->1 [3] 3/-1/-1->2->1 [4] 3/-1/-1->2->1 [5] 3/-1/-1->2->1 [6] 3/-1/-1->2->1 [7] 3/-1/-1->2->1 [8] 3/-1/-1->2->1 [9] 3/-1/-1->2->1 [10] 3/-1/-1->2->1 [11] 3/-1/-1->2->1 [12] 3/-1/-1->2->1 [13] 3/-1/-1->2->1 [14] 3/-1/-1->2->1 [15] 3/-1/-1->2->1 [16] 3/-1/-1->2->1 [17] 3/-1/-1->2->1 [18] 3/-1/-1->2->1 [19] 3/-1/-1->2->1 [20] 3/-1/-1->2->1 [21] 3/-1/-1->2->1 [22] 3/-1/-1->2->1 [23] 3/-1/-1->2->1 +ywang29-vrdb-test1-worker-0:425406:427020 [1] NCCL INFO Trees [0] 2/-1/-1->1->0 [1] 2/-1/-1->1->0 [2] 2/-1/-1->1->0 [3] 2/-1/-1->1->0 [4] 2/-1/-1->1->0 [5] 2/-1/-1->1->0 [6] 2/-1/-1->1->0 [7] 2/-1/-1->1->0 [8] 2/-1/-1->1->0 [9] 2/-1/-1->1->0 [10] 2/-1/-1->1->0 [11] 2/-1/-1->1->0 [12] 2/-1/-1->1->0 [13] 2/-1/-1->1->0 [14] 2/-1/-1->1->0 [15] 2/-1/-1->1->0 [16] 2/-1/-1->1->0 [17] 2/-1/-1->1->0 [18] 2/-1/-1->1->0 [19] 2/-1/-1->1->0 [20] 2/-1/-1->1->0 [21] 2/-1/-1->1->0 [22] 2/-1/-1->1->0 [23] 2/-1/-1->1->0 +ywang29-vrdb-test1-worker-0:425407:427024 [2] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:425406:427020 [1] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:425405:427002 [0] NCCL INFO Channel 00/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:425405:427002 [0] NCCL INFO Channel 01/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:425405:427002 [0] NCCL INFO Channel 02/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:425412:427022 [7] NCCL INFO Trees [0] -1/-1/-1->7->6 [1] -1/-1/-1->7->6 [2] -1/-1/-1->7->6 [3] -1/-1/-1->7->6 [4] -1/-1/-1->7->6 [5] -1/-1/-1->7->6 [6] -1/-1/-1->7->6 [7] -1/-1/-1->7->6 [8] -1/-1/-1->7->6 [9] -1/-1/-1->7->6 [10] -1/-1/-1->7->6 [11] -1/-1/-1->7->6 [12] -1/-1/-1->7->6 [13] -1/-1/-1->7->6 [14] -1/-1/-1->7->6 [15] -1/-1/-1->7->6 [16] -1/-1/-1->7->6 [17] -1/-1/-1->7->6 [18] -1/-1/-1->7->6 [19] -1/-1/-1->7->6 [20] -1/-1/-1->7->6 [21] -1/-1/-1->7->6 [22] -1/-1/-1->7->6 [23] -1/-1/-1->7->6 +ywang29-vrdb-test1-worker-0:425405:427002 [0] NCCL INFO Channel 03/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:425412:427022 [7] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:425405:427002 [0] NCCL INFO Channel 04/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:425405:427002 [0] NCCL INFO Channel 05/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:425409:427023 [4] NCCL INFO Trees [0] 5/-1/-1->4->3 [1] 5/-1/-1->4->3 [2] 5/-1/-1->4->3 [3] 5/-1/-1->4->3 [4] 5/-1/-1->4->3 [5] 5/-1/-1->4->3 [6] 5/-1/-1->4->3 [7] 5/-1/-1->4->3 [8] 5/-1/-1->4->3 [9] 5/-1/-1->4->3 [10] 5/-1/-1->4->3 [11] 5/-1/-1->4->3 [12] 5/-1/-1->4->3 [13] 5/-1/-1->4->3 [14] 5/-1/-1->4->3 [15] 5/-1/-1->4->3 [16] 5/-1/-1->4->3 [17] 5/-1/-1->4->3 [18] 5/-1/-1->4->3 [19] 5/-1/-1->4->3 [20] 5/-1/-1->4->3 [21] 5/-1/-1->4->3 [22] 5/-1/-1->4->3 [23] 5/-1/-1->4->3 +ywang29-vrdb-test1-worker-0:425405:427002 [0] NCCL INFO Channel 06/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:425409:427023 [4] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:425405:427002 [0] NCCL INFO Channel 07/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:425405:427002 [0] NCCL INFO Channel 08/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:425405:427002 [0] NCCL INFO Channel 09/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:425405:427002 [0] NCCL INFO Channel 10/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:425411:427007 [6] NCCL INFO Trees [0] 7/-1/-1->6->5 [1] 7/-1/-1->6->5 [2] 7/-1/-1->6->5 [3] 7/-1/-1->6->5 [4] 7/-1/-1->6->5 [5] 7/-1/-1->6->5 [6] 7/-1/-1->6->5 [7] 7/-1/-1->6->5 [8] 7/-1/-1->6->5 [9] 7/-1/-1->6->5 [10] 7/-1/-1->6->5 [11] 7/-1/-1->6->5 [12] 7/-1/-1->6->5 [13] 7/-1/-1->6->5 [14] 7/-1/-1->6->5 [15] 7/-1/-1->6->5 [16] 7/-1/-1->6->5 [17] 7/-1/-1->6->5 [18] 7/-1/-1->6->5 [19] 7/-1/-1->6->5 [20] 7/-1/-1->6->5 [21] 7/-1/-1->6->5 [22] 7/-1/-1->6->5 [23] 7/-1/-1->6->5 +ywang29-vrdb-test1-worker-0:425410:427021 [5] NCCL INFO Trees [0] 6/-1/-1->5->4 [1] 6/-1/-1->5->4 [2] 6/-1/-1->5->4 [3] 6/-1/-1->5->4 [4] 6/-1/-1->5->4 [5] 6/-1/-1->5->4 [6] 6/-1/-1->5->4 [7] 6/-1/-1->5->4 [8] 6/-1/-1->5->4 [9] 6/-1/-1->5->4 [10] 6/-1/-1->5->4 [11] 6/-1/-1->5->4 [12] 6/-1/-1->5->4 [13] 6/-1/-1->5->4 [14] 6/-1/-1->5->4 [15] 6/-1/-1->5->4 [16] 6/-1/-1->5->4 [17] 6/-1/-1->5->4 [18] 6/-1/-1->5->4 [19] 6/-1/-1->5->4 [20] 6/-1/-1->5->4 [21] 6/-1/-1->5->4 [22] 6/-1/-1->5->4 [23] 6/-1/-1->5->4 +ywang29-vrdb-test1-worker-0:425408:427025 [3] NCCL INFO Trees [0] 4/-1/-1->3->2 [1] 4/-1/-1->3->2 [2] 4/-1/-1->3->2 [3] 4/-1/-1->3->2 [4] 4/-1/-1->3->2 [5] 4/-1/-1->3->2 [6] 4/-1/-1->3->2 [7] 4/-1/-1->3->2 [8] 4/-1/-1->3->2 [9] 4/-1/-1->3->2 [10] 4/-1/-1->3->2 [11] 4/-1/-1->3->2 [12] 4/-1/-1->3->2 [13] 4/-1/-1->3->2 [14] 4/-1/-1->3->2 [15] 4/-1/-1->3->2 [16] 4/-1/-1->3->2 [17] 4/-1/-1->3->2 [18] 4/-1/-1->3->2 [19] 4/-1/-1->3->2 [20] 4/-1/-1->3->2 [21] 4/-1/-1->3->2 [22] 4/-1/-1->3->2 [23] 4/-1/-1->3->2 +ywang29-vrdb-test1-worker-0:425405:427002 [0] NCCL INFO Channel 11/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:425405:427002 [0] NCCL INFO Channel 12/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:425405:427002 [0] NCCL INFO Channel 13/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:425408:427025 [3] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:425410:427021 [5] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:425411:427007 [6] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:425405:427002 [0] NCCL INFO Channel 14/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:425405:427002 [0] NCCL INFO Channel 15/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:425405:427002 [0] NCCL INFO Channel 16/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:425405:427002 [0] NCCL INFO Channel 17/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:425405:427002 [0] NCCL INFO Channel 18/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:425405:427002 [0] NCCL INFO Channel 19/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:425405:427002 [0] NCCL INFO Channel 20/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:425405:427002 [0] NCCL INFO Channel 21/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:425405:427002 [0] NCCL INFO Channel 22/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:425405:427002 [0] NCCL INFO Channel 23/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:425405:427002 [0] NCCL INFO Trees [0] 1/-1/-1->0->-1 [1] 1/-1/-1->0->-1 [2] 1/-1/-1->0->-1 [3] 1/-1/-1->0->-1 [4] 1/-1/-1->0->-1 [5] 1/-1/-1->0->-1 [6] 1/-1/-1->0->-1 [7] 1/-1/-1->0->-1 [8] 1/-1/-1->0->-1 [9] 1/-1/-1->0->-1 [10] 1/-1/-1->0->-1 [11] 1/-1/-1->0->-1 [12] 1/-1/-1->0->-1 [13] 1/-1/-1->0->-1 [14] 1/-1/-1->0->-1 [15] 1/-1/-1->0->-1 [16] 1/-1/-1->0->-1 [17] 1/-1/-1->0->-1 [18] 1/-1/-1->0->-1 [19] 1/-1/-1->0->-1 [20] 1/-1/-1->0->-1 [21] 1/-1/-1->0->-1 [22] 1/-1/-1->0->-1 [23] 1/-1/-1->0->-1 +ywang29-vrdb-test1-worker-0:425405:427002 [0] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:425406:427020 [1] NCCL INFO Channel 00/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:427024 [2] NCCL INFO Channel 00/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:427020 [1] NCCL INFO Channel 01/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:427022 [7] NCCL INFO Channel 00/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:427024 [2] NCCL INFO Channel 01/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:427020 [1] NCCL INFO Channel 02/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:427022 [7] NCCL INFO Channel 01/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:427024 [2] NCCL INFO Channel 02/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:427020 [1] NCCL INFO Channel 03/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425405:427002 [0] NCCL INFO Channel 00/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:427025 [3] NCCL INFO Channel 00/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:427022 [7] NCCL INFO Channel 02/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:427024 [2] NCCL INFO Channel 03/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:427020 [1] NCCL INFO Channel 04/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:427021 [5] NCCL INFO Channel 00/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:427007 [6] NCCL INFO Channel 00/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425405:427002 [0] NCCL INFO Channel 01/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:427025 [3] NCCL INFO Channel 01/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:427022 [7] NCCL INFO Channel 03/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:427024 [2] NCCL INFO Channel 04/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:427020 [1] NCCL INFO Channel 05/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:427021 [5] NCCL INFO Channel 01/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:427007 [6] NCCL INFO Channel 01/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425405:427002 [0] NCCL INFO Channel 02/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:427022 [7] NCCL INFO Channel 04/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:427025 [3] NCCL INFO Channel 02/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:427020 [1] NCCL INFO Channel 06/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:427024 [2] NCCL INFO Channel 05/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:427021 [5] NCCL INFO Channel 02/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:427007 [6] NCCL INFO Channel 02/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425405:427002 [0] NCCL INFO Channel 03/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:427022 [7] NCCL INFO Channel 05/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:427025 [3] NCCL INFO Channel 03/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:427020 [1] NCCL INFO Channel 07/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:427024 [2] NCCL INFO Channel 06/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:427007 [6] NCCL INFO Channel 03/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:427021 [5] NCCL INFO Channel 03/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425405:427002 [0] NCCL INFO Channel 04/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:427022 [7] NCCL INFO Channel 06/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:427025 [3] NCCL INFO Channel 04/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:427020 [1] NCCL INFO Channel 08/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:427024 [2] NCCL INFO Channel 07/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:427007 [6] NCCL INFO Channel 04/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:427021 [5] NCCL INFO Channel 04/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425405:427002 [0] NCCL INFO Channel 05/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:427022 [7] NCCL INFO Channel 07/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:427025 [3] NCCL INFO Channel 05/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:427020 [1] NCCL INFO Channel 09/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:427024 [2] NCCL INFO Channel 08/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:427007 [6] NCCL INFO Channel 05/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425405:427002 [0] NCCL INFO Channel 06/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:427021 [5] NCCL INFO Channel 05/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:427022 [7] NCCL INFO Channel 08/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:427020 [1] NCCL INFO Channel 10/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:427025 [3] NCCL INFO Channel 06/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:427024 [2] NCCL INFO Channel 09/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:427007 [6] NCCL INFO Channel 06/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425405:427002 [0] NCCL INFO Channel 07/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:427022 [7] NCCL INFO Channel 09/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:427020 [1] NCCL INFO Channel 11/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:427021 [5] NCCL INFO Channel 06/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:427025 [3] NCCL INFO Channel 07/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:427024 [2] NCCL INFO Channel 10/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:427007 [6] NCCL INFO Channel 07/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425405:427002 [0] NCCL INFO Channel 08/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:427022 [7] NCCL INFO Channel 10/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:427020 [1] NCCL INFO Channel 12/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:427021 [5] NCCL INFO Channel 07/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:427025 [3] NCCL INFO Channel 08/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:427024 [2] NCCL INFO Channel 11/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:427007 [6] NCCL INFO Channel 08/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425405:427002 [0] NCCL INFO Channel 09/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:427022 [7] NCCL INFO Channel 11/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:427020 [1] NCCL INFO Channel 13/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:427021 [5] NCCL INFO Channel 08/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:427023 [4] NCCL INFO Channel 00/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:427024 [2] NCCL INFO Channel 12/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:427025 [3] NCCL INFO Channel 09/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:427007 [6] NCCL INFO Channel 09/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425405:427002 [0] NCCL INFO Channel 10/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:427022 [7] NCCL INFO Channel 12/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:427020 [1] NCCL INFO Channel 14/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:427021 [5] NCCL INFO Channel 09/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:427023 [4] NCCL INFO Channel 01/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:427024 [2] NCCL INFO Channel 13/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:427025 [3] NCCL INFO Channel 10/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:427007 [6] NCCL INFO Channel 10/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425405:427002 [0] NCCL INFO Channel 11/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:427022 [7] NCCL INFO Channel 13/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:427020 [1] NCCL INFO Channel 15/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:427021 [5] NCCL INFO Channel 10/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:427023 [4] NCCL INFO Channel 02/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:427024 [2] NCCL INFO Channel 14/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:427025 [3] NCCL INFO Channel 11/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:427007 [6] NCCL INFO Channel 11/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425405:427002 [0] NCCL INFO Channel 12/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:427022 [7] NCCL INFO Channel 14/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:427020 [1] NCCL INFO Channel 16/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:427024 [2] NCCL INFO Channel 15/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:427021 [5] NCCL INFO Channel 11/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:427023 [4] NCCL INFO Channel 03/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:427025 [3] NCCL INFO Channel 12/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:427007 [6] NCCL INFO Channel 12/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:427020 [1] NCCL INFO Channel 17/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425405:427002 [0] NCCL INFO Channel 13/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:427022 [7] NCCL INFO Channel 15/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:427023 [4] NCCL INFO Channel 04/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:427024 [2] NCCL INFO Channel 16/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:427021 [5] NCCL INFO Channel 12/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:427025 [3] NCCL INFO Channel 13/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:427020 [1] NCCL INFO Channel 18/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:427007 [6] NCCL INFO Channel 13/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425405:427002 [0] NCCL INFO Channel 14/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:427022 [7] NCCL INFO Channel 16/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:427023 [4] NCCL INFO Channel 05/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:427024 [2] NCCL INFO Channel 17/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:427021 [5] NCCL INFO Channel 13/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:427025 [3] NCCL INFO Channel 14/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:427020 [1] NCCL INFO Channel 19/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:427007 [6] NCCL INFO Channel 14/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:427022 [7] NCCL INFO Channel 17/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425405:427002 [0] NCCL INFO Channel 15/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:427024 [2] NCCL INFO Channel 18/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:427023 [4] NCCL INFO Channel 06/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:427025 [3] NCCL INFO Channel 15/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:427021 [5] NCCL INFO Channel 14/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:427020 [1] NCCL INFO Channel 20/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:427007 [6] NCCL INFO Channel 15/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:427022 [7] NCCL INFO Channel 18/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425405:427002 [0] NCCL INFO Channel 16/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:427024 [2] NCCL INFO Channel 19/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:427023 [4] NCCL INFO Channel 07/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:427025 [3] NCCL INFO Channel 16/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:427021 [5] NCCL INFO Channel 15/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:427020 [1] NCCL INFO Channel 21/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:427007 [6] NCCL INFO Channel 16/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:427023 [4] NCCL INFO Channel 08/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:427022 [7] NCCL INFO Channel 19/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:427024 [2] NCCL INFO Channel 20/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425405:427002 [0] NCCL INFO Channel 17/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:427025 [3] NCCL INFO Channel 17/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:427021 [5] NCCL INFO Channel 16/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:427020 [1] NCCL INFO Channel 22/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:427007 [6] NCCL INFO Channel 17/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:427023 [4] NCCL INFO Channel 09/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:427022 [7] NCCL INFO Channel 20/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:427024 [2] NCCL INFO Channel 21/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425405:427002 [0] NCCL INFO Channel 18/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:427025 [3] NCCL INFO Channel 18/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:427021 [5] NCCL INFO Channel 17/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:427020 [1] NCCL INFO Channel 23/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:427007 [6] NCCL INFO Channel 18/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:427023 [4] NCCL INFO Channel 10/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:427022 [7] NCCL INFO Channel 21/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:427024 [2] NCCL INFO Channel 22/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425405:427002 [0] NCCL INFO Channel 19/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:427025 [3] NCCL INFO Channel 19/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:427021 [5] NCCL INFO Channel 18/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:427023 [4] NCCL INFO Channel 11/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:427007 [6] NCCL INFO Channel 19/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:427022 [7] NCCL INFO Channel 22/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:427024 [2] NCCL INFO Channel 23/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425405:427002 [0] NCCL INFO Channel 20/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:427025 [3] NCCL INFO Channel 20/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:427021 [5] NCCL INFO Channel 19/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:427022 [7] NCCL INFO Channel 23/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:427023 [4] NCCL INFO Channel 12/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:427007 [6] NCCL INFO Channel 20/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425405:427002 [0] NCCL INFO Channel 21/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:427025 [3] NCCL INFO Channel 21/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:427021 [5] NCCL INFO Channel 20/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:427023 [4] NCCL INFO Channel 13/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:427007 [6] NCCL INFO Channel 21/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425405:427002 [0] NCCL INFO Channel 22/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:427025 [3] NCCL INFO Channel 22/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:427023 [4] NCCL INFO Channel 14/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:427021 [5] NCCL INFO Channel 21/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:427007 [6] NCCL INFO Channel 22/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425405:427002 [0] NCCL INFO Channel 23/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:427025 [3] NCCL INFO Channel 23/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:427023 [4] NCCL INFO Channel 15/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:427021 [5] NCCL INFO Channel 22/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:427007 [6] NCCL INFO Channel 23/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:427023 [4] NCCL INFO Channel 16/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:427021 [5] NCCL INFO Channel 23/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:427023 [4] NCCL INFO Channel 17/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:427023 [4] NCCL INFO Channel 18/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:427023 [4] NCCL INFO Channel 19/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:427023 [4] NCCL INFO Channel 20/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:427023 [4] NCCL INFO Channel 21/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:427023 [4] NCCL INFO Channel 22/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:427023 [4] NCCL INFO Channel 23/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:427020 [1] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:425405:427002 [0] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:425407:427024 [2] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:425408:427025 [3] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:425409:427023 [4] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:425412:427022 [7] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:425412:427022 [7] NCCL INFO Channel 00/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:427021 [5] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:425411:427007 [6] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:425412:427022 [7] NCCL INFO Channel 01/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:427022 [7] NCCL INFO Channel 02/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:427022 [7] NCCL INFO Channel 03/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:427022 [7] NCCL INFO Channel 04/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:427022 [7] NCCL INFO Channel 05/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:427022 [7] NCCL INFO Channel 06/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:427022 [7] NCCL INFO Channel 07/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:427022 [7] NCCL INFO Channel 08/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:427025 [3] NCCL INFO Channel 00/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:427022 [7] NCCL INFO Channel 09/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:427025 [3] NCCL INFO Channel 01/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:427022 [7] NCCL INFO Channel 10/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:427025 [3] NCCL INFO Channel 02/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:427022 [7] NCCL INFO Channel 11/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:427025 [3] NCCL INFO Channel 03/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:427025 [3] NCCL INFO Channel 04/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:427020 [1] NCCL INFO Channel 00/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:427025 [3] NCCL INFO Channel 05/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:427020 [1] NCCL INFO Channel 01/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:427025 [3] NCCL INFO Channel 06/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:427020 [1] NCCL INFO Channel 02/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:427025 [3] NCCL INFO Channel 07/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:427020 [1] NCCL INFO Channel 03/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:427025 [3] NCCL INFO Channel 08/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:427020 [1] NCCL INFO Channel 04/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:427024 [2] NCCL INFO Channel 00/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:427025 [3] NCCL INFO Channel 09/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:427020 [1] NCCL INFO Channel 05/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:427025 [3] NCCL INFO Channel 10/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:427024 [2] NCCL INFO Channel 01/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:427007 [6] NCCL INFO Channel 00/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:427020 [1] NCCL INFO Channel 06/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:427025 [3] NCCL INFO Channel 11/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:427024 [2] NCCL INFO Channel 02/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:427021 [5] NCCL INFO Channel 00/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:427007 [6] NCCL INFO Channel 01/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:427020 [1] NCCL INFO Channel 07/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:427025 [3] NCCL INFO Channel 12/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:427024 [2] NCCL INFO Channel 03/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:427021 [5] NCCL INFO Channel 01/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:427007 [6] NCCL INFO Channel 02/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:427020 [1] NCCL INFO Channel 08/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:427025 [3] NCCL INFO Channel 13/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:427024 [2] NCCL INFO Channel 04/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:427021 [5] NCCL INFO Channel 02/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:427007 [6] NCCL INFO Channel 03/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:427020 [1] NCCL INFO Channel 09/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:427025 [3] NCCL INFO Channel 14/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:427024 [2] NCCL INFO Channel 05/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:427021 [5] NCCL INFO Channel 03/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:427007 [6] NCCL INFO Channel 04/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:427020 [1] NCCL INFO Channel 10/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:427025 [3] NCCL INFO Channel 15/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:427024 [2] NCCL INFO Channel 06/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:427021 [5] NCCL INFO Channel 04/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:427007 [6] NCCL INFO Channel 05/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:427020 [1] NCCL INFO Channel 11/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:427022 [7] NCCL INFO Channel 12/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:427025 [3] NCCL INFO Channel 16/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:427024 [2] NCCL INFO Channel 07/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:427021 [5] NCCL INFO Channel 05/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:427007 [6] NCCL INFO Channel 06/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:427020 [1] NCCL INFO Channel 12/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:427022 [7] NCCL INFO Channel 13/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:427025 [3] NCCL INFO Channel 17/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:427024 [2] NCCL INFO Channel 08/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:427021 [5] NCCL INFO Channel 06/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:427007 [6] NCCL INFO Channel 07/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:427020 [1] NCCL INFO Channel 13/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:427022 [7] NCCL INFO Channel 14/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:427025 [3] NCCL INFO Channel 18/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:427024 [2] NCCL INFO Channel 09/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:427021 [5] NCCL INFO Channel 07/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:427007 [6] NCCL INFO Channel 08/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:427022 [7] NCCL INFO Channel 15/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:427020 [1] NCCL INFO Channel 14/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:427025 [3] NCCL INFO Channel 19/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:427024 [2] NCCL INFO Channel 10/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:427021 [5] NCCL INFO Channel 08/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:427007 [6] NCCL INFO Channel 09/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:427022 [7] NCCL INFO Channel 16/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:427020 [1] NCCL INFO Channel 15/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:427025 [3] NCCL INFO Channel 20/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:427024 [2] NCCL INFO Channel 11/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:427021 [5] NCCL INFO Channel 09/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:427007 [6] NCCL INFO Channel 10/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:427020 [1] NCCL INFO Channel 16/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:427025 [3] NCCL INFO Channel 21/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:427024 [2] NCCL INFO Channel 12/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:427021 [5] NCCL INFO Channel 10/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:427007 [6] NCCL INFO Channel 11/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:427020 [1] NCCL INFO Channel 17/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:427025 [3] NCCL INFO Channel 22/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:427024 [2] NCCL INFO Channel 13/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:427021 [5] NCCL INFO Channel 11/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:427007 [6] NCCL INFO Channel 12/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:427022 [7] NCCL INFO Channel 17/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:427020 [1] NCCL INFO Channel 18/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:427025 [3] NCCL INFO Channel 23/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:427024 [2] NCCL INFO Channel 14/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:427021 [5] NCCL INFO Channel 12/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:427007 [6] NCCL INFO Channel 13/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:427022 [7] NCCL INFO Channel 18/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:427020 [1] NCCL INFO Channel 19/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:427024 [2] NCCL INFO Channel 15/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:427021 [5] NCCL INFO Channel 13/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:427007 [6] NCCL INFO Channel 14/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:427022 [7] NCCL INFO Channel 19/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:427020 [1] NCCL INFO Channel 20/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:427024 [2] NCCL INFO Channel 16/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:427021 [5] NCCL INFO Channel 14/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:427007 [6] NCCL INFO Channel 15/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:427022 [7] NCCL INFO Channel 20/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:427020 [1] NCCL INFO Channel 21/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:427024 [2] NCCL INFO Channel 17/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:427021 [5] NCCL INFO Channel 15/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:427007 [6] NCCL INFO Channel 16/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:427022 [7] NCCL INFO Channel 21/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:427020 [1] NCCL INFO Channel 22/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:427024 [2] NCCL INFO Channel 18/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:427021 [5] NCCL INFO Channel 16/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:427007 [6] NCCL INFO Channel 17/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:427022 [7] NCCL INFO Channel 22/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:427020 [1] NCCL INFO Channel 23/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:427024 [2] NCCL INFO Channel 19/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:427021 [5] NCCL INFO Channel 17/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:427007 [6] NCCL INFO Channel 18/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:427022 [7] NCCL INFO Channel 23/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:427024 [2] NCCL INFO Channel 20/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:427007 [6] NCCL INFO Channel 19/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:427024 [2] NCCL INFO Channel 21/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:427007 [6] NCCL INFO Channel 20/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:427021 [5] NCCL INFO Channel 18/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:427024 [2] NCCL INFO Channel 22/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:427007 [6] NCCL INFO Channel 21/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:427023 [4] NCCL INFO Channel 00/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:427024 [2] NCCL INFO Channel 23/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:427021 [5] NCCL INFO Channel 19/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:427023 [4] NCCL INFO Channel 01/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:427007 [6] NCCL INFO Channel 22/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:427021 [5] NCCL INFO Channel 20/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:427023 [4] NCCL INFO Channel 02/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:427007 [6] NCCL INFO Channel 23/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:427023 [4] NCCL INFO Channel 03/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:427021 [5] NCCL INFO Channel 21/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:427023 [4] NCCL INFO Channel 04/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:427023 [4] NCCL INFO Channel 05/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:427021 [5] NCCL INFO Channel 22/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:427023 [4] NCCL INFO Channel 06/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:427021 [5] NCCL INFO Channel 23/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:427023 [4] NCCL INFO Channel 07/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:427023 [4] NCCL INFO Channel 08/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:427023 [4] NCCL INFO Channel 09/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:427023 [4] NCCL INFO Channel 10/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:427023 [4] NCCL INFO Channel 11/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:427023 [4] NCCL INFO Channel 12/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:427023 [4] NCCL INFO Channel 13/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:427023 [4] NCCL INFO Channel 14/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:427023 [4] NCCL INFO Channel 15/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:427023 [4] NCCL INFO Channel 16/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:427023 [4] NCCL INFO Channel 17/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:427023 [4] NCCL INFO Channel 18/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:427023 [4] NCCL INFO Channel 19/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:427023 [4] NCCL INFO Channel 20/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:427023 [4] NCCL INFO Channel 21/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:427023 [4] NCCL INFO Channel 22/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:427023 [4] NCCL INFO Channel 23/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425405:427002 [0] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:425405:427002 [0] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:425405:427002 [0] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:425406:427020 [1] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:425406:427020 [1] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:425406:427020 [1] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:425412:427022 [7] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:425412:427022 [7] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:425412:427022 [7] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:425407:427024 [2] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:425407:427024 [2] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:425407:427024 [2] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:425408:427025 [3] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:425408:427025 [3] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:425408:427025 [3] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:425409:427023 [4] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:425409:427023 [4] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:425409:427023 [4] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:425411:427007 [6] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:425410:427021 [5] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:425411:427007 [6] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:425411:427007 [6] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:425410:427021 [5] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:425410:427021 [5] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:425408:427025 [3] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:425408:427025 [3] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:425408:427025 [3] NCCL INFO ncclCommInitRank comm 0x556a6bee3f60 rank 3 nranks 8 cudaDev 3 nvmlDev 3 busId 201d0 commId 0x42b5c6f4c4b0197a - Init COMPLETE +ywang29-vrdb-test1-worker-0:425412:427022 [7] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:425412:427022 [7] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:425412:427022 [7] NCCL INFO ncclCommInitRank comm 0x561024e243a0 rank 7 nranks 8 cudaDev 7 nvmlDev 7 busId a01d0 commId 0x42b5c6f4c4b0197a - Init COMPLETE +ywang29-vrdb-test1-worker-0:425405:427002 [0] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:425405:427002 [0] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:425405:427002 [0] NCCL INFO ncclCommInitRank comm 0x56302a9f12b0 rank 0 nranks 8 cudaDev 0 nvmlDev 0 busId 101c0 commId 0x42b5c6f4c4b0197a - Init COMPLETE +ywang29-vrdb-test1-worker-0:425407:427024 [2] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:425407:427024 [2] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:425407:427024 [2] NCCL INFO ncclCommInitRank comm 0x55b55b04e0c0 rank 2 nranks 8 cudaDev 2 nvmlDev 2 busId 201c0 commId 0x42b5c6f4c4b0197a - Init COMPLETE +ywang29-vrdb-test1-worker-0:425411:427007 [6] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:425411:427007 [6] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:425411:427007 [6] NCCL INFO ncclCommInitRank comm 0x56163f032c00 rank 6 nranks 8 cudaDev 6 nvmlDev 6 busId a01c0 commId 0x42b5c6f4c4b0197a - Init COMPLETE +ywang29-vrdb-test1-worker-0:425406:427020 [1] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:425406:427020 [1] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:425410:427021 [5] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:425410:427021 [5] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:425406:427020 [1] NCCL INFO ncclCommInitRank comm 0x556b23711fc0 rank 1 nranks 8 cudaDev 1 nvmlDev 1 busId 101d0 commId 0x42b5c6f4c4b0197a - Init COMPLETE +ywang29-vrdb-test1-worker-0:425410:427021 [5] NCCL INFO ncclCommInitRank comm 0x5648ed843ca0 rank 5 nranks 8 cudaDev 5 nvmlDev 5 busId 901d0 commId 0x42b5c6f4c4b0197a - Init COMPLETE +ywang29-vrdb-test1-worker-0:425409:427023 [4] NCCL INFO TUNER/Plugin: Plugin load returned 11 : libnccl-net.so: cannot open shared object file: No such file or directory : when loading libnccl-tuner.so +ywang29-vrdb-test1-worker-0:425409:427023 [4] NCCL INFO TUNER/Plugin: Using internal tuner plugin. +ywang29-vrdb-test1-worker-0:425409:427023 [4] NCCL INFO ncclCommInitRank comm 0x55d67f3afe90 rank 4 nranks 8 cudaDev 4 nvmlDev 4 busId 901c0 commId 0x42b5c6f4c4b0197a - Init COMPLETE +[2025-10-12 15:16:29,999] [INFO] [partition_parameters.py:343:__exit__] finished initializing model - num_params = 459, num_elems = 0.99B +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +Some weights of Qwen2ForCausalLM were not initialized from the model checkpoint at /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model and are newly initialized: ['model.layers.0.mlp.down_proj.scores', 'model.layers.0.mlp.gate_proj.scores', 'model.layers.0.mlp.up_proj.scores', 'model.layers.0.self_attn.k_proj.scores', 'model.layers.0.self_attn.o_proj.scores', 'model.layers.0.self_attn.q_proj.scores', 'model.layers.0.self_attn.v_proj.scores', 'model.layers.1.mlp.down_proj.scores', 'model.layers.1.mlp.gate_proj.scores', 'model.layers.1.mlp.up_proj.scores', 'model.layers.1.self_attn.k_proj.scores', 'model.layers.1.self_attn.o_proj.scores', 'model.layers.1.self_attn.q_proj.scores', 'model.layers.1.self_attn.v_proj.scores', 'model.layers.10.mlp.down_proj.scores', 'model.layers.10.mlp.gate_proj.scores', 'model.layers.10.mlp.up_proj.scores', 'model.layers.10.self_attn.k_proj.scores', 'model.layers.10.self_attn.o_proj.scores', 'model.layers.10.self_attn.q_proj.scores', 'model.layers.10.self_attn.v_proj.scores', 'model.layers.11.mlp.down_proj.scores', 'model.layers.11.mlp.gate_proj.scores', 'model.layers.11.mlp.up_proj.scores', 'model.layers.11.self_attn.k_proj.scores', 'model.layers.11.self_attn.o_proj.scores', 'model.layers.11.self_attn.q_proj.scores', 'model.layers.11.self_attn.v_proj.scores', 'model.layers.12.mlp.down_proj.scores', 'model.layers.12.mlp.gate_proj.scores', 'model.layers.12.mlp.up_proj.scores', 'model.layers.12.self_attn.k_proj.scores', 'model.layers.12.self_attn.o_proj.scores', 'model.layers.12.self_attn.q_proj.scores', 'model.layers.12.self_attn.v_proj.scores', 'model.layers.13.mlp.down_proj.scores', 'model.layers.13.mlp.gate_proj.scores', 'model.layers.13.mlp.up_proj.scores', 'model.layers.13.self_attn.k_proj.scores', 'model.layers.13.self_attn.o_proj.scores', 'model.layers.13.self_attn.q_proj.scores', 'model.layers.13.self_attn.v_proj.scores', 'model.layers.14.mlp.down_proj.scores', 'model.layers.14.mlp.gate_proj.scores', 'model.layers.14.mlp.up_proj.scores', 'model.layers.14.self_attn.k_proj.scores', 'model.layers.14.self_attn.o_proj.scores', 'model.layers.14.self_attn.q_proj.scores', 'model.layers.14.self_attn.v_proj.scores', 'model.layers.15.mlp.down_proj.scores', 'model.layers.15.mlp.gate_proj.scores', 'model.layers.15.mlp.up_proj.scores', 'model.layers.15.self_attn.k_proj.scores', 'model.layers.15.self_attn.o_proj.scores', 'model.layers.15.self_attn.q_proj.scores', 'model.layers.15.self_attn.v_proj.scores', 'model.layers.16.mlp.down_proj.scores', 'model.layers.16.mlp.gate_proj.scores', 'model.layers.16.mlp.up_proj.scores', 'model.layers.16.self_attn.k_proj.scores', 'model.layers.16.self_attn.o_proj.scores', 'model.layers.16.self_attn.q_proj.scores', 'model.layers.16.self_attn.v_proj.scores', 'model.layers.17.mlp.down_proj.scores', 'model.layers.17.mlp.gate_proj.scores', 'model.layers.17.mlp.up_proj.scores', 'model.layers.17.self_attn.k_proj.scores', 'model.layers.17.self_attn.o_proj.scores', 'model.layers.17.self_attn.q_proj.scores', 'model.layers.17.self_attn.v_proj.scores', 'model.layers.18.mlp.down_proj.scores', 'model.layers.18.mlp.gate_proj.scores', 'model.layers.18.mlp.up_proj.scores', 'model.layers.18.self_attn.k_proj.scores', 'model.layers.18.self_attn.o_proj.scores', 'model.layers.18.self_attn.q_proj.scores', 'model.layers.18.self_attn.v_proj.scores', 'model.layers.19.mlp.down_proj.scores', 'model.layers.19.mlp.gate_proj.scores', 'model.layers.19.mlp.up_proj.scores', 'model.layers.19.self_attn.k_proj.scores', 'model.layers.19.self_attn.o_proj.scores', 'model.layers.19.self_attn.q_proj.scores', 'model.layers.19.self_attn.v_proj.scores', 'model.layers.2.mlp.down_proj.scores', 'model.layers.2.mlp.gate_proj.scores', 'model.layers.2.mlp.up_proj.scores', 'model.layers.2.self_attn.k_proj.scores', 'model.layers.2.self_attn.o_proj.scores', 'model.layers.2.self_attn.q_proj.scores', 'model.layers.2.self_attn.v_proj.scores', 'model.layers.20.mlp.down_proj.scores', 'model.layers.20.mlp.gate_proj.scores', 'model.layers.20.mlp.up_proj.scores', 'model.layers.20.self_attn.k_proj.scores', 'model.layers.20.self_attn.o_proj.scores', 'model.layers.20.self_attn.q_proj.scores', 'model.layers.20.self_attn.v_proj.scores', 'model.layers.21.mlp.down_proj.scores', 'model.layers.21.mlp.gate_proj.scores', 'model.layers.21.mlp.up_proj.scores', 'model.layers.21.self_attn.k_proj.scores', 'model.layers.21.self_attn.o_proj.scores', 'model.layers.21.self_attn.q_proj.scores', 'model.layers.21.self_attn.v_proj.scores', 'model.layers.22.mlp.down_proj.scores', 'model.layers.22.mlp.gate_proj.scores', 'model.layers.22.mlp.up_proj.scores', 'model.layers.22.self_attn.k_proj.scores', 'model.layers.22.self_attn.o_proj.scores', 'model.layers.22.self_attn.q_proj.scores', 'model.layers.22.self_attn.v_proj.scores', 'model.layers.23.mlp.down_proj.scores', 'model.layers.23.mlp.gate_proj.scores', 'model.layers.23.mlp.up_proj.scores', 'model.layers.23.self_attn.k_proj.scores', 'model.layers.23.self_attn.o_proj.scores', 'model.layers.23.self_attn.q_proj.scores', 'model.layers.23.self_attn.v_proj.scores', 'model.layers.3.mlp.down_proj.scores', 'model.layers.3.mlp.gate_proj.scores', 'model.layers.3.mlp.up_proj.scores', 'model.layers.3.self_attn.k_proj.scores', 'model.layers.3.self_attn.o_proj.scores', 'model.layers.3.self_attn.q_proj.scores', 'model.layers.3.self_attn.v_proj.scores', 'model.layers.4.mlp.down_proj.scores', 'model.layers.4.mlp.gate_proj.scores', 'model.layers.4.mlp.up_proj.scores', 'model.layers.4.self_attn.k_proj.scores', 'model.layers.4.self_attn.o_proj.scores', 'model.layers.4.self_attn.q_proj.scores', 'model.layers.4.self_attn.v_proj.scores', 'model.layers.5.mlp.down_proj.scores', 'model.layers.5.mlp.gate_proj.scores', 'model.layers.5.mlp.up_proj.scores', 'model.layers.5.self_attn.k_proj.scores', 'model.layers.5.self_attn.o_proj.scores', 'model.layers.5.self_attn.q_proj.scores', 'model.layers.5.self_attn.v_proj.scores', 'model.layers.6.mlp.down_proj.scores', 'model.layers.6.mlp.gate_proj.scores', 'model.layers.6.mlp.up_proj.scores', 'model.layers.6.self_attn.k_proj.scores', 'model.layers.6.self_attn.o_proj.scores', 'model.layers.6.self_attn.q_proj.scores', 'model.layers.6.self_attn.v_proj.scores', 'model.layers.7.mlp.down_proj.scores', 'model.layers.7.mlp.gate_proj.scores', 'model.layers.7.mlp.up_proj.scores', 'model.layers.7.self_attn.k_proj.scores', 'model.layers.7.self_attn.o_proj.scores', 'model.layers.7.self_attn.q_proj.scores', 'model.layers.7.self_attn.v_proj.scores', 'model.layers.8.mlp.down_proj.scores', 'model.layers.8.mlp.gate_proj.scores', 'model.layers.8.mlp.up_proj.scores', 'model.layers.8.self_attn.k_proj.scores', 'model.layers.8.self_attn.o_proj.scores', 'model.layers.8.self_attn.q_proj.scores', 'model.layers.8.self_attn.v_proj.scores', 'model.layers.9.mlp.down_proj.scores', 'model.layers.9.mlp.gate_proj.scores', 'model.layers.9.mlp.up_proj.scores', 'model.layers.9.self_attn.k_proj.scores', 'model.layers.9.self_attn.o_proj.scores', 'model.layers.9.self_attn.q_proj.scores', 'model.layers.9.self_attn.v_proj.scores'] +You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +loading language model from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/language_model +[2025-10-12 15:16:31,736] [INFO] [partition_parameters.py:343:__exit__] finished initializing model - num_params = 907, num_elems = 1.42B +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +Loading vision tower from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/vision_tower +Loading connector from /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain/connector/pytorch_model.bin... +TinyLlavaForConditionalGeneration( + (language_model): Qwen2ForCausalLM( + (model): Qwen2Model( + (embed_tokens): Embedding(151936, 896) + (layers): ModuleList( + (0-23): 24 x Qwen2DecoderLayer( + (self_attn): Qwen2FlashAttention2( + (q_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=True) + (k_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=128, bias=True) + (v_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=128, bias=True) + (o_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=False) + (rotary_emb): Qwen2RotaryEmbedding() + ) + (mlp): Qwen2MLP( + (gate_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=4864, bias=False) + (up_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=4864, bias=False) + (down_proj): SupermaskLinearSparsity_SoftForward_Normal(in_features=4864, out_features=896, bias=False) + (act_fn): SiLU() + ) + (input_layernorm): Qwen2RMSNorm() + (post_attention_layernorm): Qwen2RMSNorm() + ) + ) + (norm): Qwen2RMSNorm() + ) + (lm_head): Linear(in_features=896, out_features=151936, bias=False) + ) + (vision_tower): SIGLIPVisionTower( + (_vision_tower): SiglipVisionModel( + (vision_model): SiglipVisionTransformer( + (embeddings): SiglipVisionEmbeddings( + (patch_embedding): Conv2d(3, 1152, kernel_size=(14, 14), stride=(14, 14), padding=valid) + (position_embedding): Embedding(729, 1152) + ) + (encoder): SiglipEncoder( + (layers): ModuleList( + (0-26): 27 x SiglipEncoderLayer( + (self_attn): SiglipAttention( + (k_proj): Linear(in_features=1152, out_features=1152, bias=True) + (v_proj): Linear(in_features=1152, out_features=1152, bias=True) + (q_proj): Linear(in_features=1152, out_features=1152, bias=True) + (out_proj): Linear(in_features=1152, out_features=1152, bias=True) + ) + (layer_norm1): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + (layer_norm2): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + ) + ) + ) + (post_layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (head): SiglipMultiheadAttentionPoolingHead( + (attention): MultiheadAttention( + (out_proj): NonDynamicallyQuantizableLinear(in_features=1152, out_features=1152, bias=True) + ) + (layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True) + (mlp): SiglipMLP( + (activation_fn): PytorchGELUTanh() + (fc1): Linear(in_features=1152, out_features=4304, bias=True) + (fc2): Linear(in_features=4304, out_features=1152, bias=True) + ) + ) + ) + ) + ) + (connector): MLPConnector( + (_connector): Sequential( + (0): SupermaskLinearSparsity_SoftForward_Normal(in_features=1152, out_features=896, bias=True) + (1): GELU(approximate='none') + (2): SupermaskLinearSparsity_SoftForward_Normal(in_features=896, out_features=896, bias=True) + ) + ) +) +Pre-training init language_model.model.layers.0.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.0.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.0.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.0.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.0.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.0.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.0.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.1.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.2.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.3.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.4.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.5.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.6.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.7.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.8.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.9.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.10.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.11.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.12.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.13.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.14.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.15.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.16.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.17.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.18.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.19.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.20.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.21.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.22.mlp.down_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.self_attn.q_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.self_attn.k_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.self_attn.v_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.self_attn.o_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.mlp.gate_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.mlp.up_proj.scores: Mean=3.000000 +Pre-training init language_model.model.layers.23.mlp.down_proj.scores: Mean=3.000000 +Pre-training init connector._connector.0.scores: Mean=3.000005 +Pre-training init connector._connector.2.scores: Mean=2.999970 +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +Randomly sampled 66529 training samples (10.0% of 665298 total samples) +2025-10-12 15:16:49,583 | INFO: Total Parameters: 1283756736, Total Trainable Parameters: 359661568 +2025-10-12 15:16:49,587 | INFO: Trainable Parameters: +language_model.model.layers.0.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.0.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.0.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.0.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.0.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.0.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.0.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.1.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.1.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.1.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.1.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.1.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.1.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.1.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.2.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.2.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.2.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.2.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.2.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.2.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.2.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.3.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.3.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.3.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.3.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.3.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.3.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.3.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.4.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.4.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.4.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.4.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.4.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.4.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.4.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.5.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.5.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.5.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.5.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.5.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.5.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.5.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.6.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.6.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.6.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.6.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.6.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.6.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.6.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.7.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.7.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.7.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.7.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.7.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.7.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.7.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.8.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.8.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.8.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.8.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.8.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.8.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.8.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.9.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.9.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.9.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.9.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.9.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.9.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.9.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.10.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.10.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.10.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.10.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.10.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.10.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.10.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.11.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.11.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.11.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.11.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.11.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.11.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.11.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.12.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.12.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.12.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.12.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.12.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.12.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.12.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.13.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.13.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.13.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.13.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.13.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.13.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.13.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.14.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.14.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.14.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.14.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.14.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.14.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.14.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.15.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.15.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.15.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.15.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.15.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.15.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.15.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.16.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.16.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.16.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.16.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.16.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.16.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.16.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.17.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.17.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.17.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.17.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.17.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.17.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.17.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.18.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.18.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.18.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.18.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.18.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.18.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.18.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.19.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.19.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.19.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.19.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.19.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.19.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.19.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.20.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.20.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.20.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.20.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.20.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.20.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.20.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.21.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.21.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.21.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.21.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.21.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.21.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.21.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.22.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.22.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.22.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.22.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.22.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.22.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.22.mlp.down_proj.scores: 4358144 parameters +language_model.model.layers.23.self_attn.q_proj.scores: 802816 parameters +language_model.model.layers.23.self_attn.k_proj.scores: 114688 parameters +language_model.model.layers.23.self_attn.v_proj.scores: 114688 parameters +language_model.model.layers.23.self_attn.o_proj.scores: 802816 parameters +language_model.model.layers.23.mlp.gate_proj.scores: 4358144 parameters +language_model.model.layers.23.mlp.up_proj.scores: 4358144 parameters +language_model.model.layers.23.mlp.down_proj.scores: 4358144 parameters +connector._connector.0.scores: 1032192 parameters +connector._connector.2.scores: 802816 parameters +Parameter Offload: Total persistent parameters: 486464 in 403 params + 0%| | 0/520 [00:001->0 [1] 2/-1/-1->1->0 [2] 2/-1/-1->1->0 [3] 2/-1/-1->1->0 [4] 2/-1/-1->1->0 [5] 2/-1/-1->1->0 [6] 2/-1/-1->1->0 [7] 2/-1/-1->1->0 [8] 2/-1/-1->1->0 [9] 2/-1/-1->1->0 [10] 2/-1/-1->1->0 [11] 2/-1/-1->1->0 [12] 2/-1/-1->1->0 [13] 2/-1/-1->1->0 [14] 2/-1/-1->1->0 [15] 2/-1/-1->1->0 [16] 2/-1/-1->1->0 [17] 2/-1/-1->1->0 [18] 2/-1/-1->1->0 [19] 2/-1/-1->1->0 [20] 2/-1/-1->1->0 [21] 2/-1/-1->1->0 [22] 2/-1/-1->1->0 [23] 2/-1/-1->1->0 +ywang29-vrdb-test1-worker-0:425412:431986 [7] NCCL INFO Trees [0] -1/-1/-1->7->6 [1] -1/-1/-1->7->6 [2] -1/-1/-1->7->6 [3] -1/-1/-1->7->6 [4] -1/-1/-1->7->6 [5] -1/-1/-1->7->6 [6] -1/-1/-1->7->6 [7] -1/-1/-1->7->6 [8] -1/-1/-1->7->6 [9] -1/-1/-1->7->6 [10] -1/-1/-1->7->6 [11] -1/-1/-1->7->6 [12] -1/-1/-1->7->6 [13] -1/-1/-1->7->6 [14] -1/-1/-1->7->6 [15] -1/-1/-1->7->6 [16] -1/-1/-1->7->6 [17] -1/-1/-1->7->6 [18] -1/-1/-1->7->6 [19] -1/-1/-1->7->6 [20] -1/-1/-1->7->6 [21] -1/-1/-1->7->6 [22] -1/-1/-1->7->6 [23] -1/-1/-1->7->6 +ywang29-vrdb-test1-worker-0:425409:431988 [4] NCCL INFO Trees [0] 5/-1/-1->4->3 [1] 5/-1/-1->4->3 [2] 5/-1/-1->4->3 [3] 5/-1/-1->4->3 [4] 5/-1/-1->4->3 [5] 5/-1/-1->4->3 [6] 5/-1/-1->4->3 [7] 5/-1/-1->4->3 [8] 5/-1/-1->4->3 [9] 5/-1/-1->4->3 [10] 5/-1/-1->4->3 [11] 5/-1/-1->4->3 [12] 5/-1/-1->4->3 [13] 5/-1/-1->4->3 [14] 5/-1/-1->4->3 [15] 5/-1/-1->4->3 [16] 5/-1/-1->4->3 [17] 5/-1/-1->4->3 [18] 5/-1/-1->4->3 [19] 5/-1/-1->4->3 [20] 5/-1/-1->4->3 [21] 5/-1/-1->4->3 [22] 5/-1/-1->4->3 [23] 5/-1/-1->4->3 +ywang29-vrdb-test1-worker-0:425407:431985 [2] NCCL INFO Trees [0] 3/-1/-1->2->1 [1] 3/-1/-1->2->1 [2] 3/-1/-1->2->1 [3] 3/-1/-1->2->1 [4] 3/-1/-1->2->1 [5] 3/-1/-1->2->1 [6] 3/-1/-1->2->1 [7] 3/-1/-1->2->1 [8] 3/-1/-1->2->1 [9] 3/-1/-1->2->1 [10] 3/-1/-1->2->1 [11] 3/-1/-1->2->1 [12] 3/-1/-1->2->1 [13] 3/-1/-1->2->1 [14] 3/-1/-1->2->1 [15] 3/-1/-1->2->1 [16] 3/-1/-1->2->1 [17] 3/-1/-1->2->1 [18] 3/-1/-1->2->1 [19] 3/-1/-1->2->1 [20] 3/-1/-1->2->1 [21] 3/-1/-1->2->1 [22] 3/-1/-1->2->1 [23] 3/-1/-1->2->1 +ywang29-vrdb-test1-worker-0:425405:431982 [0] NCCL INFO Channel 09/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:425406:431984 [1] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:425408:431987 [3] NCCL INFO Trees [0] 4/-1/-1->3->2 [1] 4/-1/-1->3->2 [2] 4/-1/-1->3->2 [3] 4/-1/-1->3->2 [4] 4/-1/-1->3->2 [5] 4/-1/-1->3->2 [6] 4/-1/-1->3->2 [7] 4/-1/-1->3->2 [8] 4/-1/-1->3->2 [9] 4/-1/-1->3->2 [10] 4/-1/-1->3->2 [11] 4/-1/-1->3->2 [12] 4/-1/-1->3->2 [13] 4/-1/-1->3->2 [14] 4/-1/-1->3->2 [15] 4/-1/-1->3->2 [16] 4/-1/-1->3->2 [17] 4/-1/-1->3->2 [18] 4/-1/-1->3->2 [19] 4/-1/-1->3->2 [20] 4/-1/-1->3->2 [21] 4/-1/-1->3->2 [22] 4/-1/-1->3->2 [23] 4/-1/-1->3->2 +ywang29-vrdb-test1-worker-0:425412:431986 [7] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:425409:431988 [4] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:425407:431985 [2] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:425405:431982 [0] NCCL INFO Channel 10/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:425408:431987 [3] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:425405:431982 [0] NCCL INFO Channel 11/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:425405:431982 [0] NCCL INFO Channel 12/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:425405:431982 [0] NCCL INFO Channel 13/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:425405:431982 [0] NCCL INFO Channel 14/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:425405:431982 [0] NCCL INFO Channel 15/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:425405:431982 [0] NCCL INFO Channel 16/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:425405:431982 [0] NCCL INFO Channel 17/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:425405:431982 [0] NCCL INFO Channel 18/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:425405:431982 [0] NCCL INFO Channel 19/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:425405:431982 [0] NCCL INFO Channel 20/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:425405:431982 [0] NCCL INFO Channel 21/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:425405:431982 [0] NCCL INFO Channel 22/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:425405:431982 [0] NCCL INFO Channel 23/24 : 0 1 2 3 4 5 6 7 +ywang29-vrdb-test1-worker-0:425411:431983 [6] NCCL INFO Trees [0] 7/-1/-1->6->5 [1] 7/-1/-1->6->5 [2] 7/-1/-1->6->5 [3] 7/-1/-1->6->5 [4] 7/-1/-1->6->5 [5] 7/-1/-1->6->5 [6] 7/-1/-1->6->5 [7] 7/-1/-1->6->5 [8] 7/-1/-1->6->5 [9] 7/-1/-1->6->5 [10] 7/-1/-1->6->5 [11] 7/-1/-1->6->5 [12] 7/-1/-1->6->5 [13] 7/-1/-1->6->5 [14] 7/-1/-1->6->5 [15] 7/-1/-1->6->5 [16] 7/-1/-1->6->5 [17] 7/-1/-1->6->5 [18] 7/-1/-1->6->5 [19] 7/-1/-1->6->5 [20] 7/-1/-1->6->5 [21] 7/-1/-1->6->5 [22] 7/-1/-1->6->5 [23] 7/-1/-1->6->5 +ywang29-vrdb-test1-worker-0:425410:431989 [5] NCCL INFO Trees [0] 6/-1/-1->5->4 [1] 6/-1/-1->5->4 [2] 6/-1/-1->5->4 [3] 6/-1/-1->5->4 [4] 6/-1/-1->5->4 [5] 6/-1/-1->5->4 [6] 6/-1/-1->5->4 [7] 6/-1/-1->5->4 [8] 6/-1/-1->5->4 [9] 6/-1/-1->5->4 [10] 6/-1/-1->5->4 [11] 6/-1/-1->5->4 [12] 6/-1/-1->5->4 [13] 6/-1/-1->5->4 [14] 6/-1/-1->5->4 [15] 6/-1/-1->5->4 [16] 6/-1/-1->5->4 [17] 6/-1/-1->5->4 [18] 6/-1/-1->5->4 [19] 6/-1/-1->5->4 [20] 6/-1/-1->5->4 [21] 6/-1/-1->5->4 [22] 6/-1/-1->5->4 [23] 6/-1/-1->5->4 +ywang29-vrdb-test1-worker-0:425410:431989 [5] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:425411:431983 [6] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:425405:431982 [0] NCCL INFO Trees [0] 1/-1/-1->0->-1 [1] 1/-1/-1->0->-1 [2] 1/-1/-1->0->-1 [3] 1/-1/-1->0->-1 [4] 1/-1/-1->0->-1 [5] 1/-1/-1->0->-1 [6] 1/-1/-1->0->-1 [7] 1/-1/-1->0->-1 [8] 1/-1/-1->0->-1 [9] 1/-1/-1->0->-1 [10] 1/-1/-1->0->-1 [11] 1/-1/-1->0->-1 [12] 1/-1/-1->0->-1 [13] 1/-1/-1->0->-1 [14] 1/-1/-1->0->-1 [15] 1/-1/-1->0->-1 [16] 1/-1/-1->0->-1 [17] 1/-1/-1->0->-1 [18] 1/-1/-1->0->-1 [19] 1/-1/-1->0->-1 [20] 1/-1/-1->0->-1 [21] 1/-1/-1->0->-1 [22] 1/-1/-1->0->-1 [23] 1/-1/-1->0->-1 +ywang29-vrdb-test1-worker-0:425405:431982 [0] NCCL INFO P2P Chunksize set to 524288 +ywang29-vrdb-test1-worker-0:425407:431985 [2] NCCL INFO Channel 00/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:431986 [7] NCCL INFO Channel 00/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:431987 [3] NCCL INFO Channel 00/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:431985 [2] NCCL INFO Channel 01/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:431986 [7] NCCL INFO Channel 01/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:431987 [3] NCCL INFO Channel 01/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:431985 [2] NCCL INFO Channel 02/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:431984 [1] NCCL INFO Channel 00/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:431986 [7] NCCL INFO Channel 02/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:431987 [3] NCCL INFO Channel 02/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:431985 [2] NCCL INFO Channel 03/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:431984 [1] NCCL INFO Channel 01/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:431986 [7] NCCL INFO Channel 03/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:431985 [2] NCCL INFO Channel 04/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:431987 [3] NCCL INFO Channel 03/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:431984 [1] NCCL INFO Channel 02/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:431986 [7] NCCL INFO Channel 04/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:431985 [2] NCCL INFO Channel 05/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:431987 [3] NCCL INFO Channel 04/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:431984 [1] NCCL INFO Channel 03/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:431986 [7] NCCL INFO Channel 05/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:431985 [2] NCCL INFO Channel 06/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:431987 [3] NCCL INFO Channel 05/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:431984 [1] NCCL INFO Channel 04/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:431986 [7] NCCL INFO Channel 06/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:431985 [2] NCCL INFO Channel 07/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:431987 [3] NCCL INFO Channel 06/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:431984 [1] NCCL INFO Channel 05/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:431986 [7] NCCL INFO Channel 07/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:431985 [2] NCCL INFO Channel 08/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:431987 [3] NCCL INFO Channel 07/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:431984 [1] NCCL INFO Channel 06/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:431986 [7] NCCL INFO Channel 08/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:431985 [2] NCCL INFO Channel 09/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:431984 [1] NCCL INFO Channel 07/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:431987 [3] NCCL INFO Channel 08/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:431986 [7] NCCL INFO Channel 09/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:431984 [1] NCCL INFO Channel 08/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:431985 [2] NCCL INFO Channel 10/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:431987 [3] NCCL INFO Channel 09/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:431986 [7] NCCL INFO Channel 10/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:431984 [1] NCCL INFO Channel 09/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:431985 [2] NCCL INFO Channel 11/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:431987 [3] NCCL INFO Channel 10/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:431986 [7] NCCL INFO Channel 11/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:431984 [1] NCCL INFO Channel 10/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:431985 [2] NCCL INFO Channel 12/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:431987 [3] NCCL INFO Channel 11/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:431986 [7] NCCL INFO Channel 12/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:431984 [1] NCCL INFO Channel 11/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:431985 [2] NCCL INFO Channel 13/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:431987 [3] NCCL INFO Channel 12/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:431986 [7] NCCL INFO Channel 13/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:431984 [1] NCCL INFO Channel 12/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:431985 [2] NCCL INFO Channel 14/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:431987 [3] NCCL INFO Channel 13/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:431986 [7] NCCL INFO Channel 14/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:431984 [1] NCCL INFO Channel 13/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:431985 [2] NCCL INFO Channel 15/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:431987 [3] NCCL INFO Channel 14/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:431986 [7] NCCL INFO Channel 15/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:431984 [1] NCCL INFO Channel 14/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:431985 [2] NCCL INFO Channel 16/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:431987 [3] NCCL INFO Channel 15/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:431986 [7] NCCL INFO Channel 16/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:431984 [1] NCCL INFO Channel 15/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:431985 [2] NCCL INFO Channel 17/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:431987 [3] NCCL INFO Channel 16/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:431986 [7] NCCL INFO Channel 17/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:431984 [1] NCCL INFO Channel 16/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:431985 [2] NCCL INFO Channel 18/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:431987 [3] NCCL INFO Channel 17/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:431986 [7] NCCL INFO Channel 18/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:431984 [1] NCCL INFO Channel 17/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:431985 [2] NCCL INFO Channel 19/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:431987 [3] NCCL INFO Channel 18/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:431986 [7] NCCL INFO Channel 19/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:431984 [1] NCCL INFO Channel 18/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:431985 [2] NCCL INFO Channel 20/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:431987 [3] NCCL INFO Channel 19/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:431986 [7] NCCL INFO Channel 20/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:431984 [1] NCCL INFO Channel 19/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:431985 [2] NCCL INFO Channel 21/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:431987 [3] NCCL INFO Channel 20/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:431986 [7] NCCL INFO Channel 21/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:431984 [1] NCCL INFO Channel 20/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:431985 [2] NCCL INFO Channel 22/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:431987 [3] NCCL INFO Channel 21/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:431986 [7] NCCL INFO Channel 22/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:431984 [1] NCCL INFO Channel 21/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:431985 [2] NCCL INFO Channel 23/0 : 2[2] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:431987 [3] NCCL INFO Channel 22/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:431988 [4] NCCL INFO Channel 00/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:431984 [1] NCCL INFO Channel 22/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:431986 [7] NCCL INFO Channel 23/0 : 7[7] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425405:431982 [0] NCCL INFO Channel 00/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:431987 [3] NCCL INFO Channel 23/0 : 3[3] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:431988 [4] NCCL INFO Channel 01/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:431984 [1] NCCL INFO Channel 23/0 : 1[1] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425405:431982 [0] NCCL INFO Channel 01/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:431988 [4] NCCL INFO Channel 02/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425405:431982 [0] NCCL INFO Channel 02/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:431988 [4] NCCL INFO Channel 03/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425405:431982 [0] NCCL INFO Channel 03/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:431989 [5] NCCL INFO Channel 00/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:431988 [4] NCCL INFO Channel 04/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:431983 [6] NCCL INFO Channel 00/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425405:431982 [0] NCCL INFO Channel 04/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:431989 [5] NCCL INFO Channel 01/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:431988 [4] NCCL INFO Channel 05/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:431983 [6] NCCL INFO Channel 01/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425405:431982 [0] NCCL INFO Channel 05/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:431989 [5] NCCL INFO Channel 02/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:431988 [4] NCCL INFO Channel 06/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:431983 [6] NCCL INFO Channel 02/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425405:431982 [0] NCCL INFO Channel 06/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:431988 [4] NCCL INFO Channel 07/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:431989 [5] NCCL INFO Channel 03/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425405:431982 [0] NCCL INFO Channel 07/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:431983 [6] NCCL INFO Channel 03/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:431988 [4] NCCL INFO Channel 08/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:431989 [5] NCCL INFO Channel 04/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425405:431982 [0] NCCL INFO Channel 08/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:431983 [6] NCCL INFO Channel 04/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:431988 [4] NCCL INFO Channel 09/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:431989 [5] NCCL INFO Channel 05/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425405:431982 [0] NCCL INFO Channel 09/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:431988 [4] NCCL INFO Channel 10/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:431983 [6] NCCL INFO Channel 05/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425405:431982 [0] NCCL INFO Channel 10/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:431989 [5] NCCL INFO Channel 06/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:431988 [4] NCCL INFO Channel 11/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:431983 [6] NCCL INFO Channel 06/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425405:431982 [0] NCCL INFO Channel 11/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:431989 [5] NCCL INFO Channel 07/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425405:431982 [0] NCCL INFO Channel 12/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:431988 [4] NCCL INFO Channel 12/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:431989 [5] NCCL INFO Channel 08/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425405:431982 [0] NCCL INFO Channel 13/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:431988 [4] NCCL INFO Channel 13/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:431983 [6] NCCL INFO Channel 07/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425405:431982 [0] NCCL INFO Channel 14/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:431988 [4] NCCL INFO Channel 14/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:431989 [5] NCCL INFO Channel 09/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:431983 [6] NCCL INFO Channel 08/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425405:431982 [0] NCCL INFO Channel 15/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:431988 [4] NCCL INFO Channel 15/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:431989 [5] NCCL INFO Channel 10/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:431983 [6] NCCL INFO Channel 09/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425405:431982 [0] NCCL INFO Channel 16/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:431988 [4] NCCL INFO Channel 16/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:431989 [5] NCCL INFO Channel 11/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:431983 [6] NCCL INFO Channel 10/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425405:431982 [0] NCCL INFO Channel 17/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:431988 [4] NCCL INFO Channel 17/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:431989 [5] NCCL INFO Channel 12/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:431983 [6] NCCL INFO Channel 11/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425405:431982 [0] NCCL INFO Channel 18/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:431988 [4] NCCL INFO Channel 18/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:431989 [5] NCCL INFO Channel 13/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425405:431982 [0] NCCL INFO Channel 19/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:431989 [5] NCCL INFO Channel 14/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:431988 [4] NCCL INFO Channel 19/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425405:431982 [0] NCCL INFO Channel 20/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:431989 [5] NCCL INFO Channel 15/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:431983 [6] NCCL INFO Channel 12/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:431988 [4] NCCL INFO Channel 20/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425405:431982 [0] NCCL INFO Channel 21/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:431989 [5] NCCL INFO Channel 16/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:431983 [6] NCCL INFO Channel 13/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:431988 [4] NCCL INFO Channel 21/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425405:431982 [0] NCCL INFO Channel 22/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:431989 [5] NCCL INFO Channel 17/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:431983 [6] NCCL INFO Channel 14/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:431988 [4] NCCL INFO Channel 22/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425405:431982 [0] NCCL INFO Channel 23/0 : 0[0] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:431983 [6] NCCL INFO Channel 15/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:431989 [5] NCCL INFO Channel 18/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:431988 [4] NCCL INFO Channel 23/0 : 4[4] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:431983 [6] NCCL INFO Channel 16/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:431989 [5] NCCL INFO Channel 19/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:431983 [6] NCCL INFO Channel 17/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:431989 [5] NCCL INFO Channel 20/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:431983 [6] NCCL INFO Channel 18/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:431989 [5] NCCL INFO Channel 21/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:431983 [6] NCCL INFO Channel 19/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:431989 [5] NCCL INFO Channel 22/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:431983 [6] NCCL INFO Channel 20/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:431989 [5] NCCL INFO Channel 23/0 : 5[5] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:431983 [6] NCCL INFO Channel 21/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:431983 [6] NCCL INFO Channel 22/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:431983 [6] NCCL INFO Channel 23/0 : 6[6] -> 7[7] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:431985 [2] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:425408:431987 [3] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:425406:431984 [1] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:425409:431988 [4] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:425405:431982 [0] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:425412:431986 [7] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:425412:431986 [7] NCCL INFO Channel 00/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:431985 [2] NCCL INFO Channel 00/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:431989 [5] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:425411:431983 [6] NCCL INFO Connected all rings +ywang29-vrdb-test1-worker-0:425407:431985 [2] NCCL INFO Channel 01/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:431986 [7] NCCL INFO Channel 01/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:431985 [2] NCCL INFO Channel 02/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:431986 [7] NCCL INFO Channel 02/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:431985 [2] NCCL INFO Channel 03/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:431985 [2] NCCL INFO Channel 04/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:431985 [2] NCCL INFO Channel 05/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:431985 [2] NCCL INFO Channel 06/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:431985 [2] NCCL INFO Channel 07/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:431985 [2] NCCL INFO Channel 08/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:431985 [2] NCCL INFO Channel 09/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:431986 [7] NCCL INFO Channel 03/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:431985 [2] NCCL INFO Channel 10/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:431986 [7] NCCL INFO Channel 04/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:431985 [2] NCCL INFO Channel 11/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:431986 [7] NCCL INFO Channel 05/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:431985 [2] NCCL INFO Channel 12/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:431986 [7] NCCL INFO Channel 06/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:431985 [2] NCCL INFO Channel 13/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:431986 [7] NCCL INFO Channel 07/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:431985 [2] NCCL INFO Channel 14/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:431986 [7] NCCL INFO Channel 08/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:431985 [2] NCCL INFO Channel 15/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:431988 [4] NCCL INFO Channel 00/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:431986 [7] NCCL INFO Channel 09/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:431985 [2] NCCL INFO Channel 16/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:431988 [4] NCCL INFO Channel 01/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:431986 [7] NCCL INFO Channel 10/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:431985 [2] NCCL INFO Channel 17/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:431985 [2] NCCL INFO Channel 18/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:431985 [2] NCCL INFO Channel 19/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:431988 [4] NCCL INFO Channel 02/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:431985 [2] NCCL INFO Channel 20/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:431988 [4] NCCL INFO Channel 03/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:431985 [2] NCCL INFO Channel 21/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:431988 [4] NCCL INFO Channel 04/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:431985 [2] NCCL INFO Channel 22/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:431988 [4] NCCL INFO Channel 05/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425407:431985 [2] NCCL INFO Channel 23/0 : 2[2] -> 1[1] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:431987 [3] NCCL INFO Channel 00/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:431984 [1] NCCL INFO Channel 00/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:431988 [4] NCCL INFO Channel 06/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:431983 [6] NCCL INFO Channel 00/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:431987 [3] NCCL INFO Channel 01/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:431984 [1] NCCL INFO Channel 01/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:431988 [4] NCCL INFO Channel 07/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:431983 [6] NCCL INFO Channel 01/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:431987 [3] NCCL INFO Channel 02/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:431984 [1] NCCL INFO Channel 02/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:431988 [4] NCCL INFO Channel 08/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:431983 [6] NCCL INFO Channel 02/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:431987 [3] NCCL INFO Channel 03/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:431984 [1] NCCL INFO Channel 03/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:431988 [4] NCCL INFO Channel 09/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:431983 [6] NCCL INFO Channel 03/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:431987 [3] NCCL INFO Channel 04/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:431984 [1] NCCL INFO Channel 04/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:431988 [4] NCCL INFO Channel 10/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:431983 [6] NCCL INFO Channel 04/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:431987 [3] NCCL INFO Channel 05/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:431984 [1] NCCL INFO Channel 05/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:431988 [4] NCCL INFO Channel 11/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:431983 [6] NCCL INFO Channel 05/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:431987 [3] NCCL INFO Channel 06/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:431984 [1] NCCL INFO Channel 06/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:431988 [4] NCCL INFO Channel 12/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:431984 [1] NCCL INFO Channel 07/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:431983 [6] NCCL INFO Channel 06/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:431987 [3] NCCL INFO Channel 07/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:431988 [4] NCCL INFO Channel 13/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:431983 [6] NCCL INFO Channel 07/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:431984 [1] NCCL INFO Channel 08/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:431987 [3] NCCL INFO Channel 08/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:431988 [4] NCCL INFO Channel 14/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:431983 [6] NCCL INFO Channel 08/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:431984 [1] NCCL INFO Channel 09/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:431987 [3] NCCL INFO Channel 09/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:431988 [4] NCCL INFO Channel 15/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:431983 [6] NCCL INFO Channel 09/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:431984 [1] NCCL INFO Channel 10/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:431987 [3] NCCL INFO Channel 10/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:431988 [4] NCCL INFO Channel 16/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:431983 [6] NCCL INFO Channel 10/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:431984 [1] NCCL INFO Channel 11/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:431987 [3] NCCL INFO Channel 11/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:431988 [4] NCCL INFO Channel 17/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:431983 [6] NCCL INFO Channel 11/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:431984 [1] NCCL INFO Channel 12/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:431987 [3] NCCL INFO Channel 12/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:431988 [4] NCCL INFO Channel 18/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:431983 [6] NCCL INFO Channel 12/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:431984 [1] NCCL INFO Channel 13/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:431987 [3] NCCL INFO Channel 13/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:431989 [5] NCCL INFO Channel 00/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:431988 [4] NCCL INFO Channel 19/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:431983 [6] NCCL INFO Channel 13/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:431984 [1] NCCL INFO Channel 14/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:431987 [3] NCCL INFO Channel 14/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:431989 [5] NCCL INFO Channel 01/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:431988 [4] NCCL INFO Channel 20/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:431983 [6] NCCL INFO Channel 14/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:431984 [1] NCCL INFO Channel 15/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:431987 [3] NCCL INFO Channel 15/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:431988 [4] NCCL INFO Channel 21/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:431983 [6] NCCL INFO Channel 15/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:431986 [7] NCCL INFO Channel 11/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:431984 [1] NCCL INFO Channel 16/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:431987 [3] NCCL INFO Channel 16/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:431988 [4] NCCL INFO Channel 22/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:431986 [7] NCCL INFO Channel 12/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:431983 [6] NCCL INFO Channel 16/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:431984 [1] NCCL INFO Channel 17/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:431987 [3] NCCL INFO Channel 17/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425409:431988 [4] NCCL INFO Channel 23/0 : 4[4] -> 3[3] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:431986 [7] NCCL INFO Channel 13/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:431983 [6] NCCL INFO Channel 17/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:431984 [1] NCCL INFO Channel 18/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:431987 [3] NCCL INFO Channel 18/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:431986 [7] NCCL INFO Channel 14/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:431983 [6] NCCL INFO Channel 18/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:431984 [1] NCCL INFO Channel 19/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:431987 [3] NCCL INFO Channel 19/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:431986 [7] NCCL INFO Channel 15/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:431983 [6] NCCL INFO Channel 19/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:431984 [1] NCCL INFO Channel 20/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:431987 [3] NCCL INFO Channel 20/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:431986 [7] NCCL INFO Channel 16/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:431983 [6] NCCL INFO Channel 20/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:431984 [1] NCCL INFO Channel 21/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:431989 [5] NCCL INFO Channel 02/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:431987 [3] NCCL INFO Channel 21/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:431986 [7] NCCL INFO Channel 17/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:431983 [6] NCCL INFO Channel 21/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:431984 [1] NCCL INFO Channel 22/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:431989 [5] NCCL INFO Channel 03/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:431987 [3] NCCL INFO Channel 22/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:431986 [7] NCCL INFO Channel 18/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:431983 [6] NCCL INFO Channel 22/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425406:431984 [1] NCCL INFO Channel 23/0 : 1[1] -> 0[0] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:431989 [5] NCCL INFO Channel 04/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425408:431987 [3] NCCL INFO Channel 23/0 : 3[3] -> 2[2] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:431986 [7] NCCL INFO Channel 19/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425411:431983 [6] NCCL INFO Channel 23/0 : 6[6] -> 5[5] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:431989 [5] NCCL INFO Channel 05/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:431986 [7] NCCL INFO Channel 20/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:431989 [5] NCCL INFO Channel 06/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:431986 [7] NCCL INFO Channel 21/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:431989 [5] NCCL INFO Channel 07/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:431986 [7] NCCL INFO Channel 22/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:431989 [5] NCCL INFO Channel 08/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425412:431986 [7] NCCL INFO Channel 23/0 : 7[7] -> 6[6] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:431989 [5] NCCL INFO Channel 09/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:431989 [5] NCCL INFO Channel 10/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:431989 [5] NCCL INFO Channel 11/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:431989 [5] NCCL INFO Channel 12/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:431989 [5] NCCL INFO Channel 13/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:431989 [5] NCCL INFO Channel 14/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:431989 [5] NCCL INFO Channel 15/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:431989 [5] NCCL INFO Channel 16/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:431989 [5] NCCL INFO Channel 17/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:431989 [5] NCCL INFO Channel 18/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:431989 [5] NCCL INFO Channel 19/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:431989 [5] NCCL INFO Channel 20/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:431989 [5] NCCL INFO Channel 21/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:431989 [5] NCCL INFO Channel 22/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425410:431989 [5] NCCL INFO Channel 23/0 : 5[5] -> 4[4] via P2P/CUMEM/read +ywang29-vrdb-test1-worker-0:425405:431982 [0] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:425405:431982 [0] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:425405:431982 [0] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:425406:431984 [1] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:425406:431984 [1] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:425406:431984 [1] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:425407:431985 [2] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:425407:431985 [2] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:425407:431985 [2] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:425408:431987 [3] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:425408:431987 [3] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:425408:431987 [3] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:425409:431988 [4] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:425409:431988 [4] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:425409:431988 [4] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:425410:431989 [5] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:425410:431989 [5] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:425410:431989 [5] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:425412:431986 [7] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:425411:431983 [6] NCCL INFO Connected all trees +ywang29-vrdb-test1-worker-0:425412:431986 [7] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:425412:431986 [7] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:425411:431983 [6] NCCL INFO threadThresholds 8/8/64 | 64/8/64 | 512 | 512 +ywang29-vrdb-test1-worker-0:425411:431983 [6] NCCL INFO 24 coll channels, 24 collnet channels, 0 nvls channels, 32 p2p channels, 32 p2p channels per peer +ywang29-vrdb-test1-worker-0:425405:431982 [0] NCCL INFO ncclCommInitRank comm 0x7f55dc06a050 rank 0 nranks 8 cudaDev 0 nvmlDev 0 busId 101c0 commId 0xe4cbafd9aaf93f1b - Init COMPLETE +ywang29-vrdb-test1-worker-0:425407:431985 [2] NCCL INFO ncclCommInitRank comm 0x7fab6c06b2b0 rank 2 nranks 8 cudaDev 2 nvmlDev 2 busId 201c0 commId 0xe4cbafd9aaf93f1b - Init COMPLETE +ywang29-vrdb-test1-worker-0:425408:431987 [3] NCCL INFO ncclCommInitRank comm 0x7fe2c406b6f0 rank 3 nranks 8 cudaDev 3 nvmlDev 3 busId 201d0 commId 0xe4cbafd9aaf93f1b - Init COMPLETE +ywang29-vrdb-test1-worker-0:425412:431986 [7] NCCL INFO ncclCommInitRank comm 0x7ff4a406ae60 rank 7 nranks 8 cudaDev 7 nvmlDev 7 busId a01d0 commId 0xe4cbafd9aaf93f1b - Init COMPLETE +ywang29-vrdb-test1-worker-0:425411:431983 [6] NCCL INFO ncclCommInitRank comm 0x7f01b406b090 rank 6 nranks 8 cudaDev 6 nvmlDev 6 busId a01c0 commId 0xe4cbafd9aaf93f1b - Init COMPLETE +ywang29-vrdb-test1-worker-0:425409:431988 [4] NCCL INFO ncclCommInitRank comm 0x7f7d6406a710 rank 4 nranks 8 cudaDev 4 nvmlDev 4 busId 901c0 commId 0xe4cbafd9aaf93f1b - Init COMPLETE +ywang29-vrdb-test1-worker-0:425410:431989 [5] NCCL INFO ncclCommInitRank comm 0x7f2e7406b680 rank 5 nranks 8 cudaDev 5 nvmlDev 5 busId 901d0 commId 0xe4cbafd9aaf93f1b - Init COMPLETE +ywang29-vrdb-test1-worker-0:425406:431984 [1] NCCL INFO ncclCommInitRank comm 0x7f436006add0 rank 1 nranks 8 cudaDev 1 nvmlDev 1 busId 101d0 commId 0xe4cbafd9aaf93f1b - Init COMPLETE + 0%| | 1/520 [00:14<2:02:51, 14.20s/it] {'loss': 2.1086, 'grad_norm': 0.031194151500811007, 'learning_rate': 0.0125, 'epoch': 0.0} + 0%| | 1/520 [00:14<2:02:51, 14.20s/it] 0%| | 2/520 [00:18<1:10:23, 8.15s/it] {'loss': 2.1045, 'grad_norm': 0.03350677232581464, 'learning_rate': 0.025, 'epoch': 0.0} + 0%| | 2/520 [00:18<1:10:23, 8.15s/it] 1%| | 3/520 [00:22<53:32, 6.21s/it] {'loss': 1.6703, 'grad_norm': 0.013632152887619455, 'learning_rate': 0.037500000000000006, 'epoch': 0.01} + 1%| | 3/520 [00:22<53:32, 6.21s/it] 1%| | 4/520 [00:25<45:35, 5.30s/it] {'loss': 1.5811, 'grad_norm': 0.005704895026833276, 'learning_rate': 0.05, 'epoch': 0.01} + 1%| | 4/520 [00:25<45:35, 5.30s/it] 1%| | 5/520 [00:29<41:10, 4.80s/it] {'loss': 1.5997, 'grad_norm': 0.006628386043842916, 'learning_rate': 0.0625, 'epoch': 0.01} + 1%| | 5/520 [00:29<41:10, 4.80s/it] 1%| | 6/520 [00:33<38:30, 4.50s/it] {'loss': 1.39, 'grad_norm': 0.004477106105781777, 'learning_rate': 0.07500000000000001, 'epoch': 0.01} + 1%| | 6/520 [00:33<38:30, 4.50s/it] 1%|▏ | 7/520 [00:37<36:43, 4.29s/it] {'loss': 1.4133, 'grad_norm': 0.004718217292745684, 'learning_rate': 0.08750000000000001, 'epoch': 0.01} + 1%|▏ | 7/520 [00:37<36:43, 4.29s/it] 2%|▏ | 8/520 [00:42<37:15, 4.37s/it] {'loss': 1.4713, 'grad_norm': 0.003927186464134976, 'learning_rate': 0.1, 'epoch': 0.02} + 2%|▏ | 8/520 [00:42<37:15, 4.37s/it] 2%|▏ | 9/520 [00:46<36:57, 4.34s/it] {'loss': 1.534, 'grad_norm': 0.00478106272740763, 'learning_rate': 0.1125, 'epoch': 0.02} + 2%|▏ | 9/520 [00:46<36:57, 4.34s/it] 2%|▏ | 10/520 [00:50<35:04, 4.13s/it] {'loss': 1.3497, 'grad_norm': 0.004166229934797694, 'learning_rate': 0.125, 'epoch': 0.02} + 2%|▏ | 10/520 [00:50<35:04, 4.13s/it] 2%|▏ | 11/520 [00:53<34:07, 4.02s/it] {'loss': 1.4202, 'grad_norm': 0.0029237183264386657, 'learning_rate': 0.1375, 'epoch': 0.02} + 2%|▏ | 11/520 [00:53<34:07, 4.02s/it] 2%|▏ | 12/520 [00:57<33:12, 3.92s/it] {'loss': 1.3545, 'grad_norm': 0.003126621788515175, 'learning_rate': 0.15000000000000002, 'epoch': 0.02} + 2%|▏ | 12/520 [00:57<33:12, 3.92s/it][2025-10-12 15:17:55,911] [WARNING] [stage3.py:2069:step] 1 pytorch allocator cache flushes since last step. this happens when there is high memory pressure and is detrimental to performance. if this is happening frequently consider adjusting settings to reduce memory consumption. If you are unable to make the cache flushes go away consider adding get_accelerator().empty_cache() calls in your training loop to ensure that all ranks flush their caches at the same time + 2%|▎ | 13/520 [01:01<34:20, 4.06s/it] {'loss': 1.394, 'grad_norm': 0.0036510629579226315, 'learning_rate': 0.1625, 'epoch': 0.03} + 2%|▎ | 13/520 [01:01<34:20, 4.06s/it] 3%|▎ | 14/520 [01:05<33:23, 3.96s/it] {'loss': 1.4309, 'grad_norm': 0.002734431599977698, 'learning_rate': 0.17500000000000002, 'epoch': 0.03} + 3%|▎ | 14/520 [01:05<33:23, 3.96s/it] 3%|▎ | 15/520 [01:09<32:34, 3.87s/it] {'loss': 1.4133, 'grad_norm': 0.002428924506025619, 'learning_rate': 0.1875, 'epoch': 0.03} + 3%|▎ | 15/520 [01:09<32:34, 3.87s/it] 3%|▎ | 16/520 [01:12<31:58, 3.81s/it] {'loss': 1.3743, 'grad_norm': 0.0026261753603748182, 'learning_rate': 0.2, 'epoch': 0.03} + 3%|▎ | 16/520 [01:12<31:58, 3.81s/it] 3%|▎ | 17/520 [01:16<31:35, 3.77s/it] {'loss': 1.4846, 'grad_norm': 0.002480145816955923, 'learning_rate': 0.1999980572931538, 'epoch': 0.03} + 3%|▎ | 17/520 [01:16<31:35, 3.77s/it] 3%|▎ | 18/520 [01:20<31:25, 3.76s/it] {'loss': 1.357, 'grad_norm': 0.002975212563937178, 'learning_rate': 0.19999222924809748, 'epoch': 0.03} + 3%|▎ | 18/520 [01:20<31:25, 3.76s/it] 4%|▎ | 19/520 [01:24<31:09, 3.73s/it] {'loss': 1.3871, 'grad_norm': 0.0025422452658178164, 'learning_rate': 0.19998251609127465, 'epoch': 0.04} + 4%|▎ | 19/520 [01:24<31:09, 3.73s/it] 4%|▍ | 20/520 [01:27<31:01, 3.72s/it] {'loss': 1.3474, 'grad_norm': 0.002754843155179037, 'learning_rate': 0.19996891820008164, 'epoch': 0.04} + 4%|▍ | 20/520 [01:27<31:01, 3.72s/it] 4%|▍ | 21/520 [01:31<30:52, 3.71s/it] {'loss': 1.3749, 'grad_norm': 0.0027863011007674297, 'learning_rate': 0.19995143610285276, 'epoch': 0.04} + 4%|▍ | 21/520 [01:31<30:52, 3.71s/it] 4%|▍ | 22/520 [01:35<30:45, 3.71s/it] {'loss': 1.4918, 'grad_norm': 0.0022125951404648116, 'learning_rate': 0.19993007047883987, 'epoch': 0.04} + 4%|▍ | 22/520 [01:35<30:45, 3.71s/it] 4%|▍ | 23/520 [01:38<30:47, 3.72s/it] {'loss': 1.431, 'grad_norm': 0.0022510703696333114, 'learning_rate': 0.1999048221581858, 'epoch': 0.04} + 4%|▍ | 23/520 [01:38<30:47, 3.72s/it] 5%|▍ | 24/520 [01:42<30:36, 3.70s/it] {'loss': 1.3809, 'grad_norm': 0.0024636660511494326, 'learning_rate': 0.19987569212189224, 'epoch': 0.05} + 5%|▍ | 24/520 [01:42<30:36, 3.70s/it] 5%|▍ | 25/520 [01:46<30:26, 3.69s/it] {'loss': 1.4333, 'grad_norm': 0.002548065164074532, 'learning_rate': 0.19984268150178167, 'epoch': 0.05} + 5%|▍ | 25/520 [01:46<30:26, 3.69s/it] 5%|▌ | 26/520 [01:49<30:31, 3.71s/it] {'loss': 1.4142, 'grad_norm': 0.0022188609411721025, 'learning_rate': 0.1998057915804532, 'epoch': 0.05} + 5%|▌ | 26/520 [01:49<30:31, 3.71s/it] 5%|▌ | 27/520 [01:53<30:22, 3.70s/it] {'loss': 1.3379, 'grad_norm': 0.0025281564173845797, 'learning_rate': 0.1997650237912329, 'epoch': 0.05} + 5%|▌ | 27/520 [01:53<30:22, 3.70s/it] 5%|▌ | 28/520 [01:57<30:18, 3.70s/it] {'loss': 1.3491, 'grad_norm': 0.0023264049769110627, 'learning_rate': 0.199720379718118, 'epoch': 0.05} + 5%|▌ | 28/520 [01:57<30:18, 3.70s/it] 6%|▌ | 29/520 [02:01<30:14, 3.70s/it] {'loss': 1.3645, 'grad_norm': 0.00225236999006689, 'learning_rate': 0.19967186109571552, 'epoch': 0.06} + 6%|▌ | 29/520 [02:01<30:14, 3.70s/it] 6%|▌ | 30/520 [02:04<30:05, 3.68s/it] {'loss': 1.4651, 'grad_norm': 0.0023336005553027646, 'learning_rate': 0.19961946980917455, 'epoch': 0.06} + 6%|▌ | 30/520 [02:04<30:05, 3.68s/it] 6%|▌ | 31/520 [02:08<29:56, 3.67s/it] {'loss': 1.3456, 'grad_norm': 0.0020155419843538567, 'learning_rate': 0.1995632078941134, 'epoch': 0.06} + 6%|▌ | 31/520 [02:08<29:56, 3.67s/it] 6%|▌ | 32/520 [02:12<29:54, 3.68s/it] {'loss': 1.3225, 'grad_norm': 0.0025804372269817057, 'learning_rate': 0.19950307753654017, 'epoch': 0.06} + 6%|▌ | 32/520 [02:12<29:54, 3.68s/it] 6%|▋ | 33/520 [02:15<29:53, 3.68s/it] {'loss': 1.3581, 'grad_norm': 0.0023059876442798055, 'learning_rate': 0.19943908107276798, 'epoch': 0.06} + 6%|▋ | 33/520 [02:15<29:53, 3.68s/it] 7%|▋ | 34/520 [02:19<29:46, 3.68s/it] {'loss': 1.3484, 'grad_norm': 0.002699786527760384, 'learning_rate': 0.19937122098932428, 'epoch': 0.07} + 7%|▋ | 34/520 [02:19<29:46, 3.68s/it] 7%|▋ | 35/520 [02:23<29:40, 3.67s/it] {'loss': 1.362, 'grad_norm': 0.0024381658958004377, 'learning_rate': 0.19929949992285395, 'epoch': 0.07} + 7%|▋ | 35/520 [02:23<29:40, 3.67s/it] 7%|▋ | 36/520 [02:26<29:36, 3.67s/it] {'loss': 1.4727, 'grad_norm': 0.0022842027993470876, 'learning_rate': 0.19922392066001723, 'epoch': 0.07} + 7%|▋ | 36/520 [02:26<29:36, 3.67s/it] 7%|▋ | 37/520 [02:30<29:30, 3.67s/it] {'loss': 1.4524, 'grad_norm': 0.0024077890619267905, 'learning_rate': 0.19914448613738106, 'epoch': 0.07} + 7%|▋ | 37/520 [02:30<29:30, 3.67s/it] 7%|▋ | 38/520 [02:34<29:26, 3.66s/it] {'loss': 1.5449, 'grad_norm': 0.0021698896337111885, 'learning_rate': 0.1990611994413053, 'epoch': 0.07} + 7%|▋ | 38/520 [02:34<29:26, 3.66s/it] 8%|▊ | 39/520 [02:37<29:23, 3.67s/it] {'loss': 1.3938, 'grad_norm': 0.002738854026074265, 'learning_rate': 0.19897406380782262, 'epoch': 0.07} + 8%|▊ | 39/520 [02:37<29:23, 3.67s/it] 8%|▊ | 40/520 [02:41<29:22, 3.67s/it] {'loss': 1.4271, 'grad_norm': 0.002143977141375868, 'learning_rate': 0.19888308262251286, 'epoch': 0.08} + 8%|▊ | 40/520 [02:41<29:22, 3.67s/it] 8%|▊ | 41/520 [02:45<29:17, 3.67s/it] {'loss': 1.3989, 'grad_norm': 0.0025089863981856285, 'learning_rate': 0.19878825942037148, 'epoch': 0.08} + 8%|▊ | 41/520 [02:45<29:17, 3.67s/it] 8%|▊ | 42/520 [02:48<29:13, 3.67s/it] {'loss': 1.4264, 'grad_norm': 0.003131228199898253, 'learning_rate': 0.19868959788567211, 'epoch': 0.08} + 8%|▊ | 42/520 [02:48<29:13, 3.67s/it] 8%|▊ | 43/520 [02:52<29:18, 3.69s/it] {'loss': 1.3706, 'grad_norm': 0.00293633426849588, 'learning_rate': 0.1985871018518236, 'epoch': 0.08} + 8%|▊ | 43/520 [02:52<29:18, 3.69s/it] 8%|▊ | 44/520 [02:56<29:19, 3.70s/it] {'loss': 1.4597, 'grad_norm': 0.002194054829938838, 'learning_rate': 0.19848077530122082, 'epoch': 0.08} + 8%|▊ | 44/520 [02:56<29:19, 3.70s/it] 9%|▊ | 45/520 [02:59<29:11, 3.69s/it] {'loss': 1.435, 'grad_norm': 0.0024639451469707038, 'learning_rate': 0.19837062236509015, 'epoch': 0.09} + 9%|▊ | 45/520 [02:59<29:11, 3.69s/it] 9%|▉ | 46/520 [03:03<29:08, 3.69s/it] {'loss': 1.5253, 'grad_norm': 0.002117769914769972, 'learning_rate': 0.19825664732332884, 'epoch': 0.09} + 9%|▉ | 46/520 [03:03<29:08, 3.69s/it] 9%|▉ | 47/520 [03:07<29:25, 3.73s/it] {'loss': 1.4146, 'grad_norm': 0.0020967218494535637, 'learning_rate': 0.19813885460433878, 'epoch': 0.09} + 9%|▉ | 47/520 [03:07<29:25, 3.73s/it] 9%|▉ | 48/520 [03:11<29:14, 3.72s/it] {'loss': 1.3897, 'grad_norm': 0.0026288979331384275, 'learning_rate': 0.19801724878485438, 'epoch': 0.09} + 9%|▉ | 48/520 [03:11<29:14, 3.72s/it] 9%|▉ | 49/520 [03:14<29:02, 3.70s/it] {'loss': 1.4396, 'grad_norm': 0.0021887914817368035, 'learning_rate': 0.19789183458976486, 'epoch': 0.09} + 9%|▉ | 49/520 [03:14<29:02, 3.70s/it] 10%|▉ | 50/520 [03:18<28:53, 3.69s/it] {'loss': 1.4252, 'grad_norm': 0.002099695888511103, 'learning_rate': 0.19776261689193048, 'epoch': 0.1} + 10%|▉ | 50/520 [03:18<28:53, 3.69s/it] 10%|▉ | 51/520 [03:22<28:48, 3.68s/it] {'loss': 1.358, 'grad_norm': 0.0026651609216704463, 'learning_rate': 0.19762960071199334, 'epoch': 0.1} + 10%|▉ | 51/520 [03:22<28:48, 3.68s/it] 10%|█ | 52/520 [03:25<28:42, 3.68s/it] {'loss': 1.4912, 'grad_norm': 0.0027624159300054917, 'learning_rate': 0.19749279121818236, 'epoch': 0.1} + 10%|█ | 52/520 [03:25<28:42, 3.68s/it] 10%|█ | 53/520 [03:29<28:34, 3.67s/it] {'loss': 1.4804, 'grad_norm': 0.002407045615812515, 'learning_rate': 0.19735219372611235, 'epoch': 0.1} + 10%|█ | 53/520 [03:29<28:34, 3.67s/it] 10%|█ | 54/520 [03:33<28:36, 3.68s/it] {'loss': 1.3843, 'grad_norm': 0.0023407279512545685, 'learning_rate': 0.19720781369857746, 'epoch': 0.1} + 10%|█ | 54/520 [03:33<28:36, 3.68s/it] 11%|█ | 55/520 [03:36<28:37, 3.69s/it] {'loss': 1.3619, 'grad_norm': 0.0023460107324747223, 'learning_rate': 0.1970596567453391, 'epoch': 0.11} + 11%|█ | 55/520 [03:36<28:37, 3.69s/it] 11%|█ | 56/520 [03:40<28:39, 3.71s/it] {'loss': 1.4991, 'grad_norm': 0.0023787349904262314, 'learning_rate': 0.1969077286229078, 'epoch': 0.11} + 11%|█ | 56/520 [03:40<28:39, 3.71s/it] 11%|█ | 57/520 [03:44<28:28, 3.69s/it] {'loss': 1.3538, 'grad_norm': 0.0028859588074901447, 'learning_rate': 0.19675203523431964, 'epoch': 0.11} + 11%|█ | 57/520 [03:44<28:28, 3.69s/it] 11%|█ | 58/520 [03:47<28:22, 3.68s/it] {'loss': 1.5133, 'grad_norm': 0.00199095818526743, 'learning_rate': 0.19659258262890683, 'epoch': 0.11} + 11%|█ | 58/520 [03:47<28:22, 3.68s/it] 11%|█▏ | 59/520 [03:51<28:19, 3.69s/it] {'loss': 1.3428, 'grad_norm': 0.0022668351375148854, 'learning_rate': 0.19642937700206278, 'epoch': 0.11} + 11%|█▏ | 59/520 [03:51<28:19, 3.69s/it] 12%|█▏ | 60/520 [03:55<28:11, 3.68s/it] {'loss': 1.4407, 'grad_norm': 0.002918834873283174, 'learning_rate': 0.19626242469500121, 'epoch': 0.12} + 12%|█▏ | 60/520 [03:55<28:11, 3.68s/it] 12%|█▏ | 61/520 [03:58<28:09, 3.68s/it] {'loss': 1.4681, 'grad_norm': 0.0032819073349473877, 'learning_rate': 0.19609173219450998, 'epoch': 0.12} + 12%|█▏ | 61/520 [03:58<28:09, 3.68s/it] 12%|█▏ | 62/520 [04:02<28:07, 3.68s/it] {'loss': 1.4156, 'grad_norm': 0.002325654056427268, 'learning_rate': 0.19591730613269878, 'epoch': 0.12} + 12%|█▏ | 62/520 [04:02<28:07, 3.68s/it] 12%|█▏ | 63/520 [04:06<27:58, 3.67s/it] {'loss': 1.3934, 'grad_norm': 0.0025516343571187756, 'learning_rate': 0.19573915328674182, 'epoch': 0.12} + 12%|█▏ | 63/520 [04:06<27:58, 3.67s/it] 12%|█▏ | 64/520 [04:09<28:09, 3.71s/it] {'loss': 1.443, 'grad_norm': 0.002327230533832365, 'learning_rate': 0.1955572805786141, 'epoch': 0.12} + 12%|█▏ | 64/520 [04:09<28:09, 3.71s/it] 12%|█▎ | 65/520 [04:13<28:30, 3.76s/it] {'loss': 1.4293, 'grad_norm': 0.002310249105065401, 'learning_rate': 0.1953716950748227, 'epoch': 0.12} + 12%|█▎ | 65/520 [04:13<28:30, 3.76s/it] 13%|█▎ | 66/520 [04:17<28:08, 3.72s/it] {'loss': 1.3918, 'grad_norm': 0.0025025931614612268, 'learning_rate': 0.19518240398613226, 'epoch': 0.13} + 13%|█▎ | 66/520 [04:17<28:08, 3.72s/it] 13%|█▎ | 67/520 [04:21<27:55, 3.70s/it] {'loss': 1.2865, 'grad_norm': 0.0021741282336651437, 'learning_rate': 0.1949894146672846, 'epoch': 0.13} + 13%|█▎ | 67/520 [04:21<27:55, 3.70s/it] 13%|█▎ | 68/520 [04:24<27:46, 3.69s/it] {'loss': 1.3437, 'grad_norm': 0.002142908629602865, 'learning_rate': 0.1947927346167132, 'epoch': 0.13} + 13%|█▎ | 68/520 [04:24<27:46, 3.69s/it] 13%|█▎ | 69/520 [04:28<27:37, 3.68s/it] {'loss': 1.3336, 'grad_norm': 0.0025684386501272985, 'learning_rate': 0.1945923714762516, 'epoch': 0.13} + 13%|█▎ | 69/520 [04:28<27:37, 3.68s/it] 13%|█▎ | 70/520 [04:32<27:37, 3.68s/it] {'loss': 1.3793, 'grad_norm': 0.002657118545495918, 'learning_rate': 0.19438833303083677, 'epoch': 0.13} + 13%|█▎ | 70/520 [04:32<27:37, 3.68s/it] 14%|█▎ | 71/520 [04:35<27:30, 3.67s/it] {'loss': 1.299, 'grad_norm': 0.001964110970905583, 'learning_rate': 0.19418062720820636, 'epoch': 0.14} + 14%|█▎ | 71/520 [04:35<27:30, 3.67s/it] 14%|█▍ | 72/520 [04:39<27:49, 3.73s/it] {'loss': 1.4498, 'grad_norm': 0.0022891622660932416, 'learning_rate': 0.19396926207859086, 'epoch': 0.14} + 14%|█▍ | 72/520 [04:39<27:49, 3.73s/it] 14%|█▍ | 73/520 [04:43<28:02, 3.76s/it] {'loss': 1.2792, 'grad_norm': 0.0021496314719247856, 'learning_rate': 0.19375424585439993, 'epoch': 0.14} + 14%|█▍ | 73/520 [04:43<28:02, 3.76s/it] 14%|█▍ | 74/520 [04:47<28:10, 3.79s/it] {'loss': 1.3924, 'grad_norm': 0.0024081372312723948, 'learning_rate': 0.1935355868899034, 'epoch': 0.14} + 14%|█▍ | 74/520 [04:47<28:10, 3.79s/it] 14%|█▍ | 75/520 [04:51<28:15, 3.81s/it] {'loss': 1.2963, 'grad_norm': 0.0019398370366425597, 'learning_rate': 0.19331329368090666, 'epoch': 0.14} + 14%|█▍ | 75/520 [04:51<28:15, 3.81s/it] 15%|█▍ | 76/520 [04:55<28:15, 3.82s/it] {'loss': 1.4958, 'grad_norm': 0.002654909545910463, 'learning_rate': 0.19308737486442043, 'epoch': 0.15} + 15%|█▍ | 76/520 [04:55<28:15, 3.82s/it] 15%|█▍ | 77/520 [04:58<28:13, 3.82s/it] {'loss': 1.2197, 'grad_norm': 0.002185001716402873, 'learning_rate': 0.19285783921832536, 'epoch': 0.15} + 15%|█▍ | 77/520 [04:58<28:13, 3.82s/it] 15%|█▌ | 78/520 [05:02<28:20, 3.85s/it] {'loss': 1.355, 'grad_norm': 0.0022352637326024877, 'learning_rate': 0.19262469566103088, 'epoch': 0.15} + 15%|█▌ | 78/520 [05:02<28:20, 3.85s/it] 15%|█▌ | 79/520 [05:06<28:15, 3.84s/it] {'loss': 1.3263, 'grad_norm': 0.002248357883765011, 'learning_rate': 0.19238795325112867, 'epoch': 0.15} + 15%|█▌ | 79/520 [05:06<28:15, 3.84s/it] 15%|█▌ | 80/520 [05:10<28:13, 3.85s/it] {'loss': 1.4864, 'grad_norm': 0.0025701985205438776, 'learning_rate': 0.19214762118704076, 'epoch': 0.15} + 15%|█▌ | 80/520 [05:10<28:13, 3.85s/it] 16%|█▌ | 81/520 [05:14<28:06, 3.84s/it] {'loss': 1.4907, 'grad_norm': 0.002827708582024401, 'learning_rate': 0.19190370880666208, 'epoch': 0.16} + 16%|█▌ | 81/520 [05:14<28:06, 3.84s/it] 16%|█▌ | 82/520 [05:18<28:06, 3.85s/it] {'loss': 1.4125, 'grad_norm': 0.002086133636917763, 'learning_rate': 0.19165622558699763, 'epoch': 0.16} + 16%|█▌ | 82/520 [05:18<28:06, 3.85s/it] 16%|█▌ | 83/520 [05:21<27:51, 3.83s/it] {'loss': 1.4175, 'grad_norm': 0.002095657702663538, 'learning_rate': 0.19140518114379435, 'epoch': 0.16} + 16%|█▌ | 83/520 [05:21<27:51, 3.83s/it] 16%|█▌ | 84/520 [05:25<27:34, 3.79s/it] {'loss': 1.4397, 'grad_norm': 0.002264554332940102, 'learning_rate': 0.19115058523116735, 'epoch': 0.16} + 16%|█▌ | 84/520 [05:25<27:34, 3.79s/it] 16%|█▋ | 85/520 [05:29<27:16, 3.76s/it] {'loss': 1.4624, 'grad_norm': 0.001998126576900771, 'learning_rate': 0.1908924477412211, 'epoch': 0.16} + 16%|█▋ | 85/520 [05:29<27:16, 3.76s/it] 17%|█▋ | 86/520 [05:33<27:05, 3.74s/it] {'loss': 1.4634, 'grad_norm': 0.0019852600383214506, 'learning_rate': 0.19063077870366502, 'epoch': 0.17} + 17%|█▋ | 86/520 [05:33<27:05, 3.74s/it] 17%|█▋ | 87/520 [05:36<26:54, 3.73s/it] {'loss': 1.4297, 'grad_norm': 0.0020371619021210203, 'learning_rate': 0.1903655882854237, 'epoch': 0.17} + 17%|█▋ | 87/520 [05:36<26:54, 3.73s/it] 17%|█▋ | 88/520 [05:40<26:48, 3.72s/it] {'loss': 1.378, 'grad_norm': 0.0021780054967076353, 'learning_rate': 0.19009688679024192, 'epoch': 0.17} + 17%|█▋ | 88/520 [05:40<26:48, 3.72s/it] 17%|█▋ | 89/520 [05:44<26:39, 3.71s/it] {'loss': 1.4154, 'grad_norm': 0.0021613896962956584, 'learning_rate': 0.18982468465828442, 'epoch': 0.17} + 17%|█▋ | 89/520 [05:44<26:39, 3.71s/it] 17%|█▋ | 90/520 [05:47<26:29, 3.70s/it] {'loss': 1.3512, 'grad_norm': 0.0020826129480791783, 'learning_rate': 0.1895489924657301, 'epoch': 0.17} + 17%|█▋ | 90/520 [05:47<26:29, 3.70s/it] 18%|█▊ | 91/520 [05:51<26:29, 3.70s/it] {'loss': 1.4257, 'grad_norm': 0.0019745881061446037, 'learning_rate': 0.18926982092436118, 'epoch': 0.17} + 18%|█▊ | 91/520 [05:51<26:29, 3.70s/it] 18%|█▊ | 92/520 [05:55<26:16, 3.68s/it] {'loss': 1.3556, 'grad_norm': 0.002041898936280556, 'learning_rate': 0.18898718088114688, 'epoch': 0.18} + 18%|█▊ | 92/520 [05:55<26:16, 3.68s/it] 18%|█▊ | 93/520 [05:58<26:16, 3.69s/it] {'loss': 1.3694, 'grad_norm': 0.002091050190890678, 'learning_rate': 0.18870108331782218, 'epoch': 0.18} + 18%|█▊ | 93/520 [05:58<26:16, 3.69s/it] 18%|█▊ | 94/520 [06:02<26:12, 3.69s/it] {'loss': 1.462, 'grad_norm': 0.00233640178832006, 'learning_rate': 0.18841153935046098, 'epoch': 0.18} + 18%|█▊ | 94/520 [06:02<26:12, 3.69s/it] 18%|█▊ | 95/520 [06:06<26:02, 3.68s/it] {'loss': 1.3492, 'grad_norm': 0.0024942859089152046, 'learning_rate': 0.18811856022904425, 'epoch': 0.18} + 18%|█▊ | 95/520 [06:06<26:02, 3.68s/it] 18%|█▊ | 96/520 [06:10<26:24, 3.74s/it] {'loss': 1.37, 'grad_norm': 0.001958248771854655, 'learning_rate': 0.18782215733702287, 'epoch': 0.18} + 18%|█▊ | 96/520 [06:10<26:24, 3.74s/it] 19%|█▊ | 97/520 [06:13<26:38, 3.78s/it] {'loss': 1.3332, 'grad_norm': 0.0026114440890294337, 'learning_rate': 0.18752234219087538, 'epoch': 0.19} + 19%|█▊ | 97/520 [06:13<26:38, 3.78s/it] 19%|█▉ | 98/520 [06:17<26:42, 3.80s/it] {'loss': 1.3267, 'grad_norm': 0.0017534684446251767, 'learning_rate': 0.18721912643966054, 'epoch': 0.19} + 19%|█▉ | 98/520 [06:17<26:42, 3.80s/it] 19%|█▉ | 99/520 [06:21<26:27, 3.77s/it] {'loss': 1.3505, 'grad_norm': 0.0022456632054090445, 'learning_rate': 0.18691252186456464, 'epoch': 0.19} + 19%|█▉ | 99/520 [06:21<26:27, 3.77s/it] 19%|█▉ | 100/520 [06:25<26:30, 3.79s/it] {'loss': 1.3644, 'grad_norm': 0.0020139947485549306, 'learning_rate': 0.1866025403784439, 'epoch': 0.19} + 19%|█▉ | 100/520 [06:25<26:30, 3.79s/it] 19%|█▉ | 101/520 [06:28<26:08, 3.74s/it] {'loss': 1.3483, 'grad_norm': 0.0019670642395454387, 'learning_rate': 0.18628919402536132, 'epoch': 0.19} + 19%|█▉ | 101/520 [06:28<26:08, 3.74s/it] 20%|█▉ | 102/520 [06:32<25:50, 3.71s/it] {'loss': 1.362, 'grad_norm': 0.0020337282266623033, 'learning_rate': 0.18597249498011903, 'epoch': 0.2} + 20%|█▉ | 102/520 [06:32<25:50, 3.71s/it] 20%|█▉ | 103/520 [06:36<25:34, 3.68s/it] {'loss': 1.2804, 'grad_norm': 0.0018971254313697792, 'learning_rate': 0.18565245554778517, 'epoch': 0.2} + 20%|█▉ | 103/520 [06:36<25:34, 3.68s/it] 20%|██ | 104/520 [06:39<25:23, 3.66s/it] {'loss': 1.3553, 'grad_norm': 0.0018409269379648491, 'learning_rate': 0.18532908816321558, 'epoch': 0.2} + 20%|██ | 104/520 [06:39<25:23, 3.66s/it] 20%|██ | 105/520 [06:43<25:14, 3.65s/it] {'loss': 1.3422, 'grad_norm': 0.0017967443262210208, 'learning_rate': 0.18500240539057092, 'epoch': 0.2} + 20%|██ | 105/520 [06:43<25:14, 3.65s/it] 20%|██ | 106/520 [06:47<25:13, 3.66s/it] {'loss': 1.3824, 'grad_norm': 0.0019491387457120143, 'learning_rate': 0.18467241992282843, 'epoch': 0.2} + 20%|██ | 106/520 [06:47<25:13, 3.66s/it] 21%|██ | 107/520 [06:50<25:10, 3.66s/it] {'loss': 1.3515, 'grad_norm': 0.0021116576824456545, 'learning_rate': 0.18433914458128858, 'epoch': 0.21} + 21%|██ | 107/520 [06:50<25:10, 3.66s/it] 21%|██ | 108/520 [06:54<25:10, 3.67s/it] {'loss': 1.3115, 'grad_norm': 0.0020115230418772576, 'learning_rate': 0.18400259231507718, 'epoch': 0.21} + 21%|██ | 108/520 [06:54<25:10, 3.67s/it] 21%|██ | 109/520 [06:58<25:06, 3.67s/it] {'loss': 1.3381, 'grad_norm': 0.0017374992559405474, 'learning_rate': 0.18366277620064198, 'epoch': 0.21} + 21%|██ | 109/520 [06:58<25:06, 3.67s/it] 21%|██ | 110/520 [07:01<24:59, 3.66s/it] {'loss': 1.499, 'grad_norm': 0.0020564612266486416, 'learning_rate': 0.1833197094412449, 'epoch': 0.21} + 21%|██ | 110/520 [07:01<24:59, 3.66s/it] 21%|██▏ | 111/520 [07:05<24:57, 3.66s/it] {'loss': 1.4991, 'grad_norm': 0.002115181393110307, 'learning_rate': 0.18297340536644877, 'epoch': 0.21} + 21%|██▏ | 111/520 [07:05<24:57, 3.66s/it] 22%|██▏ | 112/520 [07:09<25:00, 3.68s/it] {'loss': 1.3822, 'grad_norm': 0.0018953511654355645, 'learning_rate': 0.1826238774315995, 'epoch': 0.22} + 22%|██▏ | 112/520 [07:09<25:00, 3.68s/it] 22%|██▏ | 113/520 [07:12<24:55, 3.68s/it] {'loss': 1.2616, 'grad_norm': 0.0018422405821420962, 'learning_rate': 0.18227113921730334, 'epoch': 0.22} + 22%|██▏ | 113/520 [07:12<24:55, 3.68s/it] 22%|██▏ | 114/520 [07:16<24:43, 3.65s/it] {'loss': 1.3547, 'grad_norm': 0.0017280538973868317, 'learning_rate': 0.1819152044288992, 'epoch': 0.22} + 22%|██▏ | 114/520 [07:16<24:43, 3.65s/it] 22%|██▏ | 115/520 [07:20<24:38, 3.65s/it] {'loss': 1.4753, 'grad_norm': 0.0018829754197676636, 'learning_rate': 0.18155608689592603, 'epoch': 0.22} + 22%|██▏ | 115/520 [07:20<24:38, 3.65s/it] 22%|██▏ | 116/520 [07:23<24:34, 3.65s/it] {'loss': 1.4758, 'grad_norm': 0.0018085208975791003, 'learning_rate': 0.18119380057158568, 'epoch': 0.22} + 22%|██▏ | 116/520 [07:23<24:34, 3.65s/it] 22%|██▎ | 117/520 [07:27<24:26, 3.64s/it] {'loss': 1.4509, 'grad_norm': 0.001936185723053909, 'learning_rate': 0.18082835953220056, 'epoch': 0.23} + 22%|██▎ | 117/520 [07:27<24:26, 3.64s/it] 23%|██▎ | 118/520 [07:30<24:19, 3.63s/it] {'loss': 1.334, 'grad_norm': 0.0017421103720308693, 'learning_rate': 0.18045977797666685, 'epoch': 0.23} + 23%|██▎ | 118/520 [07:30<24:19, 3.63s/it] 23%|██▎ | 119/520 [07:34<24:13, 3.63s/it] {'loss': 1.2889, 'grad_norm': 0.0018066449714540516, 'learning_rate': 0.1800880702259028, 'epoch': 0.23} + 23%|██▎ | 119/520 [07:34<24:13, 3.63s/it] 23%|██▎ | 120/520 [07:38<24:10, 3.63s/it] {'loss': 1.3127, 'grad_norm': 0.002266609478845127, 'learning_rate': 0.17971325072229227, 'epoch': 0.23} + 23%|██▎ | 120/520 [07:38<24:10, 3.63s/it] 23%|██▎ | 121/520 [07:41<24:06, 3.62s/it] {'loss': 1.3738, 'grad_norm': 0.0021324305049684384, 'learning_rate': 0.17933533402912352, 'epoch': 0.23} + 23%|██▎ | 121/520 [07:41<24:06, 3.62s/it] 23%|██▎ | 122/520 [07:45<24:07, 3.64s/it] {'loss': 1.2665, 'grad_norm': 0.0018015648795328032, 'learning_rate': 0.17895433483002354, 'epoch': 0.23} + 23%|██▎ | 122/520 [07:45<24:07, 3.64s/it] 24%|██▎ | 123/520 [07:49<24:03, 3.64s/it] {'loss': 1.4315, 'grad_norm': 0.0020567094780874606, 'learning_rate': 0.17857026792838737, 'epoch': 0.24} + 24%|██▎ | 123/520 [07:49<24:03, 3.64s/it] 24%|██▍ | 124/520 [07:52<24:23, 3.70s/it] {'loss': 1.3418, 'grad_norm': 0.0019326362968176065, 'learning_rate': 0.178183148246803, 'epoch': 0.24} + 24%|██▍ | 124/520 [07:52<24:23, 3.70s/it] 24%|██▍ | 125/520 [07:56<24:47, 3.77s/it] {'loss': 1.3362, 'grad_norm': 0.0018941311431476934, 'learning_rate': 0.1777929908264715, 'epoch': 0.24} + 24%|██▍ | 125/520 [07:56<24:47, 3.77s/it] 24%|██▍ | 126/520 [08:01<26:20, 4.01s/it] {'loss': 1.3405, 'grad_norm': 0.0016273272315640765, 'learning_rate': 0.17739981082662276, 'epoch': 0.24} + 24%|██▍ | 126/520 [08:01<26:20, 4.01s/it] 24%|██▍ | 127/520 [08:05<26:04, 3.98s/it] {'loss': 1.3216, 'grad_norm': 0.002215267373451069, 'learning_rate': 0.1770036235239263, 'epoch': 0.24} + 24%|██▍ | 127/520 [08:05<26:04, 3.98s/it] 25%|██▍ | 128/520 [08:09<25:51, 3.96s/it] {'loss': 1.3703, 'grad_norm': 0.0018618749267363894, 'learning_rate': 0.1766044443118978, 'epoch': 0.25} + 25%|██▍ | 128/520 [08:09<25:51, 3.96s/it] 25%|██▍ | 129/520 [08:13<25:38, 3.94s/it] {'loss': 1.3095, 'grad_norm': 0.0017271181017148348, 'learning_rate': 0.17620228870030108, 'epoch': 0.25} + 25%|██▍ | 129/520 [08:13<25:38, 3.94s/it] 25%|██▌ | 130/520 [08:17<25:29, 3.92s/it] {'loss': 1.3507, 'grad_norm': 0.0016417406384430526, 'learning_rate': 0.1757971723145453, 'epoch': 0.25} + 25%|██▌ | 130/520 [08:17<25:29, 3.92s/it] 25%|██▌ | 131/520 [08:20<25:18, 3.90s/it] {'loss': 1.3214, 'grad_norm': 0.0017445035675235994, 'learning_rate': 0.175389110895078, 'epoch': 0.25} + 25%|██▌ | 131/520 [08:20<25:18, 3.90s/it] 25%|██▌ | 132/520 [08:24<25:11, 3.90s/it] {'loss': 1.4039, 'grad_norm': 0.001931087064051963, 'learning_rate': 0.17497812029677343, 'epoch': 0.25} + 25%|██▌ | 132/520 [08:24<25:11, 3.90s/it] 26%|██▌ | 133/520 [08:28<25:01, 3.88s/it] {'loss': 1.3084, 'grad_norm': 0.0019534780588633606, 'learning_rate': 0.17456421648831655, 'epoch': 0.26} + 26%|██▌ | 133/520 [08:28<25:01, 3.88s/it] 26%|██▌ | 134/520 [08:32<24:36, 3.82s/it] {'loss': 1.3867, 'grad_norm': 0.0018792391466392366, 'learning_rate': 0.17414741555158267, 'epoch': 0.26} + 26%|██▌ | 134/520 [08:32<24:36, 3.82s/it] 26%|██▌ | 135/520 [08:35<24:09, 3.76s/it] {'loss': 1.4469, 'grad_norm': 0.0017835801054938611, 'learning_rate': 0.1737277336810124, 'epoch': 0.26} + 26%|██▌ | 135/520 [08:35<24:09, 3.76s/it] 26%|██▌ | 136/520 [08:39<23:52, 3.73s/it] {'loss': 1.3711, 'grad_norm': 0.0017151712116158435, 'learning_rate': 0.17330518718298263, 'epoch': 0.26} + 26%|██▌ | 136/520 [08:39<23:52, 3.73s/it] 26%|██▋ | 137/520 [08:43<23:42, 3.72s/it] {'loss': 1.2984, 'grad_norm': 0.002102177803505444, 'learning_rate': 0.17287979247517285, 'epoch': 0.26} + 26%|██▋ | 137/520 [08:43<23:42, 3.72s/it] 27%|██▋ | 138/520 [08:46<23:33, 3.70s/it] {'loss': 1.3111, 'grad_norm': 0.0016457585268884669, 'learning_rate': 0.17245156608592727, 'epoch': 0.27} + 27%|██▋ | 138/520 [08:46<23:33, 3.70s/it] 27%|██▋ | 139/520 [08:50<23:24, 3.69s/it] {'loss': 1.2171, 'grad_norm': 0.001804619165285778, 'learning_rate': 0.17202052465361267, 'epoch': 0.27} + 27%|██▋ | 139/520 [08:50<23:24, 3.69s/it] 27%|██▋ | 140/520 [08:54<23:20, 3.68s/it] {'loss': 1.3647, 'grad_norm': 0.0016684785893567912, 'learning_rate': 0.17158668492597184, 'epoch': 0.27} + 27%|██▋ | 140/520 [08:54<23:20, 3.68s/it] 27%|██▋ | 141/520 [08:57<23:12, 3.67s/it] {'loss': 1.4235, 'grad_norm': 0.0017220464963872786, 'learning_rate': 0.17115006375947303, 'epoch': 0.27} + 27%|██▋ | 141/520 [08:57<23:12, 3.67s/it] 27%|██▋ | 142/520 [09:01<23:05, 3.67s/it] {'loss': 1.3812, 'grad_norm': 0.0016826047900592898, 'learning_rate': 0.17071067811865476, 'epoch': 0.27} + 27%|██▋ | 142/520 [09:01<23:05, 3.67s/it] 28%|██▊ | 143/520 [09:05<23:02, 3.67s/it] {'loss': 1.3449, 'grad_norm': 0.001914791861649475, 'learning_rate': 0.17026854507546693, 'epoch': 0.28} + 28%|██▊ | 143/520 [09:05<23:02, 3.67s/it] 28%|██▊ | 144/520 [09:08<22:57, 3.66s/it] {'loss': 1.2916, 'grad_norm': 0.0017392820407774983, 'learning_rate': 0.1698236818086073, 'epoch': 0.28} + 28%|██▊ | 144/520 [09:08<22:57, 3.66s/it] 28%|██▊ | 145/520 [09:12<22:59, 3.68s/it] {'loss': 1.2361, 'grad_norm': 0.0016149149822789343, 'learning_rate': 0.16937610560285418, 'epoch': 0.28} + 28%|██▊ | 145/520 [09:12<22:59, 3.68s/it] 28%|██▊ | 146/520 [09:16<22:53, 3.67s/it] {'loss': 1.4441, 'grad_norm': 0.0018096172825527532, 'learning_rate': 0.1689258338483947, 'epoch': 0.28} + 28%|██▊ | 146/520 [09:16<22:53, 3.67s/it] 28%|██▊ | 147/520 [09:19<22:53, 3.68s/it] {'loss': 1.2767, 'grad_norm': 0.0018044578506761283, 'learning_rate': 0.16847288404014937, 'epoch': 0.28} + 28%|██▊ | 147/520 [09:20<22:53, 3.68s/it] 28%|██▊ | 148/520 [09:23<22:50, 3.68s/it] {'loss': 1.3115, 'grad_norm': 0.0017195038909287909, 'learning_rate': 0.16801727377709194, 'epoch': 0.28} + 28%|██▊ | 148/520 [09:23<22:50, 3.68s/it] 29%|██▊ | 149/520 [09:27<22:47, 3.69s/it] {'loss': 1.2579, 'grad_norm': 0.0018026600973313897, 'learning_rate': 0.16755902076156604, 'epoch': 0.29} + 29%|██▊ | 149/520 [09:27<22:47, 3.69s/it] 29%|██▉ | 150/520 [09:31<22:43, 3.68s/it] {'loss': 1.4877, 'grad_norm': 0.0018252571553692071, 'learning_rate': 0.16709814279859703, 'epoch': 0.29} + 29%|██▉ | 150/520 [09:31<22:43, 3.68s/it] 29%|██▉ | 151/520 [09:34<22:39, 3.68s/it] {'loss': 1.2976, 'grad_norm': 0.0016850374846970964, 'learning_rate': 0.1666346577952004, 'epoch': 0.29} + 29%|██▉ | 151/520 [09:34<22:39, 3.68s/it] 29%|██▉ | 152/520 [09:38<22:31, 3.67s/it] {'loss': 1.2705, 'grad_norm': 0.0017964188597106153, 'learning_rate': 0.16616858375968596, 'epoch': 0.29} + 29%|██▉ | 152/520 [09:38<22:31, 3.67s/it] 29%|██▉ | 153/520 [09:42<22:23, 3.66s/it] {'loss': 1.2966, 'grad_norm': 0.001711200279705858, 'learning_rate': 0.16569993880095807, 'epoch': 0.29} + 29%|██▉ | 153/520 [09:42<22:23, 3.66s/it] 30%|██▉ | 154/520 [09:45<22:18, 3.66s/it] {'loss': 1.3971, 'grad_norm': 0.001666935757828226, 'learning_rate': 0.16522874112781213, 'epoch': 0.3} + 30%|██▉ | 154/520 [09:45<22:18, 3.66s/it] 30%|██▉ | 155/520 [09:49<22:12, 3.65s/it] {'loss': 1.3017, 'grad_norm': 0.0017087649869465152, 'learning_rate': 0.16475500904822704, 'epoch': 0.3} + 30%|██▉ | 155/520 [09:49<22:12, 3.65s/it] 30%|███ | 156/520 [09:53<22:16, 3.67s/it] {'loss': 1.3213, 'grad_norm': 0.0018869458781312278, 'learning_rate': 0.16427876096865393, 'epoch': 0.3} + 30%|███ | 156/520 [09:53<22:16, 3.67s/it] 30%|███ | 157/520 [09:56<22:11, 3.67s/it] {'loss': 1.432, 'grad_norm': 0.0018188051464130927, 'learning_rate': 0.16380001539330089, 'epoch': 0.3} + 30%|███ | 157/520 [09:56<22:11, 3.67s/it] 30%|███ | 158/520 [10:00<22:09, 3.67s/it] {'loss': 1.3021, 'grad_norm': 0.0018377255603797265, 'learning_rate': 0.163318790923414, 'epoch': 0.3} + 30%|███ | 158/520 [10:00<22:09, 3.67s/it] 31%|███ | 159/520 [10:04<22:04, 3.67s/it] {'loss': 1.3375, 'grad_norm': 0.0016839844465711543, 'learning_rate': 0.16283510625655473, 'epoch': 0.31} + 31%|███ | 159/520 [10:04<22:04, 3.67s/it] 31%|███ | 160/520 [10:07<21:58, 3.66s/it] {'loss': 1.3625, 'grad_norm': 0.0018332536783403227, 'learning_rate': 0.16234898018587338, 'epoch': 0.31} + 31%|███ | 160/520 [10:07<21:58, 3.66s/it] 31%|███ | 161/520 [10:11<22:04, 3.69s/it] {'loss': 1.3432, 'grad_norm': 0.0017888535262510197, 'learning_rate': 0.16186043159937882, 'epoch': 0.31} + 31%|███ | 161/520 [10:11<22:04, 3.69s/it] 31%|███ | 162/520 [10:15<21:59, 3.69s/it] {'loss': 1.3698, 'grad_norm': 0.0018054455901105363, 'learning_rate': 0.16136947947920477, 'epoch': 0.31} + 31%|███ | 162/520 [10:15<21:59, 3.69s/it] 31%|███▏ | 163/520 [10:18<22:03, 3.71s/it] {'loss': 1.2169, 'grad_norm': 0.001967090788967594, 'learning_rate': 0.16087614290087207, 'epoch': 0.31} + 31%|███▏ | 163/520 [10:18<22:03, 3.71s/it] 32%|███▏ | 164/520 [10:22<21:54, 3.69s/it] {'loss': 1.1856, 'grad_norm': 0.001609718372244046, 'learning_rate': 0.16038044103254775, 'epoch': 0.32} + 32%|███▏ | 164/520 [10:22<21:54, 3.69s/it] 32%|███▏ | 165/520 [10:26<21:58, 3.71s/it] {'loss': 1.3345, 'grad_norm': 0.0016006146374777349, 'learning_rate': 0.15988239313430005, 'epoch': 0.32} + 32%|███▏ | 165/520 [10:26<21:58, 3.71s/it] 32%|███▏ | 166/520 [10:29<21:54, 3.71s/it] {'loss': 1.3255, 'grad_norm': 0.0019070448070864757, 'learning_rate': 0.15938201855735015, 'epoch': 0.32} + 32%|███▏ | 166/520 [10:29<21:54, 3.71s/it] 32%|███▏ | 167/520 [10:33<21:53, 3.72s/it] {'loss': 1.3162, 'grad_norm': 0.0016992626746556822, 'learning_rate': 0.15887933674332047, 'epoch': 0.32} + 32%|███▏ | 167/520 [10:33<21:53, 3.72s/it] 32%|███▏ | 168/520 [10:37<21:46, 3.71s/it] {'loss': 1.2443, 'grad_norm': 0.0017114022519649592, 'learning_rate': 0.158374367223479, 'epoch': 0.32} + 32%|███▏ | 168/520 [10:37<21:46, 3.71s/it] 32%|███▎ | 169/520 [10:41<21:35, 3.69s/it] {'loss': 1.3404, 'grad_norm': 0.0016843608408851967, 'learning_rate': 0.1578671296179806, 'epoch': 0.33} + 32%|███▎ | 169/520 [10:41<21:35, 3.69s/it] 33%|███▎ | 170/520 [10:44<21:30, 3.69s/it] {'loss': 1.3005, 'grad_norm': 0.0015170689110380535, 'learning_rate': 0.15735764363510463, 'epoch': 0.33} + 33%|███▎ | 170/520 [10:44<21:30, 3.69s/it] 33%|███▎ | 171/520 [10:48<21:37, 3.72s/it] {'loss': 1.2656, 'grad_norm': 0.001797796763515021, 'learning_rate': 0.15684592907048925, 'epoch': 0.33} + 33%|███▎ | 171/520 [10:48<21:37, 3.72s/it] 33%|███▎ | 172/520 [10:52<21:29, 3.70s/it] {'loss': 1.3335, 'grad_norm': 0.0017250980541949953, 'learning_rate': 0.1563320058063622, 'epoch': 0.33} + 33%|███▎ | 172/520 [10:52<21:29, 3.70s/it] 33%|███▎ | 173/520 [10:56<21:37, 3.74s/it] {'loss': 1.2624, 'grad_norm': 0.0015543832469232168, 'learning_rate': 0.15581589381076844, 'epoch': 0.33} + 33%|███▎ | 173/520 [10:56<21:37, 3.74s/it] 33%|███▎ | 174/520 [10:59<21:52, 3.79s/it] {'loss': 1.3328, 'grad_norm': 0.0016940390447418856, 'learning_rate': 0.15529761313679394, 'epoch': 0.33} + 33%|███▎ | 174/520 [10:59<21:52, 3.79s/it] 34%|███▎ | 175/520 [11:03<21:58, 3.82s/it] {'loss': 1.2361, 'grad_norm': 0.0015699371006784976, 'learning_rate': 0.15477718392178716, 'epoch': 0.34} + 34%|███▎ | 175/520 [11:03<21:58, 3.82s/it] 34%|███▍ | 176/520 [11:07<21:44, 3.79s/it] {'loss': 1.3909, 'grad_norm': 0.0017017137809183381, 'learning_rate': 0.15425462638657594, 'epoch': 0.34} + 34%|███▍ | 176/520 [11:07<21:44, 3.79s/it] 34%|███▍ | 177/520 [11:11<21:25, 3.75s/it] {'loss': 1.2558, 'grad_norm': 0.0016204196166920333, 'learning_rate': 0.1537299608346824, 'epoch': 0.34} + 34%|███▍ | 177/520 [11:11<21:25, 3.75s/it] 34%|███▍ | 178/520 [11:14<21:13, 3.72s/it] {'loss': 1.3068, 'grad_norm': 0.001744627984328165, 'learning_rate': 0.15320320765153367, 'epoch': 0.34} + 34%|███▍ | 178/520 [11:14<21:13, 3.72s/it] 34%|███▍ | 179/520 [11:18<21:02, 3.70s/it] {'loss': 1.3934, 'grad_norm': 0.00162364365272303, 'learning_rate': 0.15267438730367008, 'epoch': 0.34} + 34%|███▍ | 179/520 [11:18<21:02, 3.70s/it] 35%|███▍ | 180/520 [11:22<20:54, 3.69s/it] {'loss': 1.3071, 'grad_norm': 0.0017451065565429844, 'learning_rate': 0.1521435203379498, 'epoch': 0.35} + 35%|███▍ | 180/520 [11:22<20:54, 3.69s/it] 35%|███▍ | 181/520 [11:25<20:49, 3.69s/it] {'loss': 1.2802, 'grad_norm': 0.0014618349944084436, 'learning_rate': 0.15161062738075068, 'epoch': 0.35} + 35%|███▍ | 181/520 [11:25<20:49, 3.69s/it] 35%|███▌ | 182/520 [11:29<20:45, 3.69s/it] {'loss': 1.2892, 'grad_norm': 0.0016993044833262612, 'learning_rate': 0.1510757291371686, 'epoch': 0.35} + 35%|███▌ | 182/520 [11:29<20:45, 3.69s/it] 35%|███▌ | 183/520 [11:33<21:04, 3.75s/it] {'loss': 1.3191, 'grad_norm': 0.0017052351125989554, 'learning_rate': 0.1505388463902131, 'epoch': 0.35} + 35%|███▌ | 183/520 [11:33<21:04, 3.75s/it] 35%|███▌ | 184/520 [11:37<21:12, 3.79s/it] {'loss': 1.238, 'grad_norm': 0.0017516718543382193, 'learning_rate': 0.15000000000000002, 'epoch': 0.35} + 35%|███▌ | 184/520 [11:37<21:12, 3.79s/it] 36%|███▌ | 185/520 [11:41<21:20, 3.82s/it] {'loss': 1.3984, 'grad_norm': 0.001640263856381053, 'learning_rate': 0.14945921090294076, 'epoch': 0.36} + 36%|███▌ | 185/520 [11:41<21:20, 3.82s/it] 36%|███▌ | 186/520 [11:44<21:07, 3.80s/it] {'loss': 1.2639, 'grad_norm': 0.00165893657030719, 'learning_rate': 0.14891650011092894, 'epoch': 0.36} + 36%|███▌ | 186/520 [11:44<21:07, 3.80s/it] 36%|███▌ | 187/520 [11:48<20:50, 3.75s/it] {'loss': 1.2665, 'grad_norm': 0.0018039325989325557, 'learning_rate': 0.14837188871052398, 'epoch': 0.36} + 36%|███▌ | 187/520 [11:48<20:50, 3.75s/it] 36%|███▌ | 188/520 [11:52<20:45, 3.75s/it] {'loss': 1.3399, 'grad_norm': 0.0017124326854828155, 'learning_rate': 0.14782539786213184, 'epoch': 0.36} + 36%|███▌ | 188/520 [11:52<20:45, 3.75s/it] 36%|███▋ | 189/520 [11:56<20:38, 3.74s/it] {'loss': 1.3553, 'grad_norm': 0.0015712950193803173, 'learning_rate': 0.1472770487991827, 'epoch': 0.36} + 36%|███▋ | 189/520 [11:56<20:38, 3.74s/it] 37%|███▋ | 190/520 [11:59<20:27, 3.72s/it] {'loss': 1.2759, 'grad_norm': 0.0018247451945997719, 'learning_rate': 0.1467268628273062, 'epoch': 0.37} + 37%|███▋ | 190/520 [11:59<20:27, 3.72s/it] 37%|███▋ | 191/520 [12:03<20:19, 3.71s/it] {'loss': 1.2272, 'grad_norm': 0.0015394173491147048, 'learning_rate': 0.1461748613235034, 'epoch': 0.37} + 37%|███▋ | 191/520 [12:03<20:19, 3.71s/it] 37%|███▋ | 192/520 [12:07<20:15, 3.70s/it] {'loss': 1.3119, 'grad_norm': 0.0015935030375967385, 'learning_rate': 0.1456210657353163, 'epoch': 0.37} + 37%|███▋ | 192/520 [12:07<20:15, 3.70s/it] 37%|███▋ | 193/520 [12:10<20:16, 3.72s/it] {'loss': 1.3112, 'grad_norm': 0.0018831022836174117, 'learning_rate': 0.14506549757999454, 'epoch': 0.37} + 37%|███▋ | 193/520 [12:10<20:16, 3.72s/it] 37%|███▋ | 194/520 [12:14<20:08, 3.71s/it] {'loss': 1.1896, 'grad_norm': 0.0015905066631483535, 'learning_rate': 0.14450817844365924, 'epoch': 0.37} + 37%|███▋ | 194/520 [12:14<20:08, 3.71s/it] 38%|███▊ | 195/520 [12:18<19:58, 3.69s/it] {'loss': 1.3298, 'grad_norm': 0.001599174854647489, 'learning_rate': 0.1439491299804645, 'epoch': 0.38} + 38%|███▊ | 195/520 [12:18<19:58, 3.69s/it] 38%|███▊ | 196/520 [12:21<19:51, 3.68s/it] {'loss': 1.3018, 'grad_norm': 0.0017538354227314061, 'learning_rate': 0.14338837391175582, 'epoch': 0.38} + 38%|███▊ | 196/520 [12:21<19:51, 3.68s/it] 38%|███▊ | 197/520 [12:25<19:46, 3.67s/it] {'loss': 1.2536, 'grad_norm': 0.0016336438838674468, 'learning_rate': 0.14282593202522628, 'epoch': 0.38} + 38%|███▊ | 197/520 [12:25<19:46, 3.67s/it] 38%|███▊ | 198/520 [12:29<19:41, 3.67s/it] {'loss': 1.3275, 'grad_norm': 0.0017900678187643575, 'learning_rate': 0.14226182617406996, 'epoch': 0.38} + 38%|███▊ | 198/520 [12:29<19:41, 3.67s/it] 38%|███▊ | 199/520 [12:32<19:37, 3.67s/it] {'loss': 1.2416, 'grad_norm': 0.0017216212440293711, 'learning_rate': 0.14169607827613281, 'epoch': 0.38} + 38%|███▊ | 199/520 [12:32<19:37, 3.67s/it] 38%|███▊ | 200/520 [12:36<19:39, 3.68s/it] {'loss': 1.241, 'grad_norm': 0.0017024396305715427, 'learning_rate': 0.14112871031306118, 'epoch': 0.38} + 38%|███▊ | 200/520 [12:36<19:39, 3.68s/it] 39%|███▊ | 201/520 [12:40<19:42, 3.71s/it] {'loss': 1.2731, 'grad_norm': 0.0014592612573958207, 'learning_rate': 0.1405597443294475, 'epoch': 0.39} + 39%|███▊ | 201/520 [12:40<19:42, 3.71s/it] 39%|███▉ | 202/520 [12:44<19:42, 3.72s/it] {'loss': 1.2429, 'grad_norm': 0.001748916782266096, 'learning_rate': 0.13998920243197407, 'epoch': 0.39} + 39%|███▉ | 202/520 [12:44<19:42, 3.72s/it] 39%|███▉ | 203/520 [12:47<19:42, 3.73s/it] {'loss': 1.2996, 'grad_norm': 0.0017539017847371392, 'learning_rate': 0.13941710678855396, 'epoch': 0.39} + 39%|███▉ | 203/520 [12:47<19:42, 3.73s/it] 39%|███▉ | 204/520 [12:51<19:31, 3.71s/it] {'loss': 1.3163, 'grad_norm': 0.001677311963718997, 'learning_rate': 0.13884347962746948, 'epoch': 0.39} + 39%|███▉ | 204/520 [12:51<19:31, 3.71s/it] 39%|███▉ | 205/520 [12:55<19:22, 3.69s/it] {'loss': 1.2776, 'grad_norm': 0.0015949170025566016, 'learning_rate': 0.138268343236509, 'epoch': 0.39} + 39%|███▉ | 205/520 [12:55<19:22, 3.69s/it] 40%|███▉ | 206/520 [12:58<19:18, 3.69s/it] {'loss': 1.3638, 'grad_norm': 0.0016177407089315386, 'learning_rate': 0.13769171996210053, 'epoch': 0.4} + 40%|███▉ | 206/520 [12:58<19:18, 3.69s/it] 40%|███▉ | 207/520 [13:02<19:09, 3.67s/it] {'loss': 1.2589, 'grad_norm': 0.0016191207292089282, 'learning_rate': 0.1371136322084438, 'epoch': 0.4} + 40%|███▉ | 207/520 [13:02<19:09, 3.67s/it] 40%|████ | 208/520 [13:06<19:11, 3.69s/it] {'loss': 1.3352, 'grad_norm': 0.0018209622032630895, 'learning_rate': 0.13653410243663952, 'epoch': 0.4} + 40%|████ | 208/520 [13:06<19:11, 3.69s/it] 40%|████ | 209/520 [13:09<19:10, 3.70s/it] {'loss': 1.2506, 'grad_norm': 0.0016497328453244553, 'learning_rate': 0.13595315316381676, 'epoch': 0.4} + 40%|████ | 209/520 [13:09<19:10, 3.70s/it] 40%|████ | 210/520 [13:13<19:09, 3.71s/it] {'loss': 1.3283, 'grad_norm': 0.001688703267754755, 'learning_rate': 0.13537080696225814, 'epoch': 0.4} + 40%|████ | 210/520 [13:13<19:09, 3.71s/it] 41%|████ | 211/520 [13:17<19:04, 3.70s/it] {'loss': 1.3362, 'grad_norm': 0.001642369039264989, 'learning_rate': 0.13478708645852272, 'epoch': 0.41} + 41%|████ | 211/520 [13:17<19:04, 3.70s/it] 41%|████ | 212/520 [13:20<18:56, 3.69s/it] {'loss': 1.3184, 'grad_norm': 0.0015942845583275097, 'learning_rate': 0.1342020143325669, 'epoch': 0.41} + 41%|████ | 212/520 [13:20<18:56, 3.69s/it] 41%|████ | 213/520 [13:24<18:50, 3.68s/it] {'loss': 1.2663, 'grad_norm': 0.001876527068057501, 'learning_rate': 0.13361561331686309, 'epoch': 0.41} + 41%|████ | 213/520 [13:24<18:50, 3.68s/it] 41%|████ | 214/520 [13:28<18:48, 3.69s/it] {'loss': 1.2597, 'grad_norm': 0.0016668397150490651, 'learning_rate': 0.13302790619551672, 'epoch': 0.41} + 41%|████ | 214/520 [13:28<18:48, 3.69s/it] 41%|████▏ | 215/520 [13:32<18:42, 3.68s/it] {'loss': 1.2071, 'grad_norm': 0.0015637853610967478, 'learning_rate': 0.1324389158033807, 'epoch': 0.41} + 41%|████▏ | 215/520 [13:32<18:42, 3.68s/it] 42%|████▏ | 216/520 [13:35<18:48, 3.71s/it] {'loss': 1.189, 'grad_norm': 0.0016692784193216782, 'learning_rate': 0.13184866502516845, 'epoch': 0.42} + 42%|████▏ | 216/520 [13:35<18:48, 3.71s/it] 42%|████▏ | 217/520 [13:39<18:55, 3.75s/it] {'loss': 1.3095, 'grad_norm': 0.0016621707104926733, 'learning_rate': 0.13125717679456447, 'epoch': 0.42} + 42%|████▏ | 217/520 [13:39<18:55, 3.75s/it] 42%|████▏ | 218/520 [13:43<18:59, 3.77s/it] {'loss': 1.2968, 'grad_norm': 0.0016913242433944777, 'learning_rate': 0.13066447409333345, 'epoch': 0.42} + 42%|████▏ | 218/520 [13:43<18:59, 3.77s/it] 42%|████▏ | 219/520 [13:47<18:57, 3.78s/it] {'loss': 1.2868, 'grad_norm': 0.0015512884642544392, 'learning_rate': 0.1300705799504273, 'epoch': 0.42} + 42%|████▏ | 219/520 [13:47<18:57, 3.78s/it] 42%|████▏ | 220/520 [13:51<18:57, 3.79s/it] {'loss': 1.2525, 'grad_norm': 0.0015975518583919008, 'learning_rate': 0.12947551744109043, 'epoch': 0.42} + 42%|████▏ | 220/520 [13:51<18:57, 3.79s/it] 42%|████▎ | 221/520 [13:54<18:56, 3.80s/it] {'loss': 1.3003, 'grad_norm': 0.0016467619455485228, 'learning_rate': 0.128879309685963, 'epoch': 0.42} + 42%|████▎ | 221/520 [13:54<18:56, 3.80s/it] 43%|████▎ | 222/520 [13:58<18:57, 3.82s/it] {'loss': 1.2264, 'grad_norm': 0.0016146133918867732, 'learning_rate': 0.12828197985018275, 'epoch': 0.43} + 43%|████▎ | 222/520 [13:58<18:57, 3.82s/it] 43%|████▎ | 223/520 [14:02<18:54, 3.82s/it] {'loss': 1.2177, 'grad_norm': 0.0015000409759335176, 'learning_rate': 0.12768355114248495, 'epoch': 0.43} + 43%|████▎ | 223/520 [14:02<18:54, 3.82s/it] 43%|████▎ | 224/520 [14:06<18:37, 3.78s/it] {'loss': 1.357, 'grad_norm': 0.0015916588310524351, 'learning_rate': 0.12708404681430052, 'epoch': 0.43} + 43%|████▎ | 224/520 [14:06<18:37, 3.78s/it] 43%|████▎ | 225/520 [14:09<18:22, 3.74s/it] {'loss': 1.2285, 'grad_norm': 0.0015899139933749883, 'learning_rate': 0.1264834901588527, 'epoch': 0.43} + 43%|████▎ | 225/520 [14:09<18:22, 3.74s/it] 43%|████▎ | 226/520 [14:13<18:11, 3.71s/it] {'loss': 1.3314, 'grad_norm': 0.0015415380132018362, 'learning_rate': 0.12588190451025208, 'epoch': 0.43} + 43%|████▎ | 226/520 [14:13<18:11, 3.71s/it] 44%|████▎ | 227/520 [14:17<18:03, 3.70s/it] {'loss': 1.3193, 'grad_norm': 0.0014844290907610799, 'learning_rate': 0.12527931324258976, 'epoch': 0.44} + 44%|████▎ | 227/520 [14:17<18:03, 3.70s/it] 44%|████▍ | 228/520 [14:21<18:06, 3.72s/it] {'loss': 1.3761, 'grad_norm': 0.0016200509075400265, 'learning_rate': 0.12467573976902935, 'epoch': 0.44} + 44%|████▍ | 228/520 [14:21<18:06, 3.72s/it] 44%|████▍ | 229/520 [14:24<18:18, 3.77s/it] {'loss': 1.2852, 'grad_norm': 0.0014193984673377727, 'learning_rate': 0.12407120754089732, 'epoch': 0.44} + 44%|████▍ | 229/520 [14:24<18:18, 3.77s/it] 44%|████▍ | 230/520 [14:28<18:28, 3.82s/it] {'loss': 1.1765, 'grad_norm': 0.0015830138029698927, 'learning_rate': 0.12346574004677154, 'epoch': 0.44} + 44%|████▍ | 230/520 [14:28<18:28, 3.82s/it] 44%|████▍ | 231/520 [14:32<18:29, 3.84s/it] {'loss': 1.2386, 'grad_norm': 0.0014231172318313599, 'learning_rate': 0.12285936081156897, 'epoch': 0.44} + 44%|████▍ | 231/520 [14:32<18:29, 3.84s/it] 45%|████▍ | 232/520 [14:36<18:29, 3.85s/it] {'loss': 1.4057, 'grad_norm': 0.0017268123741230428, 'learning_rate': 0.12225209339563144, 'epoch': 0.45} + 45%|████▍ | 232/520 [14:36<18:29, 3.85s/it] 45%|████▍ | 233/520 [14:40<18:28, 3.86s/it] {'loss': 1.2932, 'grad_norm': 0.0017542806180499903, 'learning_rate': 0.12164396139381028, 'epoch': 0.45} + 45%|████▍ | 233/520 [14:40<18:28, 3.86s/it] 45%|████▌ | 234/520 [14:44<18:25, 3.87s/it] {'loss': 1.1827, 'grad_norm': 0.0016346766832677604, 'learning_rate': 0.12103498843454959, 'epoch': 0.45} + 45%|████▌ | 234/520 [14:44<18:25, 3.87s/it] 45%|████▌ | 235/520 [14:48<18:20, 3.86s/it] {'loss': 1.2431, 'grad_norm': 0.0017004512388848578, 'learning_rate': 0.12042519817896805, 'epoch': 0.45} + 45%|████▌ | 235/520 [14:48<18:20, 3.86s/it] 45%|████▌ | 236/520 [14:52<18:18, 3.87s/it] {'loss': 1.3345, 'grad_norm': 0.0015252047457960099, 'learning_rate': 0.11981461431993977, 'epoch': 0.45} + 45%|████▌ | 236/520 [14:52<18:18, 3.87s/it] 46%|████▌ | 237/520 [14:55<18:15, 3.87s/it] {'loss': 1.314, 'grad_norm': 0.0015466345929437625, 'learning_rate': 0.11920326058117364, 'epoch': 0.46} + 46%|████▌ | 237/520 [14:55<18:15, 3.87s/it] 46%|████▌ | 238/520 [14:59<18:13, 3.88s/it] {'loss': 1.251, 'grad_norm': 0.0016330185213755807, 'learning_rate': 0.11859116071629149, 'epoch': 0.46} + 46%|████▌ | 238/520 [14:59<18:13, 3.88s/it] 46%|████▌ | 239/520 [15:03<18:12, 3.89s/it] {'loss': 1.3395, 'grad_norm': 0.0016686785862732182, 'learning_rate': 0.11797833850790528, 'epoch': 0.46} + 46%|████▌ | 239/520 [15:03<18:12, 3.89s/it] 46%|████▌ | 240/520 [15:07<18:08, 3.89s/it] {'loss': 1.1281, 'grad_norm': 0.0015210900572858249, 'learning_rate': 0.11736481776669305, 'epoch': 0.46} + 46%|████▌ | 240/520 [15:07<18:08, 3.89s/it] 46%|████▋ | 241/520 [15:11<18:06, 3.90s/it] {'loss': 1.2135, 'grad_norm': 0.0015655923118755915, 'learning_rate': 0.11675062233047365, 'epoch': 0.46} + 46%|████▋ | 241/520 [15:11<18:06, 3.90s/it] 47%|████▋ | 242/520 [15:15<18:03, 3.90s/it] {'loss': 1.2319, 'grad_norm': 0.0014280892114468206, 'learning_rate': 0.11613577606328068, 'epoch': 0.47} + 47%|████▋ | 242/520 [15:15<18:03, 3.90s/it] 47%|████▋ | 243/520 [15:19<17:58, 3.89s/it] {'loss': 1.2245, 'grad_norm': 0.001535828322184256, 'learning_rate': 0.11552030285443515, 'epoch': 0.47} + 47%|████▋ | 243/520 [15:19<17:58, 3.89s/it] 47%|████▋ | 244/520 [15:23<17:53, 3.89s/it] {'loss': 1.35, 'grad_norm': 0.0016115511551643455, 'learning_rate': 0.11490422661761744, 'epoch': 0.47} + 47%|████▋ | 244/520 [15:23<17:53, 3.89s/it] 47%|████▋ | 245/520 [15:27<17:48, 3.89s/it] {'loss': 1.2058, 'grad_norm': 0.001642930767896609, 'learning_rate': 0.11428757128993801, 'epoch': 0.47} + 47%|████▋ | 245/520 [15:27<17:48, 3.89s/it] 47%|████▋ | 246/520 [15:31<17:47, 3.90s/it] {'loss': 1.3818, 'grad_norm': 0.0017001451696004866, 'learning_rate': 0.11367036083100734, 'epoch': 0.47} + 47%|████▋ | 246/520 [15:31<17:47, 3.90s/it] 48%|████▊ | 247/520 [15:34<17:43, 3.90s/it] {'loss': 1.3831, 'grad_norm': 0.0015486770950833466, 'learning_rate': 0.11305261922200519, 'epoch': 0.47} + 48%|████▊ | 247/520 [15:34<17:43, 3.90s/it] 48%|████▊ | 248/520 [15:38<17:40, 3.90s/it] {'loss': 1.2168, 'grad_norm': 0.0017002134557041353, 'learning_rate': 0.11243437046474854, 'epoch': 0.48} + 48%|████▊ | 248/520 [15:38<17:40, 3.90s/it] 48%|████▊ | 249/520 [15:42<17:20, 3.84s/it] {'loss': 1.3077, 'grad_norm': 0.0015978587217659547, 'learning_rate': 0.1118156385807593, 'epoch': 0.48} + 48%|████▊ | 249/520 [15:42<17:20, 3.84s/it] 48%|████▊ | 250/520 [15:46<17:01, 3.78s/it] {'loss': 1.243, 'grad_norm': 0.0016764082122263772, 'learning_rate': 0.11119644761033079, 'epoch': 0.48} + 48%|████▊ | 250/520 [15:46<17:01, 3.78s/it] 48%|████▊ | 251/520 [15:49<16:48, 3.75s/it] {'loss': 1.3097, 'grad_norm': 0.0014623463851488308, 'learning_rate': 0.1105768216115938, 'epoch': 0.48} + 48%|████▊ | 251/520 [15:49<16:48, 3.75s/it] 48%|████▊ | 252/520 [15:53<16:39, 3.73s/it] {'loss': 1.2814, 'grad_norm': 0.001562328386925392, 'learning_rate': 0.10995678465958168, 'epoch': 0.48} + 48%|████▊ | 252/520 [15:53<16:39, 3.73s/it] 49%|████▊ | 253/520 [15:57<16:28, 3.70s/it] {'loss': 1.3086, 'grad_norm': 0.001631004309435352, 'learning_rate': 0.10933636084529506, 'epoch': 0.49} + 49%|████▊ | 253/520 [15:57<16:28, 3.70s/it] 49%|████▉ | 254/520 [16:00<16:22, 3.69s/it] {'loss': 1.2381, 'grad_norm': 0.0014895893967927652, 'learning_rate': 0.10871557427476584, 'epoch': 0.49} + 49%|████▉ | 254/520 [16:00<16:22, 3.69s/it] 49%|████▉ | 255/520 [16:04<16:17, 3.69s/it] {'loss': 1.2375, 'grad_norm': 0.0016624933583568328, 'learning_rate': 0.10809444906812034, 'epoch': 0.49} + 49%|████▉ | 255/520 [16:04<16:17, 3.69s/it] 49%|████▉ | 256/520 [16:08<16:09, 3.67s/it] {'loss': 1.2881, 'grad_norm': 0.0016141400911425098, 'learning_rate': 0.10747300935864244, 'epoch': 0.49} + 49%|████▉ | 256/520 [16:08<16:09, 3.67s/it] 49%|████▉ | 257/520 [16:11<16:03, 3.66s/it] {'loss': 1.2787, 'grad_norm': 0.0016407178774487641, 'learning_rate': 0.10685127929183567, 'epoch': 0.49} + 49%|████▉ | 257/520 [16:11<16:03, 3.66s/it] 50%|████▉ | 258/520 [16:15<15:59, 3.66s/it] {'loss': 1.2881, 'grad_norm': 0.0014807912420813352, 'learning_rate': 0.10622928302448523, 'epoch': 0.5} + 50%|████▉ | 258/520 [16:15<15:59, 3.66s/it] 50%|████▉ | 259/520 [16:19<15:55, 3.66s/it] {'loss': 1.3632, 'grad_norm': 0.0017914155017553644, 'learning_rate': 0.10560704472371918, 'epoch': 0.5} + 50%|████▉ | 259/520 [16:19<15:55, 3.66s/it] 50%|█████ | 260/520 [16:22<15:51, 3.66s/it] {'loss': 1.3317, 'grad_norm': 0.00157493444164196, 'learning_rate': 0.10498458856606972, 'epoch': 0.5} + 50%|█████ | 260/520 [16:22<15:51, 3.66s/it] 50%|█████ | 261/520 [16:26<15:47, 3.66s/it] {'loss': 1.2717, 'grad_norm': 0.001593151307044813, 'learning_rate': 0.10436193873653361, 'epoch': 0.5} + 50%|█████ | 261/520 [16:26<15:47, 3.66s/it] 50%|█████ | 262/520 [16:30<15:44, 3.66s/it] {'loss': 1.2051, 'grad_norm': 0.0015741863832630094, 'learning_rate': 0.10373911942763259, 'epoch': 0.5} + 50%|█████ | 262/520 [16:30<15:44, 3.66s/it] 51%|█████ | 263/520 [16:33<15:40, 3.66s/it] {'loss': 1.2899, 'grad_norm': 0.001589573785128617, 'learning_rate': 0.10311615483847332, 'epoch': 0.51} + 51%|█████ | 263/520 [16:33<15:40, 3.66s/it] 51%|█████ | 264/520 [16:37<15:35, 3.66s/it] {'loss': 1.3134, 'grad_norm': 0.0015455953978852426, 'learning_rate': 0.1024930691738073, 'epoch': 0.51} + 51%|█████ | 264/520 [16:37<15:35, 3.66s/it] 51%|█████ | 265/520 [16:41<15:34, 3.67s/it] {'loss': 1.2224, 'grad_norm': 0.001740233332255517, 'learning_rate': 0.10186988664309023, 'epoch': 0.51} + 51%|█████ | 265/520 [16:41<15:34, 3.67s/it] 51%|█████ | 266/520 [16:44<15:29, 3.66s/it] {'loss': 1.0846, 'grad_norm': 0.0013933866187401925, 'learning_rate': 0.10124663145954152, 'epoch': 0.51} + 51%|█████ | 266/520 [16:44<15:29, 3.66s/it] 51%|█████▏ | 267/520 [16:48<15:25, 3.66s/it] {'loss': 1.2216, 'grad_norm': 0.0015028445545024337, 'learning_rate': 0.10062332783920336, 'epoch': 0.51} + 51%|█████▏ | 267/520 [16:48<15:25, 3.66s/it] 52%|█████▏ | 268/520 [16:52<15:35, 3.71s/it] {'loss': 1.4011, 'grad_norm': 0.001759912346658886, 'learning_rate': 0.1, 'epoch': 0.52} + 52%|█████▏ | 268/520 [16:52<15:35, 3.71s/it] 52%|█████▏ | 269/520 [16:56<15:44, 3.76s/it] {'loss': 1.323, 'grad_norm': 0.0016512928175357433, 'learning_rate': 0.09937667216079665, 'epoch': 0.52} + 52%|█████▏ | 269/520 [16:56<15:44, 3.76s/it] 52%|█████▏ | 270/520 [17:00<15:49, 3.80s/it] {'loss': 1.212, 'grad_norm': 0.0014734443784700118, 'learning_rate': 0.0987533685404585, 'epoch': 0.52} + 52%|█████▏ | 270/520 [17:00<15:49, 3.80s/it] 52%|█████▏ | 271/520 [17:03<15:49, 3.81s/it] {'loss': 1.3144, 'grad_norm': 0.001606219067880841, 'learning_rate': 0.0981301133569098, 'epoch': 0.52} + 52%|█████▏ | 271/520 [17:03<15:49, 3.81s/it] 52%|█████▏ | 272/520 [17:07<15:48, 3.82s/it] {'loss': 1.2326, 'grad_norm': 0.0015306779189609265, 'learning_rate': 0.09750693082619273, 'epoch': 0.52} + 52%|█████▏ | 272/520 [17:07<15:48, 3.82s/it] 52%|█████▎ | 273/520 [17:11<15:49, 3.84s/it] {'loss': 1.3559, 'grad_norm': 0.0015389748612064872, 'learning_rate': 0.0968838451615267, 'epoch': 0.53} + 52%|█████▎ | 273/520 [17:11<15:49, 3.84s/it] 53%|█████▎ | 274/520 [17:15<15:46, 3.85s/it] {'loss': 1.2756, 'grad_norm': 0.0016966805283094753, 'learning_rate': 0.09626088057236745, 'epoch': 0.53} + 53%|█████▎ | 274/520 [17:15<15:46, 3.85s/it] 53%|█████▎ | 275/520 [17:19<15:44, 3.86s/it] {'loss': 1.2285, 'grad_norm': 0.0016447602955286624, 'learning_rate': 0.09563806126346641, 'epoch': 0.53} + 53%|█████▎ | 275/520 [17:19<15:44, 3.86s/it] 53%|█████▎ | 276/520 [17:23<15:39, 3.85s/it] {'loss': 1.2956, 'grad_norm': 0.0018030286755626886, 'learning_rate': 0.09501541143393027, 'epoch': 0.53} + 53%|█████▎ | 276/520 [17:23<15:39, 3.85s/it] 53%|█████▎ | 277/520 [17:27<15:36, 3.86s/it] {'loss': 1.3499, 'grad_norm': 0.0015571987560019365, 'learning_rate': 0.09439295527628082, 'epoch': 0.53} + 53%|█████▎ | 277/520 [17:27<15:36, 3.86s/it] 53%|█████▎ | 278/520 [17:30<15:36, 3.87s/it] {'loss': 1.1679, 'grad_norm': 0.0014185225574077944, 'learning_rate': 0.0937707169755148, 'epoch': 0.53} + 53%|█████▎ | 278/520 [17:30<15:36, 3.87s/it] 54%|█████▎ | 279/520 [17:34<15:31, 3.86s/it] {'loss': 1.2306, 'grad_norm': 0.001761053705598696, 'learning_rate': 0.09314872070816434, 'epoch': 0.54} + 54%|█████▎ | 279/520 [17:34<15:31, 3.86s/it] 54%|█████▍ | 280/520 [17:38<15:25, 3.85s/it] {'loss': 1.2279, 'grad_norm': 0.0018738775318975119, 'learning_rate': 0.09252699064135758, 'epoch': 0.54} + 54%|█████▍ | 280/520 [17:38<15:25, 3.85s/it] 54%|█████▍ | 281/520 [17:42<15:17, 3.84s/it] {'loss': 1.3248, 'grad_norm': 0.0016475277144549702, 'learning_rate': 0.09190555093187967, 'epoch': 0.54} + 54%|█████▍ | 281/520 [17:42<15:17, 3.84s/it] 54%|█████▍ | 282/520 [17:46<15:13, 3.84s/it] {'loss': 1.1885, 'grad_norm': 0.001465229662081333, 'learning_rate': 0.09128442572523418, 'epoch': 0.54} + 54%|█████▍ | 282/520 [17:46<15:13, 3.84s/it] 54%|█████▍ | 283/520 [17:50<15:10, 3.84s/it] {'loss': 1.3483, 'grad_norm': 0.0017233314545674167, 'learning_rate': 0.09066363915470495, 'epoch': 0.54} + 54%|█████▍ | 283/520 [17:50<15:10, 3.84s/it] 55%|█████▍ | 284/520 [17:53<15:04, 3.83s/it] {'loss': 1.2172, 'grad_norm': 0.00161876726191544, 'learning_rate': 0.09004321534041836, 'epoch': 0.55} + 55%|█████▍ | 284/520 [17:53<15:04, 3.83s/it] 55%|█████▍ | 285/520 [17:57<15:01, 3.84s/it] {'loss': 1.2141, 'grad_norm': 0.0015214354030723904, 'learning_rate': 0.08942317838840624, 'epoch': 0.55} + 55%|█████▍ | 285/520 [17:57<15:01, 3.84s/it] 55%|█████▌ | 286/520 [18:01<14:58, 3.84s/it] {'loss': 1.0952, 'grad_norm': 0.001574707515465826, 'learning_rate': 0.08880355238966922, 'epoch': 0.55} + 55%|█████▌ | 286/520 [18:01<14:58, 3.84s/it] 55%|█████▌ | 287/520 [18:05<14:54, 3.84s/it] {'loss': 1.3274, 'grad_norm': 0.001584649369090583, 'learning_rate': 0.08818436141924073, 'epoch': 0.55} + 55%|█████▌ | 287/520 [18:05<14:54, 3.84s/it] 55%|█████▌ | 288/520 [18:09<14:55, 3.86s/it] {'loss': 1.3618, 'grad_norm': 0.001570974165350949, 'learning_rate': 0.08756562953525152, 'epoch': 0.55} + 55%|█████▌ | 288/520 [18:09<14:55, 3.86s/it] 56%|█████▌ | 289/520 [18:13<14:55, 3.88s/it] {'loss': 1.2276, 'grad_norm': 0.0014674356753106965, 'learning_rate': 0.08694738077799487, 'epoch': 0.56} + 56%|█████▌ | 289/520 [18:13<14:55, 3.88s/it] 56%|█████▌ | 290/520 [18:16<14:36, 3.81s/it] {'loss': 1.1491, 'grad_norm': 0.0014449983903059725, 'learning_rate': 0.08632963916899268, 'epoch': 0.56} + 56%|█████▌ | 290/520 [18:16<14:36, 3.81s/it] 56%|█████▌ | 291/520 [18:20<14:24, 3.77s/it] {'loss': 1.2031, 'grad_norm': 0.0015353997793876096, 'learning_rate': 0.08571242871006202, 'epoch': 0.56} + 56%|█████▌ | 291/520 [18:20<14:24, 3.77s/it] 56%|█████▌ | 292/520 [18:24<14:15, 3.75s/it] {'loss': 1.2517, 'grad_norm': 0.0015187848593020289, 'learning_rate': 0.08509577338238256, 'epoch': 0.56} + 56%|█████▌ | 292/520 [18:24<14:15, 3.75s/it] 56%|█████▋ | 293/520 [18:28<14:09, 3.74s/it] {'loss': 1.196, 'grad_norm': 0.0016375006845282495, 'learning_rate': 0.08447969714556484, 'epoch': 0.56} + 56%|█████▋ | 293/520 [18:28<14:09, 3.74s/it] 57%|█████▋ | 294/520 [18:31<14:19, 3.80s/it] {'loss': 1.2164, 'grad_norm': 0.0016151992203877232, 'learning_rate': 0.08386422393671933, 'epoch': 0.57} + 57%|█████▋ | 294/520 [18:31<14:19, 3.80s/it] 57%|█████▋ | 295/520 [18:35<14:12, 3.79s/it] {'loss': 1.279, 'grad_norm': 0.001543221836074902, 'learning_rate': 0.08324937766952638, 'epoch': 0.57} + 57%|█████▋ | 295/520 [18:35<14:12, 3.79s/it] 57%|█████▋ | 296/520 [18:39<14:02, 3.76s/it] {'loss': 1.1669, 'grad_norm': 0.001592933463367561, 'learning_rate': 0.08263518223330697, 'epoch': 0.57} + 57%|█████▋ | 296/520 [18:39<14:02, 3.76s/it] 57%|█████▋ | 297/520 [18:43<13:50, 3.73s/it] {'loss': 1.2917, 'grad_norm': 0.0016448860106684914, 'learning_rate': 0.08202166149209474, 'epoch': 0.57} + 57%|█████▋ | 297/520 [18:43<13:50, 3.73s/it] 57%|█████▋ | 298/520 [18:46<13:43, 3.71s/it] {'loss': 1.266, 'grad_norm': 0.0014398735083025794, 'learning_rate': 0.08140883928370855, 'epoch': 0.57} + 57%|█████▋ | 298/520 [18:46<13:43, 3.71s/it] 57%|█████▊ | 299/520 [18:50<13:38, 3.70s/it] {'loss': 1.3032, 'grad_norm': 0.0014593471724089026, 'learning_rate': 0.0807967394188264, 'epoch': 0.57} + 57%|█████▊ | 299/520 [18:50<13:38, 3.70s/it] 58%|█████▊ | 300/520 [18:54<13:30, 3.69s/it] {'loss': 1.3178, 'grad_norm': 0.0015140940326704145, 'learning_rate': 0.08018538568006027, 'epoch': 0.58} + 58%|█████▊ | 300/520 [18:54<13:30, 3.69s/it] 58%|█████▊ | 301/520 [18:57<13:25, 3.68s/it] {'loss': 1.2888, 'grad_norm': 0.0015107288365521889, 'learning_rate': 0.07957480182103199, 'epoch': 0.58} + 58%|█████▊ | 301/520 [18:57<13:25, 3.68s/it] 58%|█████▊ | 302/520 [19:01<13:26, 3.70s/it] {'loss': 1.3223, 'grad_norm': 0.0015731777079140174, 'learning_rate': 0.07896501156545044, 'epoch': 0.58} + 58%|█████▊ | 302/520 [19:01<13:26, 3.70s/it] 58%|█████▊ | 303/520 [19:05<13:26, 3.72s/it] {'loss': 1.2198, 'grad_norm': 0.0017228348931248834, 'learning_rate': 0.07835603860618973, 'epoch': 0.58} + 58%|█████▊ | 303/520 [19:05<13:26, 3.72s/it] 58%|█████▊ | 304/520 [19:09<13:38, 3.79s/it] {'loss': 1.2128, 'grad_norm': 0.0016040172740016977, 'learning_rate': 0.07774790660436857, 'epoch': 0.58} + 58%|█████▊ | 304/520 [19:09<13:38, 3.79s/it] 59%|█████▊ | 305/520 [19:12<13:29, 3.76s/it] {'loss': 1.3307, 'grad_norm': 0.0017599526196213572, 'learning_rate': 0.07714063918843106, 'epoch': 0.59} + 59%|█████▊ | 305/520 [19:12<13:29, 3.76s/it] 59%|█████▉ | 306/520 [19:16<13:19, 3.74s/it] {'loss': 1.2735, 'grad_norm': 0.0015616771532427182, 'learning_rate': 0.0765342599532285, 'epoch': 0.59} + 59%|█████▉ | 306/520 [19:16<13:19, 3.74s/it] 59%|█████▉ | 307/520 [19:20<13:11, 3.72s/it] {'loss': 1.2089, 'grad_norm': 0.0014682059689399077, 'learning_rate': 0.07592879245910272, 'epoch': 0.59} + 59%|█████▉ | 307/520 [19:20<13:11, 3.72s/it] 59%|█████▉ | 308/520 [19:23<13:01, 3.69s/it] {'loss': 1.328, 'grad_norm': 0.0015896577122237571, 'learning_rate': 0.07532426023097064, 'epoch': 0.59} + 59%|█████▉ | 308/520 [19:23<13:01, 3.69s/it] 59%|█████▉ | 309/520 [19:27<13:18, 3.78s/it] {'loss': 1.2005, 'grad_norm': 0.001424774299289521, 'learning_rate': 0.07472068675741024, 'epoch': 0.59} + 59%|█████▉ | 309/520 [19:27<13:18, 3.78s/it] 60%|█████▉ | 310/520 [19:31<13:06, 3.74s/it] {'loss': 1.1881, 'grad_norm': 0.00154302915596814, 'learning_rate': 0.07411809548974792, 'epoch': 0.6} + 60%|█████▉ | 310/520 [19:31<13:06, 3.74s/it] 60%|█████▉ | 311/520 [19:35<12:56, 3.71s/it] {'loss': 1.1545, 'grad_norm': 0.0014874748255236465, 'learning_rate': 0.07351650984114727, 'epoch': 0.6} + 60%|█████▉ | 311/520 [19:35<12:56, 3.71s/it] 60%|██████ | 312/520 [19:38<12:48, 3.70s/it] {'loss': 1.1508, 'grad_norm': 0.00159924970035723, 'learning_rate': 0.0729159531856995, 'epoch': 0.6} + 60%|██████ | 312/520 [19:38<12:48, 3.70s/it] 60%|██████ | 313/520 [19:42<12:39, 3.67s/it] {'loss': 1.1389, 'grad_norm': 0.0014068115754633374, 'learning_rate': 0.07231644885751508, 'epoch': 0.6} + 60%|██████ | 313/520 [19:42<12:39, 3.67s/it] 60%|██████ | 314/520 [19:46<12:53, 3.76s/it] {'loss': 1.1744, 'grad_norm': 0.0014030850560266402, 'learning_rate': 0.07171802014981725, 'epoch': 0.6} + 60%|██████ | 314/520 [19:46<12:53, 3.76s/it] 61%|██████ | 315/520 [19:50<12:43, 3.72s/it] {'loss': 1.2628, 'grad_norm': 0.0016828526287566217, 'learning_rate': 0.07112069031403703, 'epoch': 0.61} + 61%|██████ | 315/520 [19:50<12:43, 3.72s/it] 61%|██████ | 316/520 [19:54<13:02, 3.84s/it] {'loss': 1.1476, 'grad_norm': 0.0016381337970722781, 'learning_rate': 0.07052448255890957, 'epoch': 0.61} + 61%|██████ | 316/520 [19:54<13:02, 3.84s/it] 61%|██████ | 317/520 [19:57<12:47, 3.78s/it] {'loss': 1.171, 'grad_norm': 0.0013594198111006397, 'learning_rate': 0.0699294200495727, 'epoch': 0.61} + 61%|██████ | 317/520 [19:57<12:47, 3.78s/it] 61%|██████ | 318/520 [20:01<12:37, 3.75s/it] {'loss': 1.288, 'grad_norm': 0.0016248268092757666, 'learning_rate': 0.06933552590666658, 'epoch': 0.61} + 61%|██████ | 318/520 [20:01<12:37, 3.75s/it] 61%|██████▏ | 319/520 [20:05<12:54, 3.86s/it] {'loss': 1.1546, 'grad_norm': 0.0014149151235091865, 'learning_rate': 0.06874282320543557, 'epoch': 0.61} + 61%|██████▏ | 319/520 [20:05<12:54, 3.86s/it] 62%|██████▏ | 320/520 [20:09<12:38, 3.79s/it] {'loss': 1.1031, 'grad_norm': 0.0015462183001046343, 'learning_rate': 0.06815133497483157, 'epoch': 0.62} + 62%|██████▏ | 320/520 [20:09<12:38, 3.79s/it] 62%|██████▏ | 321/520 [20:12<12:27, 3.76s/it] {'loss': 1.298, 'grad_norm': 0.001578744418110095, 'learning_rate': 0.06756108419661931, 'epoch': 0.62} + 62%|██████▏ | 321/520 [20:12<12:27, 3.76s/it] 62%|██████▏ | 322/520 [20:16<12:17, 3.72s/it] {'loss': 1.1557, 'grad_norm': 0.0015221134724744548, 'learning_rate': 0.06697209380448332, 'epoch': 0.62} + 62%|██████▏ | 322/520 [20:16<12:17, 3.72s/it] 62%|██████▏ | 323/520 [20:20<12:09, 3.70s/it] {'loss': 1.2226, 'grad_norm': 0.0015095983348940665, 'learning_rate': 0.06638438668313694, 'epoch': 0.62} + 62%|██████▏ | 323/520 [20:20<12:09, 3.70s/it] 62%|██████▏ | 324/520 [20:23<12:01, 3.68s/it] {'loss': 1.2354, 'grad_norm': 0.001534285695793191, 'learning_rate': 0.06579798566743314, 'epoch': 0.62} + 62%|██████▏ | 324/520 [20:23<12:01, 3.68s/it] 62%|██████▎ | 325/520 [20:27<11:56, 3.68s/it] {'loss': 1.2444, 'grad_norm': 0.0016167467024412738, 'learning_rate': 0.06521291354147728, 'epoch': 0.62} + 62%|██████▎ | 325/520 [20:27<11:56, 3.68s/it] 63%|██████▎ | 326/520 [20:31<11:51, 3.67s/it] {'loss': 1.2318, 'grad_norm': 0.0015725776316832768, 'learning_rate': 0.06462919303774187, 'epoch': 0.63} + 63%|██████▎ | 326/520 [20:31<11:51, 3.67s/it] 63%|██████▎ | 327/520 [20:34<11:46, 3.66s/it] {'loss': 1.2838, 'grad_norm': 0.0016310411285360507, 'learning_rate': 0.06404684683618325, 'epoch': 0.63} + 63%|██████▎ | 327/520 [20:34<11:46, 3.66s/it] 63%|██████▎ | 328/520 [20:38<11:43, 3.66s/it] {'loss': 1.2862, 'grad_norm': 0.001567679691029478, 'learning_rate': 0.0634658975633605, 'epoch': 0.63} + 63%|██████▎ | 328/520 [20:38<11:43, 3.66s/it] 63%|██████▎ | 329/520 [20:42<11:41, 3.67s/it] {'loss': 1.1515, 'grad_norm': 0.0013378653030106094, 'learning_rate': 0.06288636779155621, 'epoch': 0.63} + 63%|██████▎ | 329/520 [20:42<11:41, 3.67s/it] 63%|██████▎ | 330/520 [20:46<11:48, 3.73s/it] {'loss': 1.2353, 'grad_norm': 0.0014175536740962766, 'learning_rate': 0.06230828003789948, 'epoch': 0.63} + 63%|██████▎ | 330/520 [20:46<11:48, 3.73s/it] 64%|██████▎ | 331/520 [20:49<11:49, 3.76s/it] {'loss': 1.189, 'grad_norm': 0.0015201243597339521, 'learning_rate': 0.06173165676349103, 'epoch': 0.64} + 64%|██████▎ | 331/520 [20:49<11:49, 3.76s/it] 64%|██████▍ | 332/520 [20:53<11:48, 3.77s/it] {'loss': 1.305, 'grad_norm': 0.0014411301246306388, 'learning_rate': 0.06115652037253053, 'epoch': 0.64} + 64%|██████▍ | 332/520 [20:53<11:48, 3.77s/it] 64%|██████▍ | 333/520 [20:57<11:44, 3.77s/it] {'loss': 1.3396, 'grad_norm': 0.0015972288977535581, 'learning_rate': 0.06058289321144608, 'epoch': 0.64} + 64%|██████▍ | 333/520 [20:57<11:44, 3.77s/it] 64%|██████▍ | 334/520 [21:01<11:44, 3.79s/it] {'loss': 1.2373, 'grad_norm': 0.001569539854772156, 'learning_rate': 0.06001079756802592, 'epoch': 0.64} + 64%|██████▍ | 334/520 [21:01<11:44, 3.79s/it] 64%|██████▍ | 335/520 [21:05<11:42, 3.80s/it] {'loss': 1.236, 'grad_norm': 0.0014360243166629456, 'learning_rate': 0.059440255670552514, 'epoch': 0.64} + 64%|██████▍ | 335/520 [21:05<11:42, 3.80s/it] 65%|██████▍ | 336/520 [21:08<11:36, 3.79s/it] {'loss': 1.1286, 'grad_norm': 0.0016253423630215747, 'learning_rate': 0.05887128968693887, 'epoch': 0.65} + 65%|██████▍ | 336/520 [21:08<11:36, 3.79s/it] 65%|██████▍ | 337/520 [21:12<11:33, 3.79s/it] {'loss': 1.1203, 'grad_norm': 0.0014563695302532967, 'learning_rate': 0.058303921723867225, 'epoch': 0.65} + 65%|██████▍ | 337/520 [21:12<11:33, 3.79s/it] 65%|██████▌ | 338/520 [21:16<11:28, 3.78s/it] {'loss': 1.2437, 'grad_norm': 0.001518481330189124, 'learning_rate': 0.05773817382593008, 'epoch': 0.65} + 65%|██████▌ | 338/520 [21:16<11:28, 3.78s/it] 65%|██████▌ | 339/520 [21:20<11:31, 3.82s/it] {'loss': 1.1902, 'grad_norm': 0.0014734023433507746, 'learning_rate': 0.057174067974773715, 'epoch': 0.65} + 65%|██████▌ | 339/520 [21:20<11:31, 3.82s/it] 65%|██████▌ | 340/520 [21:24<11:26, 3.81s/it] {'loss': 1.1752, 'grad_norm': 0.0014802617639813435, 'learning_rate': 0.056611626088244195, 'epoch': 0.65} + 65%|██████▌ | 340/520 [21:24<11:26, 3.81s/it] 66%|██████▌ | 341/520 [21:27<11:22, 3.81s/it] {'loss': 1.1915, 'grad_norm': 0.001546904231809461, 'learning_rate': 0.056050870019535494, 'epoch': 0.66} + 66%|██████▌ | 341/520 [21:27<11:22, 3.81s/it] 66%|██████▌ | 342/520 [21:31<11:13, 3.78s/it] {'loss': 1.2664, 'grad_norm': 0.001729090286722236, 'learning_rate': 0.05549182155634076, 'epoch': 0.66} + 66%|██████▌ | 342/520 [21:31<11:13, 3.78s/it] 66%|██████▌ | 343/520 [21:35<11:01, 3.74s/it] {'loss': 1.2255, 'grad_norm': 0.0014869920258992074, 'learning_rate': 0.054934502420005464, 'epoch': 0.66} + 66%|██████▌ | 343/520 [21:35<11:01, 3.74s/it] 66%|██████▌ | 344/520 [21:38<10:51, 3.70s/it] {'loss': 1.1555, 'grad_norm': 0.0014099125111800227, 'learning_rate': 0.0543789342646837, 'epoch': 0.66} + 66%|██████▌ | 344/520 [21:38<10:51, 3.70s/it] 66%|██████▋ | 345/520 [21:42<10:44, 3.68s/it] {'loss': 1.2704, 'grad_norm': 0.0016257143869129037, 'learning_rate': 0.05382513867649663, 'epoch': 0.66} + 66%|██████▋ | 345/520 [21:42<10:44, 3.68s/it] 67%|██████▋ | 346/520 [21:46<10:40, 3.68s/it] {'loss': 1.2341, 'grad_norm': 0.0014710412442301799, 'learning_rate': 0.0532731371726938, 'epoch': 0.67} + 67%|██████▋ | 346/520 [21:46<10:40, 3.68s/it] 67%|██████▋ | 347/520 [21:49<10:33, 3.66s/it] {'loss': 1.1638, 'grad_norm': 0.0014135430137599005, 'learning_rate': 0.05272295120081732, 'epoch': 0.67} + 67%|██████▋ | 347/520 [21:49<10:33, 3.66s/it]Token indices sequence length is longer than the specified maximum sequence length for this model (2778 > 2048). Running this sequence through the model will result in indexing errors + 67%|██████▋ | 348/520 [21:53<10:28, 3.66s/it] {'loss': 1.1244, 'grad_norm': 0.001675121193635658, 'learning_rate': 0.05217460213786821, 'epoch': 0.67} + 67%|██████▋ | 348/520 [21:53<10:28, 3.66s/it] 67%|██████▋ | 349/520 [21:57<10:26, 3.66s/it] {'loss': 1.1662, 'grad_norm': 0.0015307689960357882, 'learning_rate': 0.051628111289476024, 'epoch': 0.67} + 67%|██████▋ | 349/520 [21:57<10:26, 3.66s/it] 67%|██████▋ | 350/520 [22:00<10:22, 3.66s/it] {'loss': 1.2114, 'grad_norm': 0.001560597078171092, 'learning_rate': 0.051083499889071106, 'epoch': 0.67} + 67%|██████▋ | 350/520 [22:00<10:22, 3.66s/it] 68%|██████▊ | 351/520 [22:04<10:17, 3.66s/it] {'loss': 1.1224, 'grad_norm': 0.0014472868013713505, 'learning_rate': 0.05054078909705926, 'epoch': 0.68} + 68%|██████▊ | 351/520 [22:04<10:17, 3.66s/it] 68%|██████▊ | 352/520 [22:08<10:13, 3.65s/it] {'loss': 1.2423, 'grad_norm': 0.0015414601415523054, 'learning_rate': 0.050000000000000024, 'epoch': 0.68} + 68%|██████▊ | 352/520 [22:08<10:13, 3.65s/it] 68%|██████▊ | 353/520 [22:11<10:17, 3.70s/it] {'loss': 1.1853, 'grad_norm': 0.0015444716030581823, 'learning_rate': 0.04946115360978696, 'epoch': 0.68} + 68%|██████▊ | 353/520 [22:11<10:17, 3.70s/it] 68%|██████▊ | 354/520 [22:15<10:22, 3.75s/it] {'loss': 1.3123, 'grad_norm': 0.0014356773881222962, 'learning_rate': 0.048924270862831465, 'epoch': 0.68} + 68%|██████▊ | 354/520 [22:15<10:22, 3.75s/it] 68%|██████▊ | 355/520 [22:19<10:23, 3.78s/it] {'loss': 1.1767, 'grad_norm': 0.001460523926673069, 'learning_rate': 0.04838937261924933, 'epoch': 0.68} + 68%|██████▊ | 355/520 [22:19<10:23, 3.78s/it] 68%|██████▊ | 356/520 [22:23<10:23, 3.80s/it] {'loss': 1.1756, 'grad_norm': 0.0015125665771912476, 'learning_rate': 0.0478564796620502, 'epoch': 0.68} + 68%|██████▊ | 356/520 [22:23<10:23, 3.80s/it] 69%|██████▊ | 357/520 [22:27<10:20, 3.81s/it] {'loss': 1.205, 'grad_norm': 0.0014216703598608662, 'learning_rate': 0.04732561269632993, 'epoch': 0.69} + 69%|██████▊ | 357/520 [22:27<10:20, 3.81s/it] 69%|██████▉ | 358/520 [22:31<10:19, 3.82s/it] {'loss': 1.1393, 'grad_norm': 0.0014251898289984963, 'learning_rate': 0.04679679234846636, 'epoch': 0.69} + 69%|██████▉ | 358/520 [22:31<10:19, 3.82s/it] 69%|██████▉ | 359/520 [22:35<10:18, 3.84s/it] {'loss': 1.2391, 'grad_norm': 0.0014990713528969105, 'learning_rate': 0.046270039165317606, 'epoch': 0.69} + 69%|██████▉ | 359/520 [22:35<10:18, 3.84s/it] 69%|██████▉ | 360/520 [22:38<10:15, 3.84s/it] {'loss': 1.2526, 'grad_norm': 0.001574902499692622, 'learning_rate': 0.04574537361342407, 'epoch': 0.69} + 69%|██████▉ | 360/520 [22:38<10:15, 3.84s/it] 69%|██████▉ | 361/520 [22:42<10:12, 3.85s/it] {'loss': 1.2557, 'grad_norm': 0.0014111530190382255, 'learning_rate': 0.04522281607821288, 'epoch': 0.69} + 69%|██████▉ | 361/520 [22:42<10:12, 3.85s/it] 70%|██████▉ | 362/520 [22:46<10:08, 3.85s/it] {'loss': 1.199, 'grad_norm': 0.0016219017860656748, 'learning_rate': 0.04470238686320606, 'epoch': 0.7} + 70%|██████▉ | 362/520 [22:46<10:08, 3.85s/it] 70%|██████▉ | 363/520 [22:50<10:05, 3.85s/it] {'loss': 1.2163, 'grad_norm': 0.0014843669898169526, 'learning_rate': 0.044184106189231624, 'epoch': 0.7} + 70%|██████▉ | 363/520 [22:50<10:05, 3.85s/it] 70%|███████ | 364/520 [22:54<10:02, 3.86s/it] {'loss': 1.2766, 'grad_norm': 0.0015303624763658374, 'learning_rate': 0.043667994193637795, 'epoch': 0.7} + 70%|███████ | 364/520 [22:54<10:02, 3.86s/it] 70%|███████ | 365/520 [22:58<09:58, 3.86s/it] {'loss': 1.2799, 'grad_norm': 0.0015920629353325159, 'learning_rate': 0.043154070929510784, 'epoch': 0.7} + 70%|███████ | 365/520 [22:58<09:58, 3.86s/it] 70%|███████ | 366/520 [23:02<09:54, 3.86s/it] {'loss': 1.2373, 'grad_norm': 0.0014675012620823386, 'learning_rate': 0.04264235636489542, 'epoch': 0.7} + 70%|███████ | 366/520 [23:02<09:54, 3.86s/it] 71%|███████ | 367/520 [23:05<09:40, 3.79s/it] {'loss': 1.2327, 'grad_norm': 0.0015160905606539591, 'learning_rate': 0.04213287038201943, 'epoch': 0.71} + 71%|███████ | 367/520 [23:05<09:40, 3.79s/it] 71%|███████ | 368/520 [23:09<09:28, 3.74s/it] {'loss': 1.0843, 'grad_norm': 0.001543310657402249, 'learning_rate': 0.04162563277652104, 'epoch': 0.71} + 71%|███████ | 368/520 [23:09<09:28, 3.74s/it] 71%|███████ | 369/520 [23:12<09:19, 3.71s/it] {'loss': 1.2263, 'grad_norm': 0.00135619907325759, 'learning_rate': 0.04112066325667954, 'epoch': 0.71} + 71%|███████ | 369/520 [23:12<09:19, 3.71s/it] 71%|███████ | 370/520 [23:16<09:13, 3.69s/it] {'loss': 1.1471, 'grad_norm': 0.0014173599929203763, 'learning_rate': 0.04061798144264986, 'epoch': 0.71} + 71%|███████ | 370/520 [23:16<09:13, 3.69s/it] 71%|███████▏ | 371/520 [23:20<09:06, 3.67s/it] {'loss': 1.1465, 'grad_norm': 0.0015439997627980932, 'learning_rate': 0.04011760686569998, 'epoch': 0.71} + 71%|███████▏ | 371/520 [23:20<09:06, 3.67s/it] 72%|███████▏ | 372/520 [23:23<09:01, 3.66s/it] {'loss': 1.3146, 'grad_norm': 0.0013852966498121868, 'learning_rate': 0.03961955896745224, 'epoch': 0.72} + 72%|███████▏ | 372/520 [23:23<09:01, 3.66s/it] 72%|███████▏ | 373/520 [23:27<08:57, 3.66s/it] {'loss': 1.1959, 'grad_norm': 0.001594689511911395, 'learning_rate': 0.03912385709912794, 'epoch': 0.72} + 72%|███████▏ | 373/520 [23:27<08:57, 3.66s/it] 72%|███████▏ | 374/520 [23:31<08:53, 3.66s/it] {'loss': 1.2285, 'grad_norm': 0.0014787222407117906, 'learning_rate': 0.038630520520795276, 'epoch': 0.72} + 72%|███████▏ | 374/520 [23:31<08:53, 3.66s/it] 72%|███████▏ | 375/520 [23:34<08:50, 3.66s/it] {'loss': 1.1484, 'grad_norm': 0.0014485063833115698, 'learning_rate': 0.03813956840062119, 'epoch': 0.72} + 72%|███████▏ | 375/520 [23:34<08:50, 3.66s/it] 72%|███████▏ | 376/520 [23:38<08:48, 3.67s/it] {'loss': 1.2593, 'grad_norm': 0.0014057753258129853, 'learning_rate': 0.037651019814126656, 'epoch': 0.72} + 72%|███████▏ | 376/520 [23:38<08:48, 3.67s/it] 72%|███████▎ | 377/520 [23:42<08:44, 3.67s/it] {'loss': 1.2029, 'grad_norm': 0.0015507081943963894, 'learning_rate': 0.037164893743445275, 'epoch': 0.72} + 72%|███████▎ | 377/520 [23:42<08:44, 3.67s/it] 73%|███████▎ | 378/520 [23:45<08:40, 3.67s/it] {'loss': 1.2494, 'grad_norm': 0.0014414946595691083, 'learning_rate': 0.03668120907658603, 'epoch': 0.73} + 73%|███████▎ | 378/520 [23:45<08:40, 3.67s/it] 73%|███████▎ | 379/520 [23:49<08:35, 3.66s/it] {'loss': 1.2336, 'grad_norm': 0.0014308390016426532, 'learning_rate': 0.036199984606699154, 'epoch': 0.73} + 73%|███████▎ | 379/520 [23:49<08:35, 3.66s/it] 73%|███████▎ | 380/520 [23:53<08:44, 3.74s/it] {'loss': 1.2783, 'grad_norm': 0.0014890835226033941, 'learning_rate': 0.035721239031346066, 'epoch': 0.73} + 73%|███████▎ | 380/520 [23:53<08:44, 3.74s/it] 73%|███████▎ | 381/520 [23:57<08:46, 3.79s/it] {'loss': 1.2371, 'grad_norm': 0.001420540278395758, 'learning_rate': 0.03524499095177297, 'epoch': 0.73} + 73%|███████▎ | 381/520 [23:57<08:46, 3.79s/it] 73%|███████▎ | 382/520 [24:01<08:47, 3.82s/it] {'loss': 1.2391, 'grad_norm': 0.0014415551662949962, 'learning_rate': 0.03477125887218792, 'epoch': 0.73} + 73%|███████▎ | 382/520 [24:01<08:47, 3.82s/it] 74%|███████▎ | 383/520 [24:05<08:45, 3.83s/it] {'loss': 1.0726, 'grad_norm': 0.0015420850705481824, 'learning_rate': 0.03430006119904196, 'epoch': 0.74} + 74%|███████▎ | 383/520 [24:05<08:45, 3.83s/it] 74%|███████▍ | 384/520 [24:08<08:34, 3.78s/it] {'loss': 1.3053, 'grad_norm': 0.0014074037258201904, 'learning_rate': 0.033831416240314084, 'epoch': 0.74} + 74%|███████▍ | 384/520 [24:08<08:34, 3.78s/it] 74%|███████▍ | 385/520 [24:12<08:25, 3.75s/it] {'loss': 1.2059, 'grad_norm': 0.0013440852766598847, 'learning_rate': 0.03336534220479961, 'epoch': 0.74} + 74%|███████▍ | 385/520 [24:12<08:25, 3.75s/it] 74%|███████▍ | 386/520 [24:16<08:24, 3.76s/it] {'loss': 1.1596, 'grad_norm': 0.0012785693759925104, 'learning_rate': 0.032901857201403005, 'epoch': 0.74} + 74%|███████▍ | 386/520 [24:16<08:24, 3.76s/it] 74%|███████▍ | 387/520 [24:19<08:18, 3.75s/it] {'loss': 1.3141, 'grad_norm': 0.0015011699905131523, 'learning_rate': 0.032440979238433976, 'epoch': 0.74} + 74%|███████▍ | 387/520 [24:19<08:18, 3.75s/it] 75%|███████▍ | 388/520 [24:23<08:12, 3.73s/it] {'loss': 1.1108, 'grad_norm': 0.0013355142238180991, 'learning_rate': 0.03198272622290804, 'epoch': 0.75} + 75%|███████▍ | 388/520 [24:23<08:12, 3.73s/it] 75%|███████▍ | 389/520 [24:27<08:07, 3.72s/it] {'loss': 1.166, 'grad_norm': 0.0016814626384427862, 'learning_rate': 0.03152711595985065, 'epoch': 0.75} + 75%|███████▍ | 389/520 [24:27<08:07, 3.72s/it] 75%|███████▌ | 390/520 [24:30<08:03, 3.72s/it] {'loss': 1.2292, 'grad_norm': 0.0014099011354660332, 'learning_rate': 0.031074166151605298, 'epoch': 0.75} + 75%|███████▌ | 390/520 [24:30<08:03, 3.72s/it] 75%|███████▌ | 391/520 [24:34<07:56, 3.70s/it] {'loss': 1.3131, 'grad_norm': 0.0015488359352336882, 'learning_rate': 0.030623894397145836, 'epoch': 0.75} + 75%|███████▌ | 391/520 [24:34<07:56, 3.70s/it] 75%|███████▌ | 392/520 [24:38<07:52, 3.69s/it] {'loss': 1.12, 'grad_norm': 0.0014107546949614065, 'learning_rate': 0.03017631819139273, 'epoch': 0.75} + 75%|███████▌ | 392/520 [24:38<07:52, 3.69s/it] 76%|███████▌ | 393/520 [24:41<07:47, 3.68s/it] {'loss': 1.1391, 'grad_norm': 0.0012910324166432, 'learning_rate': 0.029731454924533086, 'epoch': 0.76} + 76%|███████▌ | 393/520 [24:41<07:47, 3.68s/it] 76%|███████▌ | 394/520 [24:45<07:41, 3.66s/it] {'loss': 1.182, 'grad_norm': 0.001504031388856611, 'learning_rate': 0.029289321881345254, 'epoch': 0.76} + 76%|███████▌ | 394/520 [24:45<07:41, 3.66s/it] 76%|███████▌ | 395/520 [24:49<07:35, 3.65s/it] {'loss': 1.1505, 'grad_norm': 0.0015603479103389808, 'learning_rate': 0.028849936240527008, 'epoch': 0.76} + 76%|███████▌ | 395/520 [24:49<07:35, 3.65s/it] 76%|███████▌ | 396/520 [24:52<07:31, 3.64s/it] {'loss': 1.233, 'grad_norm': 0.0015489713642375534, 'learning_rate': 0.028413315074028157, 'epoch': 0.76} + 76%|███████▌ | 396/520 [24:52<07:31, 3.64s/it] 76%|███████▋ | 397/520 [24:56<07:27, 3.64s/it] {'loss': 1.212, 'grad_norm': 0.0014305907955873677, 'learning_rate': 0.027979475346387363, 'epoch': 0.76} + 76%|███████▋ | 397/520 [24:56<07:27, 3.64s/it] 77%|███████▋ | 398/520 [25:00<07:23, 3.64s/it] {'loss': 1.2146, 'grad_norm': 0.0015630420854158165, 'learning_rate': 0.027548433914072735, 'epoch': 0.77} + 77%|███████▋ | 398/520 [25:00<07:23, 3.64s/it] 77%|███████▋ | 399/520 [25:03<07:19, 3.64s/it] {'loss': 1.1846, 'grad_norm': 0.0014667207580271705, 'learning_rate': 0.027120207524827168, 'epoch': 0.77} + 77%|███████▋ | 399/520 [25:03<07:19, 3.64s/it] 77%|███████▋ | 400/520 [25:07<07:17, 3.65s/it] {'loss': 1.2177, 'grad_norm': 0.0013601929864416973, 'learning_rate': 0.02669481281701739, 'epoch': 0.77} + 77%|███████▋ | 400/520 [25:07<07:17, 3.65s/it] 77%|███████▋ | 401/520 [25:11<07:14, 3.65s/it] {'loss': 1.0382, 'grad_norm': 0.0015893894685949203, 'learning_rate': 0.026272266318987603, 'epoch': 0.77} + 77%|███████▋ | 401/520 [25:11<07:14, 3.65s/it] 77%|███████▋ | 402/520 [25:14<07:10, 3.65s/it] {'loss': 1.1587, 'grad_norm': 0.001488417039515335, 'learning_rate': 0.02585258444841733, 'epoch': 0.77} + 77%|███████▋ | 402/520 [25:14<07:10, 3.65s/it] 78%|███████▊ | 403/520 [25:18<07:08, 3.66s/it] {'loss': 1.1933, 'grad_norm': 0.0016182417739092662, 'learning_rate': 0.025435783511683442, 'epoch': 0.78} + 78%|███████▊ | 403/520 [25:18<07:08, 3.66s/it] 78%|███████▊ | 404/520 [25:22<07:08, 3.70s/it] {'loss': 1.0928, 'grad_norm': 0.0017073967932922224, 'learning_rate': 0.02502187970322657, 'epoch': 0.78} + 78%|███████▊ | 404/520 [25:22<07:08, 3.70s/it] 78%|███████▊ | 405/520 [25:26<07:11, 3.76s/it] {'loss': 1.1919, 'grad_norm': 0.0014214116603882931, 'learning_rate': 0.02461088910492202, 'epoch': 0.78} + 78%|███████▊ | 405/520 [25:26<07:11, 3.76s/it] 78%|███████▊ | 406/520 [25:29<07:13, 3.80s/it] {'loss': 1.1172, 'grad_norm': 0.001659354661348191, 'learning_rate': 0.02420282768545469, 'epoch': 0.78} + 78%|███████▊ | 406/520 [25:29<07:13, 3.80s/it] 78%|███████▊ | 407/520 [25:33<07:13, 3.84s/it] {'loss': 1.2806, 'grad_norm': 0.0015641917583073025, 'learning_rate': 0.02379771129969892, 'epoch': 0.78} + 78%|███████▊ | 407/520 [25:33<07:13, 3.84s/it] 78%|███████▊ | 408/520 [25:37<07:11, 3.85s/it] {'loss': 1.1743, 'grad_norm': 0.0016535506743662813, 'learning_rate': 0.023395555688102213, 'epoch': 0.78} + 78%|███████▊ | 408/520 [25:37<07:11, 3.85s/it] 79%|███████▊ | 409/520 [25:41<07:09, 3.87s/it] {'loss': 1.2981, 'grad_norm': 0.0015661153103296657, 'learning_rate': 0.02299637647607372, 'epoch': 0.79} + 79%|███████▊ | 409/520 [25:41<07:09, 3.87s/it] 79%|███████▉ | 410/520 [25:45<07:05, 3.87s/it] {'loss': 1.0257, 'grad_norm': 0.001460305660031482, 'learning_rate': 0.022600189173377264, 'epoch': 0.79} + 79%|███████▉ | 410/520 [25:45<07:05, 3.87s/it] 79%|███████▉ | 411/520 [25:49<06:54, 3.81s/it] {'loss': 1.2832, 'grad_norm': 0.0017471208614681525, 'learning_rate': 0.022207009173528525, 'epoch': 0.79} + 79%|███████▉ | 411/520 [25:49<06:54, 3.81s/it] 79%|███████▉ | 412/520 [25:52<06:47, 3.77s/it] {'loss': 1.1869, 'grad_norm': 0.0015599384905121132, 'learning_rate': 0.02181685175319702, 'epoch': 0.79} + 79%|███████▉ | 412/520 [25:52<06:47, 3.77s/it] 79%|███████▉ | 413/520 [25:56<06:40, 3.74s/it] {'loss': 1.2134, 'grad_norm': 0.0014533872501581986, 'learning_rate': 0.021429732071612653, 'epoch': 0.79} + 79%|███████▉ | 413/520 [25:56<06:40, 3.74s/it] 80%|███████▉ | 414/520 [26:00<06:33, 3.71s/it] {'loss': 1.0164, 'grad_norm': 0.0012595806101444271, 'learning_rate': 0.02104566516997647, 'epoch': 0.8} + 80%|███████▉ | 414/520 [26:00<06:33, 3.71s/it] 80%|███████▉ | 415/520 [26:03<06:28, 3.70s/it] {'loss': 1.1643, 'grad_norm': 0.0014404357623926767, 'learning_rate': 0.020664665970876496, 'epoch': 0.8} + 80%|███████▉ | 415/520 [26:03<06:28, 3.70s/it] 80%|████████ | 416/520 [26:07<06:23, 3.68s/it] {'loss': 1.0841, 'grad_norm': 0.0015900758611214017, 'learning_rate': 0.020286749277707784, 'epoch': 0.8} + 80%|████████ | 416/520 [26:07<06:23, 3.68s/it] 80%|████████ | 417/520 [26:11<06:18, 3.67s/it] {'loss': 1.2455, 'grad_norm': 0.0015520076578950403, 'learning_rate': 0.019911929774097215, 'epoch': 0.8} + 80%|████████ | 417/520 [26:11<06:18, 3.67s/it] 80%|████████ | 418/520 [26:14<06:12, 3.66s/it] {'loss': 1.2308, 'grad_norm': 0.0013775028905990009, 'learning_rate': 0.019540222023333165, 'epoch': 0.8} + 80%|████████ | 418/520 [26:14<06:12, 3.66s/it] 81%|████████ | 419/520 [26:18<06:09, 3.65s/it] {'loss': 1.2223, 'grad_norm': 0.001609374795792937, 'learning_rate': 0.01917164046779948, 'epoch': 0.81} + 81%|████████ | 419/520 [26:18<06:09, 3.65s/it] 81%|████████ | 420/520 [26:22<06:05, 3.66s/it] {'loss': 1.1079, 'grad_norm': 0.0015441927655296737, 'learning_rate': 0.018806199428414352, 'epoch': 0.81} + 81%|████████ | 420/520 [26:22<06:05, 3.66s/it] 81%|████████ | 421/520 [26:25<06:01, 3.66s/it] {'loss': 1.0442, 'grad_norm': 0.0016669253114015208, 'learning_rate': 0.018443913104073985, 'epoch': 0.81} + 81%|████████ | 421/520 [26:25<06:01, 3.66s/it] 81%|████████ | 422/520 [26:29<05:57, 3.65s/it] {'loss': 1.1652, 'grad_norm': 0.0015387417754256741, 'learning_rate': 0.01808479557110081, 'epoch': 0.81} + 81%|████████ | 422/520 [26:29<05:57, 3.65s/it] 81%|████████▏ | 423/520 [26:33<05:55, 3.66s/it] {'loss': 1.1525, 'grad_norm': 0.0016224065033854217, 'learning_rate': 0.017728860782696667, 'epoch': 0.81} + 81%|████████▏ | 423/520 [26:33<05:55, 3.66s/it] 82%|████████▏ | 424/520 [26:36<05:52, 3.68s/it] {'loss': 1.2949, 'grad_norm': 0.0014872226874024696, 'learning_rate': 0.017376122568400532, 'epoch': 0.82} + 82%|████████▏ | 424/520 [26:36<05:52, 3.68s/it] 82%|████████▏ | 425/520 [26:40<05:50, 3.69s/it] {'loss': 1.1572, 'grad_norm': 0.0014385036319551894, 'learning_rate': 0.017026594633551252, 'epoch': 0.82} + 82%|████████▏ | 425/520 [26:40<05:50, 3.69s/it] 82%|████████▏ | 426/520 [26:44<05:44, 3.67s/it] {'loss': 1.1792, 'grad_norm': 0.0018581508650537917, 'learning_rate': 0.01668029055875512, 'epoch': 0.82} + 82%|████████▏ | 426/520 [26:44<05:44, 3.67s/it] 82%|████████▏ | 427/520 [26:47<05:40, 3.66s/it] {'loss': 1.1014, 'grad_norm': 0.001452582018796891, 'learning_rate': 0.016337223799358026, 'epoch': 0.82} + 82%|████████▏ | 427/520 [26:47<05:40, 3.66s/it] 82%|████████▏ | 428/520 [26:51<05:36, 3.66s/it] {'loss': 1.0736, 'grad_norm': 0.0015045517277063947, 'learning_rate': 0.01599740768492286, 'epoch': 0.82} + 82%|████████▏ | 428/520 [26:51<05:36, 3.66s/it] 82%|████████▎ | 429/520 [26:55<05:32, 3.65s/it] {'loss': 1.1722, 'grad_norm': 0.001459252425414527, 'learning_rate': 0.015660855418711452, 'epoch': 0.82} + 82%|████████▎ | 429/520 [26:55<05:32, 3.65s/it]Token indices sequence length is longer than the specified maximum sequence length for this model (2076 > 2048). Running this sequence through the model will result in indexing errors + 83%|████████▎ | 430/520 [26:58<05:28, 3.65s/it] {'loss': 1.168, 'grad_norm': 0.0013754620097715408, 'learning_rate': 0.015327580077171589, 'epoch': 0.83} + 83%|████████▎ | 430/520 [26:58<05:28, 3.65s/it] 83%|████████▎ | 431/520 [27:02<05:26, 3.67s/it] {'loss': 1.1834, 'grad_norm': 0.0015440704678463682, 'learning_rate': 0.014997594609429088, 'epoch': 0.83} + 83%|████████▎ | 431/520 [27:02<05:26, 3.67s/it] 83%|████████▎ | 432/520 [27:06<05:22, 3.67s/it] {'loss': 1.082, 'grad_norm': 0.0015051529740221633, 'learning_rate': 0.01467091183678444, 'epoch': 0.83} + 83%|████████▎ | 432/520 [27:06<05:22, 3.67s/it] 83%|████████▎ | 433/520 [27:09<05:19, 3.67s/it] {'loss': 1.2129, 'grad_norm': 0.0014142108900096005, 'learning_rate': 0.014347544452214867, 'epoch': 0.83} + 83%|████████▎ | 433/520 [27:09<05:19, 3.67s/it] 83%|████████▎ | 434/520 [27:13<05:15, 3.67s/it] {'loss': 0.9572, 'grad_norm': 0.0014366998048082678, 'learning_rate': 0.014027505019880971, 'epoch': 0.83} + 83%|████████▎ | 434/520 [27:13<05:15, 3.67s/it] 84%|████████▎ | 435/520 [27:17<05:12, 3.67s/it] {'loss': 1.2497, 'grad_norm': 0.0016275707595824714, 'learning_rate': 0.013710805974638696, 'epoch': 0.84} + 84%|████████▎ | 435/520 [27:17<05:12, 3.67s/it] 84%|████████▍ | 436/520 [27:20<05:08, 3.67s/it] {'loss': 1.046, 'grad_norm': 0.001503294624354917, 'learning_rate': 0.01339745962155613, 'epoch': 0.84} + 84%|████████▍ | 436/520 [27:20<05:08, 3.67s/it] 84%|████████▍ | 437/520 [27:24<05:05, 3.68s/it] {'loss': 1.274, 'grad_norm': 0.0015032544078932958, 'learning_rate': 0.01308747813543536, 'epoch': 0.84} + 84%|████████▍ | 437/520 [27:24<05:05, 3.68s/it] 84%|████████▍ | 438/520 [27:28<05:02, 3.69s/it] {'loss': 1.0883, 'grad_norm': 0.0014526322542679126, 'learning_rate': 0.012780873560339467, 'epoch': 0.84} + 84%|████████▍ | 438/520 [27:28<05:02, 3.69s/it] 84%|████████▍ | 439/520 [27:31<04:58, 3.69s/it] {'loss': 1.1606, 'grad_norm': 0.0012970530459106187, 'learning_rate': 0.012477657809124632, 'epoch': 0.84} + 84%|████████▍ | 439/520 [27:31<04:58, 3.69s/it] 85%|████████▍ | 440/520 [27:35<04:54, 3.68s/it] {'loss': 1.1282, 'grad_norm': 0.001493350404149568, 'learning_rate': 0.012177842662977134, 'epoch': 0.85} + 85%|████████▍ | 440/520 [27:35<04:54, 3.68s/it] 85%|████████▍ | 441/520 [27:39<04:53, 3.72s/it] {'loss': 1.1737, 'grad_norm': 0.0014895119309937018, 'learning_rate': 0.01188143977095576, 'epoch': 0.85} + 85%|████████▍ | 441/520 [27:39<04:53, 3.72s/it] 85%|████████▌ | 442/520 [27:43<04:52, 3.75s/it] {'loss': 1.1892, 'grad_norm': 0.001586828458076632, 'learning_rate': 0.011588460649539035, 'epoch': 0.85} + 85%|████████▌ | 442/520 [27:43<04:52, 3.75s/it] 85%|████████▌ | 443/520 [27:47<04:51, 3.78s/it] {'loss': 1.2072, 'grad_norm': 0.0014389270756175333, 'learning_rate': 0.011298916682177829, 'epoch': 0.85} + 85%|████████▌ | 443/520 [27:47<04:51, 3.78s/it] 85%|████████▌ | 444/520 [27:50<04:48, 3.79s/it] {'loss': 1.1691, 'grad_norm': 0.0013725717266378267, 'learning_rate': 0.011012819118853146, 'epoch': 0.85} + 85%|████████▌ | 444/520 [27:50<04:48, 3.79s/it] 86%|████████▌ | 445/520 [27:54<04:44, 3.80s/it] {'loss': 1.0963, 'grad_norm': 0.001424541798109955, 'learning_rate': 0.01073017907563887, 'epoch': 0.86} + 86%|████████▌ | 445/520 [27:54<04:44, 3.80s/it] 86%|████████▌ | 446/520 [27:58<04:41, 3.81s/it] {'loss': 1.2551, 'grad_norm': 0.0014190514813943898, 'learning_rate': 0.010451007534269908, 'epoch': 0.86} + 86%|████████▌ | 446/520 [27:58<04:41, 3.81s/it] 86%|████████▌ | 447/520 [28:02<04:39, 3.82s/it] {'loss': 1.181, 'grad_norm': 0.001502580344573424, 'learning_rate': 0.010175315341715598, 'epoch': 0.86} + 86%|████████▌ | 447/520 [28:02<04:39, 3.82s/it] 86%|████████▌ | 448/520 [28:06<04:35, 3.83s/it] {'loss': 1.1633, 'grad_norm': 0.0015208432313101844, 'learning_rate': 0.009903113209758098, 'epoch': 0.86} + 86%|████████▌ | 448/520 [28:06<04:35, 3.83s/it] 86%|████████▋ | 449/520 [28:10<04:32, 3.84s/it] {'loss': 1.2104, 'grad_norm': 0.0015171616853639131, 'learning_rate': 0.009634411714576352, 'epoch': 0.86} + 86%|████████▋ | 449/520 [28:10<04:32, 3.84s/it] 87%|████████▋ | 450/520 [28:13<04:28, 3.84s/it] {'loss': 1.2021, 'grad_norm': 0.0015531287893260626, 'learning_rate': 0.009369221296335007, 'epoch': 0.87} + 87%|████████▋ | 450/520 [28:13<04:28, 3.84s/it] 87%|████████▋ | 451/520 [28:17<04:24, 3.83s/it] {'loss': 1.1932, 'grad_norm': 0.0014818070694645469, 'learning_rate': 0.009107552258778906, 'epoch': 0.87} + 87%|████████▋ | 451/520 [28:17<04:24, 3.83s/it] 87%|████████▋ | 452/520 [28:21<04:20, 3.83s/it] {'loss': 1.2491, 'grad_norm': 0.0013953506695247957, 'learning_rate': 0.008849414768832687, 'epoch': 0.87} + 87%|████████▋ | 452/520 [28:21<04:20, 3.83s/it] 87%|████████▋ | 453/520 [28:25<04:15, 3.82s/it] {'loss': 1.2264, 'grad_norm': 0.0014230855536509315, 'learning_rate': 0.008594818856205699, 'epoch': 0.87} + 87%|████████▋ | 453/520 [28:25<04:15, 3.82s/it] 87%|████████▋ | 454/520 [28:29<04:11, 3.82s/it] {'loss': 1.1027, 'grad_norm': 0.001569929496018057, 'learning_rate': 0.00834377441300238, 'epoch': 0.87} + 87%|████████▋ | 454/520 [28:29<04:11, 3.82s/it] 88%|████████▊ | 455/520 [28:32<04:08, 3.82s/it] {'loss': 1.2458, 'grad_norm': 0.0014839874407294887, 'learning_rate': 0.008096291193337934, 'epoch': 0.88} + 88%|████████▊ | 455/520 [28:32<04:08, 3.82s/it] 88%|████████▊ | 456/520 [28:36<04:04, 3.82s/it] {'loss': 1.1646, 'grad_norm': 0.001494283490116101, 'learning_rate': 0.007852378812959226, 'epoch': 0.88} + 88%|████████▊ | 456/520 [28:36<04:04, 3.82s/it] 88%|████████▊ | 457/520 [28:40<03:59, 3.81s/it] {'loss': 1.1512, 'grad_norm': 0.0013323979109914688, 'learning_rate': 0.007612046748871327, 'epoch': 0.88} + 88%|████████▊ | 457/520 [28:40<03:59, 3.81s/it] 88%|████████▊ | 458/520 [28:44<03:56, 3.81s/it] {'loss': 1.3049, 'grad_norm': 0.0016708995796752312, 'learning_rate': 0.007375304338969136, 'epoch': 0.88} + 88%|████████▊ | 458/520 [28:44<03:56, 3.81s/it] 88%|████████▊ | 459/520 [28:48<03:52, 3.82s/it] {'loss': 1.2298, 'grad_norm': 0.0014673392578915367, 'learning_rate': 0.007142160781674645, 'epoch': 0.88} + 88%|████████▊ | 459/520 [28:48<03:52, 3.82s/it] 88%|████████▊ | 460/520 [28:52<03:49, 3.82s/it] {'loss': 1.1115, 'grad_norm': 0.0014968457371749106, 'learning_rate': 0.006912625135579587, 'epoch': 0.88} + 88%|████████▊ | 460/520 [28:52<03:49, 3.82s/it] 89%|████████▊ | 461/520 [28:55<03:45, 3.83s/it] {'loss': 1.2377, 'grad_norm': 0.0012181331610182805, 'learning_rate': 0.0066867063190933496, 'epoch': 0.89} + 89%|████████▊ | 461/520 [28:55<03:45, 3.83s/it] 89%|████████▉ | 462/520 [28:59<03:41, 3.82s/it] {'loss': 1.2992, 'grad_norm': 0.0014625786554414676, 'learning_rate': 0.006464413110096601, 'epoch': 0.89} + 89%|████████▉ | 462/520 [28:59<03:41, 3.82s/it] 89%|████████▉ | 463/520 [29:03<03:37, 3.82s/it] {'loss': 1.0737, 'grad_norm': 0.001501568873187397, 'learning_rate': 0.006245754145600091, 'epoch': 0.89} + 89%|████████▉ | 463/520 [29:03<03:37, 3.82s/it] 89%|████████▉ | 464/520 [29:07<03:34, 3.82s/it] {'loss': 1.2157, 'grad_norm': 0.0016047728557644892, 'learning_rate': 0.006030737921409169, 'epoch': 0.89} + 89%|████████▉ | 464/520 [29:07<03:34, 3.82s/it] 89%|████████▉ | 465/520 [29:11<03:30, 3.83s/it] {'loss': 1.323, 'grad_norm': 0.0015772334678356812, 'learning_rate': 0.005819372791793654, 'epoch': 0.89} + 89%|████████▉ | 465/520 [29:11<03:30, 3.83s/it] 90%|████████▉ | 466/520 [29:15<03:27, 3.84s/it] {'loss': 1.1966, 'grad_norm': 0.001345586528407217, 'learning_rate': 0.005611666969163243, 'epoch': 0.9} + 90%|████████▉ | 466/520 [29:15<03:27, 3.84s/it] 90%|████████▉ | 467/520 [29:18<03:23, 3.84s/it] {'loss': 1.1889, 'grad_norm': 0.001363189518309568, 'learning_rate': 0.005407628523748398, 'epoch': 0.9} + 90%|████████▉ | 467/520 [29:18<03:23, 3.84s/it] 90%|█████████ | 468/520 [29:22<03:19, 3.84s/it] {'loss': 1.18, 'grad_norm': 0.0017157811094304378, 'learning_rate': 0.00520726538328683, 'epoch': 0.9} + 90%|█████████ | 468/520 [29:22<03:19, 3.84s/it] 90%|█████████ | 469/520 [29:26<03:15, 3.83s/it] {'loss': 1.2363, 'grad_norm': 0.00160284006912858, 'learning_rate': 0.005010585332715401, 'epoch': 0.9} + 90%|█████████ | 469/520 [29:26<03:15, 3.83s/it] 90%|█████████ | 470/520 [29:30<03:10, 3.80s/it] {'loss': 1.1155, 'grad_norm': 0.0013489108564726496, 'learning_rate': 0.004817596013867765, 'epoch': 0.9} + 90%|█████████ | 470/520 [29:30<03:10, 3.80s/it] 91%|█████████ | 471/520 [29:33<03:03, 3.75s/it] {'loss': 1.1372, 'grad_norm': 0.0015250609827655451, 'learning_rate': 0.004628304925177318, 'epoch': 0.91} + 91%|█████████ | 471/520 [29:33<03:03, 3.75s/it] 91%|█████████ | 472/520 [29:37<02:58, 3.73s/it] {'loss': 1.1065, 'grad_norm': 0.0014561191948262752, 'learning_rate': 0.004442719421385921, 'epoch': 0.91} + 91%|█████████ | 472/520 [29:37<02:58, 3.73s/it] 91%|█████████ | 473/520 [29:41<02:53, 3.70s/it] {'loss': 1.1679, 'grad_norm': 0.0015107733933541986, 'learning_rate': 0.004260846713258193, 'epoch': 0.91} + 91%|█████████ | 473/520 [29:41<02:53, 3.70s/it] 91%|█████████ | 474/520 [29:44<02:51, 3.72s/it] {'loss': 1.2215, 'grad_norm': 0.00138713423655468, 'learning_rate': 0.004082693867301224, 'epoch': 0.91} + 91%|█████████ | 474/520 [29:44<02:51, 3.72s/it] 91%|█████████▏| 475/520 [29:48<02:49, 3.76s/it] {'loss': 1.1395, 'grad_norm': 0.0013871853488330113, 'learning_rate': 0.003908267805490051, 'epoch': 0.91} + 91%|█████████▏| 475/520 [29:48<02:49, 3.76s/it] 92%|█████████▏| 476/520 [29:52<02:46, 3.79s/it] {'loss': 1.1623, 'grad_norm': 0.0015374503386658637, 'learning_rate': 0.003737575304998797, 'epoch': 0.92} + 92%|█████████▏| 476/520 [29:52<02:46, 3.79s/it] 92%|█████████▏| 477/520 [29:56<02:43, 3.80s/it] {'loss': 1.1473, 'grad_norm': 0.0015959376771076787, 'learning_rate': 0.003570622997937234, 'epoch': 0.92} + 92%|█████████▏| 477/520 [29:56<02:43, 3.80s/it] 92%|█████████▏| 478/520 [30:00<02:40, 3.82s/it] {'loss': 1.1024, 'grad_norm': 0.0014890271075108927, 'learning_rate': 0.00340741737109318, 'epoch': 0.92} + 92%|█████████▏| 478/520 [30:00<02:40, 3.82s/it] 92%|█████████▏| 479/520 [30:04<02:36, 3.83s/it] {'loss': 1.189, 'grad_norm': 0.001606225106920096, 'learning_rate': 0.003247964765680389, 'epoch': 0.92} + 92%|█████████▏| 479/520 [30:04<02:36, 3.83s/it] 92%|█████████▏| 480/520 [30:08<02:33, 3.83s/it] {'loss': 1.2113, 'grad_norm': 0.0013827150702183295, 'learning_rate': 0.0030922713770922153, 'epoch': 0.92} + 92%|█████████▏| 480/520 [30:08<02:33, 3.83s/it] 92%|█████████▎| 481/520 [30:11<02:30, 3.85s/it] {'loss': 1.2011, 'grad_norm': 0.0013208289414254938, 'learning_rate': 0.0029403432546609046, 'epoch': 0.93} + 92%|█████████▎| 481/520 [30:11<02:30, 3.85s/it] 93%|█████████▎| 482/520 [30:15<02:26, 3.84s/it] {'loss': 1.2183, 'grad_norm': 0.0014122125950340552, 'learning_rate': 0.0027921863014225504, 'epoch': 0.93} + 93%|█████████▎| 482/520 [30:15<02:26, 3.84s/it] 93%|█████████▎| 483/520 [30:19<02:22, 3.84s/it] {'loss': 1.1727, 'grad_norm': 0.001532021583507493, 'learning_rate': 0.002647806273887665, 'epoch': 0.93} + 93%|█████████▎| 483/520 [30:19<02:22, 3.84s/it] 93%|█████████▎| 484/520 [30:23<02:17, 3.82s/it] {'loss': 1.1758, 'grad_norm': 0.0014973007069770904, 'learning_rate': 0.0025072087818176383, 'epoch': 0.93} + 93%|█████████▎| 484/520 [30:23<02:17, 3.82s/it] 93%|█████████▎| 485/520 [30:27<02:12, 3.78s/it] {'loss': 1.1309, 'grad_norm': 0.0014116783905802966, 'learning_rate': 0.002370399288006664, 'epoch': 0.93} + 93%|█████████▎| 485/520 [30:27<02:12, 3.78s/it] 93%|█████████▎| 486/520 [30:30<02:07, 3.74s/it] {'loss': 1.2514, 'grad_norm': 0.0015027780739366651, 'learning_rate': 0.0022373831080695463, 'epoch': 0.93} + 93%|█████████▎| 486/520 [30:30<02:07, 3.74s/it] 94%|█████████▎| 487/520 [30:34<02:03, 3.73s/it] {'loss': 1.105, 'grad_norm': 0.0014387125203500876, 'learning_rate': 0.0021081654102351635, 'epoch': 0.94} + 94%|█████████▎| 487/520 [30:34<02:03, 3.73s/it] 94%|█████████▍| 488/520 [30:38<01:59, 3.72s/it] {'loss': 1.0516, 'grad_norm': 0.001515829741039417, 'learning_rate': 0.0019827512151456175, 'epoch': 0.94} + 94%|█████████▍| 488/520 [30:38<01:59, 3.72s/it] 94%|█████████▍| 489/520 [30:41<01:54, 3.70s/it] {'loss': 1.2204, 'grad_norm': 0.0012712652323557908, 'learning_rate': 0.0018611453956612345, 'epoch': 0.94} + 94%|█████████▍| 489/520 [30:41<01:54, 3.70s/it] 94%|█████████▍| 490/520 [30:45<01:50, 3.68s/it] {'loss': 1.1728, 'grad_norm': 0.001534915272290847, 'learning_rate': 0.0017433526766711727, 'epoch': 0.94} + 94%|█████████▍| 490/520 [30:45<01:50, 3.68s/it] 94%|█████████▍| 491/520 [30:49<01:46, 3.66s/it] {'loss': 1.1336, 'grad_norm': 0.0015512613384427382, 'learning_rate': 0.0016293776349098677, 'epoch': 0.94} + 94%|█████████▍| 491/520 [30:49<01:46, 3.66s/it] 95%|█████████▍| 492/520 [30:52<01:42, 3.67s/it] {'loss': 1.2475, 'grad_norm': 0.0015327276401406317, 'learning_rate': 0.0015192246987791981, 'epoch': 0.95} + 95%|█████████▍| 492/520 [30:52<01:42, 3.67s/it] 95%|█████████▍| 493/520 [30:56<01:38, 3.66s/it] {'loss': 1.2335, 'grad_norm': 0.0015361585801653844, 'learning_rate': 0.0014128981481764114, 'epoch': 0.95} + 95%|█████████▍| 493/520 [30:56<01:38, 3.66s/it] 95%|█████████▌| 494/520 [30:59<01:34, 3.65s/it] {'loss': 1.1913, 'grad_norm': 0.0014219961091645067, 'learning_rate': 0.0013104021143278911, 'epoch': 0.95} + 95%|█████████▌| 494/520 [30:59<01:34, 3.65s/it] 95%|█████████▌| 495/520 [31:03<01:31, 3.65s/it] {'loss': 1.1465, 'grad_norm': 0.0014638596319297634, 'learning_rate': 0.0012117405796285285, 'epoch': 0.95} + 95%|█████████▌| 495/520 [31:03<01:31, 3.65s/it] 95%|█████████▌| 496/520 [31:07<01:27, 3.66s/it] {'loss': 1.0696, 'grad_norm': 0.0014996486964154252, 'learning_rate': 0.0011169173774871477, 'epoch': 0.95} + 95%|█████████▌| 496/520 [31:07<01:27, 3.66s/it] 96%|█████████▌| 497/520 [31:10<01:24, 3.65s/it] {'loss': 1.1463, 'grad_norm': 0.0012760215611671158, 'learning_rate': 0.0010259361921774012, 'epoch': 0.96} + 96%|█████████▌| 497/520 [31:10<01:24, 3.65s/it] 96%|█████████▌| 498/520 [31:14<01:20, 3.65s/it] {'loss': 1.1471, 'grad_norm': 0.0014521135606189773, 'learning_rate': 0.000938800558694719, 'epoch': 0.96} + 96%|█████████▌| 498/520 [31:14<01:20, 3.65s/it] 96%|█████████▌| 499/520 [31:18<01:16, 3.66s/it] {'loss': 1.2887, 'grad_norm': 0.0014705559640795683, 'learning_rate': 0.0008555138626189618, 'epoch': 0.96} + 96%|█████████▌| 499/520 [31:18<01:16, 3.66s/it] 96%|█████████▌| 500/520 [31:21<01:13, 3.65s/it] {'loss': 1.267, 'grad_norm': 0.0018784892902699263, 'learning_rate': 0.0007760793399827937, 'epoch': 0.96} + 96%|█████████▌| 500/520 [31:21<01:13, 3.65s/it] 96%|█████████▋| 501/520 [31:25<01:09, 3.65s/it] {'loss': 1.1976, 'grad_norm': 0.0015518019407411354, 'learning_rate': 0.000700500077146038, 'epoch': 0.96} + 96%|█████████▋| 501/520 [31:25<01:09, 3.65s/it] 97%|█████████▋| 502/520 [31:29<01:05, 3.65s/it] {'loss': 1.1896, 'grad_norm': 0.0014393259784890712, 'learning_rate': 0.0006287790106757397, 'epoch': 0.97} + 97%|█████████▋| 502/520 [31:29<01:05, 3.65s/it] 97%|█████████▋| 503/520 [31:32<01:02, 3.65s/it] {'loss': 1.1803, 'grad_norm': 0.0015193460921063965, 'learning_rate': 0.0005609189272320237, 'epoch': 0.97} + 97%|█████████▋| 503/520 [31:32<01:02, 3.65s/it] 97%|█████████▋| 504/520 [31:36<00:58, 3.65s/it] {'loss': 1.1894, 'grad_norm': 0.0017837156373864517, 'learning_rate': 0.000496922463459859, 'epoch': 0.97} + 97%|█████████▋| 504/520 [31:36<00:58, 3.65s/it] 97%|█████████▋| 505/520 [31:40<00:54, 3.65s/it] {'loss': 1.2258, 'grad_norm': 0.001526507917252772, 'learning_rate': 0.0004367921058866187, 'epoch': 0.97} + 97%|█████████▋| 505/520 [31:40<00:54, 3.65s/it] 97%|█████████▋| 506/520 [31:43<00:51, 3.67s/it] {'loss': 1.1439, 'grad_norm': 0.0015629269917579965, 'learning_rate': 0.0003805301908254455, 'epoch': 0.97} + 97%|█████████▋| 506/520 [31:43<00:51, 3.67s/it] 98%|█████████▊| 507/520 [31:47<00:47, 3.66s/it] {'loss': 1.335, 'grad_norm': 0.0014264001484257229, 'learning_rate': 0.0003281389042844918, 'epoch': 0.97} + 98%|█████████▊| 507/520 [31:47<00:47, 3.66s/it] 98%|█████████▊| 508/520 [31:51<00:43, 3.65s/it] {'loss': 1.2587, 'grad_norm': 0.0015555469655359577, 'learning_rate': 0.00027962028188198705, 'epoch': 0.98} + 98%|█████████▊| 508/520 [31:51<00:43, 3.65s/it] 98%|█████████▊| 509/520 [31:54<00:40, 3.65s/it] {'loss': 1.2272, 'grad_norm': 0.001436405746317977, 'learning_rate': 0.00023497620876711256, 'epoch': 0.98} + 98%|█████████▊| 509/520 [31:54<00:40, 3.65s/it] 98%|█████████▊| 510/520 [31:58<00:36, 3.65s/it] {'loss': 1.1816, 'grad_norm': 0.0014754934112811778, 'learning_rate': 0.00019420841954681523, 'epoch': 0.98} + 98%|█████████▊| 510/520 [31:58<00:36, 3.65s/it] 98%|█████████▊| 511/520 [32:02<00:32, 3.64s/it] {'loss': 1.1554, 'grad_norm': 0.001424765353420618, 'learning_rate': 0.00015731849821833956, 'epoch': 0.98} + 98%|█████████▊| 511/520 [32:02<00:32, 3.64s/it] 98%|█████████▊| 512/520 [32:05<00:29, 3.65s/it] {'loss': 1.0409, 'grad_norm': 0.0015133964589877263, 'learning_rate': 0.00012430787810776555, 'epoch': 0.98} + 98%|█████████▊| 512/520 [32:05<00:29, 3.65s/it] 99%|█████████▊| 513/520 [32:09<00:25, 3.65s/it] {'loss': 1.2408, 'grad_norm': 0.0016713641686166112, 'learning_rate': 9.517784181422018e-05, 'epoch': 0.99} + 99%|█████████▊| 513/520 [32:09<00:25, 3.65s/it] 99%|█████████▉| 514/520 [32:13<00:21, 3.65s/it] {'loss': 1.2068, 'grad_norm': 0.0013572992527845302, 'learning_rate': 6.992952116013917e-05, 'epoch': 0.99} + 99%|█████████▉| 514/520 [32:13<00:21, 3.65s/it] 99%|█████████▉| 515/520 [32:16<00:18, 3.64s/it] {'loss': 1.2549, 'grad_norm': 0.0016713990638754343, 'learning_rate': 4.856389714723575e-05, 'epoch': 0.99} + 99%|█████████▉| 515/520 [32:16<00:18, 3.64s/it] 99%|█████████▉| 516/520 [32:20<00:14, 3.64s/it] {'loss': 1.1569, 'grad_norm': 0.0014279464147993514, 'learning_rate': 3.108179991837545e-05, 'epoch': 0.99} + 99%|█████████▉| 516/520 [32:20<00:14, 3.64s/it] 99%|█████████▉| 517/520 [32:23<00:10, 3.62s/it] {'loss': 1.2243, 'grad_norm': 0.0014289807257993664, 'learning_rate': 1.7483908725357544e-05, 'epoch': 0.99} + 99%|█████████▉| 517/520 [32:23<00:10, 3.62s/it] 100%|█████████▉| 518/520 [32:27<00:07, 3.60s/it] {'loss': 1.1746, 'grad_norm': 0.0015877077381205198, 'learning_rate': 7.770751902513862e-06, 'epoch': 1.0} + 100%|█████████▉| 518/520 [32:27<00:07, 3.60s/it] 100%|█████████▉| 519/520 [32:31<00:03, 3.60s/it] {'loss': 1.1911, 'grad_norm': 0.0014771510270581987, 'learning_rate': 1.9427068461808086e-06, 'epoch': 1.0} + 100%|█████████▉| 519/520 [32:31<00:03, 3.60s/it] 100%|██████████| 520/520 [32:35<00:00, 3.86s/it] {'loss': 1.2053, 'grad_norm': 0.0014531522196770997, 'learning_rate': 0.0, 'epoch': 1.0} + 100%|██████████| 520/520 [32:35<00:00, 3.86s/it] {'train_runtime': 1955.496, 'train_samples_per_second': 34.022, 'train_steps_per_second': 0.266, 'train_loss': 1.2783002635607352, 'epoch': 1.0} + 100%|██████████| 520/520 [32:35<00:00, 3.86s/it] 100%|██████████| 520/520 [32:35<00:00, 3.76s/it] +[2025-10-12 15:49:35,531] [INFO] [launch.py:348:main] Process 425411 exits successfully. +[2025-10-12 15:49:35,531] [INFO] [launch.py:348:main] Process 425406 exits successfully. +[2025-10-12 15:49:35,532] [INFO] [launch.py:348:main] Process 425410 exits successfully. +[2025-10-12 15:49:35,532] [INFO] [launch.py:348:main] Process 425407 exits successfully. +[2025-10-12 15:49:36,533] [INFO] [launch.py:348:main] Process 425409 exits successfully. +[2025-10-12 15:49:36,534] [INFO] [launch.py:348:main] Process 425412 exits successfully. +[2025-10-12 15:49:36,534] [INFO] [launch.py:348:main] Process 425408 exits successfully. +[2025-10-12 15:49:40,539] [INFO] [launch.py:348:main] Process 425405 exits successfully. +==== EXPERIMENT COMPLETED: qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.1_2e-1_connector-3.0_1.1_2e-1_ablation ==== +Log File: qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.1_2e-1_connector-3.0_1.1_2e-1_ablation_20251012_151526.log +Timestamp: 2025-10-12 15:49:42 +===================================== diff --git a/logs_oct11/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.3_2e-1_connector-3.0_1.3_2e-1_ablation_20251012_154943.log b/logs_oct11/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.3_2e-1_connector-3.0_1.3_2e-1_ablation_20251012_154943.log new file mode 100644 index 0000000000000000000000000000000000000000..a3c6392d998e1d09c69e12d991d6b21e6d945736 --- /dev/null +++ b/logs_oct11/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.3_2e-1_connector-3.0_1.3_2e-1_ablation_20251012_154943.log @@ -0,0 +1,169 @@ +==== STARTING EXPERIMENT: qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.3_2e-1_connector-3.0_1.3_2e-1_ablation ==== +Log File: qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.3_2e-1_connector-3.0_1.3_2e-1_ablation_20251012_154943.log +Timestamp: 2025-10-12 15:49:43 +===================================== +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 15:49:45,676] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 15:49:48,368] [WARNING] [runner.py:202:fetch_hostfile] Unable to find hostfile, will proceed with training with local resources only. +[2025-10-12 15:49:48,370] [INFO] [runner.py:568:main] cmd = /opt/conda/envs/tinyllava/bin/python3.10 -u -m deepspeed.launcher.launch --world_info=eyJsb2NhbGhvc3QiOiBbMCwgMSwgMiwgMywgNCwgNSwgNiwgN119 --master_addr=127.0.0.1 --master_port=29501 --enable_each_rank_log=None tinyllava/train/train.py --deepspeed ./scripts/zero3.json --data_path /s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json --image_folder /s3-code/ywang29/datasets/tinyllava --is_multimodal True --conv_version qwen2_base --model_name_or_path Qwen/Qwen2.5-0.5B --vision_tower google/siglip-so400m-patch14-384 --vision_tower2 --connector_type mlp2x_gelu --mm_vision_select_layer -2 --image_aspect_ratio square --attn_implementation flash_attention_2 --bf16 True --training_recipe common --tune_type_llm full --tune_type_vision_tower frozen --tune_vision_tower_from_layer 0 --tune_type_connector full --group_by_modality_length True --pretrained_model_path /nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain --output_dir /nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.3_2e-1_connector-3.0_1.3_2e-1_ablation --num_train_epochs 1 --per_device_train_batch_size 4 --per_device_eval_batch_size 4 --gradient_accumulation_steps 4 --evaluation_strategy no --learning_rate 2e-1 --weight_decay 0. --warmup_ratio 0.03 --lr_scheduler_type cosine --logging_steps 1 --tf32 False --model_max_length 2048 --gradient_checkpointing True --dataloader_num_workers 8 --lazy_preprocess True --report_to tensorboard --tokenizer_use_fast False --run_name tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune --subnet_mode_text both --subnet_type_text None --mask_type_text soft --init_mean_text 3.0 --temperature_attn_text 1.3 --temperature_mlp_text 1.3 --backward_type_text normal --masked_layers_text all --subnet_mode_vision both --subnet_type_vision None --mask_type_vision soft --init_mean_vision 3.0 --temperature_attn_vision 1.3 --temperature_mlp_vision 1.3 --backward_type_vision normal --masked_layers_vision all --subnet_type_connector global --mask_type_connector soft --init_mean_connector 3.0 --temperature_connector 1.3 --backward_type_connector normal --mm_projector_lr 2e-1 --seed 42 --mask_model llm-connector --save_strategy steps --save_steps 50000 --save_total_limit 1 --train_data_ratio 0.1 +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 15:49:50,940] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 15:49:52,009] [INFO] [launch.py:138:main] 0 NCCL_VERSION=2.21.5 +[2025-10-12 15:49:52,009] [INFO] [launch.py:138:main] 0 NCCL_SOCKET_IFNAME=eth +[2025-10-12 15:49:52,009] [INFO] [launch.py:145:main] WORLD INFO DICT: {'localhost': [0, 1, 2, 3, 4, 5, 6, 7]} +[2025-10-12 15:49:52,009] [INFO] [launch.py:151:main] nnodes=1, num_local_procs=8, node_rank=0 +[2025-10-12 15:49:52,009] [INFO] [launch.py:162:main] global_rank_mapping=defaultdict(, {'localhost': [0, 1, 2, 3, 4, 5, 6, 7]}) +[2025-10-12 15:49:52,009] [INFO] [launch.py:163:main] dist_world_size=8 +[2025-10-12 15:49:52,009] [INFO] [launch.py:165:main] Setting CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 +[2025-10-12 15:49:52,012] [INFO] [launch.py:253:main] process 445249 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=0', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.3_2e-1_connector-3.0_1.3_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '1.3', '--temperature_mlp_text', '1.3', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '1.3', '--temperature_mlp_vision', '1.3', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '1.3', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 15:49:52,014] [INFO] [launch.py:253:main] process 445250 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=1', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.3_2e-1_connector-3.0_1.3_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '1.3', '--temperature_mlp_text', '1.3', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '1.3', '--temperature_mlp_vision', '1.3', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '1.3', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 15:49:52,016] [INFO] [launch.py:253:main] process 445251 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=2', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.3_2e-1_connector-3.0_1.3_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '1.3', '--temperature_mlp_text', '1.3', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '1.3', '--temperature_mlp_vision', '1.3', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '1.3', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 15:49:52,018] [INFO] [launch.py:253:main] process 445252 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=3', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.3_2e-1_connector-3.0_1.3_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '1.3', '--temperature_mlp_text', '1.3', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '1.3', '--temperature_mlp_vision', '1.3', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '1.3', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 15:49:52,021] [INFO] [launch.py:253:main] process 445253 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=4', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.3_2e-1_connector-3.0_1.3_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '1.3', '--temperature_mlp_text', '1.3', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '1.3', '--temperature_mlp_vision', '1.3', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '1.3', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 15:49:52,023] [INFO] [launch.py:253:main] process 445254 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=5', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.3_2e-1_connector-3.0_1.3_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '1.3', '--temperature_mlp_text', '1.3', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '1.3', '--temperature_mlp_vision', '1.3', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '1.3', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 15:49:52,025] [INFO] [launch.py:253:main] process 445255 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=6', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.3_2e-1_connector-3.0_1.3_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '1.3', '--temperature_mlp_text', '1.3', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '1.3', '--temperature_mlp_vision', '1.3', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '1.3', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +[2025-10-12 15:49:52,027] [INFO] [launch.py:253:main] process 445256 spawned with command: ['/opt/conda/envs/tinyllava/bin/python3.10', '-u', 'tinyllava/train/train.py', '--local_rank=7', '--deepspeed', './scripts/zero3.json', '--data_path', '/s3-code/ywang29/datasets/tinyllava/text_files/llava_v1_5_mix665k.json', '--image_folder', '/s3-code/ywang29/datasets/tinyllava', '--is_multimodal', 'True', '--conv_version', 'qwen2_base', '--model_name_or_path', 'Qwen/Qwen2.5-0.5B', '--vision_tower', 'google/siglip-so400m-patch14-384', '--vision_tower2', '', '--connector_type', 'mlp2x_gelu', '--mm_vision_select_layer', '-2', '--image_aspect_ratio', 'square', '--attn_implementation', 'flash_attention_2', '--bf16', 'True', '--training_recipe', 'common', '--tune_type_llm', 'full', '--tune_type_vision_tower', 'frozen', '--tune_vision_tower_from_layer', '0', '--tune_type_connector', 'full', '--group_by_modality_length', 'True', '--pretrained_model_path', '/nfs/ywang29/TinyLLaVA/checkpoints/tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-pretrain', '--output_dir', '/nfs/ywang29/TinyLLaVA/checkpoints/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.3_2e-1_connector-3.0_1.3_2e-1_ablation', '--num_train_epochs', '1', '--per_device_train_batch_size', '4', '--per_device_eval_batch_size', '4', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--learning_rate', '2e-1', '--weight_decay', '0.', '--warmup_ratio', '0.03', '--lr_scheduler_type', 'cosine', '--logging_steps', '1', '--tf32', 'False', '--model_max_length', '2048', '--gradient_checkpointing', 'True', '--dataloader_num_workers', '8', '--lazy_preprocess', 'True', '--report_to', 'tensorboard', '--tokenizer_use_fast', 'False', '--run_name', 'tiny-llava-Qwen2.5-0.5B-siglip-so400m-patch14-384-qwen2.5-0_5b_base-masktune', '--subnet_mode_text', 'both', '--subnet_type_text', 'None', '--mask_type_text', 'soft', '--init_mean_text', '3.0', '--temperature_attn_text', '1.3', '--temperature_mlp_text', '1.3', '--backward_type_text', 'normal', '--masked_layers_text', 'all', '--subnet_mode_vision', 'both', '--subnet_type_vision', 'None', '--mask_type_vision', 'soft', '--init_mean_vision', '3.0', '--temperature_attn_vision', '1.3', '--temperature_mlp_vision', '1.3', '--backward_type_vision', 'normal', '--masked_layers_vision', 'all', '--subnet_type_connector', 'global', '--mask_type_connector', 'soft', '--init_mean_connector', '3.0', '--temperature_connector', '1.3', '--backward_type_connector', 'normal', '--mm_projector_lr', '2e-1', '--seed', '42', '--mask_model', 'llm-connector', '--save_strategy', 'steps', '--save_steps', '50000', '--save_total_limit', '1', '--train_data_ratio', '0.1'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/torch/cuda/__init__.py:51: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you. + import pynvml # type: ignore[import] +[2025-10-12 15:49:58,674] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 15:49:58,716] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 15:49:58,770] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 15:49:58,770] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 15:49:58,780] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 15:49:58,780] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 15:49:58,785] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 15:49:58,787] [INFO] [real_accelerator.py:191:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2025-10-12 15:49:59,093] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 15:49:59,093] [INFO] [comm.py:668:init_distributed] Initializing TorchBackend in DeepSpeed with backend nccl +[2025-10-12 15:49:59,124] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 15:49:59,171] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 15:49:59,173] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 15:49:59,182] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 15:49:59,185] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 15:49:59,189] [INFO] [comm.py:637:init_distributed] cdb=None +[2025-10-12 15:49:59,190] [INFO] [comm.py:637:init_distributed] cdb=None +Apply masks for the following modules: ['llm', 'connector'] +{'llm': {'model_name_or_path': 'Qwen/Qwen2.5-0.5B', 'cache_dir': None, 'attn_implementation': 'flash_attention_2', 'subnet_mode': 'both', 'subnet_type': 'None', 'sparsity_attn': None, 'sparsity_mlp': None, 'threshold_attn': None, 'threshold_mlp': None, 'temperature_attn': 1.3, 'temperature_mlp': 1.3, 'masked_layers': 'all', 'mask_type': 'soft', 'backward_type': 'normal'}, 'vision_tower': {'model_name_or_path': 'google/siglip-so400m-patch14-384'}, 'connector': {'connector_type': 'mlp2x_gelu', 'subnet_type': 'global', 'threshold': None, 'sparsity': None, 'temperature': 1.3, 'mask_type': 'soft', 'backward_type': 'normal'}} +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +Apply masks for the following modules: ['llm', 'connector'] +/opt/conda/envs/tinyllava/lib/python3.10/site-packages/huggingface_hub/file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`. + warnings.warn( +TinyLlavaConfig { + "backward_type_connector": "normal", + "cache_dir": null, + "connector_type": "mlp2x_gelu", + "hidden_size": 896, + "ignore_index": -100, + "image_aspect_ratio": "square", + "image_token_index": -200, + "llm_model_name_or_path": "Qwen/Qwen2.5-0.5B", + "mask_model": [ + "llm", + "connector" + ], + "mask_type_connector": "soft", + "model_type": "tinyllava", + "num_queries": 128, + "num_resampler_layers": 3, + "pad_token": null, + "resampler_hidden_size": 768, + "sparsity_connector": null, + "subnet_type_connector": "global", + "temperature_connector": 1.3, + "text_config": { + "_name_or_path": "Qwen/Qwen2.5-0.5B", + "architectures": [ + "Qwen2ForCausalLM" + ], + "backward_type": "normal", + "bos_token_id": 151643, + "eos_token_id": 151643, + "hidden_size": 896, + "intermediate_size": 4864, + "mask_type": "soft", + "masked_layers": "all", + "max_position_embeddings": 32768, + "max_window_layers": 24, + "model_type": "qwen2", + "num_attention_heads": 14, + "num_hidden_layers": 24, + "num_key_value_heads": 2, + "rope_theta": 1000000.0, + "sliding_window": 32768, + "subnet_mode": "both", + "subnet_type": "None", + "temperature_attn": 1.3, + "temperature_mlp": 1.3, + "tie_word_embeddings": true, + "torch_dtype": "bfloat16", + "use_mrope": false, + "use_sliding_window": false, + "vocab_size": 151936 + }, + "threshold_connector": null, + "tokenizer_model_max_length": 2048, + "tokenizer_name_or_path": "Qwen/Qwen2.5-0.5B", + "tokenizer_padding_side": "right", + "tokenizer_use_fast": false, + "transformers_version": "4.40.1", + "tune_type_connector": "frozen", + "tune_type_llm": "frozen", + "tune_type_vision_tower": "frozen", + "tune_vision_tower_from_layer": -1, + "use_cache": false, + "vision_config": { + "hidden_act": "gelu_pytorch_tanh", + "hidden_size": 1152, + "image_size": 384, + "intermediate_size": 4304, + "layer_norm_eps": 1e-06, + "model_name_or_path": "google/siglip-so400m-patch14-384", + "model_name_or_path2": "", + "model_type": "siglip_vision_model", + "num_attention_heads": 16, + "num_hidden_layers": 27, + "patch_size": 14 + }, + "vision_feature_layer": -2, + "vision_feature_select_strategy": "patch", + "vision_hidden_size": 1152, + "vision_model_name_or_path": "google/siglip-so400m-patch14-384", + "vision_model_name_or_path2": "", + "vocab_size": 151936 +} +