diff --git a/.gitattributes b/.gitattributes index 415ed7dbdf7fcdfeff86d24bb55784b7d2cbb7b4..183a89f936ae14fb182851ee4fbd2a099ea10e02 100644 --- a/.gitattributes +++ b/.gitattributes @@ -43,3 +43,21 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text 0219110932_cosmos_mip_object/videos/mem_mipnerf360_575f06997f088dc5824bec5a960459303edd0154e65b3594e29810380ac7ea50.mp4 filter=lfs diff=lfs merge=lfs -text 0219110932_cosmos_mip_object/videos/mem_mipnerf360_5f84c46d31489bd1a6b57b7c5c8cbc15280e279f8eb03122abd9362ac7333d13.mp4 filter=lfs diff=lfs merge=lfs -text 0219110932_cosmos_mip_object/videos/mem_mipnerf360_a8f137cd302ac277409b1dcd26e4c8a03d8600bc4f6a80af40c71a5a980a98c5.mp4 filter=lfs diff=lfs merge=lfs -text +0219122156_cosmos_mip_spatial/mem_mipnerf360_575f06997f088dc5824bec5a960459303edd0154e65b3594e29810380ac7ea50/segment_0.mp4 filter=lfs diff=lfs merge=lfs -text +0219122156_cosmos_mip_spatial/mem_mipnerf360_575f06997f088dc5824bec5a960459303edd0154e65b3594e29810380ac7ea50/segment_1.mp4 filter=lfs diff=lfs merge=lfs -text +0219122156_cosmos_mip_spatial/mem_mipnerf360_575f06997f088dc5824bec5a960459303edd0154e65b3594e29810380ac7ea50/segment_2.mp4 filter=lfs diff=lfs merge=lfs -text +0219122156_cosmos_mip_spatial/mem_mipnerf360_575f06997f088dc5824bec5a960459303edd0154e65b3594e29810380ac7ea50/segment_3.mp4 filter=lfs diff=lfs merge=lfs -text +0219122156_cosmos_mip_spatial/mem_mipnerf360_575f06997f088dc5824bec5a960459303edd0154e65b3594e29810380ac7ea50/segment_4.mp4 filter=lfs diff=lfs merge=lfs -text +0219122156_cosmos_mip_spatial/mem_mipnerf360_5f84c46d31489bd1a6b57b7c5c8cbc15280e279f8eb03122abd9362ac7333d13/segment_0.mp4 filter=lfs diff=lfs merge=lfs -text +0219122156_cosmos_mip_spatial/mem_mipnerf360_5f84c46d31489bd1a6b57b7c5c8cbc15280e279f8eb03122abd9362ac7333d13/segment_1.mp4 filter=lfs diff=lfs merge=lfs -text +0219122156_cosmos_mip_spatial/mem_mipnerf360_5f84c46d31489bd1a6b57b7c5c8cbc15280e279f8eb03122abd9362ac7333d13/segment_2.mp4 filter=lfs diff=lfs merge=lfs -text +0219122156_cosmos_mip_spatial/mem_mipnerf360_5f84c46d31489bd1a6b57b7c5c8cbc15280e279f8eb03122abd9362ac7333d13/segment_3.mp4 filter=lfs diff=lfs merge=lfs -text +0219122156_cosmos_mip_spatial/mem_mipnerf360_5f84c46d31489bd1a6b57b7c5c8cbc15280e279f8eb03122abd9362ac7333d13/segment_4.mp4 filter=lfs diff=lfs merge=lfs -text +0219122156_cosmos_mip_spatial/mem_mipnerf360_a8f137cd302ac277409b1dcd26e4c8a03d8600bc4f6a80af40c71a5a980a98c5/segment_0.mp4 filter=lfs diff=lfs merge=lfs -text +0219122156_cosmos_mip_spatial/mem_mipnerf360_a8f137cd302ac277409b1dcd26e4c8a03d8600bc4f6a80af40c71a5a980a98c5/segment_1.mp4 filter=lfs diff=lfs merge=lfs -text +0219122156_cosmos_mip_spatial/mem_mipnerf360_a8f137cd302ac277409b1dcd26e4c8a03d8600bc4f6a80af40c71a5a980a98c5/segment_2.mp4 filter=lfs diff=lfs merge=lfs -text +0219122156_cosmos_mip_spatial/mem_mipnerf360_a8f137cd302ac277409b1dcd26e4c8a03d8600bc4f6a80af40c71a5a980a98c5/segment_3.mp4 filter=lfs diff=lfs merge=lfs -text +0219122156_cosmos_mip_spatial/mem_mipnerf360_a8f137cd302ac277409b1dcd26e4c8a03d8600bc4f6a80af40c71a5a980a98c5/segment_4.mp4 filter=lfs diff=lfs merge=lfs -text +0219122156_cosmos_mip_spatial/videos/mem_mipnerf360_575f06997f088dc5824bec5a960459303edd0154e65b3594e29810380ac7ea50.mp4 filter=lfs diff=lfs merge=lfs -text +0219122156_cosmos_mip_spatial/videos/mem_mipnerf360_5f84c46d31489bd1a6b57b7c5c8cbc15280e279f8eb03122abd9362ac7333d13.mp4 filter=lfs diff=lfs merge=lfs -text +0219122156_cosmos_mip_spatial/videos/mem_mipnerf360_a8f137cd302ac277409b1dcd26e4c8a03d8600bc4f6a80af40c71a5a980a98c5.mp4 filter=lfs diff=lfs merge=lfs -text diff --git a/0219122156_cosmos_mip_spatial/.locks/mem_mipnerf360_575f06997f088dc5824bec5a960459303edd0154e65b3594e29810380ac7ea50.lock b/0219122156_cosmos_mip_spatial/.locks/mem_mipnerf360_575f06997f088dc5824bec5a960459303edd0154e65b3594e29810380ac7ea50.lock new file mode 100644 index 0000000000000000000000000000000000000000..df3819b48cde9c9d481f0484a831ee417f2cdb9c --- /dev/null +++ b/0219122156_cosmos_mip_spatial/.locks/mem_mipnerf360_575f06997f088dc5824bec5a960459303edd0154e65b3594e29810380ac7ea50.lock @@ -0,0 +1 @@ +done:bravo-tongclass-a800:2448340 \ No newline at end of file diff --git a/0219122156_cosmos_mip_spatial/.locks/mem_mipnerf360_5f84c46d31489bd1a6b57b7c5c8cbc15280e279f8eb03122abd9362ac7333d13.lock b/0219122156_cosmos_mip_spatial/.locks/mem_mipnerf360_5f84c46d31489bd1a6b57b7c5c8cbc15280e279f8eb03122abd9362ac7333d13.lock new file mode 100644 index 0000000000000000000000000000000000000000..a59f003a540829f90cebc86a543927da1d2b1470 --- /dev/null +++ b/0219122156_cosmos_mip_spatial/.locks/mem_mipnerf360_5f84c46d31489bd1a6b57b7c5c8cbc15280e279f8eb03122abd9362ac7333d13.lock @@ -0,0 +1 @@ +done:bravo-tongclass-a800:2448342 \ No newline at end of file diff --git a/0219122156_cosmos_mip_spatial/.locks/mem_mipnerf360_a8f137cd302ac277409b1dcd26e4c8a03d8600bc4f6a80af40c71a5a980a98c5.lock b/0219122156_cosmos_mip_spatial/.locks/mem_mipnerf360_a8f137cd302ac277409b1dcd26e4c8a03d8600bc4f6a80af40c71a5a980a98c5.lock new file mode 100644 index 0000000000000000000000000000000000000000..6db6b4cf2c7e0a2b7467dbf9f28ee0b3aa69c77d --- /dev/null +++ b/0219122156_cosmos_mip_spatial/.locks/mem_mipnerf360_a8f137cd302ac277409b1dcd26e4c8a03d8600bc4f6a80af40c71a5a980a98c5.lock @@ -0,0 +1 @@ +done:bravo-tongclass-a800:2448341 \ No newline at end of file diff --git a/0219122156_cosmos_mip_spatial/gpu_0.log b/0219122156_cosmos_mip_spatial/gpu_0.log new file mode 100644 index 0000000000000000000000000000000000000000..96084be984c7eb4b14ff2c2d594f05ebdcc55132 --- /dev/null +++ b/0219122156_cosmos_mip_spatial/gpu_0.log @@ -0,0 +1,2 @@ +[lock] Claimed: mem_mipnerf360_5f84c46d31489bd1a6b57b7c5c8cbc15280e279f8eb03122abd9362ac7333d13 (worker=bravo-tongclass-a800:2448342) +[lock] Done: mem_mipnerf360_5f84c46d31489bd1a6b57b7c5c8cbc15280e279f8eb03122abd9362ac7333d13 diff --git a/0219122156_cosmos_mip_spatial/gpu_0_instances.txt b/0219122156_cosmos_mip_spatial/gpu_0_instances.txt new file mode 100644 index 0000000000000000000000000000000000000000..fcc814358f45a39be60a1d1e31a2d9afa4c9b38b --- /dev/null +++ b/0219122156_cosmos_mip_spatial/gpu_0_instances.txt @@ -0,0 +1 @@ +mem_mipnerf360_5f84c46d31489bd1a6b57b7c5c8cbc15280e279f8eb03122abd9362ac7333d13 diff --git a/0219122156_cosmos_mip_spatial/gpu_1.log b/0219122156_cosmos_mip_spatial/gpu_1.log new file mode 100644 index 0000000000000000000000000000000000000000..c87c45966fc0c475cdb035ee68270a270d97657f --- /dev/null +++ b/0219122156_cosmos_mip_spatial/gpu_1.log @@ -0,0 +1,2 @@ +[lock] Claimed: mem_mipnerf360_a8f137cd302ac277409b1dcd26e4c8a03d8600bc4f6a80af40c71a5a980a98c5 (worker=bravo-tongclass-a800:2448341) +[lock] Done: mem_mipnerf360_a8f137cd302ac277409b1dcd26e4c8a03d8600bc4f6a80af40c71a5a980a98c5 diff --git a/0219122156_cosmos_mip_spatial/gpu_1_instances.txt b/0219122156_cosmos_mip_spatial/gpu_1_instances.txt new file mode 100644 index 0000000000000000000000000000000000000000..28b3b4222bc40f55c2fb0875fc39c9577e9e0182 --- /dev/null +++ b/0219122156_cosmos_mip_spatial/gpu_1_instances.txt @@ -0,0 +1 @@ +mem_mipnerf360_a8f137cd302ac277409b1dcd26e4c8a03d8600bc4f6a80af40c71a5a980a98c5 diff --git a/0219122156_cosmos_mip_spatial/gpu_2.log b/0219122156_cosmos_mip_spatial/gpu_2.log new file mode 100644 index 0000000000000000000000000000000000000000..f75d0427785c3c3e41a6f32df0016f64b700b323 --- /dev/null +++ b/0219122156_cosmos_mip_spatial/gpu_2.log @@ -0,0 +1,2 @@ +[lock] Claimed: mem_mipnerf360_575f06997f088dc5824bec5a960459303edd0154e65b3594e29810380ac7ea50 (worker=bravo-tongclass-a800:2448340) +[lock] Done: mem_mipnerf360_575f06997f088dc5824bec5a960459303edd0154e65b3594e29810380ac7ea50 diff --git a/0219122156_cosmos_mip_spatial/gpu_2_instances.txt b/0219122156_cosmos_mip_spatial/gpu_2_instances.txt new file mode 100644 index 0000000000000000000000000000000000000000..2b84f58faf6a3c80815ae75cfb6013f75a1bc5f1 --- /dev/null +++ b/0219122156_cosmos_mip_spatial/gpu_2_instances.txt @@ -0,0 +1 @@ +mem_mipnerf360_575f06997f088dc5824bec5a960459303edd0154e65b3594e29810380ac7ea50 diff --git a/0219122156_cosmos_mip_spatial/mapping.json b/0219122156_cosmos_mip_spatial/mapping.json new file mode 100644 index 0000000000000000000000000000000000000000..9a3bdd46de30c9cd97b034b32716ba2615676364 --- /dev/null +++ b/0219122156_cosmos_mip_spatial/mapping.json @@ -0,0 +1,92 @@ +[ + { + "instance_id": "mem_mipnerf360_575f06997f088dc5824bec5a960459303edd0154e65b3594e29810380ac7ea50", + "fps": 16.0, + "height": 704, + "width": 1280, + "model_name": "cosmos", + "level": "level_3", + "prompts_mapping": { + "prompt_1": [ + 0, + 76 + ], + "prompt_2": [ + 77, + 153 + ], + "prompt_3": [ + 154, + 230 + ], + "prompt_4": [ + 231, + 307 + ], + "prompt_5": [ + 308, + 384 + ] + } + }, + { + "instance_id": "mem_mipnerf360_5f84c46d31489bd1a6b57b7c5c8cbc15280e279f8eb03122abd9362ac7333d13", + "fps": 16.0, + "height": 704, + "width": 1280, + "model_name": "cosmos", + "level": "level_3", + "prompts_mapping": { + "prompt_1": [ + 0, + 76 + ], + "prompt_2": [ + 77, + 153 + ], + "prompt_3": [ + 154, + 230 + ], + "prompt_4": [ + 231, + 307 + ], + "prompt_5": [ + 308, + 384 + ] + } + }, + { + "instance_id": "mem_mipnerf360_a8f137cd302ac277409b1dcd26e4c8a03d8600bc4f6a80af40c71a5a980a98c5", + "fps": 16.0, + "height": 704, + "width": 1280, + "model_name": "cosmos", + "level": "level_3", + "prompts_mapping": { + "prompt_1": [ + 0, + 76 + ], + "prompt_2": [ + 77, + 153 + ], + "prompt_3": [ + 154, + 230 + ], + "prompt_4": [ + 231, + 307 + ], + "prompt_5": [ + 308, + 384 + ] + } + } +] \ No newline at end of file diff --git a/0219122156_cosmos_mip_spatial/mem_mipnerf360_575f06997f088dc5824bec5a960459303edd0154e65b3594e29810380ac7ea50/config.yaml b/0219122156_cosmos_mip_spatial/mem_mipnerf360_575f06997f088dc5824bec5a960459303edd0154e65b3594e29810380ac7ea50/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..52a9cded596b4121df986b5b811aa3e7a80feddb --- /dev/null +++ b/0219122156_cosmos_mip_spatial/mem_mipnerf360_575f06997f088dc5824bec5a960459303edd0154e65b3594e29810380ac7ea50/config.yaml @@ -0,0 +1,760 @@ +checkpoint: + broadcast_via_filesystem: 'False' + dcp_allow_mismatched_size: 'False' + dcp_async_mode_enabled: 'False' + enable_gcs_patch_in_boto3: 'False' + jit: + device: cuda + dtype: bfloat16 + enabled: 'False' + input_shape: null + strict: 'True' + keys_not_to_resume: [] + load_ema_to_reg: 'False' + load_from_object_store: + bucket: bucket + credentials: credentials/s3_checkpoint.secret + enabled: 'True' + load_path: /home/bravo/zzhangg/MemBench/generation/pipelines/cosmos-predict2.5/models/Cosmos-Predict2.5-2B/base/post-trained/81edfebe-bd6a-4039-8c1d-737df1a790bf_ema_bf16.pt + load_training_state: 'False' + only_load_scheduler_state: 'False' + save_iter: '1000' + save_to_object_store: + bucket: bucket + credentials: credentials/s3_checkpoint.secret + enabled: 'True' + strict_resume: 'True' + type: + _target_: + callbacks: null + disable_async: 'False' + verbose: 'True' +dataloader_train: + dataloaders: + image_data: + dataloader: + batch_size: '12' + cache_augment_fn: functools.partial(, + n=1) + cache_size: '8' + concat_size: '1' + dataset: + augmentor_name: image_basic_augmentor_without_embeddings + caption_type: qwen2p5_7b_v4 + dataset_resolution_type: gt720p + embedding_type: null + resolution: '720' + num_workers: '6' + use_cache: 'False' + ratio: '1' + video_data: + dataloader: + batch_size: '1' + cache_augment_fn: functools.partial(, n=1.8) + cache_size: '16' + concat_size: '1' + dataset: + augmentor_name: video_basic_augmentor_v2 + caption_type: t2w_qwen2p5_7b + dataset_resolution_type: all + embedding_type: null + max_fps_thres: '60' + min_fps_thres: '10' + num_video_frames: '93' + resolution: '720' + use_native_fps: 'True' + video_decoder_name: video_naive_bytes + use_cache: 'False' + ratio: '3' +dataloader_val: + _target_: + dataloaders: + image_data: + dataloader: + _target_: + batch_size: '2' + cache_augment_fn: null + cache_replay_name: image_dataloader + cache_size: '32' + concat_size: '1' + dataset: + _target_: + len_t5: '512' + resolution: '512' + t5_dim: '1024' + num_workers: '8' + pin_memory: 'True' + shuffle: 'False' + use_cache: 'False' + webdataset: 'False' + ratio: '1' + video_data: + dataloader: + _target_: + batch_size: '1' + cache_augment_fn: null + cache_replay_name: video_dataloader + cache_size: '32' + concat_size: '1' + dataset: + _target_: + len_t5: '512' + num_video_frames: '136' + resolution: '512' + t5_dim: '1024' + num_workers: '8' + pin_memory: 'True' + shuffle: 'False' + use_cache: 'False' + webdataset: 'False' + ratio: '1' +defaults: +- _self_ +- data_train: mock +- data_val: mock +- optimizer: fusedadamw +- scheduler: lambdalinear +- model: ddp +- callbacks: basic +- net: null +- conditioner: video_prediction_conditioner +- ema: power +- tokenizer: cosmos_tokenizer_causal_cv8x8x8_c16_res720_t121_it121_v1_0 +- checkpoint: s3 +- ckpt_type: dummy +- experiment: null +job: + cluster: null + group: official_runs_vid2vid + name: Stage-c_pt_4-Index-2-Size-2B-Res-720-Fps-16-Note-rf_with_edm_ckpt + project: cosmos_diffusion_v2 + wandb_mode: online +model: + _recursive_: 'False' + _target_: + config: + conditional_frame_timestep: 0.1 + conditional_frames_probs: + 0: 0.5 + 1: 0.25 + 2: 0.25 + conditioner: + _target_: + fps: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: fps + output_key: fps + padding_mask: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: padding_mask + output_key: padding_mask + text: + _target_: + credential_path: credentials/s3_training.secret + dropout_rate: '0.2' + empty_string_embeddings_path: s3://bucket/predict2_assets/reason1_empty_string_embeddings.pt + input_key: + - t5_text_embeddings + use_empty_string: 'False' + use_video_condition: + _target_: + dropout_rate: '0.0' + input_key: fps + output_key: use_video_condition + conditioning_strategy: frame_replace + denoise_replace_gt_frames: true + ema: + enabled: false + iteration_shift: 0 + rate: 0.1 + fsdp_shard_size: 1 + high_sigma_ratio: 0.05 + high_sigma_timesteps_max: 1000 + high_sigma_timesteps_min: 980 + init_lora_weights: true + input_caption_key: ai_caption + input_data_key: video + input_image_key: images + lora_alpha: 32 + lora_rank: 32 + lora_target_modules: q_proj,k_proj,v_proj,output_proj,mlp.layer1,mlp.layer2 + max_num_conditional_frames: 2 + min_num_conditional_frames: 0 + net: + _target_: + adaln_lora_dim: '256' + atten_backend: minimal_a2a + concat_padding_mask: 'True' + crossattn_emb_channels: '1024' + crossattn_proj_in_channels: '100352' + extra_per_block_abs_pos_emb: 'False' + in_channels: '16' + max_frames: '128' + max_img_h: '240' + max_img_w: '240' + model_channels: '2048' + num_blocks: '28' + num_heads: '16' + out_channels: '16' + patch_spatial: '2' + patch_temporal: '1' + pos_emb_cls: rope3d + pos_emb_interpolation: crop + pos_emb_learnable: 'True' + rope_enable_fps_modulation: 'False' + rope_h_extrapolation_ratio: '3.0' + rope_t_extrapolation_ratio: '1.0' + rope_w_extrapolation_ratio: '3.0' + sac_config: + every_n_blocks: 1 + mode: predict2_2b_720_aggressive + timestep_scale: '0.001' + use_adaln_lora: 'True' + use_crossattn_projection: 'True' + use_wan_fp32_strategy: 'True' + precision: bfloat16 + resolution: '720' + shift: 5 + state_ch: 16 + state_t: 24 + text_encoder_class: reason1p1_7B + text_encoder_config: + ckpt_path: s3://bucket/cosmos_reasoning1/sft_exp700/sft_exp721-1_qwen7b_tl_721_5vs5_s3_balanced_n32_resume_16k/checkpoints/iter_000016000/model/ + compute_online: true + embedding_concat_strategy: full_concat + model_config: + _target_: + model_config: + _target_: cosmos_predict2._src.reason1.configs.default.model_config_qwen.QwenModelConfig + activation_checkpoint: + mode: selective + models: vlm + selective_ac_option: op + add_answer_tag: 'True' + add_cross_attention: 'False' + add_image_start_end_tag: 'False' + add_tile_tag: 'False' + architectures: + - Qwen2_5_VLForConditionalGeneration + attention_dropout: '0.0' + attn_implementation: flash_attention_2 + attn_implementation_autoset: 'True' + aux_loss_coeff: '0.0' + bad_words_ids: null + begin_suppress_tokens: null + bos_token_id: '151643' + cache_dir: null + checkpoint: + async_mode: disabled + create_seed_checkpoint: false + enable_checkpoint: false + export_dtype: float32 + folder: checkpoint + interval: 500 + interval_type: steps + model_weights_only: false + chunk_size_feed_forward: '0' + ckpt_dir: null + ckpt_path: null + comm: + init_timeout_seconds: 300 + trace_buf_size: 20000 + train_timeout_seconds: 100 + cp_size: null + cross_attention_hidden_size: null + decoder_start_token_id: null + deterministic: 'False' + diversity_penalty: '0.0' + do_sample: 'False' + early_stopping: 'False' + encoder_no_repeat_ngram_size: '0' + eos_token_id: '151645' + ep_size: null + experimental: + enable_async_tensor_parallel: false + enable_compiled_autograd: false + pipeline_parallel_degree: 1 + exponential_decay_length_penalty: null + finetuning_task: null + float8: + enable_float8_linear: false + forced_bos_token_id: null + forced_eos_token_id: null + freeze_llm: 'False' + freeze_mm_projector: 'False' + freeze_vision_encoder: 'False' + fsdp_enabled: 'False' + hidden_act: silu + hidden_size: '3584' + id2label: + 0: LABEL_0 + 1: LABEL_1 + image_token_id: '151655' + initializer_range: '0.02' + intermediate_size: '18944' + is_decoder: 'False' + is_encoder_decoder: 'False' + label2id: + LABEL_0: '0' + LABEL_1: '1' + length_penalty: '1.0' + loss_per_token: 'True' + max_batch_size: '1' + max_length: '20' + max_position_embeddings: '128000' + max_seq_len: '128000' + max_window_layers: '28' + min_length: '0' + mm_projector: null + model_type: qwen2_5_vl + name_or_path: Qwen/Qwen2.5-VL-7B-Instruct + no_repeat_ngram_size: '0' + num_attention_heads: '28' + num_beam_groups: '1' + num_beams: '1' + num_hidden_layers: '28' + num_key_value_heads: '4' + num_return_sequences: '1' + num_tiles: '1' + optimizer: + early_step_in_backward: false + end_lr: 2.5e-05 + fused: false + init_lr: 1.0e-05 + lr: 0.0003 + lr_multiplier_llm: 1.0 + lr_multiplier_mm_projector: 1.0 + lr_multiplier_vision_encoder: 0.1 + name: AdamW + output_attentions: 'False' + output_hidden_states: 'True' + output_scores: 'False' + pad_token_id: null + precision: bfloat16 + prefix: null + prepend_padding: 'False' + problem_type: null + pruned_heads: _Nothing.NOTHING + remove_invalid_values: 'False' + repetition_penalty: '1.0' + return_dict: 'True' + return_dict_in_generate: 'False' + rms_norm_eps: 1e-06 + rope_scaling: + mrope_section: + - '16' + - '24' + - '24' + rope_type: default + type: default + rope_theta: '1000000.0' + s3_credential_path: credentials/pbss_dir.secret + seed: '0' + sep_token_id: null + sliding_window: '32768' + suppress_tokens: null + task_specific_params: null + temperature: '1.0' + tf_legacy_loss: 'False' + tie_encoder_decoder: 'False' + tie_word_embeddings: 'False' + tile_tag_type: space_separated + tokenizer_class: null + tokenizer_type: Qwen/Qwen2.5-VL-7B-Instruct + top_k: '50' + top_p: '1.0' + torch_dtype: bfloat16 + torchscript: 'False' + training: + compile: false + context_parallel_degree: 1 + data_parallel_replicate_degree: 1 + data_parallel_shard_degree: -1 + disable_loss_parallel: false + enable_cpu_offload: false + fsdp_reshard_after_forward: default + mixed_precision_param: bfloat16 + mixed_precision_reduce: float32 + steps: 400000 + tensor_parallel_degree: 1 + use_cosine_decay: false + use_linear_decay: true + warmup_steps: 1000 + training_seq_len: '4096' + transformers_version: 4.51.0.dev0 + typical_p: '1.0' + use_bfloat16: 'False' + use_cache: 'False' + use_fsdp2: 'True' + use_return_dict: 'True' + use_rope_from_torchtitan: 'False' + use_sliding_window: 'False' + video_token_id: '151656' + vision_config: + _target_: cosmos_predict2._src.reason1.configs.default.model_config_qwen.QwenVisionConfig + add_cross_attention: 'False' + architectures: null + attn_implementation: flash_attention_2 + attn_implementation_autoset: 'True' + bad_words_ids: null + begin_suppress_tokens: null + bos_token_id: null + chunk_size_feed_forward: '0' + cross_attention_hidden_size: null + decoder_start_token_id: null + depth: '32' + diversity_penalty: '0.0' + do_sample: 'False' + early_stopping: 'False' + embed_dim: null + encoder_no_repeat_ngram_size: '0' + eos_token_id: null + exponential_decay_length_penalty: null + finetuning_task: null + forced_bos_token_id: null + forced_eos_token_id: null + fullatt_block_indexes: + - '7' + - '15' + - '23' + - '31' + hidden_act: silu + hidden_size: '1280' + id2label: + 0: LABEL_0 + 1: LABEL_1 + in_channels: '3' + in_chans: '3' + intermediate_size: '3420' + is_decoder: 'False' + is_encoder_decoder: 'False' + label2id: + LABEL_0: '0' + LABEL_1: '1' + length_penalty: '1.0' + max_length: '20' + min_length: '0' + mlp_ratio: null + model_type: qwen2_5_vl + name_or_path: '' + no_repeat_ngram_size: '0' + num_beam_groups: '1' + num_beams: '1' + num_heads: '16' + num_return_sequences: '1' + out_hidden_size: '3584' + output_attentions: 'False' + output_hidden_states: 'False' + output_scores: 'False' + pad_token_id: null + patch_size: '14' + prefix: null + problem_type: null + pruned_heads: _Nothing.NOTHING + remove_invalid_values: 'False' + repetition_penalty: '1.0' + return_dict: 'True' + return_dict_in_generate: 'False' + sep_token_id: null + spatial_merge_size: '2' + spatial_patch_size: '14' + suppress_tokens: null + task_specific_params: null + temperature: '1.0' + temporal_patch_size: '2' + tf_legacy_loss: 'False' + tie_encoder_decoder: 'False' + tie_word_embeddings: 'True' + tokenizer_class: null + tokens_per_second: '2' + top_k: '50' + top_p: '1.0' + torch_dtype: bfloat16 + torchscript: 'False' + typical_p: '1.0' + use_bfloat16: 'False' + window_size: '112' + vision_encoder: openai/clip-vit-base-patch32 + vision_encoder_config: + depth_init: true + dim: 1024 + ffn_dim_multiplier: null + head_dim: null + hidden_act: null + hidden_dim: 4096 + image_size: 1024 + image_token_id: null + multiple_of: null + n_heads: 16 + n_kv_heads: null + n_layers: 24 + norm_eps: 1.0e-05 + norm_type: rmsnorm + num_channels: 3 + patch_size: 16 + proj_bias: null + qkv_bias: null + rope_theta: 10000.0 + use_cache: false + use_rope_from_torchtitan: false + vision_encoder_in_channels: '3' + vision_end_token_id: '151653' + vision_start_token_id: '151652' + vision_token_id: '151654' + vocab_size: '152064' + z_loss_coeff: '0.0' + tokenizer: + _target_: + cache_dir: null + tokenizer_type: Qwen/Qwen2.5-VL-7B-Instruct + n_layers_per_group: 5 + s3_credential_path: credentials/s3_checkpoint.secret + tokenizer: + _target_: + chunk_duration: '81' + load_mean_std: 'False' + name: wan2pt1_tokenizer + temporal_window: '16' + train_time_distribution: logitnormal + train_time_weight: reweighting + use_dora: false + use_dynamic_shift: false + use_high_sigma_strategy: false + use_kerras_sigma_at_inference: true + use_lora: false + use_torch_compile: false +model_parallel: + _cpu_offloading_context: null + async_tensor_model_parallel_allreduce: false + autocast_dtype: torch.float32 + barrier_with_L1_time: true + batch_p2p_comm: true + batch_p2p_sync: true + bf16: false + context_parallel_size: 2 + cpu_offloading: false + cpu_offloading_activations: true + cpu_offloading_double_buffering: false + cpu_offloading_num_layers: 0 + cpu_offloading_weights: true + cross_entropy_fusion_impl: native + cross_entropy_loss_fusion: false + deallocate_pipeline_outputs: false + defer_embedding_wgrad_compute: false + delay_wgrad_compute: false + deterministic_mode: false + enable_autocast: false + expert_model_parallel_size: 1 + expert_tensor_parallel_size: 1 + finalize_model_grads_func: null + fp16: false + grad_scale_func: null + grad_sync_func: null + gradient_accumulation_fusion: false + hierarchical_context_parallel_sizes: null + microbatch_group_size_per_vp_stage: 1 + moe_extended_tp: false + no_sync_func: null + num_microbatches_with_partial_activation_checkpoints: null + overlap_moe_expert_parallel_comm: false + overlap_p2p_comm: false + overlap_p2p_comm_warmup_flush: false + param_sync_func: null + params_dtype: torch.float32 + perform_initialization: true + pipeline_dtype: null + pipeline_model_parallel_comm_backend: null + pipeline_model_parallel_size: 1 + sequence_parallel: false + tensor_model_parallel_size: 1 + timers: null + tp_comm_atomic_ag: false + tp_comm_atomic_rs: false + tp_comm_bootstrap_backend: nccl + tp_comm_bulk_dgrad: true + tp_comm_bulk_wgrad: true + tp_comm_overlap: false + tp_comm_overlap_ag: true + tp_comm_overlap_disable_fc1: false + tp_comm_overlap_disable_qkv: false + tp_comm_overlap_rs: true + tp_comm_overlap_rs_dgrad: false + tp_comm_split_ag: true + tp_comm_split_rs: true + use_cpu_initialization: false + use_ring_exchange_p2p: false + use_te_rng_tracker: false + variable_seq_lengths: false + virtual_pipeline_model_parallel_size: null + wgrad_deferral_limit: 0 +optimizer: + _target_: + betas: + - '0.9' + - '0.999' + eps: 1e-08 + fused: 'True' + lr: 3e-05 + model: null + optim_type: adamw + weight_decay: '0.001' +scheduler: + _target_: + cycle_lengths: + - '400000' + f_max: + - '0.99' + f_min: + - '0.4' + f_start: + - 1e-06 + verbosity_interval: '0' + warm_up_steps: + - '100' +trainer: + callbacks: + compile_tokenizer: + _target_: + compile_after_iterations: '4' + dynamic: 'False' + enabled: 'True' + dataloader_speed: + _target_: + every_n: '200' + save_s3: 'True' + step_size: '1' + device_monitor: + _target_: + every_n: '200' + log_memory_detail: 'True' + save_s3: 'True' + step_size: '1' + upload_every_n_mul: '10' + every_n_sample_ema: + _target_: + do_x0_prediction: 'False' + every_n: '1000000000000' + fps: '16' + guidance: + - '0.0' + - '3.0' + - '7.0' + is_ema: 'True' + n_sample_to_save: '128' + n_sigmas_for_x0_prediction: '4' + n_viz_sample: '3' + num_sampling_step: '35' + prompt_type: t5_xxl + run_at_start: 'False' + save_s3: 'True' + step_size: '1' + use_negative_prompt: 'False' + every_n_sample_reg: + _target_: + do_x0_prediction: 'False' + every_n: '1000000000000' + fps: '16' + guidance: + - '0.0' + - '3.0' + - '7.0' + is_ema: 'False' + n_sample_to_save: '128' + n_sigmas_for_x0_prediction: '4' + n_viz_sample: '3' + num_sampling_step: '35' + prompt_type: t5_xxl + run_at_start: 'False' + save_s3: 'True' + step_size: '1' + use_negative_prompt: 'False' + grad_clip: + _target_: + clip_norm: '0.1' + force_finite: 'True' + heart_beat: + _target_: + every_n: '10' + save_s3: 'True' + step_size: '1' + update_interval_in_minute: '20' + iter_speed: + _target_: + every_n: '200' + hit_thres: '5' + save_s3: 'True' + save_s3_every_log_n: '10' + low_prec: + _target_: + config: null + trainer: null + update_iter: '1' + manual_gc: + _target_: + every_n: '200' + warm_up: '5' + wandb: + _target_: + logging_iter_multipler: '1' + save_logging_iter_multipler: '10' + save_s3: 'True' + wandb_10x: + _target_: + logging_iter_multipler: '10' + save_logging_iter_multipler: '1' + save_s3: 'True' + cudnn: + benchmark: 'True' + deterministic: 'False' + ddp: + broadcast_buffers: 'True' + find_unused_parameters: 'False' + static_graph: 'True' + distributed_parallelism: fsdp + grad_accum_iter: '1' + grad_scaler_args: + enabled: 'False' + logging_iter: '200' + max_iter: '150000' + max_val_iter: null + memory_format: torch.preserve_format + profiling: + enable_memory_snapshot: 'False' + enable_profiling: 'False' + profile_freq: '1' + profile_memory: 'False' + record_shape: 'False' + save_s3: 'False' + target_ranks: + - '0' + - '1' + - '2' + - '3' + - '4' + - '5' + - '6' + - '7' + with_modules: 'True' + with_stack: 'True' + run_validation: 'False' + run_validation_on_start: 'False' + seed: '0' + straggler_detection: + analyze_backward: 'True' + analyze_dataloading: 'True' + analyze_forward: 'True' + analyze_optimizer: 'True' + enabled: 'True' + max_diff: '1.5' + profile_freq: '1' + raise_error: 'True' + report_freq: '100' + timeout_period: '999999999' + type: + validation_iter: '100' +upload_reproducible_setup: 'True' diff --git a/0219122156_cosmos_mip_spatial/mem_mipnerf360_575f06997f088dc5824bec5a960459303edd0154e65b3594e29810380ac7ea50/console.log b/0219122156_cosmos_mip_spatial/mem_mipnerf360_575f06997f088dc5824bec5a960459303edd0154e65b3594e29810380ac7ea50/console.log new file mode 100644 index 0000000000000000000000000000000000000000..cafc1d34beab3f2db4dbc2755f2d9f42aeb9b5ea --- /dev/null +++ b/0219122156_cosmos_mip_spatial/mem_mipnerf360_575f06997f088dc5824bec5a960459303edd0154e65b3594e29810380ac7ea50/console.log @@ -0,0 +1,22 @@ +[02-19 13:19:38|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpoint_db.py:181:path] Downloading checkpoint nvidia/Cosmos-Predict2.5-2B/base/post-trained(81edfebe-bd6a-4039-8c1d-737df1a790bf) +[02-19 13:19:38|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpoint_db.py:99:_download] Downloading checkpoint file from Hugging Face with {'repo_id': 'nvidia/Cosmos-Predict2.5-2B', 'repo_type': 'model', 'revision': '15a82a2ec231bc318692aa0456a36537c806e7d4', 'filename': 'base/post-trained/81edfebe-bd6a-4039-8c1d-737df1a790bf_ema_bf16.pt'} +[02-19 13:19:38|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpoint_db.py:181:path] Downloading checkpoint nvidia/Cosmos-Predict2.5-2B/base/pre-trained(d20b7120-df3e-4911-919d-db6e08bad31c) +[02-19 13:19:38|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpoint_db.py:99:_download] Downloading checkpoint file from Hugging Face with {'repo_id': 'nvidia/Cosmos-Predict2.5-2B', 'repo_type': 'model', 'revision': '15a82a2ec231bc318692aa0456a36537c806e7d4', 'filename': 'base/pre-trained/d20b7120-df3e-4911-919d-db6e08bad31c_ema_bf16.pt'} +[02-19 13:19:38|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpoint_db.py:181:path] Downloading checkpoint nvidia/Cosmos-Predict2.5-14B/base/pre-trained(54937b8c-29de-4f04-862c-e67b04ec41e8) +[02-19 13:19:38|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpoint_db.py:99:_download] Downloading checkpoint file from Hugging Face with {'repo_id': 'nvidia/Cosmos-Predict2.5-14B', 'repo_type': 'model', 'revision': '03eb354f35eae0d6e0c1be3c9f94d8551e125570', 'filename': 'base/pre-trained/54937b8c-29de-4f04-862c-e67b04ec41e8_ema_bf16.pt'} +[02-19 13:19:40|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpoint_db.py:181:path] Downloading checkpoint Wan2.1/vae(685afcaa-4de2-42fe-b7b9-69f7a2dee4d8) +[02-19 13:19:40|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpoint_db.py:99:_download] Downloading checkpoint file from Hugging Face with {'repo_id': 'nvidia/Cosmos-Predict2.5-2B', 'repo_type': 'model', 'revision': '6787e176dce74a101d922174a95dba29fa5f0c55', 'filename': 'tokenizer.pth'} +[02-19 13:19:42|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpoint_db.py:181:path] Downloading checkpoint Qwen/Qwen2.5-VL-7B-Instruct(7219c6c7-f878-4137-bbdb-76842ea85e70) +[02-19 13:19:42|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpoint_db.py:134:_download] Downloading checkpoint from Hugging Face with {'repo_id': 'nvidia/Cosmos-Reason1-7B', 'repo_type': 'model', 'revision': '3210bec0495fdc7a8d3dbb8d58da5711eab4b423', 'allow_patterns': ['*']} +[02-19 13:19:43|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpoint_db.py:181:path] Downloading checkpoint nvidia/Cosmos-Reason1.1-7B(cb3e3ffa-7b08-4c34-822d-61c7aa31a14f) +[02-19 13:19:43|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpoint_db.py:134:_download] Downloading checkpoint from Hugging Face with {'repo_id': 'nvidia/Cosmos-Reason1-7B', 'repo_type': 'model', 'revision': '3210bec0495fdc7a8d3dbb8d58da5711eab4b423', 'allow_patterns': ['*']} +[02-19 13:20:13|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/inference.py:60:__init__] Saved config to /home/bravo/zzhangg/MemBench/generation/outputs/cosmos/0219122156_cosmos_mip_spatial/mem_mipnerf360_575f06997f088dc5824bec5a960459303edd0154e65b3594e29810380ac7ea50/config.yaml +[02-19 13:20:13|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/inference.py:82:generate] Generating 1 samples: ['segment_4'] +[02-19 13:20:13|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/inference.py:86:generate] [1/1] Processing sample segment_4 +[02-19 13:20:13|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/inference.py:99:_generate_sample] Saved arguments to /home/bravo/zzhangg/MemBench/generation/outputs/cosmos/0219122156_cosmos_mip_spatial/mem_mipnerf360_575f06997f088dc5824bec5a960459303edd0154e65b3594e29810380ac7ea50/segment_4.json +[02-19 13:20:13|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/inference.py:113:_generate_sample] Guardrail checks on prompt are disabled +[02-19 13:20:13|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/inference.py:118:_generate_sample] Generating video with autoregressive mode... +[02-19 13:33:34|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/inference.py:134:_generate_sample] Autoregressive generation completed, preparing to save video... +[02-19 13:33:34|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/inference.py:172:_generate_sample] Guardrail checks on video are disabled +[02-19 13:33:34|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/inference.py:175:_generate_sample] Saving video to %s.mp4 ... +[02-19 13:33:36|SUCCESS|generation/pipelines/cosmos-predict2.5/cosmos_predict2/inference.py:178:_generate_sample] Saved video to /home/bravo/zzhangg/MemBench/generation/outputs/cosmos/0219122156_cosmos_mip_spatial/mem_mipnerf360_575f06997f088dc5824bec5a960459303edd0154e65b3594e29810380ac7ea50/segment_4.mp4 diff --git a/0219122156_cosmos_mip_spatial/mem_mipnerf360_575f06997f088dc5824bec5a960459303edd0154e65b3594e29810380ac7ea50/debug.log b/0219122156_cosmos_mip_spatial/mem_mipnerf360_575f06997f088dc5824bec5a960459303edd0154e65b3594e29810380ac7ea50/debug.log new file mode 100644 index 0000000000000000000000000000000000000000..dc1a95e70c7ed8a61a9d921a636aa00bab430759 --- /dev/null +++ b/0219122156_cosmos_mip_spatial/mem_mipnerf360_575f06997f088dc5824bec5a960459303edd0154e65b3594e29810380ac7ea50/debug.log @@ -0,0 +1,353 @@ +[02-19 13:19:25|DEBUG|generation/pipelines/cosmos-predict2.5/packages/cosmos-oss/cosmos_oss/init.py:176:init_output_dir] Flags(internal=False, smoke=False, verbose=False, experimental_checkpoints=False) +[02-19 13:19:33|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/env_parsers/env_parser.py:86:get_val_dict] getting val dict of CredentialEnvParser +[02-19 13:19:33|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/env_parsers/env_parser.py:86:get_val_dict] getting val dict of CredentialEnvParser +[02-19 13:19:35|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/visualize/video.py:31:] No module named 'ffmpegcv' +[02-19 13:19:35|CRITICAL|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/checkpointer/dcp.py:188:] for the back comptiable pytorch! New DefaultLoadPlanner class is created. +[02-19 13:19:35|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/inference.py:32:__init__] SetupArguments(output_dir=PosixPath('/home/bravo/zzhangg/MemBench/generation/outputs/cosmos/0219122156_cosmos_mip_spatial/mem_mipnerf360_575f06997f088dc5824bec5a960459303edd0154e65b3594e29810380ac7ea50') model='2B/post-trained' checkpoint_path='/home/bravo/zzhangg/MemBench/generation/pipelines/cosmos-predict2.5/models/Cosmos-Predict2.5-2B/base/post-trained/81edfebe-bd6a-4039-8c1d-737df1a790bf_ema_bf16.pt' experiment='Stage-c_pt_4-Index-2-Size-2B-Res-720-Fps-16-Note-rf_with_edm_ckpt' config_file='cosmos_predict2/_src/predict2/configs/video2world/config.py' context_parallel_size=1 offload_diffusion_model=False offload_tokenizer=False offload_text_encoder=False disable_guardrails=True offload_guardrail_models=True keep_going=True profile=False) +[02-19 13:19:35|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/datasets/augmentor_provider.py:94:augmentor_register] registering video_basic_augmentor_v1... +[02-19 13:19:35|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/datasets/augmentor_provider.py:94:augmentor_register] registering video_basic_augmentor_v2... +[02-19 13:19:35|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/datasets/augmentor_provider.py:94:augmentor_register] registering noframedrop_nocameramove_video_augmentor_v1... +[02-19 13:19:35|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/datasets/augmentor_provider.py:94:augmentor_register] registering nocameramove_video_augmentor_v1... +[02-19 13:19:35|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/datasets/augmentor_provider.py:94:augmentor_register] registering image_basic_augmentor... +[02-19 13:19:35|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/datasets/augmentor_provider.py:94:augmentor_register] registering image_basic_augmentor_without_embeddings... +[02-19 13:19:35|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/attention/utils/safe_log.py:35:debug] Flash Attention v3 is not supported because the Python package was not found. +[02-19 13:19:35|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/attention/utils/safe_log.py:35:debug] NATTEN Attention is not supported due to insufficient NATTEN version natten.__version__='0.21.0', expected at least NATTEN_MIN_RELEASE_VERSION=[0, 21, 5], or NATTEN_MIN_DEV_VERSION=([0, 21, 5], 9). +[02-19 13:19:35|CRITICAL|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:192:import_all_modules_from_package] Reloading all modules from package cosmos_predict2._src.predict2.configs.video2world.experiment +[02-19 13:19:35|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_predict2._src.predict2.configs.video2world.experiment.reason_embeddings +[02-19 13:19:35|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_predict2._src.predict2.configs.video2world.experiment.reason_embeddings.model_14b_reason_1p1 +[02-19 13:19:35|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_predict2._src.predict2.configs.video2world.experiment.reason_embeddings.model_14b_reason_1p1_rectified_flow +[02-19 13:19:35|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_predict2._src.predict2.configs.video2world.experiment.reason_embeddings.model_2B_reason_1p1 +[02-19 13:19:35|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_predict2._src.predict2.configs.video2world.experiment.reason_embeddings.model_2B_reason_1p1_rectified_flow +[02-19 13:19:36|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_predict2._src.predict2.configs.video2world.experiment.reason_embeddings.model_7b_reason_1p1 +[02-19 13:19:36|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_predict2._src.predict2.configs.video2world.experiment.reason_embeddings.stage3_14B_index_3 +[02-19 13:19:36|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_predict2._src.predict2.configs.video2world.experiment.reason_embeddings.stage3_2B +[02-19 13:19:36|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_predict2._src.predict2.configs.video2world.experiment.resume_text2world +[02-19 13:19:36|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_predict2._src.predict2.configs.video2world.experiment.resume_text2world.sparse_14B +[02-19 13:19:37|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_predict2._src.predict2.configs.video2world.experiment.resume_text2world.sparse_2B +[02-19 13:19:37|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_predict2._src.predict2.configs.video2world.experiment.resume_text2world.stage3 +[02-19 13:19:37|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_predict2._src.predict2.configs.video2world.experiment.resume_text2world.stage3_2B_ablation +[02-19 13:19:37|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_predict2._src.predict2.configs.video2world.experiment.resume_text2world.state3_14B_index_3 +[02-19 13:19:38|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_predict2._src.predict2.configs.video2world.experiment.specialized_model +[02-19 13:19:38|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_predict2._src.predict2.configs.video2world.experiment.specialized_model.SFT_14B_RF +[02-19 13:19:38|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_predict2._src.predict2.configs.video2world.experiment.specialized_model.SFT_2B +[02-19 13:19:38|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_predict2._src.predict2.configs.video2world.experiment.specialized_model.SFT_2B_RF +[02-19 13:19:38|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_face +[02-19 13:19:38|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_crowded +[02-19 13:19:38|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_high_motion +[02-19 13:19:38|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_robotics +[02-19 13:19:38|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_av +[02-19 13:19:38|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_physical_ai +[02-19 13:19:38|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_4k_cooldown +[02-19 13:19:38|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_face_only2 +[02-19 13:19:38|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_crowded_only2 +[02-19 13:19:38|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_high_motion_only2 +[02-19 13:19:38|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_robotics_only2 +[02-19 13:19:38|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_av_only2 +[02-19 13:19:38|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_physical_ai_only2 +[02-19 13:19:38|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_4k_cooldown_only2 +[02-19 13:19:38|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_face_high_sigma +[02-19 13:19:38|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_crowded_high_sigma +[02-19 13:19:38|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_high_motion_high_sigma +[02-19 13:19:38|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_robotics_high_sigma +[02-19 13:19:38|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_av_high_sigma +[02-19 13:19:38|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_physical_ai_high_sigma +[02-19 13:19:38|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_4k_cooldown_high_sigma +[02-19 13:19:38|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_face_high_sigma2 +[02-19 13:19:38|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_crowded_high_sigma2 +[02-19 13:19:38|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_robotics_high_sigma2 +[02-19 13:19:38|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-2-Size-2B-Res-720-Fps-16-Note-rf_with_edm_ckpt +[02-19 13:19:38|CRITICAL|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:192:import_all_modules_from_package] Reloading all modules from package cosmos_predict2.experiments +[02-19 13:19:38|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_predict2.experiments.base +[02-19 13:19:38|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_predict2.experiments.base.action +[02-19 13:19:38|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpoint_db.py:181:path] Downloading checkpoint nvidia/Cosmos-Predict2.5-2B/base/post-trained(81edfebe-bd6a-4039-8c1d-737df1a790bf) +[02-19 13:19:38|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpoint_db.py:99:_download] Downloading checkpoint file from Hugging Face with {'repo_id': 'nvidia/Cosmos-Predict2.5-2B', 'repo_type': 'model', 'revision': '15a82a2ec231bc318692aa0456a36537c806e7d4', 'filename': 'base/post-trained/81edfebe-bd6a-4039-8c1d-737df1a790bf_ema_bf16.pt'} +[02-19 13:19:38|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_predict2.experiments.base.cosmos_nemo_assets +[02-19 13:19:38|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpoint_db.py:181:path] Downloading checkpoint nvidia/Cosmos-Predict2.5-2B/base/pre-trained(d20b7120-df3e-4911-919d-db6e08bad31c) +[02-19 13:19:38|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpoint_db.py:99:_download] Downloading checkpoint file from Hugging Face with {'repo_id': 'nvidia/Cosmos-Predict2.5-2B', 'repo_type': 'model', 'revision': '15a82a2ec231bc318692aa0456a36537c806e7d4', 'filename': 'base/pre-trained/d20b7120-df3e-4911-919d-db6e08bad31c_ema_bf16.pt'} +[02-19 13:19:38|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_predict2.experiments.base.cosmos_nemo_assets_lora +[02-19 13:19:38|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_predict2.experiments.base.groot +[02-19 13:19:38|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpoint_db.py:181:path] Downloading checkpoint nvidia/Cosmos-Predict2.5-14B/base/pre-trained(54937b8c-29de-4f04-862c-e67b04ec41e8) +[02-19 13:19:38|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpoint_db.py:99:_download] Downloading checkpoint file from Hugging Face with {'repo_id': 'nvidia/Cosmos-Predict2.5-14B', 'repo_type': 'model', 'revision': '03eb354f35eae0d6e0c1be3c9f94d8551e125570', 'filename': 'base/pre-trained/54937b8c-29de-4f04-862c-e67b04ec41e8_ema_bf16.pt'} +[02-19 13:19:40|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/utils/model_loader.py:91:load_model_from_checkpoint] Overriding config checkpoint path with: /home/bravo/zzhangg/MemBench/generation/pipelines/cosmos-predict2.5/models/Cosmos-Predict2.5-2B/base/post-trained/81edfebe-bd6a-4039-8c1d-737df1a790bf_ema_bf16.pt +[02-19 13:19:40|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/misc.py:152:set_random_seed] Using random seed 0. +[02-19 13:19:40|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/utils/model_loader.py:111:load_model_from_checkpoint] Loading model from /home/bravo/zzhangg/MemBench/generation/pipelines/cosmos-predict2.5/models/Cosmos-Predict2.5-2B/base/post-trained/81edfebe-bd6a-4039-8c1d-737df1a790bf_ema_bf16.pt +[02-19 13:19:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/models/text2world_model_rectified_flow.py:132:__init__] DiffusionModel: precision torch.bfloat16 +[02-19 13:19:40|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpoint_db.py:181:path] Downloading checkpoint Wan2.1/vae(685afcaa-4de2-42fe-b7b9-69f7a2dee4d8) +[02-19 13:19:40|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpoint_db.py:99:_download] Downloading checkpoint file from Hugging Face with {'repo_id': 'nvidia/Cosmos-Predict2.5-2B', 'repo_type': 'model', 'revision': '6787e176dce74a101d922174a95dba29fa5f0c55', 'filename': 'tokenizer.pth'} +[02-19 13:19:41|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/tokenizers/wan2pt1.py:669:_video_vae] loading /home/bravo/.cache/huggingface/hub/models--nvidia--Cosmos-Predict2.5-2B/snapshots/6787e176dce74a101d922174a95dba29fa5f0c55/tokenizer.pth +[02-19 13:19:41|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/distributed.py:415:sync_model_states] Synchronizing model states from rank 0 to all ranks in process group [0]. +[02-19 13:19:41|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/timer.py:150:_log] Time spent on DiffusionModel: set_up_tokenizer: 1.12 s +[02-19 13:19:41|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #0-fps: + ReMapkey + input key: fps + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: fps + Dtype: None +[02-19 13:19:41|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #1-padding_mask: + ReMapkey + input key: padding_mask + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: padding_mask + Dtype: None +[02-19 13:19:41|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #2-text: + TextAttr + input key: ['t5_text_embeddings'] + Param count: 0 + Trainable: None + Dropout rate: 0.2 + Output key: [crossattn_emb] +[02-19 13:19:41|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #3-use_video_condition: + BooleanFlag + input key: fps + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: use_video_condition + This is a boolean flag +[02-19 13:19:41|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1624:build_pos_embed] Building positional embedding with rope3d class, impl +[02-19 13:19:41|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:886:__init__] Using AdaLN LoRA Flag: True. We enable bias if no AdaLN LoRA for backward compatibility. +[02-19 13:19:41|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:19:41|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:19:41|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:19:41|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:19:41|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:19:41|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:19:41|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:19:41|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:19:41|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:19:41|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:19:41|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:19:41|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:19:41|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:19:41|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:19:41|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:19:41|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:19:41|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:19:41|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:19:41|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:19:41|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:19:41|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:19:41|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:19:41|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:19:41|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:19:41|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:19:41|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:19:41|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:19:41|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:19:41|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:19:41|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:19:41|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:19:41|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:19:41|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:19:41|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:19:41|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:19:41|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:19:41|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:19:41|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:19:41|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:19:41|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:19:41|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:19:41|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:19:41|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:19:41|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:19:41|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:19:41|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:19:41|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:19:41|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:19:41|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:19:41|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:19:41|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:19:41|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:19:41|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:19:41|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:19:41|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:19:41|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:19:42|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1804:enable_selective_checkpoint] Enable selective checkpoint with predict2_2b_720_aggressive, for every 1 blocks. Total blocks: 28 +[02-19 13:19:42|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 0 +[02-19 13:19:42|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 1 +[02-19 13:19:42|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 2 +[02-19 13:19:42|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 3 +[02-19 13:19:42|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 4 +[02-19 13:19:42|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 5 +[02-19 13:19:42|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 6 +[02-19 13:19:42|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 7 +[02-19 13:19:42|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 8 +[02-19 13:19:42|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 9 +[02-19 13:19:42|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 10 +[02-19 13:19:42|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 11 +[02-19 13:19:42|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 12 +[02-19 13:19:42|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 13 +[02-19 13:19:42|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 14 +[02-19 13:19:42|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 15 +[02-19 13:19:42|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 16 +[02-19 13:19:42|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 17 +[02-19 13:19:42|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 18 +[02-19 13:19:42|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 19 +[02-19 13:19:42|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 20 +[02-19 13:19:42|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 21 +[02-19 13:19:42|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 22 +[02-19 13:19:42|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 23 +[02-19 13:19:42|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 24 +[02-19 13:19:42|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 25 +[02-19 13:19:42|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 26 +[02-19 13:19:42|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 27 +[02-19 13:19:42|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/timer.py:150:_log] Time spent on meta to cuda and broadcast model states: 0.14 s +[02-19 13:19:42|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/timer.py:150:_log] Time spent on Creating PyTorch model: 0.65 s +[02-19 13:19:42|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/timer.py:150:_log] Time spent on Creating PyTorch model and ema if enabled: 0.65 s +[02-19 13:19:42|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/timer.py:150:_log] Time spent on DiffusionModel: set_up_model: 0.65 s +[02-19 13:19:42|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/text_encoders/text_encoder.py:74:__init__] Instantiating text encoder model... +[02-19 13:19:42|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpoint_db.py:181:path] Downloading checkpoint Qwen/Qwen2.5-VL-7B-Instruct(7219c6c7-f878-4137-bbdb-76842ea85e70) +[02-19 13:19:42|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpoint_db.py:134:_download] Downloading checkpoint from Hugging Face with {'repo_id': 'nvidia/Cosmos-Reason1-7B', 'repo_type': 'model', 'revision': '3210bec0495fdc7a8d3dbb8d58da5711eab4b423', 'allow_patterns': ['*']} +[02-19 13:19:42|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/reason1/tokenizer/processor.py:75:__init__] Successfully loaded processor from local cache +[02-19 13:19:42|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/reason1/models/vlm_base.py:116:__init__] Setting torch default dtype from torch.float32 to torch.bfloat16 +[02-19 13:19:43|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/reason1/parallelisms/parallel_dims.py:50:_validate] dp_shard is set to -1, will be automatically determined based on world_size 1 // 1. +[02-19 13:19:43|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/reason1/parallelisms/parallel_dims.py:54:_validate] dp_shard is set to 1. +[02-19 13:19:43|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/reason1/parallelisms/parallel_dims.py:75:build_mesh] Building 0-D device mesh with [], [] +[02-19 13:19:43|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/reason1/parallelisms/parallel_dims.py:105:build_mesh] mesh: DeviceMesh('cuda', 0) +[02-19 13:19:43|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/reason1/parallelisms/parallelize_qwen.py:336:apply_ac] Applied selective activation checkpointing to the model +[02-19 13:19:43|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/reason1/models/vlm_base.py:123:__init__] Reset torch default dtype to torch.float32 +[02-19 13:19:43|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/text_encoders/text_encoder.py:84:__init__] Loading checkpoint from s3://bucket/cosmos_reasoning1/sft_exp700/sft_exp721-1_qwen7b_tl_721_5vs5_s3_balanced_n32_resume_16k/checkpoints/iter_000016000/model/. +[02-19 13:19:43|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpoint_db.py:181:path] Downloading checkpoint nvidia/Cosmos-Reason1.1-7B(cb3e3ffa-7b08-4c34-822d-61c7aa31a14f) +[02-19 13:19:43|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpoint_db.py:134:_download] Downloading checkpoint from Hugging Face with {'repo_id': 'nvidia/Cosmos-Reason1-7B', 'repo_type': 'model', 'revision': '3210bec0495fdc7a8d3dbb8d58da5711eab4b423', 'allow_patterns': ['*']} +[02-19 13:20:08|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/text_encoders/text_encoder.py:113:__init__] Finished loading checkpoint from /home/bravo/.cache/huggingface/hub/models--nvidia--Cosmos-Reason1-7B/snapshots/3210bec0495fdc7a8d3dbb8d58da5711eab4b423. +[02-19 13:20:08|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/text_encoders/text_encoder.py:116:__init__] Text encoder model instantiated +[02-19 13:20:08|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/timer.py:150:_log] Time spent on instantiate model: 27.69 s +[02-19 13:20:08|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/utils/model_loader.py:227:load_model_state_dict_from_checkpoint] Loading model cached locally from /home/bravo/zzhangg/MemBench/generation/pipelines/cosmos-predict2.5/models/Cosmos-Predict2.5-2B/base/post-trained/81edfebe-bd6a-4039-8c1d-737df1a790bf_ema_bf16.pt +[02-19 13:20:12|CRITICAL|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/models/text2world_model_rectified_flow.py:919:load_state_dict] load model in non-strict mode +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:12|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key t_embedding_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:13|CRITICAL|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/models/text2world_model_rectified_flow.py:920:load_state_dict] [RANK 0] _IncompatibleKeys(missing_keys=[], unexpected_keys=[], incorrect_shapes=[]) +[02-19 13:20:13|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/distributed.py:415:sync_model_states] Synchronizing model states from rank 0 to all ranks in process group [0]. +[02-19 13:20:13|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/inference.py:60:__init__] Saved config to /home/bravo/zzhangg/MemBench/generation/outputs/cosmos/0219122156_cosmos_mip_spatial/mem_mipnerf360_575f06997f088dc5824bec5a960459303edd0154e65b3594e29810380ac7ea50/config.yaml +[02-19 13:20:13|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/inference.py:82:generate] Generating 1 samples: ['segment_4'] +[02-19 13:20:13|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/inference.py:86:generate] [1/1] Processing sample segment_4 +[02-19 13:20:13|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/inference.py:93:_generate_sample] InferenceArguments(name='segment_4' prompt_path=None prompt='80-100%: Toward the end, the camera tilts upward, offering a top-down view of the table and vase. The surrounding garden and background elements are fully visible once more, completing the loop of camera movement. The video concludes with a stable shot of the table, ensuring spatial and environmental consistency throughout the entire sequence.' negative_prompt='' seed=46 guidance=7 inference_type= input_path=PosixPath('/home/bravo/zzhangg/MemBench/generation/outputs/cosmos/0219122156_cosmos_mip_spatial/mem_mipnerf360_575f06997f088dc5824bec5a960459303edd0154e65b3594e29810380ac7ea50/segment_3.mp4') resolution='none' num_output_frames=77 num_steps=35 enable_autoregressive=True chunk_size=77 chunk_overlap=1) +[02-19 13:20:13|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/inference.py:99:_generate_sample] Saved arguments to /home/bravo/zzhangg/MemBench/generation/outputs/cosmos/0219122156_cosmos_mip_spatial/mem_mipnerf360_575f06997f088dc5824bec5a960459303edd0154e65b3594e29810380ac7ea50/segment_4.json +[02-19 13:20:13|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/inference.py:113:_generate_sample] Guardrail checks on prompt are disabled +[02-19 13:20:13|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/inference.py:118:_generate_sample] Generating video with autoregressive mode... +[02-19 13:20:13|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/inference/video2world.py:730:generate_autoregressive_from_batch] Processing video input for autoregressive: /home/bravo/zzhangg/MemBench/generation/outputs/cosmos/0219122156_cosmos_mip_spatial/mem_mipnerf360_575f06997f088dc5824bec5a960459303edd0154e65b3594e29810380ac7ea50/segment_3.mp4 +[02-19 13:20:14|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/inference/video2world.py:798:generate_autoregressive_from_batch] Generating 1 chunks with chunk_size=77, chunk_overlap=1 for 77 total frames +[02-19 13:20:14|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/inference/video2world.py:816:generate_autoregressive_from_batch] Processing chunk 1/1, frames 0-77 +[02-19 13:20:15|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/inference/video2world.py:556:generate_vid2world] GPU memory usage after getting data_batch: 19.79 GB +[02-19 13:20:15|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/inference/video2world.py:594:generate_vid2world] [Memory Optimization] Starting latent sample generation +[02-19 13:33:34|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/inference/video2world.py:878:generate_autoregressive_from_batch] Generated final video with shape torch.Size([1, 3, 77, 704, 1280]) +[02-19 13:33:34|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/inference.py:134:_generate_sample] Autoregressive generation completed, preparing to save video... +[02-19 13:33:34|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/inference.py:172:_generate_sample] Guardrail checks on video are disabled +[02-19 13:33:34|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/inference.py:175:_generate_sample] Saving video to %s.mp4 ... +[02-19 13:33:35|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/easy_io/handlers/imageio_video_handler.py:163:dump_to_fileobj] mimsave_kwargs: {'fps': 16, 'quality': 5, 'macro_block_size': 1, 'ffmpeg_params': ['-s', '1280x704'], 'output_params': ['-f', 'mp4']} +[02-19 13:33:36|SUCCESS|generation/pipelines/cosmos-predict2.5/cosmos_predict2/inference.py:178:_generate_sample] Saved video to /home/bravo/zzhangg/MemBench/generation/outputs/cosmos/0219122156_cosmos_mip_spatial/mem_mipnerf360_575f06997f088dc5824bec5a960459303edd0154e65b3594e29810380ac7ea50/segment_4.mp4 diff --git a/0219122156_cosmos_mip_spatial/mem_mipnerf360_575f06997f088dc5824bec5a960459303edd0154e65b3594e29810380ac7ea50/segment_0.json b/0219122156_cosmos_mip_spatial/mem_mipnerf360_575f06997f088dc5824bec5a960459303edd0154e65b3594e29810380ac7ea50/segment_0.json new file mode 100644 index 0000000000000000000000000000000000000000..85173371f6fda000a2f43c41f5707f1779dc48c9 --- /dev/null +++ b/0219122156_cosmos_mip_spatial/mem_mipnerf360_575f06997f088dc5824bec5a960459303edd0154e65b3594e29810380ac7ea50/segment_0.json @@ -0,0 +1 @@ +{"name":"segment_0","prompt_path":null,"prompt":"0-20%: The video opens with a close-up view of the round wooden table with a central vase of dried leaves, placed on hexagonal stone paving. The surrounding garden features lush green grass and an ivy-covered brick wall. The camera begins a slow horizontal pan to the right, starting to reveal more of the garden's background.","negative_prompt":"","seed":42,"guidance":7,"inference_type":"text2world","input_path":null,"resolution":"none","num_output_frames":77,"num_steps":35,"enable_autoregressive":true,"chunk_size":77,"chunk_overlap":1} \ No newline at end of file diff --git a/0219122156_cosmos_mip_spatial/mem_mipnerf360_575f06997f088dc5824bec5a960459303edd0154e65b3594e29810380ac7ea50/segment_0.mp4 b/0219122156_cosmos_mip_spatial/mem_mipnerf360_575f06997f088dc5824bec5a960459303edd0154e65b3594e29810380ac7ea50/segment_0.mp4 new file mode 100644 index 0000000000000000000000000000000000000000..a0bf15fde8ac756405a8323c4786fee1e5a682b6 --- /dev/null +++ b/0219122156_cosmos_mip_spatial/mem_mipnerf360_575f06997f088dc5824bec5a960459303edd0154e65b3594e29810380ac7ea50/segment_0.mp4 @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:41745849c8889b46227265f7f67aa77c500e5e9f42745590d436bbf5e1d07e43 +size 927741 diff --git a/0219122156_cosmos_mip_spatial/mem_mipnerf360_575f06997f088dc5824bec5a960459303edd0154e65b3594e29810380ac7ea50/segment_1.json b/0219122156_cosmos_mip_spatial/mem_mipnerf360_575f06997f088dc5824bec5a960459303edd0154e65b3594e29810380ac7ea50/segment_1.json new file mode 100644 index 0000000000000000000000000000000000000000..4c33a86ce191a639f59db455dd830dcafd4329a4 --- /dev/null +++ b/0219122156_cosmos_mip_spatial/mem_mipnerf360_575f06997f088dc5824bec5a960459303edd0154e65b3594e29810380ac7ea50/segment_1.json @@ -0,0 +1 @@ +{"name":"segment_1","prompt_path":null,"prompt":"20-40%: As the camera continues its horizontal pan, it captures the table from different angles, showcasing the intricate design of the wooden slats and the central vase. The background elements, such as the ivy-covered wall and garden plants, begin to shift in and out of the frame. The camera then starts to tilt upward, changing the vertical perspective.","negative_prompt":"","seed":43,"guidance":7,"inference_type":"video2world","input_path":"/home/bravo/zzhangg/MemBench/generation/outputs/cosmos/0219122156_cosmos_mip_spatial/mem_mipnerf360_575f06997f088dc5824bec5a960459303edd0154e65b3594e29810380ac7ea50/segment_0.mp4","resolution":"none","num_output_frames":77,"num_steps":35,"enable_autoregressive":true,"chunk_size":77,"chunk_overlap":1} \ No newline at end of file diff --git a/0219122156_cosmos_mip_spatial/mem_mipnerf360_575f06997f088dc5824bec5a960459303edd0154e65b3594e29810380ac7ea50/segment_1.mp4 b/0219122156_cosmos_mip_spatial/mem_mipnerf360_575f06997f088dc5824bec5a960459303edd0154e65b3594e29810380ac7ea50/segment_1.mp4 new file mode 100644 index 0000000000000000000000000000000000000000..0e004a99812ad196a69efb87949b88aa41d7cd63 --- /dev/null +++ b/0219122156_cosmos_mip_spatial/mem_mipnerf360_575f06997f088dc5824bec5a960459303edd0154e65b3594e29810380ac7ea50/segment_1.mp4 @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bfe9b8cf152867619dfeb63d39cfdfca307e2144bf00bcedaf866b867ca64eca +size 1073159 diff --git a/0219122156_cosmos_mip_spatial/mem_mipnerf360_575f06997f088dc5824bec5a960459303edd0154e65b3594e29810380ac7ea50/segment_2.json b/0219122156_cosmos_mip_spatial/mem_mipnerf360_575f06997f088dc5824bec5a960459303edd0154e65b3594e29810380ac7ea50/segment_2.json new file mode 100644 index 0000000000000000000000000000000000000000..ce8a74a2932d6b6a0cae14ca96e60dcb11147426 --- /dev/null +++ b/0219122156_cosmos_mip_spatial/mem_mipnerf360_575f06997f088dc5824bec5a960459303edd0154e65b3594e29810380ac7ea50/segment_2.json @@ -0,0 +1 @@ +{"name":"segment_2","prompt_path":null,"prompt":"40-60%: The camera completes its horizontal pan, returning to a similar angle as the beginning, but now with a slightly adjusted vertical perspective. The focus remains on the table and vase, with the background elements consistently visible. The camera then tilts downward, providing a closer view of the table's base and the stone paving.","negative_prompt":"","seed":44,"guidance":7,"inference_type":"video2world","input_path":"/home/bravo/zzhangg/MemBench/generation/outputs/cosmos/0219122156_cosmos_mip_spatial/mem_mipnerf360_575f06997f088dc5824bec5a960459303edd0154e65b3594e29810380ac7ea50/segment_1.mp4","resolution":"none","num_output_frames":77,"num_steps":35,"enable_autoregressive":true,"chunk_size":77,"chunk_overlap":1} \ No newline at end of file diff --git a/0219122156_cosmos_mip_spatial/mem_mipnerf360_575f06997f088dc5824bec5a960459303edd0154e65b3594e29810380ac7ea50/segment_2.mp4 b/0219122156_cosmos_mip_spatial/mem_mipnerf360_575f06997f088dc5824bec5a960459303edd0154e65b3594e29810380ac7ea50/segment_2.mp4 new file mode 100644 index 0000000000000000000000000000000000000000..921d2d5adebffa1db72cabf997b511582b9b502a --- /dev/null +++ b/0219122156_cosmos_mip_spatial/mem_mipnerf360_575f06997f088dc5824bec5a960459303edd0154e65b3594e29810380ac7ea50/segment_2.mp4 @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bbd5ec4e95db70377d3c465101d7205d25e9289c42aeea146df0c9e9da35b6ac +size 1446517 diff --git a/0219122156_cosmos_mip_spatial/mem_mipnerf360_575f06997f088dc5824bec5a960459303edd0154e65b3594e29810380ac7ea50/segment_3.json b/0219122156_cosmos_mip_spatial/mem_mipnerf360_575f06997f088dc5824bec5a960459303edd0154e65b3594e29810380ac7ea50/segment_3.json new file mode 100644 index 0000000000000000000000000000000000000000..c9373cd06be80f855dcc1d2e82761fde2ccfb0ae --- /dev/null +++ b/0219122156_cosmos_mip_spatial/mem_mipnerf360_575f06997f088dc5824bec5a960459303edd0154e65b3594e29810380ac7ea50/segment_3.json @@ -0,0 +1 @@ +{"name":"segment_3","prompt_path":null,"prompt":"60-80%: The camera continues to tilt downward, emphasizing the details of the table's construction and the texture of the stone paving. The background elements, including the ivy-covered wall and garden plants, remain consistent as they re-enter the frame. The camera then begins to tilt upward again, preparing for a final vertical perspective change.","negative_prompt":"","seed":45,"guidance":7,"inference_type":"video2world","input_path":"/home/bravo/zzhangg/MemBench/generation/outputs/cosmos/0219122156_cosmos_mip_spatial/mem_mipnerf360_575f06997f088dc5824bec5a960459303edd0154e65b3594e29810380ac7ea50/segment_2.mp4","resolution":"none","num_output_frames":77,"num_steps":35,"enable_autoregressive":true,"chunk_size":77,"chunk_overlap":1} \ No newline at end of file diff --git a/0219122156_cosmos_mip_spatial/mem_mipnerf360_575f06997f088dc5824bec5a960459303edd0154e65b3594e29810380ac7ea50/segment_3.mp4 b/0219122156_cosmos_mip_spatial/mem_mipnerf360_575f06997f088dc5824bec5a960459303edd0154e65b3594e29810380ac7ea50/segment_3.mp4 new file mode 100644 index 0000000000000000000000000000000000000000..5f8bbd1b98dd0935023ca3696c4a4b8a948fcc89 --- /dev/null +++ b/0219122156_cosmos_mip_spatial/mem_mipnerf360_575f06997f088dc5824bec5a960459303edd0154e65b3594e29810380ac7ea50/segment_3.mp4 @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ef1727b0f6e2a4837a4a32131c0fb3a652d950e45bc84a421be763ee170e1cdf +size 1100231 diff --git a/0219122156_cosmos_mip_spatial/mem_mipnerf360_575f06997f088dc5824bec5a960459303edd0154e65b3594e29810380ac7ea50/segment_4.json b/0219122156_cosmos_mip_spatial/mem_mipnerf360_575f06997f088dc5824bec5a960459303edd0154e65b3594e29810380ac7ea50/segment_4.json new file mode 100644 index 0000000000000000000000000000000000000000..46bc22d4d94815c19e6db8de779489f26ce940c6 --- /dev/null +++ b/0219122156_cosmos_mip_spatial/mem_mipnerf360_575f06997f088dc5824bec5a960459303edd0154e65b3594e29810380ac7ea50/segment_4.json @@ -0,0 +1 @@ +{"name":"segment_4","prompt_path":null,"prompt":"80-100%: Toward the end, the camera tilts upward, offering a top-down view of the table and vase. The surrounding garden and background elements are fully visible once more, completing the loop of camera movement. The video concludes with a stable shot of the table, ensuring spatial and environmental consistency throughout the entire sequence.","negative_prompt":"","seed":46,"guidance":7,"inference_type":"video2world","input_path":"/home/bravo/zzhangg/MemBench/generation/outputs/cosmos/0219122156_cosmos_mip_spatial/mem_mipnerf360_575f06997f088dc5824bec5a960459303edd0154e65b3594e29810380ac7ea50/segment_3.mp4","resolution":"none","num_output_frames":77,"num_steps":35,"enable_autoregressive":true,"chunk_size":77,"chunk_overlap":1} \ No newline at end of file diff --git a/0219122156_cosmos_mip_spatial/mem_mipnerf360_575f06997f088dc5824bec5a960459303edd0154e65b3594e29810380ac7ea50/segment_4.mp4 b/0219122156_cosmos_mip_spatial/mem_mipnerf360_575f06997f088dc5824bec5a960459303edd0154e65b3594e29810380ac7ea50/segment_4.mp4 new file mode 100644 index 0000000000000000000000000000000000000000..6499999bb2b72b462b063df1a142d8f0076e1738 --- /dev/null +++ b/0219122156_cosmos_mip_spatial/mem_mipnerf360_575f06997f088dc5824bec5a960459303edd0154e65b3594e29810380ac7ea50/segment_4.mp4 @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cc90364d5e6dcfefeb851a7610919d281f3f26a08c1a482eff9744ba1319cbee +size 1162361 diff --git a/0219122156_cosmos_mip_spatial/mem_mipnerf360_5f84c46d31489bd1a6b57b7c5c8cbc15280e279f8eb03122abd9362ac7333d13/config.yaml b/0219122156_cosmos_mip_spatial/mem_mipnerf360_5f84c46d31489bd1a6b57b7c5c8cbc15280e279f8eb03122abd9362ac7333d13/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..9a39eee54ea6efbbedd8ae0ae81f622fd73dbca0 --- /dev/null +++ b/0219122156_cosmos_mip_spatial/mem_mipnerf360_5f84c46d31489bd1a6b57b7c5c8cbc15280e279f8eb03122abd9362ac7333d13/config.yaml @@ -0,0 +1,760 @@ +checkpoint: + broadcast_via_filesystem: 'False' + dcp_allow_mismatched_size: 'False' + dcp_async_mode_enabled: 'False' + enable_gcs_patch_in_boto3: 'False' + jit: + device: cuda + dtype: bfloat16 + enabled: 'False' + input_shape: null + strict: 'True' + keys_not_to_resume: [] + load_ema_to_reg: 'False' + load_from_object_store: + bucket: bucket + credentials: credentials/s3_checkpoint.secret + enabled: 'True' + load_path: /home/bravo/zzhangg/MemBench/generation/pipelines/cosmos-predict2.5/models/Cosmos-Predict2.5-2B/base/post-trained/81edfebe-bd6a-4039-8c1d-737df1a790bf_ema_bf16.pt + load_training_state: 'False' + only_load_scheduler_state: 'False' + save_iter: '1000' + save_to_object_store: + bucket: bucket + credentials: credentials/s3_checkpoint.secret + enabled: 'True' + strict_resume: 'True' + type: + _target_: + callbacks: null + disable_async: 'False' + verbose: 'True' +dataloader_train: + dataloaders: + image_data: + dataloader: + batch_size: '12' + cache_augment_fn: functools.partial(, + n=1) + cache_size: '8' + concat_size: '1' + dataset: + augmentor_name: image_basic_augmentor_without_embeddings + caption_type: qwen2p5_7b_v4 + dataset_resolution_type: gt720p + embedding_type: null + resolution: '720' + num_workers: '6' + use_cache: 'False' + ratio: '1' + video_data: + dataloader: + batch_size: '1' + cache_augment_fn: functools.partial(, n=1.8) + cache_size: '16' + concat_size: '1' + dataset: + augmentor_name: video_basic_augmentor_v2 + caption_type: t2w_qwen2p5_7b + dataset_resolution_type: all + embedding_type: null + max_fps_thres: '60' + min_fps_thres: '10' + num_video_frames: '93' + resolution: '720' + use_native_fps: 'True' + video_decoder_name: video_naive_bytes + use_cache: 'False' + ratio: '3' +dataloader_val: + _target_: + dataloaders: + image_data: + dataloader: + _target_: + batch_size: '2' + cache_augment_fn: null + cache_replay_name: image_dataloader + cache_size: '32' + concat_size: '1' + dataset: + _target_: + len_t5: '512' + resolution: '512' + t5_dim: '1024' + num_workers: '8' + pin_memory: 'True' + shuffle: 'False' + use_cache: 'False' + webdataset: 'False' + ratio: '1' + video_data: + dataloader: + _target_: + batch_size: '1' + cache_augment_fn: null + cache_replay_name: video_dataloader + cache_size: '32' + concat_size: '1' + dataset: + _target_: + len_t5: '512' + num_video_frames: '136' + resolution: '512' + t5_dim: '1024' + num_workers: '8' + pin_memory: 'True' + shuffle: 'False' + use_cache: 'False' + webdataset: 'False' + ratio: '1' +defaults: +- _self_ +- data_train: mock +- data_val: mock +- optimizer: fusedadamw +- scheduler: lambdalinear +- model: ddp +- callbacks: basic +- net: null +- conditioner: video_prediction_conditioner +- ema: power +- tokenizer: cosmos_tokenizer_causal_cv8x8x8_c16_res720_t121_it121_v1_0 +- checkpoint: s3 +- ckpt_type: dummy +- experiment: null +job: + cluster: null + group: official_runs_vid2vid + name: Stage-c_pt_4-Index-2-Size-2B-Res-720-Fps-16-Note-rf_with_edm_ckpt + project: cosmos_diffusion_v2 + wandb_mode: online +model: + _recursive_: 'False' + _target_: + config: + conditional_frame_timestep: 0.1 + conditional_frames_probs: + 0: 0.5 + 1: 0.25 + 2: 0.25 + conditioner: + _target_: + fps: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: fps + output_key: fps + padding_mask: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: padding_mask + output_key: padding_mask + text: + _target_: + credential_path: credentials/s3_training.secret + dropout_rate: '0.2' + empty_string_embeddings_path: s3://bucket/predict2_assets/reason1_empty_string_embeddings.pt + input_key: + - t5_text_embeddings + use_empty_string: 'False' + use_video_condition: + _target_: + dropout_rate: '0.0' + input_key: fps + output_key: use_video_condition + conditioning_strategy: frame_replace + denoise_replace_gt_frames: true + ema: + enabled: false + iteration_shift: 0 + rate: 0.1 + fsdp_shard_size: 1 + high_sigma_ratio: 0.05 + high_sigma_timesteps_max: 1000 + high_sigma_timesteps_min: 980 + init_lora_weights: true + input_caption_key: ai_caption + input_data_key: video + input_image_key: images + lora_alpha: 32 + lora_rank: 32 + lora_target_modules: q_proj,k_proj,v_proj,output_proj,mlp.layer1,mlp.layer2 + max_num_conditional_frames: 2 + min_num_conditional_frames: 0 + net: + _target_: + adaln_lora_dim: '256' + atten_backend: minimal_a2a + concat_padding_mask: 'True' + crossattn_emb_channels: '1024' + crossattn_proj_in_channels: '100352' + extra_per_block_abs_pos_emb: 'False' + in_channels: '16' + max_frames: '128' + max_img_h: '240' + max_img_w: '240' + model_channels: '2048' + num_blocks: '28' + num_heads: '16' + out_channels: '16' + patch_spatial: '2' + patch_temporal: '1' + pos_emb_cls: rope3d + pos_emb_interpolation: crop + pos_emb_learnable: 'True' + rope_enable_fps_modulation: 'False' + rope_h_extrapolation_ratio: '3.0' + rope_t_extrapolation_ratio: '1.0' + rope_w_extrapolation_ratio: '3.0' + sac_config: + every_n_blocks: 1 + mode: predict2_2b_720_aggressive + timestep_scale: '0.001' + use_adaln_lora: 'True' + use_crossattn_projection: 'True' + use_wan_fp32_strategy: 'True' + precision: bfloat16 + resolution: '720' + shift: 5 + state_ch: 16 + state_t: 24 + text_encoder_class: reason1p1_7B + text_encoder_config: + ckpt_path: s3://bucket/cosmos_reasoning1/sft_exp700/sft_exp721-1_qwen7b_tl_721_5vs5_s3_balanced_n32_resume_16k/checkpoints/iter_000016000/model/ + compute_online: true + embedding_concat_strategy: full_concat + model_config: + _target_: + model_config: + _target_: cosmos_predict2._src.reason1.configs.default.model_config_qwen.QwenModelConfig + activation_checkpoint: + mode: selective + models: vlm + selective_ac_option: op + add_answer_tag: 'True' + add_cross_attention: 'False' + add_image_start_end_tag: 'False' + add_tile_tag: 'False' + architectures: + - Qwen2_5_VLForConditionalGeneration + attention_dropout: '0.0' + attn_implementation: flash_attention_2 + attn_implementation_autoset: 'True' + aux_loss_coeff: '0.0' + bad_words_ids: null + begin_suppress_tokens: null + bos_token_id: '151643' + cache_dir: null + checkpoint: + async_mode: disabled + create_seed_checkpoint: false + enable_checkpoint: false + export_dtype: float32 + folder: checkpoint + interval: 500 + interval_type: steps + model_weights_only: false + chunk_size_feed_forward: '0' + ckpt_dir: null + ckpt_path: null + comm: + init_timeout_seconds: 300 + trace_buf_size: 20000 + train_timeout_seconds: 100 + cp_size: null + cross_attention_hidden_size: null + decoder_start_token_id: null + deterministic: 'False' + diversity_penalty: '0.0' + do_sample: 'False' + early_stopping: 'False' + encoder_no_repeat_ngram_size: '0' + eos_token_id: '151645' + ep_size: null + experimental: + enable_async_tensor_parallel: false + enable_compiled_autograd: false + pipeline_parallel_degree: 1 + exponential_decay_length_penalty: null + finetuning_task: null + float8: + enable_float8_linear: false + forced_bos_token_id: null + forced_eos_token_id: null + freeze_llm: 'False' + freeze_mm_projector: 'False' + freeze_vision_encoder: 'False' + fsdp_enabled: 'False' + hidden_act: silu + hidden_size: '3584' + id2label: + 0: LABEL_0 + 1: LABEL_1 + image_token_id: '151655' + initializer_range: '0.02' + intermediate_size: '18944' + is_decoder: 'False' + is_encoder_decoder: 'False' + label2id: + LABEL_0: '0' + LABEL_1: '1' + length_penalty: '1.0' + loss_per_token: 'True' + max_batch_size: '1' + max_length: '20' + max_position_embeddings: '128000' + max_seq_len: '128000' + max_window_layers: '28' + min_length: '0' + mm_projector: null + model_type: qwen2_5_vl + name_or_path: Qwen/Qwen2.5-VL-7B-Instruct + no_repeat_ngram_size: '0' + num_attention_heads: '28' + num_beam_groups: '1' + num_beams: '1' + num_hidden_layers: '28' + num_key_value_heads: '4' + num_return_sequences: '1' + num_tiles: '1' + optimizer: + early_step_in_backward: false + end_lr: 2.5e-05 + fused: false + init_lr: 1.0e-05 + lr: 0.0003 + lr_multiplier_llm: 1.0 + lr_multiplier_mm_projector: 1.0 + lr_multiplier_vision_encoder: 0.1 + name: AdamW + output_attentions: 'False' + output_hidden_states: 'True' + output_scores: 'False' + pad_token_id: null + precision: bfloat16 + prefix: null + prepend_padding: 'False' + problem_type: null + pruned_heads: _Nothing.NOTHING + remove_invalid_values: 'False' + repetition_penalty: '1.0' + return_dict: 'True' + return_dict_in_generate: 'False' + rms_norm_eps: 1e-06 + rope_scaling: + mrope_section: + - '16' + - '24' + - '24' + rope_type: default + type: default + rope_theta: '1000000.0' + s3_credential_path: credentials/pbss_dir.secret + seed: '0' + sep_token_id: null + sliding_window: '32768' + suppress_tokens: null + task_specific_params: null + temperature: '1.0' + tf_legacy_loss: 'False' + tie_encoder_decoder: 'False' + tie_word_embeddings: 'False' + tile_tag_type: space_separated + tokenizer_class: null + tokenizer_type: Qwen/Qwen2.5-VL-7B-Instruct + top_k: '50' + top_p: '1.0' + torch_dtype: bfloat16 + torchscript: 'False' + training: + compile: false + context_parallel_degree: 1 + data_parallel_replicate_degree: 1 + data_parallel_shard_degree: -1 + disable_loss_parallel: false + enable_cpu_offload: false + fsdp_reshard_after_forward: default + mixed_precision_param: bfloat16 + mixed_precision_reduce: float32 + steps: 400000 + tensor_parallel_degree: 1 + use_cosine_decay: false + use_linear_decay: true + warmup_steps: 1000 + training_seq_len: '4096' + transformers_version: 4.51.0.dev0 + typical_p: '1.0' + use_bfloat16: 'False' + use_cache: 'False' + use_fsdp2: 'True' + use_return_dict: 'True' + use_rope_from_torchtitan: 'False' + use_sliding_window: 'False' + video_token_id: '151656' + vision_config: + _target_: cosmos_predict2._src.reason1.configs.default.model_config_qwen.QwenVisionConfig + add_cross_attention: 'False' + architectures: null + attn_implementation: flash_attention_2 + attn_implementation_autoset: 'True' + bad_words_ids: null + begin_suppress_tokens: null + bos_token_id: null + chunk_size_feed_forward: '0' + cross_attention_hidden_size: null + decoder_start_token_id: null + depth: '32' + diversity_penalty: '0.0' + do_sample: 'False' + early_stopping: 'False' + embed_dim: null + encoder_no_repeat_ngram_size: '0' + eos_token_id: null + exponential_decay_length_penalty: null + finetuning_task: null + forced_bos_token_id: null + forced_eos_token_id: null + fullatt_block_indexes: + - '7' + - '15' + - '23' + - '31' + hidden_act: silu + hidden_size: '1280' + id2label: + 0: LABEL_0 + 1: LABEL_1 + in_channels: '3' + in_chans: '3' + intermediate_size: '3420' + is_decoder: 'False' + is_encoder_decoder: 'False' + label2id: + LABEL_0: '0' + LABEL_1: '1' + length_penalty: '1.0' + max_length: '20' + min_length: '0' + mlp_ratio: null + model_type: qwen2_5_vl + name_or_path: '' + no_repeat_ngram_size: '0' + num_beam_groups: '1' + num_beams: '1' + num_heads: '16' + num_return_sequences: '1' + out_hidden_size: '3584' + output_attentions: 'False' + output_hidden_states: 'False' + output_scores: 'False' + pad_token_id: null + patch_size: '14' + prefix: null + problem_type: null + pruned_heads: _Nothing.NOTHING + remove_invalid_values: 'False' + repetition_penalty: '1.0' + return_dict: 'True' + return_dict_in_generate: 'False' + sep_token_id: null + spatial_merge_size: '2' + spatial_patch_size: '14' + suppress_tokens: null + task_specific_params: null + temperature: '1.0' + temporal_patch_size: '2' + tf_legacy_loss: 'False' + tie_encoder_decoder: 'False' + tie_word_embeddings: 'True' + tokenizer_class: null + tokens_per_second: '2' + top_k: '50' + top_p: '1.0' + torch_dtype: bfloat16 + torchscript: 'False' + typical_p: '1.0' + use_bfloat16: 'False' + window_size: '112' + vision_encoder: openai/clip-vit-base-patch32 + vision_encoder_config: + depth_init: true + dim: 1024 + ffn_dim_multiplier: null + head_dim: null + hidden_act: null + hidden_dim: 4096 + image_size: 1024 + image_token_id: null + multiple_of: null + n_heads: 16 + n_kv_heads: null + n_layers: 24 + norm_eps: 1.0e-05 + norm_type: rmsnorm + num_channels: 3 + patch_size: 16 + proj_bias: null + qkv_bias: null + rope_theta: 10000.0 + use_cache: false + use_rope_from_torchtitan: false + vision_encoder_in_channels: '3' + vision_end_token_id: '151653' + vision_start_token_id: '151652' + vision_token_id: '151654' + vocab_size: '152064' + z_loss_coeff: '0.0' + tokenizer: + _target_: + cache_dir: null + tokenizer_type: Qwen/Qwen2.5-VL-7B-Instruct + n_layers_per_group: 5 + s3_credential_path: credentials/s3_checkpoint.secret + tokenizer: + _target_: + chunk_duration: '81' + load_mean_std: 'False' + name: wan2pt1_tokenizer + temporal_window: '16' + train_time_distribution: logitnormal + train_time_weight: reweighting + use_dora: false + use_dynamic_shift: false + use_high_sigma_strategy: false + use_kerras_sigma_at_inference: true + use_lora: false + use_torch_compile: false +model_parallel: + _cpu_offloading_context: null + async_tensor_model_parallel_allreduce: false + autocast_dtype: torch.float32 + barrier_with_L1_time: true + batch_p2p_comm: true + batch_p2p_sync: true + bf16: false + context_parallel_size: 2 + cpu_offloading: false + cpu_offloading_activations: true + cpu_offloading_double_buffering: false + cpu_offloading_num_layers: 0 + cpu_offloading_weights: true + cross_entropy_fusion_impl: native + cross_entropy_loss_fusion: false + deallocate_pipeline_outputs: false + defer_embedding_wgrad_compute: false + delay_wgrad_compute: false + deterministic_mode: false + enable_autocast: false + expert_model_parallel_size: 1 + expert_tensor_parallel_size: 1 + finalize_model_grads_func: null + fp16: false + grad_scale_func: null + grad_sync_func: null + gradient_accumulation_fusion: false + hierarchical_context_parallel_sizes: null + microbatch_group_size_per_vp_stage: 1 + moe_extended_tp: false + no_sync_func: null + num_microbatches_with_partial_activation_checkpoints: null + overlap_moe_expert_parallel_comm: false + overlap_p2p_comm: false + overlap_p2p_comm_warmup_flush: false + param_sync_func: null + params_dtype: torch.float32 + perform_initialization: true + pipeline_dtype: null + pipeline_model_parallel_comm_backend: null + pipeline_model_parallel_size: 1 + sequence_parallel: false + tensor_model_parallel_size: 1 + timers: null + tp_comm_atomic_ag: false + tp_comm_atomic_rs: false + tp_comm_bootstrap_backend: nccl + tp_comm_bulk_dgrad: true + tp_comm_bulk_wgrad: true + tp_comm_overlap: false + tp_comm_overlap_ag: true + tp_comm_overlap_disable_fc1: false + tp_comm_overlap_disable_qkv: false + tp_comm_overlap_rs: true + tp_comm_overlap_rs_dgrad: false + tp_comm_split_ag: true + tp_comm_split_rs: true + use_cpu_initialization: false + use_ring_exchange_p2p: false + use_te_rng_tracker: false + variable_seq_lengths: false + virtual_pipeline_model_parallel_size: null + wgrad_deferral_limit: 0 +optimizer: + _target_: + betas: + - '0.9' + - '0.999' + eps: 1e-08 + fused: 'True' + lr: 3e-05 + model: null + optim_type: adamw + weight_decay: '0.001' +scheduler: + _target_: + cycle_lengths: + - '400000' + f_max: + - '0.99' + f_min: + - '0.4' + f_start: + - 1e-06 + verbosity_interval: '0' + warm_up_steps: + - '100' +trainer: + callbacks: + compile_tokenizer: + _target_: + compile_after_iterations: '4' + dynamic: 'False' + enabled: 'True' + dataloader_speed: + _target_: + every_n: '200' + save_s3: 'True' + step_size: '1' + device_monitor: + _target_: + every_n: '200' + log_memory_detail: 'True' + save_s3: 'True' + step_size: '1' + upload_every_n_mul: '10' + every_n_sample_ema: + _target_: + do_x0_prediction: 'False' + every_n: '1000000000000' + fps: '16' + guidance: + - '0.0' + - '3.0' + - '7.0' + is_ema: 'True' + n_sample_to_save: '128' + n_sigmas_for_x0_prediction: '4' + n_viz_sample: '3' + num_sampling_step: '35' + prompt_type: t5_xxl + run_at_start: 'False' + save_s3: 'True' + step_size: '1' + use_negative_prompt: 'False' + every_n_sample_reg: + _target_: + do_x0_prediction: 'False' + every_n: '1000000000000' + fps: '16' + guidance: + - '0.0' + - '3.0' + - '7.0' + is_ema: 'False' + n_sample_to_save: '128' + n_sigmas_for_x0_prediction: '4' + n_viz_sample: '3' + num_sampling_step: '35' + prompt_type: t5_xxl + run_at_start: 'False' + save_s3: 'True' + step_size: '1' + use_negative_prompt: 'False' + grad_clip: + _target_: + clip_norm: '0.1' + force_finite: 'True' + heart_beat: + _target_: + every_n: '10' + save_s3: 'True' + step_size: '1' + update_interval_in_minute: '20' + iter_speed: + _target_: + every_n: '200' + hit_thres: '5' + save_s3: 'True' + save_s3_every_log_n: '10' + low_prec: + _target_: + config: null + trainer: null + update_iter: '1' + manual_gc: + _target_: + every_n: '200' + warm_up: '5' + wandb: + _target_: + logging_iter_multipler: '1' + save_logging_iter_multipler: '10' + save_s3: 'True' + wandb_10x: + _target_: + logging_iter_multipler: '10' + save_logging_iter_multipler: '1' + save_s3: 'True' + cudnn: + benchmark: 'True' + deterministic: 'False' + ddp: + broadcast_buffers: 'True' + find_unused_parameters: 'False' + static_graph: 'True' + distributed_parallelism: fsdp + grad_accum_iter: '1' + grad_scaler_args: + enabled: 'False' + logging_iter: '200' + max_iter: '150000' + max_val_iter: null + memory_format: torch.preserve_format + profiling: + enable_memory_snapshot: 'False' + enable_profiling: 'False' + profile_freq: '1' + profile_memory: 'False' + record_shape: 'False' + save_s3: 'False' + target_ranks: + - '0' + - '1' + - '2' + - '3' + - '4' + - '5' + - '6' + - '7' + with_modules: 'True' + with_stack: 'True' + run_validation: 'False' + run_validation_on_start: 'False' + seed: '0' + straggler_detection: + analyze_backward: 'True' + analyze_dataloading: 'True' + analyze_forward: 'True' + analyze_optimizer: 'True' + enabled: 'True' + max_diff: '1.5' + profile_freq: '1' + raise_error: 'True' + report_freq: '100' + timeout_period: '999999999' + type: + validation_iter: '100' +upload_reproducible_setup: 'True' diff --git a/0219122156_cosmos_mip_spatial/mem_mipnerf360_5f84c46d31489bd1a6b57b7c5c8cbc15280e279f8eb03122abd9362ac7333d13/console.log b/0219122156_cosmos_mip_spatial/mem_mipnerf360_5f84c46d31489bd1a6b57b7c5c8cbc15280e279f8eb03122abd9362ac7333d13/console.log new file mode 100644 index 0000000000000000000000000000000000000000..b774bf6d487f4729ab57558f1bff3d5eaf7d1cbd --- /dev/null +++ b/0219122156_cosmos_mip_spatial/mem_mipnerf360_5f84c46d31489bd1a6b57b7c5c8cbc15280e279f8eb03122abd9362ac7333d13/console.log @@ -0,0 +1,22 @@ +[02-19 13:19:52|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpoint_db.py:181:path] Downloading checkpoint nvidia/Cosmos-Predict2.5-2B/base/post-trained(81edfebe-bd6a-4039-8c1d-737df1a790bf) +[02-19 13:19:52|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpoint_db.py:99:_download] Downloading checkpoint file from Hugging Face with {'repo_id': 'nvidia/Cosmos-Predict2.5-2B', 'repo_type': 'model', 'revision': '15a82a2ec231bc318692aa0456a36537c806e7d4', 'filename': 'base/post-trained/81edfebe-bd6a-4039-8c1d-737df1a790bf_ema_bf16.pt'} +[02-19 13:19:52|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpoint_db.py:181:path] Downloading checkpoint nvidia/Cosmos-Predict2.5-2B/base/pre-trained(d20b7120-df3e-4911-919d-db6e08bad31c) +[02-19 13:19:52|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpoint_db.py:99:_download] Downloading checkpoint file from Hugging Face with {'repo_id': 'nvidia/Cosmos-Predict2.5-2B', 'repo_type': 'model', 'revision': '15a82a2ec231bc318692aa0456a36537c806e7d4', 'filename': 'base/pre-trained/d20b7120-df3e-4911-919d-db6e08bad31c_ema_bf16.pt'} +[02-19 13:19:52|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpoint_db.py:181:path] Downloading checkpoint nvidia/Cosmos-Predict2.5-14B/base/pre-trained(54937b8c-29de-4f04-862c-e67b04ec41e8) +[02-19 13:19:52|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpoint_db.py:99:_download] Downloading checkpoint file from Hugging Face with {'repo_id': 'nvidia/Cosmos-Predict2.5-14B', 'repo_type': 'model', 'revision': '03eb354f35eae0d6e0c1be3c9f94d8551e125570', 'filename': 'base/pre-trained/54937b8c-29de-4f04-862c-e67b04ec41e8_ema_bf16.pt'} +[02-19 13:19:53|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpoint_db.py:181:path] Downloading checkpoint Wan2.1/vae(685afcaa-4de2-42fe-b7b9-69f7a2dee4d8) +[02-19 13:19:53|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpoint_db.py:99:_download] Downloading checkpoint file from Hugging Face with {'repo_id': 'nvidia/Cosmos-Predict2.5-2B', 'repo_type': 'model', 'revision': '6787e176dce74a101d922174a95dba29fa5f0c55', 'filename': 'tokenizer.pth'} +[02-19 13:19:55|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpoint_db.py:181:path] Downloading checkpoint Qwen/Qwen2.5-VL-7B-Instruct(7219c6c7-f878-4137-bbdb-76842ea85e70) +[02-19 13:19:55|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpoint_db.py:134:_download] Downloading checkpoint from Hugging Face with {'repo_id': 'nvidia/Cosmos-Reason1-7B', 'repo_type': 'model', 'revision': '3210bec0495fdc7a8d3dbb8d58da5711eab4b423', 'allow_patterns': ['*']} +[02-19 13:19:56|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpoint_db.py:181:path] Downloading checkpoint nvidia/Cosmos-Reason1.1-7B(cb3e3ffa-7b08-4c34-822d-61c7aa31a14f) +[02-19 13:19:56|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpoint_db.py:134:_download] Downloading checkpoint from Hugging Face with {'repo_id': 'nvidia/Cosmos-Reason1-7B', 'repo_type': 'model', 'revision': '3210bec0495fdc7a8d3dbb8d58da5711eab4b423', 'allow_patterns': ['*']} +[02-19 13:20:26|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/inference.py:60:__init__] Saved config to /home/bravo/zzhangg/MemBench/generation/outputs/cosmos/0219122156_cosmos_mip_spatial/mem_mipnerf360_5f84c46d31489bd1a6b57b7c5c8cbc15280e279f8eb03122abd9362ac7333d13/config.yaml +[02-19 13:20:26|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/inference.py:82:generate] Generating 1 samples: ['segment_4'] +[02-19 13:20:26|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/inference.py:86:generate] [1/1] Processing sample segment_4 +[02-19 13:20:26|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/inference.py:99:_generate_sample] Saved arguments to /home/bravo/zzhangg/MemBench/generation/outputs/cosmos/0219122156_cosmos_mip_spatial/mem_mipnerf360_5f84c46d31489bd1a6b57b7c5c8cbc15280e279f8eb03122abd9362ac7333d13/segment_4.json +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/inference.py:113:_generate_sample] Guardrail checks on prompt are disabled +[02-19 13:20:26|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/inference.py:118:_generate_sample] Generating video with autoregressive mode... +[02-19 13:33:50|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/inference.py:134:_generate_sample] Autoregressive generation completed, preparing to save video... +[02-19 13:33:50|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/inference.py:172:_generate_sample] Guardrail checks on video are disabled +[02-19 13:33:50|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/inference.py:175:_generate_sample] Saving video to %s.mp4 ... +[02-19 13:33:52|SUCCESS|generation/pipelines/cosmos-predict2.5/cosmos_predict2/inference.py:178:_generate_sample] Saved video to /home/bravo/zzhangg/MemBench/generation/outputs/cosmos/0219122156_cosmos_mip_spatial/mem_mipnerf360_5f84c46d31489bd1a6b57b7c5c8cbc15280e279f8eb03122abd9362ac7333d13/segment_4.mp4 diff --git a/0219122156_cosmos_mip_spatial/mem_mipnerf360_5f84c46d31489bd1a6b57b7c5c8cbc15280e279f8eb03122abd9362ac7333d13/debug.log b/0219122156_cosmos_mip_spatial/mem_mipnerf360_5f84c46d31489bd1a6b57b7c5c8cbc15280e279f8eb03122abd9362ac7333d13/debug.log new file mode 100644 index 0000000000000000000000000000000000000000..90f49097e9aabd7fb164c4bce138747f4339bec6 --- /dev/null +++ b/0219122156_cosmos_mip_spatial/mem_mipnerf360_5f84c46d31489bd1a6b57b7c5c8cbc15280e279f8eb03122abd9362ac7333d13/debug.log @@ -0,0 +1,353 @@ +[02-19 13:19:39|DEBUG|generation/pipelines/cosmos-predict2.5/packages/cosmos-oss/cosmos_oss/init.py:176:init_output_dir] Flags(internal=False, smoke=False, verbose=False, experimental_checkpoints=False) +[02-19 13:19:46|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/env_parsers/env_parser.py:86:get_val_dict] getting val dict of CredentialEnvParser +[02-19 13:19:46|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/env_parsers/env_parser.py:86:get_val_dict] getting val dict of CredentialEnvParser +[02-19 13:19:48|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/visualize/video.py:31:] No module named 'ffmpegcv' +[02-19 13:19:48|CRITICAL|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/checkpointer/dcp.py:188:] for the back comptiable pytorch! New DefaultLoadPlanner class is created. +[02-19 13:19:48|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/inference.py:32:__init__] SetupArguments(output_dir=PosixPath('/home/bravo/zzhangg/MemBench/generation/outputs/cosmos/0219122156_cosmos_mip_spatial/mem_mipnerf360_5f84c46d31489bd1a6b57b7c5c8cbc15280e279f8eb03122abd9362ac7333d13') model='2B/post-trained' checkpoint_path='/home/bravo/zzhangg/MemBench/generation/pipelines/cosmos-predict2.5/models/Cosmos-Predict2.5-2B/base/post-trained/81edfebe-bd6a-4039-8c1d-737df1a790bf_ema_bf16.pt' experiment='Stage-c_pt_4-Index-2-Size-2B-Res-720-Fps-16-Note-rf_with_edm_ckpt' config_file='cosmos_predict2/_src/predict2/configs/video2world/config.py' context_parallel_size=1 offload_diffusion_model=False offload_tokenizer=False offload_text_encoder=False disable_guardrails=True offload_guardrail_models=True keep_going=True profile=False) +[02-19 13:19:48|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/datasets/augmentor_provider.py:94:augmentor_register] registering video_basic_augmentor_v1... +[02-19 13:19:48|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/datasets/augmentor_provider.py:94:augmentor_register] registering video_basic_augmentor_v2... +[02-19 13:19:48|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/datasets/augmentor_provider.py:94:augmentor_register] registering noframedrop_nocameramove_video_augmentor_v1... +[02-19 13:19:48|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/datasets/augmentor_provider.py:94:augmentor_register] registering nocameramove_video_augmentor_v1... +[02-19 13:19:48|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/datasets/augmentor_provider.py:94:augmentor_register] registering image_basic_augmentor... +[02-19 13:19:48|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/datasets/augmentor_provider.py:94:augmentor_register] registering image_basic_augmentor_without_embeddings... +[02-19 13:19:48|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/attention/utils/safe_log.py:35:debug] Flash Attention v3 is not supported because the Python package was not found. +[02-19 13:19:48|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/attention/utils/safe_log.py:35:debug] NATTEN Attention is not supported due to insufficient NATTEN version natten.__version__='0.21.0', expected at least NATTEN_MIN_RELEASE_VERSION=[0, 21, 5], or NATTEN_MIN_DEV_VERSION=([0, 21, 5], 9). +[02-19 13:19:48|CRITICAL|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:192:import_all_modules_from_package] Reloading all modules from package cosmos_predict2._src.predict2.configs.video2world.experiment +[02-19 13:19:48|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_predict2._src.predict2.configs.video2world.experiment.reason_embeddings +[02-19 13:19:48|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_predict2._src.predict2.configs.video2world.experiment.reason_embeddings.model_14b_reason_1p1 +[02-19 13:19:48|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_predict2._src.predict2.configs.video2world.experiment.reason_embeddings.model_14b_reason_1p1_rectified_flow +[02-19 13:19:49|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_predict2._src.predict2.configs.video2world.experiment.reason_embeddings.model_2B_reason_1p1 +[02-19 13:19:49|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_predict2._src.predict2.configs.video2world.experiment.reason_embeddings.model_2B_reason_1p1_rectified_flow +[02-19 13:19:49|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_predict2._src.predict2.configs.video2world.experiment.reason_embeddings.model_7b_reason_1p1 +[02-19 13:19:49|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_predict2._src.predict2.configs.video2world.experiment.reason_embeddings.stage3_14B_index_3 +[02-19 13:19:49|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_predict2._src.predict2.configs.video2world.experiment.reason_embeddings.stage3_2B +[02-19 13:19:49|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_predict2._src.predict2.configs.video2world.experiment.resume_text2world +[02-19 13:19:49|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_predict2._src.predict2.configs.video2world.experiment.resume_text2world.sparse_14B +[02-19 13:19:50|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_predict2._src.predict2.configs.video2world.experiment.resume_text2world.sparse_2B +[02-19 13:19:50|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_predict2._src.predict2.configs.video2world.experiment.resume_text2world.stage3 +[02-19 13:19:50|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_predict2._src.predict2.configs.video2world.experiment.resume_text2world.stage3_2B_ablation +[02-19 13:19:51|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_predict2._src.predict2.configs.video2world.experiment.resume_text2world.state3_14B_index_3 +[02-19 13:19:51|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_predict2._src.predict2.configs.video2world.experiment.specialized_model +[02-19 13:19:51|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_predict2._src.predict2.configs.video2world.experiment.specialized_model.SFT_14B_RF +[02-19 13:19:51|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_predict2._src.predict2.configs.video2world.experiment.specialized_model.SFT_2B +[02-19 13:19:51|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_predict2._src.predict2.configs.video2world.experiment.specialized_model.SFT_2B_RF +[02-19 13:19:51|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_face +[02-19 13:19:51|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_crowded +[02-19 13:19:51|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_high_motion +[02-19 13:19:51|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_robotics +[02-19 13:19:51|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_av +[02-19 13:19:51|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_physical_ai +[02-19 13:19:51|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_4k_cooldown +[02-19 13:19:51|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_face_only2 +[02-19 13:19:51|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_crowded_only2 +[02-19 13:19:51|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_high_motion_only2 +[02-19 13:19:51|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_robotics_only2 +[02-19 13:19:51|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_av_only2 +[02-19 13:19:51|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_physical_ai_only2 +[02-19 13:19:51|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_4k_cooldown_only2 +[02-19 13:19:51|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_face_high_sigma +[02-19 13:19:51|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_crowded_high_sigma +[02-19 13:19:51|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_high_motion_high_sigma +[02-19 13:19:51|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_robotics_high_sigma +[02-19 13:19:51|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_av_high_sigma +[02-19 13:19:51|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_physical_ai_high_sigma +[02-19 13:19:51|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_4k_cooldown_high_sigma +[02-19 13:19:52|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_face_high_sigma2 +[02-19 13:19:52|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_crowded_high_sigma2 +[02-19 13:19:52|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_robotics_high_sigma2 +[02-19 13:19:52|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-2-Size-2B-Res-720-Fps-16-Note-rf_with_edm_ckpt +[02-19 13:19:52|CRITICAL|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:192:import_all_modules_from_package] Reloading all modules from package cosmos_predict2.experiments +[02-19 13:19:52|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_predict2.experiments.base +[02-19 13:19:52|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_predict2.experiments.base.action +[02-19 13:19:52|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpoint_db.py:181:path] Downloading checkpoint nvidia/Cosmos-Predict2.5-2B/base/post-trained(81edfebe-bd6a-4039-8c1d-737df1a790bf) +[02-19 13:19:52|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpoint_db.py:99:_download] Downloading checkpoint file from Hugging Face with {'repo_id': 'nvidia/Cosmos-Predict2.5-2B', 'repo_type': 'model', 'revision': '15a82a2ec231bc318692aa0456a36537c806e7d4', 'filename': 'base/post-trained/81edfebe-bd6a-4039-8c1d-737df1a790bf_ema_bf16.pt'} +[02-19 13:19:52|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_predict2.experiments.base.cosmos_nemo_assets +[02-19 13:19:52|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpoint_db.py:181:path] Downloading checkpoint nvidia/Cosmos-Predict2.5-2B/base/pre-trained(d20b7120-df3e-4911-919d-db6e08bad31c) +[02-19 13:19:52|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpoint_db.py:99:_download] Downloading checkpoint file from Hugging Face with {'repo_id': 'nvidia/Cosmos-Predict2.5-2B', 'repo_type': 'model', 'revision': '15a82a2ec231bc318692aa0456a36537c806e7d4', 'filename': 'base/pre-trained/d20b7120-df3e-4911-919d-db6e08bad31c_ema_bf16.pt'} +[02-19 13:19:52|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_predict2.experiments.base.cosmos_nemo_assets_lora +[02-19 13:19:52|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_predict2.experiments.base.groot +[02-19 13:19:52|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpoint_db.py:181:path] Downloading checkpoint nvidia/Cosmos-Predict2.5-14B/base/pre-trained(54937b8c-29de-4f04-862c-e67b04ec41e8) +[02-19 13:19:52|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpoint_db.py:99:_download] Downloading checkpoint file from Hugging Face with {'repo_id': 'nvidia/Cosmos-Predict2.5-14B', 'repo_type': 'model', 'revision': '03eb354f35eae0d6e0c1be3c9f94d8551e125570', 'filename': 'base/pre-trained/54937b8c-29de-4f04-862c-e67b04ec41e8_ema_bf16.pt'} +[02-19 13:19:53|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/utils/model_loader.py:91:load_model_from_checkpoint] Overriding config checkpoint path with: /home/bravo/zzhangg/MemBench/generation/pipelines/cosmos-predict2.5/models/Cosmos-Predict2.5-2B/base/post-trained/81edfebe-bd6a-4039-8c1d-737df1a790bf_ema_bf16.pt +[02-19 13:19:53|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/misc.py:152:set_random_seed] Using random seed 0. +[02-19 13:19:53|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/utils/model_loader.py:111:load_model_from_checkpoint] Loading model from /home/bravo/zzhangg/MemBench/generation/pipelines/cosmos-predict2.5/models/Cosmos-Predict2.5-2B/base/post-trained/81edfebe-bd6a-4039-8c1d-737df1a790bf_ema_bf16.pt +[02-19 13:19:53|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/models/text2world_model_rectified_flow.py:132:__init__] DiffusionModel: precision torch.bfloat16 +[02-19 13:19:53|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpoint_db.py:181:path] Downloading checkpoint Wan2.1/vae(685afcaa-4de2-42fe-b7b9-69f7a2dee4d8) +[02-19 13:19:53|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpoint_db.py:99:_download] Downloading checkpoint file from Hugging Face with {'repo_id': 'nvidia/Cosmos-Predict2.5-2B', 'repo_type': 'model', 'revision': '6787e176dce74a101d922174a95dba29fa5f0c55', 'filename': 'tokenizer.pth'} +[02-19 13:19:54|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/tokenizers/wan2pt1.py:669:_video_vae] loading /home/bravo/.cache/huggingface/hub/models--nvidia--Cosmos-Predict2.5-2B/snapshots/6787e176dce74a101d922174a95dba29fa5f0c55/tokenizer.pth +[02-19 13:19:54|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/distributed.py:415:sync_model_states] Synchronizing model states from rank 0 to all ranks in process group [0]. +[02-19 13:19:54|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/timer.py:150:_log] Time spent on DiffusionModel: set_up_tokenizer: 1.05 s +[02-19 13:19:54|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #0-fps: + ReMapkey + input key: fps + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: fps + Dtype: None +[02-19 13:19:54|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #1-padding_mask: + ReMapkey + input key: padding_mask + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: padding_mask + Dtype: None +[02-19 13:19:54|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #2-text: + TextAttr + input key: ['t5_text_embeddings'] + Param count: 0 + Trainable: None + Dropout rate: 0.2 + Output key: [crossattn_emb] +[02-19 13:19:54|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #3-use_video_condition: + BooleanFlag + input key: fps + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: use_video_condition + This is a boolean flag +[02-19 13:19:54|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1624:build_pos_embed] Building positional embedding with rope3d class, impl +[02-19 13:19:54|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:886:__init__] Using AdaLN LoRA Flag: True. We enable bias if no AdaLN LoRA for backward compatibility. +[02-19 13:19:54|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:19:54|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:19:54|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:19:54|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:19:54|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:19:54|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:19:54|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:19:54|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:19:54|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:19:54|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:19:54|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:19:54|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:19:54|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:19:54|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:19:54|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:19:54|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:19:54|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:19:54|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:19:54|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:19:54|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:19:54|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:19:54|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:19:54|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:19:54|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:19:54|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:19:54|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:19:54|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:19:54|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:19:54|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:19:54|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:19:54|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:19:54|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:19:54|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:19:54|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:19:54|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:19:54|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:19:54|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:19:54|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:19:54|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:19:54|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:19:54|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:19:54|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:19:54|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:19:54|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:19:54|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:19:54|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:19:54|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:19:54|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:19:54|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:19:54|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:19:54|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:19:54|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:19:54|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:19:54|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:19:54|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:19:54|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:19:55|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1804:enable_selective_checkpoint] Enable selective checkpoint with predict2_2b_720_aggressive, for every 1 blocks. Total blocks: 28 +[02-19 13:19:55|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 0 +[02-19 13:19:55|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 1 +[02-19 13:19:55|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 2 +[02-19 13:19:55|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 3 +[02-19 13:19:55|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 4 +[02-19 13:19:55|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 5 +[02-19 13:19:55|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 6 +[02-19 13:19:55|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 7 +[02-19 13:19:55|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 8 +[02-19 13:19:55|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 9 +[02-19 13:19:55|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 10 +[02-19 13:19:55|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 11 +[02-19 13:19:55|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 12 +[02-19 13:19:55|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 13 +[02-19 13:19:55|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 14 +[02-19 13:19:55|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 15 +[02-19 13:19:55|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 16 +[02-19 13:19:55|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 17 +[02-19 13:19:55|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 18 +[02-19 13:19:55|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 19 +[02-19 13:19:55|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 20 +[02-19 13:19:55|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 21 +[02-19 13:19:55|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 22 +[02-19 13:19:55|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 23 +[02-19 13:19:55|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 24 +[02-19 13:19:55|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 25 +[02-19 13:19:55|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 26 +[02-19 13:19:55|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 27 +[02-19 13:19:55|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/timer.py:150:_log] Time spent on meta to cuda and broadcast model states: 0.13 s +[02-19 13:19:55|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/timer.py:150:_log] Time spent on Creating PyTorch model: 0.61 s +[02-19 13:19:55|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/timer.py:150:_log] Time spent on Creating PyTorch model and ema if enabled: 0.62 s +[02-19 13:19:55|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/timer.py:150:_log] Time spent on DiffusionModel: set_up_model: 0.62 s +[02-19 13:19:55|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/text_encoders/text_encoder.py:74:__init__] Instantiating text encoder model... +[02-19 13:19:55|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpoint_db.py:181:path] Downloading checkpoint Qwen/Qwen2.5-VL-7B-Instruct(7219c6c7-f878-4137-bbdb-76842ea85e70) +[02-19 13:19:55|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpoint_db.py:134:_download] Downloading checkpoint from Hugging Face with {'repo_id': 'nvidia/Cosmos-Reason1-7B', 'repo_type': 'model', 'revision': '3210bec0495fdc7a8d3dbb8d58da5711eab4b423', 'allow_patterns': ['*']} +[02-19 13:19:55|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/reason1/tokenizer/processor.py:75:__init__] Successfully loaded processor from local cache +[02-19 13:19:55|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/reason1/models/vlm_base.py:116:__init__] Setting torch default dtype from torch.float32 to torch.bfloat16 +[02-19 13:19:56|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/reason1/parallelisms/parallel_dims.py:50:_validate] dp_shard is set to -1, will be automatically determined based on world_size 1 // 1. +[02-19 13:19:56|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/reason1/parallelisms/parallel_dims.py:54:_validate] dp_shard is set to 1. +[02-19 13:19:56|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/reason1/parallelisms/parallel_dims.py:75:build_mesh] Building 0-D device mesh with [], [] +[02-19 13:19:56|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/reason1/parallelisms/parallel_dims.py:105:build_mesh] mesh: DeviceMesh('cuda', 0) +[02-19 13:19:56|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/reason1/parallelisms/parallelize_qwen.py:336:apply_ac] Applied selective activation checkpointing to the model +[02-19 13:19:56|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/reason1/models/vlm_base.py:123:__init__] Reset torch default dtype to torch.float32 +[02-19 13:19:56|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/text_encoders/text_encoder.py:84:__init__] Loading checkpoint from s3://bucket/cosmos_reasoning1/sft_exp700/sft_exp721-1_qwen7b_tl_721_5vs5_s3_balanced_n32_resume_16k/checkpoints/iter_000016000/model/. +[02-19 13:19:56|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpoint_db.py:181:path] Downloading checkpoint nvidia/Cosmos-Reason1.1-7B(cb3e3ffa-7b08-4c34-822d-61c7aa31a14f) +[02-19 13:19:56|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpoint_db.py:134:_download] Downloading checkpoint from Hugging Face with {'repo_id': 'nvidia/Cosmos-Reason1-7B', 'repo_type': 'model', 'revision': '3210bec0495fdc7a8d3dbb8d58da5711eab4b423', 'allow_patterns': ['*']} +[02-19 13:20:21|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/text_encoders/text_encoder.py:113:__init__] Finished loading checkpoint from /home/bravo/.cache/huggingface/hub/models--nvidia--Cosmos-Reason1-7B/snapshots/3210bec0495fdc7a8d3dbb8d58da5711eab4b423. +[02-19 13:20:21|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/text_encoders/text_encoder.py:116:__init__] Text encoder model instantiated +[02-19 13:20:21|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/timer.py:150:_log] Time spent on instantiate model: 27.72 s +[02-19 13:20:21|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/utils/model_loader.py:227:load_model_state_dict_from_checkpoint] Loading model cached locally from /home/bravo/zzhangg/MemBench/generation/pipelines/cosmos-predict2.5/models/Cosmos-Predict2.5-2B/base/post-trained/81edfebe-bd6a-4039-8c1d-737df1a790bf_ema_bf16.pt +[02-19 13:20:26|CRITICAL|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/models/text2world_model_rectified_flow.py:919:load_state_dict] load model in non-strict mode +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key t_embedding_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:26|CRITICAL|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/models/text2world_model_rectified_flow.py:920:load_state_dict] [RANK 0] _IncompatibleKeys(missing_keys=[], unexpected_keys=[], incorrect_shapes=[]) +[02-19 13:20:26|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/distributed.py:415:sync_model_states] Synchronizing model states from rank 0 to all ranks in process group [0]. +[02-19 13:20:26|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/inference.py:60:__init__] Saved config to /home/bravo/zzhangg/MemBench/generation/outputs/cosmos/0219122156_cosmos_mip_spatial/mem_mipnerf360_5f84c46d31489bd1a6b57b7c5c8cbc15280e279f8eb03122abd9362ac7333d13/config.yaml +[02-19 13:20:26|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/inference.py:82:generate] Generating 1 samples: ['segment_4'] +[02-19 13:20:26|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/inference.py:86:generate] [1/1] Processing sample segment_4 +[02-19 13:20:26|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/inference.py:93:_generate_sample] InferenceArguments(name='segment_4' prompt_path=None prompt='80-100%: Toward the end, the camera completes its loop by returning to the starting point on the left side of the countertop. It provides a final overview of the entire scene, ensuring that all spatial relationships and background elements remain consistent with the earlier views.' negative_prompt='' seed=46 guidance=7 inference_type= input_path=PosixPath('/home/bravo/zzhangg/MemBench/generation/outputs/cosmos/0219122156_cosmos_mip_spatial/mem_mipnerf360_5f84c46d31489bd1a6b57b7c5c8cbc15280e279f8eb03122abd9362ac7333d13/segment_3.mp4') resolution='none' num_output_frames=77 num_steps=35 enable_autoregressive=True chunk_size=77 chunk_overlap=1) +[02-19 13:20:26|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/inference.py:99:_generate_sample] Saved arguments to /home/bravo/zzhangg/MemBench/generation/outputs/cosmos/0219122156_cosmos_mip_spatial/mem_mipnerf360_5f84c46d31489bd1a6b57b7c5c8cbc15280e279f8eb03122abd9362ac7333d13/segment_4.json +[02-19 13:20:26|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/inference.py:113:_generate_sample] Guardrail checks on prompt are disabled +[02-19 13:20:26|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/inference.py:118:_generate_sample] Generating video with autoregressive mode... +[02-19 13:20:26|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/inference/video2world.py:730:generate_autoregressive_from_batch] Processing video input for autoregressive: /home/bravo/zzhangg/MemBench/generation/outputs/cosmos/0219122156_cosmos_mip_spatial/mem_mipnerf360_5f84c46d31489bd1a6b57b7c5c8cbc15280e279f8eb03122abd9362ac7333d13/segment_3.mp4 +[02-19 13:20:27|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/inference/video2world.py:798:generate_autoregressive_from_batch] Generating 1 chunks with chunk_size=77, chunk_overlap=1 for 77 total frames +[02-19 13:20:27|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/inference/video2world.py:816:generate_autoregressive_from_batch] Processing chunk 1/1, frames 0-77 +[02-19 13:20:28|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/inference/video2world.py:556:generate_vid2world] GPU memory usage after getting data_batch: 19.79 GB +[02-19 13:20:28|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/inference/video2world.py:594:generate_vid2world] [Memory Optimization] Starting latent sample generation +[02-19 13:33:50|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/inference/video2world.py:878:generate_autoregressive_from_batch] Generated final video with shape torch.Size([1, 3, 77, 704, 1280]) +[02-19 13:33:50|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/inference.py:134:_generate_sample] Autoregressive generation completed, preparing to save video... +[02-19 13:33:50|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/inference.py:172:_generate_sample] Guardrail checks on video are disabled +[02-19 13:33:50|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/inference.py:175:_generate_sample] Saving video to %s.mp4 ... +[02-19 13:33:51|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/easy_io/handlers/imageio_video_handler.py:163:dump_to_fileobj] mimsave_kwargs: {'fps': 16, 'quality': 5, 'macro_block_size': 1, 'ffmpeg_params': ['-s', '1280x704'], 'output_params': ['-f', 'mp4']} +[02-19 13:33:52|SUCCESS|generation/pipelines/cosmos-predict2.5/cosmos_predict2/inference.py:178:_generate_sample] Saved video to /home/bravo/zzhangg/MemBench/generation/outputs/cosmos/0219122156_cosmos_mip_spatial/mem_mipnerf360_5f84c46d31489bd1a6b57b7c5c8cbc15280e279f8eb03122abd9362ac7333d13/segment_4.mp4 diff --git a/0219122156_cosmos_mip_spatial/mem_mipnerf360_5f84c46d31489bd1a6b57b7c5c8cbc15280e279f8eb03122abd9362ac7333d13/segment_0.json b/0219122156_cosmos_mip_spatial/mem_mipnerf360_5f84c46d31489bd1a6b57b7c5c8cbc15280e279f8eb03122abd9362ac7333d13/segment_0.json new file mode 100644 index 0000000000000000000000000000000000000000..47d95789ea14e3bba55dfc16ec63ab314fa48806 --- /dev/null +++ b/0219122156_cosmos_mip_spatial/mem_mipnerf360_5f84c46d31489bd1a6b57b7c5c8cbc15280e279f8eb03122abd9362ac7333d13/segment_0.json @@ -0,0 +1 @@ +{"name":"segment_0","prompt_path":null,"prompt":"0-20%: The video opens with a close-up view of the left side of the countertop, focusing on a large metal mixing bowl, several butternut squashes, and a blue oven mitt. The camera begins its horizontal pan to the right, capturing the initial layout of the scene.","negative_prompt":"","seed":42,"guidance":7,"inference_type":"text2world","input_path":null,"resolution":"none","num_output_frames":77,"num_steps":35,"enable_autoregressive":true,"chunk_size":77,"chunk_overlap":1} \ No newline at end of file diff --git a/0219122156_cosmos_mip_spatial/mem_mipnerf360_5f84c46d31489bd1a6b57b7c5c8cbc15280e279f8eb03122abd9362ac7333d13/segment_0.mp4 b/0219122156_cosmos_mip_spatial/mem_mipnerf360_5f84c46d31489bd1a6b57b7c5c8cbc15280e279f8eb03122abd9362ac7333d13/segment_0.mp4 new file mode 100644 index 0000000000000000000000000000000000000000..4b5242399d0e597fad8d8ceaff93ada9d8de573f --- /dev/null +++ b/0219122156_cosmos_mip_spatial/mem_mipnerf360_5f84c46d31489bd1a6b57b7c5c8cbc15280e279f8eb03122abd9362ac7333d13/segment_0.mp4 @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e2accc9cf0bcfc65002dfd5137f4a1d97912ea3080525439f2921edf33469dc7 +size 859614 diff --git a/0219122156_cosmos_mip_spatial/mem_mipnerf360_5f84c46d31489bd1a6b57b7c5c8cbc15280e279f8eb03122abd9362ac7333d13/segment_1.json b/0219122156_cosmos_mip_spatial/mem_mipnerf360_5f84c46d31489bd1a6b57b7c5c8cbc15280e279f8eb03122abd9362ac7333d13/segment_1.json new file mode 100644 index 0000000000000000000000000000000000000000..4c24fe5a02e7a40f104ce2931fbe70239e75ec2f --- /dev/null +++ b/0219122156_cosmos_mip_spatial/mem_mipnerf360_5f84c46d31489bd1a6b57b7c5c8cbc15280e279f8eb03122abd9362ac7333d13/segment_1.json @@ -0,0 +1 @@ +{"name":"segment_1","prompt_path":null,"prompt":"20-40%: As the camera continues its pan, it reveals more items on the countertop, including an egg carton, a bag of chips, and a blender. The focus shifts toward the center of the countertop, showcasing additional objects like a wooden bowl with fruits and a can of soda.","negative_prompt":"","seed":43,"guidance":7,"inference_type":"video2world","input_path":"/home/bravo/zzhangg/MemBench/generation/outputs/cosmos/0219122156_cosmos_mip_spatial/mem_mipnerf360_5f84c46d31489bd1a6b57b7c5c8cbc15280e279f8eb03122abd9362ac7333d13/segment_0.mp4","resolution":"none","num_output_frames":77,"num_steps":35,"enable_autoregressive":true,"chunk_size":77,"chunk_overlap":1} \ No newline at end of file diff --git a/0219122156_cosmos_mip_spatial/mem_mipnerf360_5f84c46d31489bd1a6b57b7c5c8cbc15280e279f8eb03122abd9362ac7333d13/segment_1.mp4 b/0219122156_cosmos_mip_spatial/mem_mipnerf360_5f84c46d31489bd1a6b57b7c5c8cbc15280e279f8eb03122abd9362ac7333d13/segment_1.mp4 new file mode 100644 index 0000000000000000000000000000000000000000..e5cddeaed90a9631bb4494e65de93acf9fc35259 --- /dev/null +++ b/0219122156_cosmos_mip_spatial/mem_mipnerf360_5f84c46d31489bd1a6b57b7c5c8cbc15280e279f8eb03122abd9362ac7333d13/segment_1.mp4 @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:44835fc66214124a1bcd24eede0c486d59054bcd3743f14efed6130a69d64c93 +size 1010870 diff --git a/0219122156_cosmos_mip_spatial/mem_mipnerf360_5f84c46d31489bd1a6b57b7c5c8cbc15280e279f8eb03122abd9362ac7333d13/segment_2.json b/0219122156_cosmos_mip_spatial/mem_mipnerf360_5f84c46d31489bd1a6b57b7c5c8cbc15280e279f8eb03122abd9362ac7333d13/segment_2.json new file mode 100644 index 0000000000000000000000000000000000000000..90727a3782c58c4b2fca86471ab8b6a75a6f8268 --- /dev/null +++ b/0219122156_cosmos_mip_spatial/mem_mipnerf360_5f84c46d31489bd1a6b57b7c5c8cbc15280e279f8eb03122abd9362ac7333d13/segment_2.json @@ -0,0 +1 @@ +{"name":"segment_2","prompt_path":null,"prompt":"40-60%: The camera reaches the right side of the countertop, displaying a variety of items such as a loaf of bread in a plastic bag, a knife, and a bag of pasta. It then begins to retrace its path, moving back toward the left side of the countertop.","negative_prompt":"","seed":44,"guidance":7,"inference_type":"video2world","input_path":"/home/bravo/zzhangg/MemBench/generation/outputs/cosmos/0219122156_cosmos_mip_spatial/mem_mipnerf360_5f84c46d31489bd1a6b57b7c5c8cbc15280e279f8eb03122abd9362ac7333d13/segment_1.mp4","resolution":"none","num_output_frames":77,"num_steps":35,"enable_autoregressive":true,"chunk_size":77,"chunk_overlap":1} \ No newline at end of file diff --git a/0219122156_cosmos_mip_spatial/mem_mipnerf360_5f84c46d31489bd1a6b57b7c5c8cbc15280e279f8eb03122abd9362ac7333d13/segment_2.mp4 b/0219122156_cosmos_mip_spatial/mem_mipnerf360_5f84c46d31489bd1a6b57b7c5c8cbc15280e279f8eb03122abd9362ac7333d13/segment_2.mp4 new file mode 100644 index 0000000000000000000000000000000000000000..de0f6bdc69aa119b77b59f70e3bebc56ef68041a --- /dev/null +++ b/0219122156_cosmos_mip_spatial/mem_mipnerf360_5f84c46d31489bd1a6b57b7c5c8cbc15280e279f8eb03122abd9362ac7333d13/segment_2.mp4 @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8d6003b85306430ce07e2c17aa4ad238a371fc105c9d7d9fa65984af659566c7 +size 909739 diff --git a/0219122156_cosmos_mip_spatial/mem_mipnerf360_5f84c46d31489bd1a6b57b7c5c8cbc15280e279f8eb03122abd9362ac7333d13/segment_3.json b/0219122156_cosmos_mip_spatial/mem_mipnerf360_5f84c46d31489bd1a6b57b7c5c8cbc15280e279f8eb03122abd9362ac7333d13/segment_3.json new file mode 100644 index 0000000000000000000000000000000000000000..08573999bf7d155310b3838abde400b2a6bce0f0 --- /dev/null +++ b/0219122156_cosmos_mip_spatial/mem_mipnerf360_5f84c46d31489bd1a6b57b7c5c8cbc15280e279f8eb03122abd9362ac7333d13/segment_3.json @@ -0,0 +1 @@ +{"name":"segment_3","prompt_path":null,"prompt":"60-80%: The camera retraces its earlier path, revisiting the center and left sections of the countertop. It captures the same objects from slightly different angles, emphasizing the consistency of their positions and the overall layout of the scene.","negative_prompt":"","seed":45,"guidance":7,"inference_type":"video2world","input_path":"/home/bravo/zzhangg/MemBench/generation/outputs/cosmos/0219122156_cosmos_mip_spatial/mem_mipnerf360_5f84c46d31489bd1a6b57b7c5c8cbc15280e279f8eb03122abd9362ac7333d13/segment_2.mp4","resolution":"none","num_output_frames":77,"num_steps":35,"enable_autoregressive":true,"chunk_size":77,"chunk_overlap":1} \ No newline at end of file diff --git a/0219122156_cosmos_mip_spatial/mem_mipnerf360_5f84c46d31489bd1a6b57b7c5c8cbc15280e279f8eb03122abd9362ac7333d13/segment_3.mp4 b/0219122156_cosmos_mip_spatial/mem_mipnerf360_5f84c46d31489bd1a6b57b7c5c8cbc15280e279f8eb03122abd9362ac7333d13/segment_3.mp4 new file mode 100644 index 0000000000000000000000000000000000000000..febc45f4bdd2ce0e5d345a6803f9fd5ab8afed55 --- /dev/null +++ b/0219122156_cosmos_mip_spatial/mem_mipnerf360_5f84c46d31489bd1a6b57b7c5c8cbc15280e279f8eb03122abd9362ac7333d13/segment_3.mp4 @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2243baf307624764aa2371f131c16608d4d09785a15059b023bac3e210e4f04e +size 1052164 diff --git a/0219122156_cosmos_mip_spatial/mem_mipnerf360_5f84c46d31489bd1a6b57b7c5c8cbc15280e279f8eb03122abd9362ac7333d13/segment_4.json b/0219122156_cosmos_mip_spatial/mem_mipnerf360_5f84c46d31489bd1a6b57b7c5c8cbc15280e279f8eb03122abd9362ac7333d13/segment_4.json new file mode 100644 index 0000000000000000000000000000000000000000..38b9bf82c5ed283f3a17c0a2b980260df7167d68 --- /dev/null +++ b/0219122156_cosmos_mip_spatial/mem_mipnerf360_5f84c46d31489bd1a6b57b7c5c8cbc15280e279f8eb03122abd9362ac7333d13/segment_4.json @@ -0,0 +1 @@ +{"name":"segment_4","prompt_path":null,"prompt":"80-100%: Toward the end, the camera completes its loop by returning to the starting point on the left side of the countertop. It provides a final overview of the entire scene, ensuring that all spatial relationships and background elements remain consistent with the earlier views.","negative_prompt":"","seed":46,"guidance":7,"inference_type":"video2world","input_path":"/home/bravo/zzhangg/MemBench/generation/outputs/cosmos/0219122156_cosmos_mip_spatial/mem_mipnerf360_5f84c46d31489bd1a6b57b7c5c8cbc15280e279f8eb03122abd9362ac7333d13/segment_3.mp4","resolution":"none","num_output_frames":77,"num_steps":35,"enable_autoregressive":true,"chunk_size":77,"chunk_overlap":1} \ No newline at end of file diff --git a/0219122156_cosmos_mip_spatial/mem_mipnerf360_5f84c46d31489bd1a6b57b7c5c8cbc15280e279f8eb03122abd9362ac7333d13/segment_4.mp4 b/0219122156_cosmos_mip_spatial/mem_mipnerf360_5f84c46d31489bd1a6b57b7c5c8cbc15280e279f8eb03122abd9362ac7333d13/segment_4.mp4 new file mode 100644 index 0000000000000000000000000000000000000000..59eb3b2546ba2674b28c5f979b9967d4764a4072 --- /dev/null +++ b/0219122156_cosmos_mip_spatial/mem_mipnerf360_5f84c46d31489bd1a6b57b7c5c8cbc15280e279f8eb03122abd9362ac7333d13/segment_4.mp4 @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:db41a89b799da334bde1240c54be5cb9679ac757a5e1f9dda33962ed86d95400 +size 1246519 diff --git a/0219122156_cosmos_mip_spatial/mem_mipnerf360_a8f137cd302ac277409b1dcd26e4c8a03d8600bc4f6a80af40c71a5a980a98c5/config.yaml b/0219122156_cosmos_mip_spatial/mem_mipnerf360_a8f137cd302ac277409b1dcd26e4c8a03d8600bc4f6a80af40c71a5a980a98c5/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..400637c15f73cbb3df929fe2dec18cd588b7f277 --- /dev/null +++ b/0219122156_cosmos_mip_spatial/mem_mipnerf360_a8f137cd302ac277409b1dcd26e4c8a03d8600bc4f6a80af40c71a5a980a98c5/config.yaml @@ -0,0 +1,760 @@ +checkpoint: + broadcast_via_filesystem: 'False' + dcp_allow_mismatched_size: 'False' + dcp_async_mode_enabled: 'False' + enable_gcs_patch_in_boto3: 'False' + jit: + device: cuda + dtype: bfloat16 + enabled: 'False' + input_shape: null + strict: 'True' + keys_not_to_resume: [] + load_ema_to_reg: 'False' + load_from_object_store: + bucket: bucket + credentials: credentials/s3_checkpoint.secret + enabled: 'True' + load_path: /home/bravo/zzhangg/MemBench/generation/pipelines/cosmos-predict2.5/models/Cosmos-Predict2.5-2B/base/post-trained/81edfebe-bd6a-4039-8c1d-737df1a790bf_ema_bf16.pt + load_training_state: 'False' + only_load_scheduler_state: 'False' + save_iter: '1000' + save_to_object_store: + bucket: bucket + credentials: credentials/s3_checkpoint.secret + enabled: 'True' + strict_resume: 'True' + type: + _target_: + callbacks: null + disable_async: 'False' + verbose: 'True' +dataloader_train: + dataloaders: + image_data: + dataloader: + batch_size: '12' + cache_augment_fn: functools.partial(, + n=1) + cache_size: '8' + concat_size: '1' + dataset: + augmentor_name: image_basic_augmentor_without_embeddings + caption_type: qwen2p5_7b_v4 + dataset_resolution_type: gt720p + embedding_type: null + resolution: '720' + num_workers: '6' + use_cache: 'False' + ratio: '1' + video_data: + dataloader: + batch_size: '1' + cache_augment_fn: functools.partial(, n=1.8) + cache_size: '16' + concat_size: '1' + dataset: + augmentor_name: video_basic_augmentor_v2 + caption_type: t2w_qwen2p5_7b + dataset_resolution_type: all + embedding_type: null + max_fps_thres: '60' + min_fps_thres: '10' + num_video_frames: '93' + resolution: '720' + use_native_fps: 'True' + video_decoder_name: video_naive_bytes + use_cache: 'False' + ratio: '3' +dataloader_val: + _target_: + dataloaders: + image_data: + dataloader: + _target_: + batch_size: '2' + cache_augment_fn: null + cache_replay_name: image_dataloader + cache_size: '32' + concat_size: '1' + dataset: + _target_: + len_t5: '512' + resolution: '512' + t5_dim: '1024' + num_workers: '8' + pin_memory: 'True' + shuffle: 'False' + use_cache: 'False' + webdataset: 'False' + ratio: '1' + video_data: + dataloader: + _target_: + batch_size: '1' + cache_augment_fn: null + cache_replay_name: video_dataloader + cache_size: '32' + concat_size: '1' + dataset: + _target_: + len_t5: '512' + num_video_frames: '136' + resolution: '512' + t5_dim: '1024' + num_workers: '8' + pin_memory: 'True' + shuffle: 'False' + use_cache: 'False' + webdataset: 'False' + ratio: '1' +defaults: +- _self_ +- data_train: mock +- data_val: mock +- optimizer: fusedadamw +- scheduler: lambdalinear +- model: ddp +- callbacks: basic +- net: null +- conditioner: video_prediction_conditioner +- ema: power +- tokenizer: cosmos_tokenizer_causal_cv8x8x8_c16_res720_t121_it121_v1_0 +- checkpoint: s3 +- ckpt_type: dummy +- experiment: null +job: + cluster: null + group: official_runs_vid2vid + name: Stage-c_pt_4-Index-2-Size-2B-Res-720-Fps-16-Note-rf_with_edm_ckpt + project: cosmos_diffusion_v2 + wandb_mode: online +model: + _recursive_: 'False' + _target_: + config: + conditional_frame_timestep: 0.1 + conditional_frames_probs: + 0: 0.5 + 1: 0.25 + 2: 0.25 + conditioner: + _target_: + fps: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: fps + output_key: fps + padding_mask: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: padding_mask + output_key: padding_mask + text: + _target_: + credential_path: credentials/s3_training.secret + dropout_rate: '0.2' + empty_string_embeddings_path: s3://bucket/predict2_assets/reason1_empty_string_embeddings.pt + input_key: + - t5_text_embeddings + use_empty_string: 'False' + use_video_condition: + _target_: + dropout_rate: '0.0' + input_key: fps + output_key: use_video_condition + conditioning_strategy: frame_replace + denoise_replace_gt_frames: true + ema: + enabled: false + iteration_shift: 0 + rate: 0.1 + fsdp_shard_size: 1 + high_sigma_ratio: 0.05 + high_sigma_timesteps_max: 1000 + high_sigma_timesteps_min: 980 + init_lora_weights: true + input_caption_key: ai_caption + input_data_key: video + input_image_key: images + lora_alpha: 32 + lora_rank: 32 + lora_target_modules: q_proj,k_proj,v_proj,output_proj,mlp.layer1,mlp.layer2 + max_num_conditional_frames: 2 + min_num_conditional_frames: 0 + net: + _target_: + adaln_lora_dim: '256' + atten_backend: minimal_a2a + concat_padding_mask: 'True' + crossattn_emb_channels: '1024' + crossattn_proj_in_channels: '100352' + extra_per_block_abs_pos_emb: 'False' + in_channels: '16' + max_frames: '128' + max_img_h: '240' + max_img_w: '240' + model_channels: '2048' + num_blocks: '28' + num_heads: '16' + out_channels: '16' + patch_spatial: '2' + patch_temporal: '1' + pos_emb_cls: rope3d + pos_emb_interpolation: crop + pos_emb_learnable: 'True' + rope_enable_fps_modulation: 'False' + rope_h_extrapolation_ratio: '3.0' + rope_t_extrapolation_ratio: '1.0' + rope_w_extrapolation_ratio: '3.0' + sac_config: + every_n_blocks: 1 + mode: predict2_2b_720_aggressive + timestep_scale: '0.001' + use_adaln_lora: 'True' + use_crossattn_projection: 'True' + use_wan_fp32_strategy: 'True' + precision: bfloat16 + resolution: '720' + shift: 5 + state_ch: 16 + state_t: 24 + text_encoder_class: reason1p1_7B + text_encoder_config: + ckpt_path: s3://bucket/cosmos_reasoning1/sft_exp700/sft_exp721-1_qwen7b_tl_721_5vs5_s3_balanced_n32_resume_16k/checkpoints/iter_000016000/model/ + compute_online: true + embedding_concat_strategy: full_concat + model_config: + _target_: + model_config: + _target_: cosmos_predict2._src.reason1.configs.default.model_config_qwen.QwenModelConfig + activation_checkpoint: + mode: selective + models: vlm + selective_ac_option: op + add_answer_tag: 'True' + add_cross_attention: 'False' + add_image_start_end_tag: 'False' + add_tile_tag: 'False' + architectures: + - Qwen2_5_VLForConditionalGeneration + attention_dropout: '0.0' + attn_implementation: flash_attention_2 + attn_implementation_autoset: 'True' + aux_loss_coeff: '0.0' + bad_words_ids: null + begin_suppress_tokens: null + bos_token_id: '151643' + cache_dir: null + checkpoint: + async_mode: disabled + create_seed_checkpoint: false + enable_checkpoint: false + export_dtype: float32 + folder: checkpoint + interval: 500 + interval_type: steps + model_weights_only: false + chunk_size_feed_forward: '0' + ckpt_dir: null + ckpt_path: null + comm: + init_timeout_seconds: 300 + trace_buf_size: 20000 + train_timeout_seconds: 100 + cp_size: null + cross_attention_hidden_size: null + decoder_start_token_id: null + deterministic: 'False' + diversity_penalty: '0.0' + do_sample: 'False' + early_stopping: 'False' + encoder_no_repeat_ngram_size: '0' + eos_token_id: '151645' + ep_size: null + experimental: + enable_async_tensor_parallel: false + enable_compiled_autograd: false + pipeline_parallel_degree: 1 + exponential_decay_length_penalty: null + finetuning_task: null + float8: + enable_float8_linear: false + forced_bos_token_id: null + forced_eos_token_id: null + freeze_llm: 'False' + freeze_mm_projector: 'False' + freeze_vision_encoder: 'False' + fsdp_enabled: 'False' + hidden_act: silu + hidden_size: '3584' + id2label: + 0: LABEL_0 + 1: LABEL_1 + image_token_id: '151655' + initializer_range: '0.02' + intermediate_size: '18944' + is_decoder: 'False' + is_encoder_decoder: 'False' + label2id: + LABEL_0: '0' + LABEL_1: '1' + length_penalty: '1.0' + loss_per_token: 'True' + max_batch_size: '1' + max_length: '20' + max_position_embeddings: '128000' + max_seq_len: '128000' + max_window_layers: '28' + min_length: '0' + mm_projector: null + model_type: qwen2_5_vl + name_or_path: Qwen/Qwen2.5-VL-7B-Instruct + no_repeat_ngram_size: '0' + num_attention_heads: '28' + num_beam_groups: '1' + num_beams: '1' + num_hidden_layers: '28' + num_key_value_heads: '4' + num_return_sequences: '1' + num_tiles: '1' + optimizer: + early_step_in_backward: false + end_lr: 2.5e-05 + fused: false + init_lr: 1.0e-05 + lr: 0.0003 + lr_multiplier_llm: 1.0 + lr_multiplier_mm_projector: 1.0 + lr_multiplier_vision_encoder: 0.1 + name: AdamW + output_attentions: 'False' + output_hidden_states: 'True' + output_scores: 'False' + pad_token_id: null + precision: bfloat16 + prefix: null + prepend_padding: 'False' + problem_type: null + pruned_heads: _Nothing.NOTHING + remove_invalid_values: 'False' + repetition_penalty: '1.0' + return_dict: 'True' + return_dict_in_generate: 'False' + rms_norm_eps: 1e-06 + rope_scaling: + mrope_section: + - '16' + - '24' + - '24' + rope_type: default + type: default + rope_theta: '1000000.0' + s3_credential_path: credentials/pbss_dir.secret + seed: '0' + sep_token_id: null + sliding_window: '32768' + suppress_tokens: null + task_specific_params: null + temperature: '1.0' + tf_legacy_loss: 'False' + tie_encoder_decoder: 'False' + tie_word_embeddings: 'False' + tile_tag_type: space_separated + tokenizer_class: null + tokenizer_type: Qwen/Qwen2.5-VL-7B-Instruct + top_k: '50' + top_p: '1.0' + torch_dtype: bfloat16 + torchscript: 'False' + training: + compile: false + context_parallel_degree: 1 + data_parallel_replicate_degree: 1 + data_parallel_shard_degree: -1 + disable_loss_parallel: false + enable_cpu_offload: false + fsdp_reshard_after_forward: default + mixed_precision_param: bfloat16 + mixed_precision_reduce: float32 + steps: 400000 + tensor_parallel_degree: 1 + use_cosine_decay: false + use_linear_decay: true + warmup_steps: 1000 + training_seq_len: '4096' + transformers_version: 4.51.0.dev0 + typical_p: '1.0' + use_bfloat16: 'False' + use_cache: 'False' + use_fsdp2: 'True' + use_return_dict: 'True' + use_rope_from_torchtitan: 'False' + use_sliding_window: 'False' + video_token_id: '151656' + vision_config: + _target_: cosmos_predict2._src.reason1.configs.default.model_config_qwen.QwenVisionConfig + add_cross_attention: 'False' + architectures: null + attn_implementation: flash_attention_2 + attn_implementation_autoset: 'True' + bad_words_ids: null + begin_suppress_tokens: null + bos_token_id: null + chunk_size_feed_forward: '0' + cross_attention_hidden_size: null + decoder_start_token_id: null + depth: '32' + diversity_penalty: '0.0' + do_sample: 'False' + early_stopping: 'False' + embed_dim: null + encoder_no_repeat_ngram_size: '0' + eos_token_id: null + exponential_decay_length_penalty: null + finetuning_task: null + forced_bos_token_id: null + forced_eos_token_id: null + fullatt_block_indexes: + - '7' + - '15' + - '23' + - '31' + hidden_act: silu + hidden_size: '1280' + id2label: + 0: LABEL_0 + 1: LABEL_1 + in_channels: '3' + in_chans: '3' + intermediate_size: '3420' + is_decoder: 'False' + is_encoder_decoder: 'False' + label2id: + LABEL_0: '0' + LABEL_1: '1' + length_penalty: '1.0' + max_length: '20' + min_length: '0' + mlp_ratio: null + model_type: qwen2_5_vl + name_or_path: '' + no_repeat_ngram_size: '0' + num_beam_groups: '1' + num_beams: '1' + num_heads: '16' + num_return_sequences: '1' + out_hidden_size: '3584' + output_attentions: 'False' + output_hidden_states: 'False' + output_scores: 'False' + pad_token_id: null + patch_size: '14' + prefix: null + problem_type: null + pruned_heads: _Nothing.NOTHING + remove_invalid_values: 'False' + repetition_penalty: '1.0' + return_dict: 'True' + return_dict_in_generate: 'False' + sep_token_id: null + spatial_merge_size: '2' + spatial_patch_size: '14' + suppress_tokens: null + task_specific_params: null + temperature: '1.0' + temporal_patch_size: '2' + tf_legacy_loss: 'False' + tie_encoder_decoder: 'False' + tie_word_embeddings: 'True' + tokenizer_class: null + tokens_per_second: '2' + top_k: '50' + top_p: '1.0' + torch_dtype: bfloat16 + torchscript: 'False' + typical_p: '1.0' + use_bfloat16: 'False' + window_size: '112' + vision_encoder: openai/clip-vit-base-patch32 + vision_encoder_config: + depth_init: true + dim: 1024 + ffn_dim_multiplier: null + head_dim: null + hidden_act: null + hidden_dim: 4096 + image_size: 1024 + image_token_id: null + multiple_of: null + n_heads: 16 + n_kv_heads: null + n_layers: 24 + norm_eps: 1.0e-05 + norm_type: rmsnorm + num_channels: 3 + patch_size: 16 + proj_bias: null + qkv_bias: null + rope_theta: 10000.0 + use_cache: false + use_rope_from_torchtitan: false + vision_encoder_in_channels: '3' + vision_end_token_id: '151653' + vision_start_token_id: '151652' + vision_token_id: '151654' + vocab_size: '152064' + z_loss_coeff: '0.0' + tokenizer: + _target_: + cache_dir: null + tokenizer_type: Qwen/Qwen2.5-VL-7B-Instruct + n_layers_per_group: 5 + s3_credential_path: credentials/s3_checkpoint.secret + tokenizer: + _target_: + chunk_duration: '81' + load_mean_std: 'False' + name: wan2pt1_tokenizer + temporal_window: '16' + train_time_distribution: logitnormal + train_time_weight: reweighting + use_dora: false + use_dynamic_shift: false + use_high_sigma_strategy: false + use_kerras_sigma_at_inference: true + use_lora: false + use_torch_compile: false +model_parallel: + _cpu_offloading_context: null + async_tensor_model_parallel_allreduce: false + autocast_dtype: torch.float32 + barrier_with_L1_time: true + batch_p2p_comm: true + batch_p2p_sync: true + bf16: false + context_parallel_size: 2 + cpu_offloading: false + cpu_offloading_activations: true + cpu_offloading_double_buffering: false + cpu_offloading_num_layers: 0 + cpu_offloading_weights: true + cross_entropy_fusion_impl: native + cross_entropy_loss_fusion: false + deallocate_pipeline_outputs: false + defer_embedding_wgrad_compute: false + delay_wgrad_compute: false + deterministic_mode: false + enable_autocast: false + expert_model_parallel_size: 1 + expert_tensor_parallel_size: 1 + finalize_model_grads_func: null + fp16: false + grad_scale_func: null + grad_sync_func: null + gradient_accumulation_fusion: false + hierarchical_context_parallel_sizes: null + microbatch_group_size_per_vp_stage: 1 + moe_extended_tp: false + no_sync_func: null + num_microbatches_with_partial_activation_checkpoints: null + overlap_moe_expert_parallel_comm: false + overlap_p2p_comm: false + overlap_p2p_comm_warmup_flush: false + param_sync_func: null + params_dtype: torch.float32 + perform_initialization: true + pipeline_dtype: null + pipeline_model_parallel_comm_backend: null + pipeline_model_parallel_size: 1 + sequence_parallel: false + tensor_model_parallel_size: 1 + timers: null + tp_comm_atomic_ag: false + tp_comm_atomic_rs: false + tp_comm_bootstrap_backend: nccl + tp_comm_bulk_dgrad: true + tp_comm_bulk_wgrad: true + tp_comm_overlap: false + tp_comm_overlap_ag: true + tp_comm_overlap_disable_fc1: false + tp_comm_overlap_disable_qkv: false + tp_comm_overlap_rs: true + tp_comm_overlap_rs_dgrad: false + tp_comm_split_ag: true + tp_comm_split_rs: true + use_cpu_initialization: false + use_ring_exchange_p2p: false + use_te_rng_tracker: false + variable_seq_lengths: false + virtual_pipeline_model_parallel_size: null + wgrad_deferral_limit: 0 +optimizer: + _target_: + betas: + - '0.9' + - '0.999' + eps: 1e-08 + fused: 'True' + lr: 3e-05 + model: null + optim_type: adamw + weight_decay: '0.001' +scheduler: + _target_: + cycle_lengths: + - '400000' + f_max: + - '0.99' + f_min: + - '0.4' + f_start: + - 1e-06 + verbosity_interval: '0' + warm_up_steps: + - '100' +trainer: + callbacks: + compile_tokenizer: + _target_: + compile_after_iterations: '4' + dynamic: 'False' + enabled: 'True' + dataloader_speed: + _target_: + every_n: '200' + save_s3: 'True' + step_size: '1' + device_monitor: + _target_: + every_n: '200' + log_memory_detail: 'True' + save_s3: 'True' + step_size: '1' + upload_every_n_mul: '10' + every_n_sample_ema: + _target_: + do_x0_prediction: 'False' + every_n: '1000000000000' + fps: '16' + guidance: + - '0.0' + - '3.0' + - '7.0' + is_ema: 'True' + n_sample_to_save: '128' + n_sigmas_for_x0_prediction: '4' + n_viz_sample: '3' + num_sampling_step: '35' + prompt_type: t5_xxl + run_at_start: 'False' + save_s3: 'True' + step_size: '1' + use_negative_prompt: 'False' + every_n_sample_reg: + _target_: + do_x0_prediction: 'False' + every_n: '1000000000000' + fps: '16' + guidance: + - '0.0' + - '3.0' + - '7.0' + is_ema: 'False' + n_sample_to_save: '128' + n_sigmas_for_x0_prediction: '4' + n_viz_sample: '3' + num_sampling_step: '35' + prompt_type: t5_xxl + run_at_start: 'False' + save_s3: 'True' + step_size: '1' + use_negative_prompt: 'False' + grad_clip: + _target_: + clip_norm: '0.1' + force_finite: 'True' + heart_beat: + _target_: + every_n: '10' + save_s3: 'True' + step_size: '1' + update_interval_in_minute: '20' + iter_speed: + _target_: + every_n: '200' + hit_thres: '5' + save_s3: 'True' + save_s3_every_log_n: '10' + low_prec: + _target_: + config: null + trainer: null + update_iter: '1' + manual_gc: + _target_: + every_n: '200' + warm_up: '5' + wandb: + _target_: + logging_iter_multipler: '1' + save_logging_iter_multipler: '10' + save_s3: 'True' + wandb_10x: + _target_: + logging_iter_multipler: '10' + save_logging_iter_multipler: '1' + save_s3: 'True' + cudnn: + benchmark: 'True' + deterministic: 'False' + ddp: + broadcast_buffers: 'True' + find_unused_parameters: 'False' + static_graph: 'True' + distributed_parallelism: fsdp + grad_accum_iter: '1' + grad_scaler_args: + enabled: 'False' + logging_iter: '200' + max_iter: '150000' + max_val_iter: null + memory_format: torch.preserve_format + profiling: + enable_memory_snapshot: 'False' + enable_profiling: 'False' + profile_freq: '1' + profile_memory: 'False' + record_shape: 'False' + save_s3: 'False' + target_ranks: + - '0' + - '1' + - '2' + - '3' + - '4' + - '5' + - '6' + - '7' + with_modules: 'True' + with_stack: 'True' + run_validation: 'False' + run_validation_on_start: 'False' + seed: '0' + straggler_detection: + analyze_backward: 'True' + analyze_dataloading: 'True' + analyze_forward: 'True' + analyze_optimizer: 'True' + enabled: 'True' + max_diff: '1.5' + profile_freq: '1' + raise_error: 'True' + report_freq: '100' + timeout_period: '999999999' + type: + validation_iter: '100' +upload_reproducible_setup: 'True' diff --git a/0219122156_cosmos_mip_spatial/mem_mipnerf360_a8f137cd302ac277409b1dcd26e4c8a03d8600bc4f6a80af40c71a5a980a98c5/console.log b/0219122156_cosmos_mip_spatial/mem_mipnerf360_a8f137cd302ac277409b1dcd26e4c8a03d8600bc4f6a80af40c71a5a980a98c5/console.log new file mode 100644 index 0000000000000000000000000000000000000000..2f3d86961143523b3cbc60103d3f418f508b60ae --- /dev/null +++ b/0219122156_cosmos_mip_spatial/mem_mipnerf360_a8f137cd302ac277409b1dcd26e4c8a03d8600bc4f6a80af40c71a5a980a98c5/console.log @@ -0,0 +1,22 @@ +[02-19 13:20:05|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpoint_db.py:181:path] Downloading checkpoint nvidia/Cosmos-Predict2.5-2B/base/post-trained(81edfebe-bd6a-4039-8c1d-737df1a790bf) +[02-19 13:20:05|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpoint_db.py:99:_download] Downloading checkpoint file from Hugging Face with {'repo_id': 'nvidia/Cosmos-Predict2.5-2B', 'repo_type': 'model', 'revision': '15a82a2ec231bc318692aa0456a36537c806e7d4', 'filename': 'base/post-trained/81edfebe-bd6a-4039-8c1d-737df1a790bf_ema_bf16.pt'} +[02-19 13:20:05|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpoint_db.py:181:path] Downloading checkpoint nvidia/Cosmos-Predict2.5-2B/base/pre-trained(d20b7120-df3e-4911-919d-db6e08bad31c) +[02-19 13:20:05|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpoint_db.py:99:_download] Downloading checkpoint file from Hugging Face with {'repo_id': 'nvidia/Cosmos-Predict2.5-2B', 'repo_type': 'model', 'revision': '15a82a2ec231bc318692aa0456a36537c806e7d4', 'filename': 'base/pre-trained/d20b7120-df3e-4911-919d-db6e08bad31c_ema_bf16.pt'} +[02-19 13:20:05|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpoint_db.py:181:path] Downloading checkpoint nvidia/Cosmos-Predict2.5-14B/base/pre-trained(54937b8c-29de-4f04-862c-e67b04ec41e8) +[02-19 13:20:05|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpoint_db.py:99:_download] Downloading checkpoint file from Hugging Face with {'repo_id': 'nvidia/Cosmos-Predict2.5-14B', 'repo_type': 'model', 'revision': '03eb354f35eae0d6e0c1be3c9f94d8551e125570', 'filename': 'base/pre-trained/54937b8c-29de-4f04-862c-e67b04ec41e8_ema_bf16.pt'} +[02-19 13:20:06|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpoint_db.py:181:path] Downloading checkpoint Wan2.1/vae(685afcaa-4de2-42fe-b7b9-69f7a2dee4d8) +[02-19 13:20:06|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpoint_db.py:99:_download] Downloading checkpoint file from Hugging Face with {'repo_id': 'nvidia/Cosmos-Predict2.5-2B', 'repo_type': 'model', 'revision': '6787e176dce74a101d922174a95dba29fa5f0c55', 'filename': 'tokenizer.pth'} +[02-19 13:20:08|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpoint_db.py:181:path] Downloading checkpoint Qwen/Qwen2.5-VL-7B-Instruct(7219c6c7-f878-4137-bbdb-76842ea85e70) +[02-19 13:20:08|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpoint_db.py:134:_download] Downloading checkpoint from Hugging Face with {'repo_id': 'nvidia/Cosmos-Reason1-7B', 'repo_type': 'model', 'revision': '3210bec0495fdc7a8d3dbb8d58da5711eab4b423', 'allow_patterns': ['*']} +[02-19 13:20:09|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpoint_db.py:181:path] Downloading checkpoint nvidia/Cosmos-Reason1.1-7B(cb3e3ffa-7b08-4c34-822d-61c7aa31a14f) +[02-19 13:20:09|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpoint_db.py:134:_download] Downloading checkpoint from Hugging Face with {'repo_id': 'nvidia/Cosmos-Reason1-7B', 'repo_type': 'model', 'revision': '3210bec0495fdc7a8d3dbb8d58da5711eab4b423', 'allow_patterns': ['*']} +[02-19 13:20:40|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/inference.py:60:__init__] Saved config to /home/bravo/zzhangg/MemBench/generation/outputs/cosmos/0219122156_cosmos_mip_spatial/mem_mipnerf360_a8f137cd302ac277409b1dcd26e4c8a03d8600bc4f6a80af40c71a5a980a98c5/config.yaml +[02-19 13:20:40|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/inference.py:82:generate] Generating 1 samples: ['segment_4'] +[02-19 13:20:40|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/inference.py:86:generate] [1/1] Processing sample segment_4 +[02-19 13:20:40|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/inference.py:99:_generate_sample] Saved arguments to /home/bravo/zzhangg/MemBench/generation/outputs/cosmos/0219122156_cosmos_mip_spatial/mem_mipnerf360_a8f137cd302ac277409b1dcd26e4c8a03d8600bc4f6a80af40c71a5a980a98c5/segment_4.json +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/inference.py:113:_generate_sample] Guardrail checks on prompt are disabled +[02-19 13:20:40|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/inference.py:118:_generate_sample] Generating video with autoregressive mode... +[02-19 13:34:07|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/inference.py:134:_generate_sample] Autoregressive generation completed, preparing to save video... +[02-19 13:34:07|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/inference.py:172:_generate_sample] Guardrail checks on video are disabled +[02-19 13:34:07|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/inference.py:175:_generate_sample] Saving video to %s.mp4 ... +[02-19 13:34:09|SUCCESS|generation/pipelines/cosmos-predict2.5/cosmos_predict2/inference.py:178:_generate_sample] Saved video to /home/bravo/zzhangg/MemBench/generation/outputs/cosmos/0219122156_cosmos_mip_spatial/mem_mipnerf360_a8f137cd302ac277409b1dcd26e4c8a03d8600bc4f6a80af40c71a5a980a98c5/segment_4.mp4 diff --git a/0219122156_cosmos_mip_spatial/mem_mipnerf360_a8f137cd302ac277409b1dcd26e4c8a03d8600bc4f6a80af40c71a5a980a98c5/debug.log b/0219122156_cosmos_mip_spatial/mem_mipnerf360_a8f137cd302ac277409b1dcd26e4c8a03d8600bc4f6a80af40c71a5a980a98c5/debug.log new file mode 100644 index 0000000000000000000000000000000000000000..ecbec14c913788625e1692651178eba02a3b6950 --- /dev/null +++ b/0219122156_cosmos_mip_spatial/mem_mipnerf360_a8f137cd302ac277409b1dcd26e4c8a03d8600bc4f6a80af40c71a5a980a98c5/debug.log @@ -0,0 +1,353 @@ +[02-19 13:19:52|DEBUG|generation/pipelines/cosmos-predict2.5/packages/cosmos-oss/cosmos_oss/init.py:176:init_output_dir] Flags(internal=False, smoke=False, verbose=False, experimental_checkpoints=False) +[02-19 13:19:59|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/env_parsers/env_parser.py:86:get_val_dict] getting val dict of CredentialEnvParser +[02-19 13:19:59|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/env_parsers/env_parser.py:86:get_val_dict] getting val dict of CredentialEnvParser +[02-19 13:20:01|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/visualize/video.py:31:] No module named 'ffmpegcv' +[02-19 13:20:01|CRITICAL|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/checkpointer/dcp.py:188:] for the back comptiable pytorch! New DefaultLoadPlanner class is created. +[02-19 13:20:01|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/inference.py:32:__init__] SetupArguments(output_dir=PosixPath('/home/bravo/zzhangg/MemBench/generation/outputs/cosmos/0219122156_cosmos_mip_spatial/mem_mipnerf360_a8f137cd302ac277409b1dcd26e4c8a03d8600bc4f6a80af40c71a5a980a98c5') model='2B/post-trained' checkpoint_path='/home/bravo/zzhangg/MemBench/generation/pipelines/cosmos-predict2.5/models/Cosmos-Predict2.5-2B/base/post-trained/81edfebe-bd6a-4039-8c1d-737df1a790bf_ema_bf16.pt' experiment='Stage-c_pt_4-Index-2-Size-2B-Res-720-Fps-16-Note-rf_with_edm_ckpt' config_file='cosmos_predict2/_src/predict2/configs/video2world/config.py' context_parallel_size=1 offload_diffusion_model=False offload_tokenizer=False offload_text_encoder=False disable_guardrails=True offload_guardrail_models=True keep_going=True profile=False) +[02-19 13:20:01|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/datasets/augmentor_provider.py:94:augmentor_register] registering video_basic_augmentor_v1... +[02-19 13:20:01|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/datasets/augmentor_provider.py:94:augmentor_register] registering video_basic_augmentor_v2... +[02-19 13:20:01|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/datasets/augmentor_provider.py:94:augmentor_register] registering noframedrop_nocameramove_video_augmentor_v1... +[02-19 13:20:01|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/datasets/augmentor_provider.py:94:augmentor_register] registering nocameramove_video_augmentor_v1... +[02-19 13:20:01|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/datasets/augmentor_provider.py:94:augmentor_register] registering image_basic_augmentor... +[02-19 13:20:01|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/datasets/augmentor_provider.py:94:augmentor_register] registering image_basic_augmentor_without_embeddings... +[02-19 13:20:01|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/attention/utils/safe_log.py:35:debug] Flash Attention v3 is not supported because the Python package was not found. +[02-19 13:20:01|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/attention/utils/safe_log.py:35:debug] NATTEN Attention is not supported due to insufficient NATTEN version natten.__version__='0.21.0', expected at least NATTEN_MIN_RELEASE_VERSION=[0, 21, 5], or NATTEN_MIN_DEV_VERSION=([0, 21, 5], 9). +[02-19 13:20:01|CRITICAL|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:192:import_all_modules_from_package] Reloading all modules from package cosmos_predict2._src.predict2.configs.video2world.experiment +[02-19 13:20:01|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_predict2._src.predict2.configs.video2world.experiment.reason_embeddings +[02-19 13:20:01|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_predict2._src.predict2.configs.video2world.experiment.reason_embeddings.model_14b_reason_1p1 +[02-19 13:20:02|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_predict2._src.predict2.configs.video2world.experiment.reason_embeddings.model_14b_reason_1p1_rectified_flow +[02-19 13:20:02|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_predict2._src.predict2.configs.video2world.experiment.reason_embeddings.model_2B_reason_1p1 +[02-19 13:20:02|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_predict2._src.predict2.configs.video2world.experiment.reason_embeddings.model_2B_reason_1p1_rectified_flow +[02-19 13:20:02|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_predict2._src.predict2.configs.video2world.experiment.reason_embeddings.model_7b_reason_1p1 +[02-19 13:20:02|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_predict2._src.predict2.configs.video2world.experiment.reason_embeddings.stage3_14B_index_3 +[02-19 13:20:02|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_predict2._src.predict2.configs.video2world.experiment.reason_embeddings.stage3_2B +[02-19 13:20:02|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_predict2._src.predict2.configs.video2world.experiment.resume_text2world +[02-19 13:20:02|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_predict2._src.predict2.configs.video2world.experiment.resume_text2world.sparse_14B +[02-19 13:20:03|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_predict2._src.predict2.configs.video2world.experiment.resume_text2world.sparse_2B +[02-19 13:20:03|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_predict2._src.predict2.configs.video2world.experiment.resume_text2world.stage3 +[02-19 13:20:03|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_predict2._src.predict2.configs.video2world.experiment.resume_text2world.stage3_2B_ablation +[02-19 13:20:04|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_predict2._src.predict2.configs.video2world.experiment.resume_text2world.state3_14B_index_3 +[02-19 13:20:04|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_predict2._src.predict2.configs.video2world.experiment.specialized_model +[02-19 13:20:04|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_predict2._src.predict2.configs.video2world.experiment.specialized_model.SFT_14B_RF +[02-19 13:20:04|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_predict2._src.predict2.configs.video2world.experiment.specialized_model.SFT_2B +[02-19 13:20:04|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_predict2._src.predict2.configs.video2world.experiment.specialized_model.SFT_2B_RF +[02-19 13:20:04|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_face +[02-19 13:20:04|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_crowded +[02-19 13:20:04|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_high_motion +[02-19 13:20:04|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_robotics +[02-19 13:20:04|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_av +[02-19 13:20:04|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_physical_ai +[02-19 13:20:04|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_4k_cooldown +[02-19 13:20:04|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_face_only2 +[02-19 13:20:04|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_crowded_only2 +[02-19 13:20:04|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_high_motion_only2 +[02-19 13:20:04|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_robotics_only2 +[02-19 13:20:04|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_av_only2 +[02-19 13:20:04|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_physical_ai_only2 +[02-19 13:20:04|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_4k_cooldown_only2 +[02-19 13:20:04|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_face_high_sigma +[02-19 13:20:04|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_crowded_high_sigma +[02-19 13:20:04|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_high_motion_high_sigma +[02-19 13:20:04|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_robotics_high_sigma +[02-19 13:20:04|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_av_high_sigma +[02-19 13:20:04|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_physical_ai_high_sigma +[02-19 13:20:04|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_4k_cooldown_high_sigma +[02-19 13:20:05|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_face_high_sigma2 +[02-19 13:20:05|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_crowded_high_sigma2 +[02-19 13:20:05|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_robotics_high_sigma2 +[02-19 13:20:05|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-2-Size-2B-Res-720-Fps-16-Note-rf_with_edm_ckpt +[02-19 13:20:05|CRITICAL|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:192:import_all_modules_from_package] Reloading all modules from package cosmos_predict2.experiments +[02-19 13:20:05|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_predict2.experiments.base +[02-19 13:20:05|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_predict2.experiments.base.action +[02-19 13:20:05|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpoint_db.py:181:path] Downloading checkpoint nvidia/Cosmos-Predict2.5-2B/base/post-trained(81edfebe-bd6a-4039-8c1d-737df1a790bf) +[02-19 13:20:05|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpoint_db.py:99:_download] Downloading checkpoint file from Hugging Face with {'repo_id': 'nvidia/Cosmos-Predict2.5-2B', 'repo_type': 'model', 'revision': '15a82a2ec231bc318692aa0456a36537c806e7d4', 'filename': 'base/post-trained/81edfebe-bd6a-4039-8c1d-737df1a790bf_ema_bf16.pt'} +[02-19 13:20:05|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_predict2.experiments.base.cosmos_nemo_assets +[02-19 13:20:05|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpoint_db.py:181:path] Downloading checkpoint nvidia/Cosmos-Predict2.5-2B/base/pre-trained(d20b7120-df3e-4911-919d-db6e08bad31c) +[02-19 13:20:05|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpoint_db.py:99:_download] Downloading checkpoint file from Hugging Face with {'repo_id': 'nvidia/Cosmos-Predict2.5-2B', 'repo_type': 'model', 'revision': '15a82a2ec231bc318692aa0456a36537c806e7d4', 'filename': 'base/pre-trained/d20b7120-df3e-4911-919d-db6e08bad31c_ema_bf16.pt'} +[02-19 13:20:05|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_predict2.experiments.base.cosmos_nemo_assets_lora +[02-19 13:20:05|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_predict2.experiments.base.groot +[02-19 13:20:05|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpoint_db.py:181:path] Downloading checkpoint nvidia/Cosmos-Predict2.5-14B/base/pre-trained(54937b8c-29de-4f04-862c-e67b04ec41e8) +[02-19 13:20:05|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpoint_db.py:99:_download] Downloading checkpoint file from Hugging Face with {'repo_id': 'nvidia/Cosmos-Predict2.5-14B', 'repo_type': 'model', 'revision': '03eb354f35eae0d6e0c1be3c9f94d8551e125570', 'filename': 'base/pre-trained/54937b8c-29de-4f04-862c-e67b04ec41e8_ema_bf16.pt'} +[02-19 13:20:06|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/utils/model_loader.py:91:load_model_from_checkpoint] Overriding config checkpoint path with: /home/bravo/zzhangg/MemBench/generation/pipelines/cosmos-predict2.5/models/Cosmos-Predict2.5-2B/base/post-trained/81edfebe-bd6a-4039-8c1d-737df1a790bf_ema_bf16.pt +[02-19 13:20:06|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/misc.py:152:set_random_seed] Using random seed 0. +[02-19 13:20:06|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/utils/model_loader.py:111:load_model_from_checkpoint] Loading model from /home/bravo/zzhangg/MemBench/generation/pipelines/cosmos-predict2.5/models/Cosmos-Predict2.5-2B/base/post-trained/81edfebe-bd6a-4039-8c1d-737df1a790bf_ema_bf16.pt +[02-19 13:20:06|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/models/text2world_model_rectified_flow.py:132:__init__] DiffusionModel: precision torch.bfloat16 +[02-19 13:20:06|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpoint_db.py:181:path] Downloading checkpoint Wan2.1/vae(685afcaa-4de2-42fe-b7b9-69f7a2dee4d8) +[02-19 13:20:06|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpoint_db.py:99:_download] Downloading checkpoint file from Hugging Face with {'repo_id': 'nvidia/Cosmos-Predict2.5-2B', 'repo_type': 'model', 'revision': '6787e176dce74a101d922174a95dba29fa5f0c55', 'filename': 'tokenizer.pth'} +[02-19 13:20:07|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/tokenizers/wan2pt1.py:669:_video_vae] loading /home/bravo/.cache/huggingface/hub/models--nvidia--Cosmos-Predict2.5-2B/snapshots/6787e176dce74a101d922174a95dba29fa5f0c55/tokenizer.pth +[02-19 13:20:07|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/distributed.py:415:sync_model_states] Synchronizing model states from rank 0 to all ranks in process group [0]. +[02-19 13:20:07|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/timer.py:150:_log] Time spent on DiffusionModel: set_up_tokenizer: 1.11 s +[02-19 13:20:07|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #0-fps: + ReMapkey + input key: fps + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: fps + Dtype: None +[02-19 13:20:07|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #1-padding_mask: + ReMapkey + input key: padding_mask + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: padding_mask + Dtype: None +[02-19 13:20:07|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #2-text: + TextAttr + input key: ['t5_text_embeddings'] + Param count: 0 + Trainable: None + Dropout rate: 0.2 + Output key: [crossattn_emb] +[02-19 13:20:07|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #3-use_video_condition: + BooleanFlag + input key: fps + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: use_video_condition + This is a boolean flag +[02-19 13:20:07|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1624:build_pos_embed] Building positional embedding with rope3d class, impl +[02-19 13:20:07|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:886:__init__] Using AdaLN LoRA Flag: True. We enable bias if no AdaLN LoRA for backward compatibility. +[02-19 13:20:07|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:20:07|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:20:07|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:20:07|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:20:07|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:20:07|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:20:07|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:20:07|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:20:07|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:20:07|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:20:07|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:20:07|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:20:07|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:20:07|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:20:07|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:20:07|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:20:07|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:20:07|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:20:07|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:20:07|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:20:07|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:20:07|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:20:07|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:20:07|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:20:07|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:20:07|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:20:07|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:20:07|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:20:07|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:20:07|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:20:07|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:20:07|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:20:07|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:20:07|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:20:07|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:20:07|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:20:07|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:20:07|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:20:07|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:20:07|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:20:07|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:20:07|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:20:07|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:20:07|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:20:07|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:20:07|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:20:07|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:20:07|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:20:07|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:20:07|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:20:07|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:20:07|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:20:08|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:20:08|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:20:08|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[02-19 13:20:08|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[02-19 13:20:08|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1804:enable_selective_checkpoint] Enable selective checkpoint with predict2_2b_720_aggressive, for every 1 blocks. Total blocks: 28 +[02-19 13:20:08|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 0 +[02-19 13:20:08|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 1 +[02-19 13:20:08|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 2 +[02-19 13:20:08|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 3 +[02-19 13:20:08|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 4 +[02-19 13:20:08|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 5 +[02-19 13:20:08|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 6 +[02-19 13:20:08|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 7 +[02-19 13:20:08|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 8 +[02-19 13:20:08|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 9 +[02-19 13:20:08|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 10 +[02-19 13:20:08|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 11 +[02-19 13:20:08|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 12 +[02-19 13:20:08|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 13 +[02-19 13:20:08|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 14 +[02-19 13:20:08|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 15 +[02-19 13:20:08|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 16 +[02-19 13:20:08|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 17 +[02-19 13:20:08|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 18 +[02-19 13:20:08|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 19 +[02-19 13:20:08|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 20 +[02-19 13:20:08|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 21 +[02-19 13:20:08|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 22 +[02-19 13:20:08|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 23 +[02-19 13:20:08|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 24 +[02-19 13:20:08|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 25 +[02-19 13:20:08|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 26 +[02-19 13:20:08|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 27 +[02-19 13:20:08|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/timer.py:150:_log] Time spent on meta to cuda and broadcast model states: 0.13 s +[02-19 13:20:08|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/timer.py:150:_log] Time spent on Creating PyTorch model: 0.61 s +[02-19 13:20:08|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/timer.py:150:_log] Time spent on Creating PyTorch model and ema if enabled: 0.62 s +[02-19 13:20:08|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/timer.py:150:_log] Time spent on DiffusionModel: set_up_model: 0.62 s +[02-19 13:20:08|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/text_encoders/text_encoder.py:74:__init__] Instantiating text encoder model... +[02-19 13:20:08|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpoint_db.py:181:path] Downloading checkpoint Qwen/Qwen2.5-VL-7B-Instruct(7219c6c7-f878-4137-bbdb-76842ea85e70) +[02-19 13:20:08|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpoint_db.py:134:_download] Downloading checkpoint from Hugging Face with {'repo_id': 'nvidia/Cosmos-Reason1-7B', 'repo_type': 'model', 'revision': '3210bec0495fdc7a8d3dbb8d58da5711eab4b423', 'allow_patterns': ['*']} +[02-19 13:20:09|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/reason1/tokenizer/processor.py:75:__init__] Successfully loaded processor from local cache +[02-19 13:20:09|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/reason1/models/vlm_base.py:116:__init__] Setting torch default dtype from torch.float32 to torch.bfloat16 +[02-19 13:20:09|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/reason1/parallelisms/parallel_dims.py:50:_validate] dp_shard is set to -1, will be automatically determined based on world_size 1 // 1. +[02-19 13:20:09|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/reason1/parallelisms/parallel_dims.py:54:_validate] dp_shard is set to 1. +[02-19 13:20:09|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/reason1/parallelisms/parallel_dims.py:75:build_mesh] Building 0-D device mesh with [], [] +[02-19 13:20:09|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/reason1/parallelisms/parallel_dims.py:105:build_mesh] mesh: DeviceMesh('cuda', 0) +[02-19 13:20:09|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/reason1/parallelisms/parallelize_qwen.py:336:apply_ac] Applied selective activation checkpointing to the model +[02-19 13:20:09|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/reason1/models/vlm_base.py:123:__init__] Reset torch default dtype to torch.float32 +[02-19 13:20:09|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/text_encoders/text_encoder.py:84:__init__] Loading checkpoint from s3://bucket/cosmos_reasoning1/sft_exp700/sft_exp721-1_qwen7b_tl_721_5vs5_s3_balanced_n32_resume_16k/checkpoints/iter_000016000/model/. +[02-19 13:20:09|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpoint_db.py:181:path] Downloading checkpoint nvidia/Cosmos-Reason1.1-7B(cb3e3ffa-7b08-4c34-822d-61c7aa31a14f) +[02-19 13:20:09|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpoint_db.py:134:_download] Downloading checkpoint from Hugging Face with {'repo_id': 'nvidia/Cosmos-Reason1-7B', 'repo_type': 'model', 'revision': '3210bec0495fdc7a8d3dbb8d58da5711eab4b423', 'allow_patterns': ['*']} +[02-19 13:20:35|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/text_encoders/text_encoder.py:113:__init__] Finished loading checkpoint from /home/bravo/.cache/huggingface/hub/models--nvidia--Cosmos-Reason1-7B/snapshots/3210bec0495fdc7a8d3dbb8d58da5711eab4b423. +[02-19 13:20:35|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/text_encoders/text_encoder.py:116:__init__] Text encoder model instantiated +[02-19 13:20:35|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/timer.py:150:_log] Time spent on instantiate model: 28.57 s +[02-19 13:20:35|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/utils/model_loader.py:227:load_model_state_dict_from_checkpoint] Loading model cached locally from /home/bravo/zzhangg/MemBench/generation/pipelines/cosmos-predict2.5/models/Cosmos-Predict2.5-2B/base/post-trained/81edfebe-bd6a-4039-8c1d-737df1a790bf_ema_bf16.pt +[02-19 13:20:39|CRITICAL|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/models/text2world_model_rectified_flow.py:919:load_state_dict] load model in non-strict mode +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key t_embedding_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[02-19 13:20:40|CRITICAL|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/models/text2world_model_rectified_flow.py:920:load_state_dict] [RANK 0] _IncompatibleKeys(missing_keys=[], unexpected_keys=[], incorrect_shapes=[]) +[02-19 13:20:40|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/distributed.py:415:sync_model_states] Synchronizing model states from rank 0 to all ranks in process group [0]. +[02-19 13:20:40|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/inference.py:60:__init__] Saved config to /home/bravo/zzhangg/MemBench/generation/outputs/cosmos/0219122156_cosmos_mip_spatial/mem_mipnerf360_a8f137cd302ac277409b1dcd26e4c8a03d8600bc4f6a80af40c71a5a980a98c5/config.yaml +[02-19 13:20:40|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/inference.py:82:generate] Generating 1 samples: ['segment_4'] +[02-19 13:20:40|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/inference.py:86:generate] [1/1] Processing sample segment_4 +[02-19 13:20:40|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/inference.py:93:_generate_sample] InferenceArguments(name='segment_4' prompt_path=None prompt='In the final moments, the camera completes its pan and tilting motion, returning to a view centered on the bonsai tree. The red bicycle is once again visible in the background, and the wooden floor and patterned rug are clearly seen. The camera makes slight adjustments in angle and zoom, ensuring the spatial consistency of the scene is maintained as it revisits familiar elements.' negative_prompt='' seed=46 guidance=7 inference_type= input_path=PosixPath('/home/bravo/zzhangg/MemBench/generation/outputs/cosmos/0219122156_cosmos_mip_spatial/mem_mipnerf360_a8f137cd302ac277409b1dcd26e4c8a03d8600bc4f6a80af40c71a5a980a98c5/segment_3.mp4') resolution='none' num_output_frames=77 num_steps=35 enable_autoregressive=True chunk_size=77 chunk_overlap=1) +[02-19 13:20:40|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/inference.py:99:_generate_sample] Saved arguments to /home/bravo/zzhangg/MemBench/generation/outputs/cosmos/0219122156_cosmos_mip_spatial/mem_mipnerf360_a8f137cd302ac277409b1dcd26e4c8a03d8600bc4f6a80af40c71a5a980a98c5/segment_4.json +[02-19 13:20:40|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/inference.py:113:_generate_sample] Guardrail checks on prompt are disabled +[02-19 13:20:40|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/inference.py:118:_generate_sample] Generating video with autoregressive mode... +[02-19 13:20:40|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/inference/video2world.py:730:generate_autoregressive_from_batch] Processing video input for autoregressive: /home/bravo/zzhangg/MemBench/generation/outputs/cosmos/0219122156_cosmos_mip_spatial/mem_mipnerf360_a8f137cd302ac277409b1dcd26e4c8a03d8600bc4f6a80af40c71a5a980a98c5/segment_3.mp4 +[02-19 13:20:41|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/inference/video2world.py:798:generate_autoregressive_from_batch] Generating 1 chunks with chunk_size=77, chunk_overlap=1 for 77 total frames +[02-19 13:20:41|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/inference/video2world.py:816:generate_autoregressive_from_batch] Processing chunk 1/1, frames 0-77 +[02-19 13:20:42|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/inference/video2world.py:556:generate_vid2world] GPU memory usage after getting data_batch: 19.79 GB +[02-19 13:20:42|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/inference/video2world.py:594:generate_vid2world] [Memory Optimization] Starting latent sample generation +[02-19 13:34:07|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/predict2/inference/video2world.py:878:generate_autoregressive_from_batch] Generated final video with shape torch.Size([1, 3, 77, 704, 1280]) +[02-19 13:34:07|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/inference.py:134:_generate_sample] Autoregressive generation completed, preparing to save video... +[02-19 13:34:07|WARNING|generation/pipelines/cosmos-predict2.5/cosmos_predict2/inference.py:172:_generate_sample] Guardrail checks on video are disabled +[02-19 13:34:07|INFO|generation/pipelines/cosmos-predict2.5/cosmos_predict2/inference.py:175:_generate_sample] Saving video to %s.mp4 ... +[02-19 13:34:09|DEBUG|generation/pipelines/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/utils/easy_io/handlers/imageio_video_handler.py:163:dump_to_fileobj] mimsave_kwargs: {'fps': 16, 'quality': 5, 'macro_block_size': 1, 'ffmpeg_params': ['-s', '1280x704'], 'output_params': ['-f', 'mp4']} +[02-19 13:34:09|SUCCESS|generation/pipelines/cosmos-predict2.5/cosmos_predict2/inference.py:178:_generate_sample] Saved video to /home/bravo/zzhangg/MemBench/generation/outputs/cosmos/0219122156_cosmos_mip_spatial/mem_mipnerf360_a8f137cd302ac277409b1dcd26e4c8a03d8600bc4f6a80af40c71a5a980a98c5/segment_4.mp4 diff --git a/0219122156_cosmos_mip_spatial/mem_mipnerf360_a8f137cd302ac277409b1dcd26e4c8a03d8600bc4f6a80af40c71a5a980a98c5/segment_0.json b/0219122156_cosmos_mip_spatial/mem_mipnerf360_a8f137cd302ac277409b1dcd26e4c8a03d8600bc4f6a80af40c71a5a980a98c5/segment_0.json new file mode 100644 index 0000000000000000000000000000000000000000..2d11009eadc791974ec040d605f876e53a0e8f2c --- /dev/null +++ b/0219122156_cosmos_mip_spatial/mem_mipnerf360_a8f137cd302ac277409b1dcd26e4c8a03d8600bc4f6a80af40c71a5a980a98c5/segment_0.json @@ -0,0 +1 @@ +{"name":"segment_0","prompt_path":null,"prompt":"The video opens with a tight focus on the bonsai tree, highlighting its pink blossoms and the purple cloth beneath it. The red bicycle and Yamaha keyboard are partially visible in the background. The camera begins a slow horizontal pan to the left, revealing a window that allows natural light to enter the room, along with a computer monitor and various items like boxes and a patterned rug.","negative_prompt":"","seed":42,"guidance":7,"inference_type":"text2world","input_path":null,"resolution":"none","num_output_frames":77,"num_steps":35,"enable_autoregressive":true,"chunk_size":77,"chunk_overlap":1} \ No newline at end of file diff --git a/0219122156_cosmos_mip_spatial/mem_mipnerf360_a8f137cd302ac277409b1dcd26e4c8a03d8600bc4f6a80af40c71a5a980a98c5/segment_0.mp4 b/0219122156_cosmos_mip_spatial/mem_mipnerf360_a8f137cd302ac277409b1dcd26e4c8a03d8600bc4f6a80af40c71a5a980a98c5/segment_0.mp4 new file mode 100644 index 0000000000000000000000000000000000000000..69f095a702d2a027593418576d92aa787a4e9f8f --- /dev/null +++ b/0219122156_cosmos_mip_spatial/mem_mipnerf360_a8f137cd302ac277409b1dcd26e4c8a03d8600bc4f6a80af40c71a5a980a98c5/segment_0.mp4 @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:204c7c41ef5c14893738b5a8dd7166043122fe8dc3c12573a685ae21982d4706 +size 1748517 diff --git a/0219122156_cosmos_mip_spatial/mem_mipnerf360_a8f137cd302ac277409b1dcd26e4c8a03d8600bc4f6a80af40c71a5a980a98c5/segment_1.json b/0219122156_cosmos_mip_spatial/mem_mipnerf360_a8f137cd302ac277409b1dcd26e4c8a03d8600bc4f6a80af40c71a5a980a98c5/segment_1.json new file mode 100644 index 0000000000000000000000000000000000000000..5b2020ba999e932d2cf0a725077b5bcca8301f5c --- /dev/null +++ b/0219122156_cosmos_mip_spatial/mem_mipnerf360_a8f137cd302ac277409b1dcd26e4c8a03d8600bc4f6a80af40c71a5a980a98c5/segment_1.json @@ -0,0 +1 @@ +{"name":"segment_1","prompt_path":null,"prompt":"As the camera continues its pan, it moves further left, showcasing more of the room's contents, including a plastic bag filled with items and additional boxes. The natural light from the window becomes more apparent, casting shadows across the wooden floor. The camera then starts to tilt upward, shifting the focus toward the upper part of the room, including the computer monitor and ceiling.","negative_prompt":"","seed":43,"guidance":7,"inference_type":"video2world","input_path":"/home/bravo/zzhangg/MemBench/generation/outputs/cosmos/0219122156_cosmos_mip_spatial/mem_mipnerf360_a8f137cd302ac277409b1dcd26e4c8a03d8600bc4f6a80af40c71a5a980a98c5/segment_0.mp4","resolution":"none","num_output_frames":77,"num_steps":35,"enable_autoregressive":true,"chunk_size":77,"chunk_overlap":1} \ No newline at end of file diff --git a/0219122156_cosmos_mip_spatial/mem_mipnerf360_a8f137cd302ac277409b1dcd26e4c8a03d8600bc4f6a80af40c71a5a980a98c5/segment_1.mp4 b/0219122156_cosmos_mip_spatial/mem_mipnerf360_a8f137cd302ac277409b1dcd26e4c8a03d8600bc4f6a80af40c71a5a980a98c5/segment_1.mp4 new file mode 100644 index 0000000000000000000000000000000000000000..370b8cdcbf5a57ca1a9d41e946061b34618decb4 --- /dev/null +++ b/0219122156_cosmos_mip_spatial/mem_mipnerf360_a8f137cd302ac277409b1dcd26e4c8a03d8600bc4f6a80af40c71a5a980a98c5/segment_1.mp4 @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:160e7241ae47afe59ceb7759610aefeed42a2708ec9b33c46f3cdec916d9e0bb +size 1430288 diff --git a/0219122156_cosmos_mip_spatial/mem_mipnerf360_a8f137cd302ac277409b1dcd26e4c8a03d8600bc4f6a80af40c71a5a980a98c5/segment_2.json b/0219122156_cosmos_mip_spatial/mem_mipnerf360_a8f137cd302ac277409b1dcd26e4c8a03d8600bc4f6a80af40c71a5a980a98c5/segment_2.json new file mode 100644 index 0000000000000000000000000000000000000000..d88cac6fe2bb9f1483eec770ac7bd64897a06556 --- /dev/null +++ b/0219122156_cosmos_mip_spatial/mem_mipnerf360_a8f137cd302ac277409b1dcd26e4c8a03d8600bc4f6a80af40c71a5a980a98c5/segment_2.json @@ -0,0 +1 @@ +{"name":"segment_2","prompt_path":null,"prompt":"The camera reaches its highest point, capturing the top of the room and the ceiling before beginning to tilt back down. It then starts to pan back to the right, retracing its path and bringing the bonsai tree back into the center of the frame. The red bicycle and Yamaha keyboard reappear in the background as the camera returns to a familiar angle.","negative_prompt":"","seed":44,"guidance":7,"inference_type":"video2world","input_path":"/home/bravo/zzhangg/MemBench/generation/outputs/cosmos/0219122156_cosmos_mip_spatial/mem_mipnerf360_a8f137cd302ac277409b1dcd26e4c8a03d8600bc4f6a80af40c71a5a980a98c5/segment_1.mp4","resolution":"none","num_output_frames":77,"num_steps":35,"enable_autoregressive":true,"chunk_size":77,"chunk_overlap":1} \ No newline at end of file diff --git a/0219122156_cosmos_mip_spatial/mem_mipnerf360_a8f137cd302ac277409b1dcd26e4c8a03d8600bc4f6a80af40c71a5a980a98c5/segment_2.mp4 b/0219122156_cosmos_mip_spatial/mem_mipnerf360_a8f137cd302ac277409b1dcd26e4c8a03d8600bc4f6a80af40c71a5a980a98c5/segment_2.mp4 new file mode 100644 index 0000000000000000000000000000000000000000..f6b2806a19a3dd1bc73b0314878d8f3a9c38e630 --- /dev/null +++ b/0219122156_cosmos_mip_spatial/mem_mipnerf360_a8f137cd302ac277409b1dcd26e4c8a03d8600bc4f6a80af40c71a5a980a98c5/segment_2.mp4 @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bfb9cb78cbe769f4bb501c2a01bb44b6cd9853ac9622e7b54019c8781eab1af3 +size 1012866 diff --git a/0219122156_cosmos_mip_spatial/mem_mipnerf360_a8f137cd302ac277409b1dcd26e4c8a03d8600bc4f6a80af40c71a5a980a98c5/segment_3.json b/0219122156_cosmos_mip_spatial/mem_mipnerf360_a8f137cd302ac277409b1dcd26e4c8a03d8600bc4f6a80af40c71a5a980a98c5/segment_3.json new file mode 100644 index 0000000000000000000000000000000000000000..32adbb9c44574eadb112534e10d0a75d5e3f90ca --- /dev/null +++ b/0219122156_cosmos_mip_spatial/mem_mipnerf360_a8f137cd302ac277409b1dcd26e4c8a03d8600bc4f6a80af40c71a5a980a98c5/segment_3.json @@ -0,0 +1 @@ +{"name":"segment_3","prompt_path":null,"prompt":"As the camera continues its pan back to the right, it focuses once again on the bonsai tree, with the red bicycle and Yamaha keyboard clearly visible. The wooden floor and patterned rug are also in view. The camera makes slight adjustments in angle and zoom, ensuring the spatial consistency of the scene is maintained as it revisits familiar elements.","negative_prompt":"","seed":45,"guidance":7,"inference_type":"video2world","input_path":"/home/bravo/zzhangg/MemBench/generation/outputs/cosmos/0219122156_cosmos_mip_spatial/mem_mipnerf360_a8f137cd302ac277409b1dcd26e4c8a03d8600bc4f6a80af40c71a5a980a98c5/segment_2.mp4","resolution":"none","num_output_frames":77,"num_steps":35,"enable_autoregressive":true,"chunk_size":77,"chunk_overlap":1} \ No newline at end of file diff --git a/0219122156_cosmos_mip_spatial/mem_mipnerf360_a8f137cd302ac277409b1dcd26e4c8a03d8600bc4f6a80af40c71a5a980a98c5/segment_3.mp4 b/0219122156_cosmos_mip_spatial/mem_mipnerf360_a8f137cd302ac277409b1dcd26e4c8a03d8600bc4f6a80af40c71a5a980a98c5/segment_3.mp4 new file mode 100644 index 0000000000000000000000000000000000000000..0de62acfd2e8f59952d1a7f57c182715a03444c8 --- /dev/null +++ b/0219122156_cosmos_mip_spatial/mem_mipnerf360_a8f137cd302ac277409b1dcd26e4c8a03d8600bc4f6a80af40c71a5a980a98c5/segment_3.mp4 @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c6df6ba11736eb031fd65d130faa034c9d4c19f15d0ee6cb807c8cd6bea77da6 +size 1731798 diff --git a/0219122156_cosmos_mip_spatial/mem_mipnerf360_a8f137cd302ac277409b1dcd26e4c8a03d8600bc4f6a80af40c71a5a980a98c5/segment_4.json b/0219122156_cosmos_mip_spatial/mem_mipnerf360_a8f137cd302ac277409b1dcd26e4c8a03d8600bc4f6a80af40c71a5a980a98c5/segment_4.json new file mode 100644 index 0000000000000000000000000000000000000000..9eac5ba9a13dd764201300a49e2000472f72f77a --- /dev/null +++ b/0219122156_cosmos_mip_spatial/mem_mipnerf360_a8f137cd302ac277409b1dcd26e4c8a03d8600bc4f6a80af40c71a5a980a98c5/segment_4.json @@ -0,0 +1 @@ +{"name":"segment_4","prompt_path":null,"prompt":"In the final moments, the camera completes its pan and tilting motion, returning to a view centered on the bonsai tree. The red bicycle is once again visible in the background, and the wooden floor and patterned rug are clearly seen. The camera makes slight adjustments in angle and zoom, ensuring the spatial consistency of the scene is maintained as it revisits familiar elements.","negative_prompt":"","seed":46,"guidance":7,"inference_type":"video2world","input_path":"/home/bravo/zzhangg/MemBench/generation/outputs/cosmos/0219122156_cosmos_mip_spatial/mem_mipnerf360_a8f137cd302ac277409b1dcd26e4c8a03d8600bc4f6a80af40c71a5a980a98c5/segment_3.mp4","resolution":"none","num_output_frames":77,"num_steps":35,"enable_autoregressive":true,"chunk_size":77,"chunk_overlap":1} \ No newline at end of file diff --git a/0219122156_cosmos_mip_spatial/mem_mipnerf360_a8f137cd302ac277409b1dcd26e4c8a03d8600bc4f6a80af40c71a5a980a98c5/segment_4.mp4 b/0219122156_cosmos_mip_spatial/mem_mipnerf360_a8f137cd302ac277409b1dcd26e4c8a03d8600bc4f6a80af40c71a5a980a98c5/segment_4.mp4 new file mode 100644 index 0000000000000000000000000000000000000000..dae107772fb946263a38a60082a340ebe949ec36 --- /dev/null +++ b/0219122156_cosmos_mip_spatial/mem_mipnerf360_a8f137cd302ac277409b1dcd26e4c8a03d8600bc4f6a80af40c71a5a980a98c5/segment_4.mp4 @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6714a0716d67c0e9bc3b5bd649cd75fa735766dab0d71e86b591fd585467c500 +size 1378349 diff --git a/0219122156_cosmos_mip_spatial/videos/mem_mipnerf360_575f06997f088dc5824bec5a960459303edd0154e65b3594e29810380ac7ea50.mp4 b/0219122156_cosmos_mip_spatial/videos/mem_mipnerf360_575f06997f088dc5824bec5a960459303edd0154e65b3594e29810380ac7ea50.mp4 new file mode 100644 index 0000000000000000000000000000000000000000..c12da25b51d96cf1686b9c7154fd82cf3ef05f79 --- /dev/null +++ b/0219122156_cosmos_mip_spatial/videos/mem_mipnerf360_575f06997f088dc5824bec5a960459303edd0154e65b3594e29810380ac7ea50.mp4 @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f159c17e9271db1a9b018eb11b062bca8932188c62f33809ea4c1913f9e5fd20 +size 5706848 diff --git a/0219122156_cosmos_mip_spatial/videos/mem_mipnerf360_5f84c46d31489bd1a6b57b7c5c8cbc15280e279f8eb03122abd9362ac7333d13.mp4 b/0219122156_cosmos_mip_spatial/videos/mem_mipnerf360_5f84c46d31489bd1a6b57b7c5c8cbc15280e279f8eb03122abd9362ac7333d13.mp4 new file mode 100644 index 0000000000000000000000000000000000000000..2677bb2d94b116ec9f6048a6d29476fd761aa3fc --- /dev/null +++ b/0219122156_cosmos_mip_spatial/videos/mem_mipnerf360_5f84c46d31489bd1a6b57b7c5c8cbc15280e279f8eb03122abd9362ac7333d13.mp4 @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8bf015f7fdc6ef1fd22f06abed509ff4ef44eace7679bb62a030d22019953ede +size 5075761 diff --git a/0219122156_cosmos_mip_spatial/videos/mem_mipnerf360_a8f137cd302ac277409b1dcd26e4c8a03d8600bc4f6a80af40c71a5a980a98c5.mp4 b/0219122156_cosmos_mip_spatial/videos/mem_mipnerf360_a8f137cd302ac277409b1dcd26e4c8a03d8600bc4f6a80af40c71a5a980a98c5.mp4 new file mode 100644 index 0000000000000000000000000000000000000000..8694c7622a9e4f4ec19605f36d596f6880642a1d --- /dev/null +++ b/0219122156_cosmos_mip_spatial/videos/mem_mipnerf360_a8f137cd302ac277409b1dcd26e4c8a03d8600bc4f6a80af40c71a5a980a98c5.mp4 @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6131b508afe3f0d06c087799020114345ae834e67c9ed7d6fc51eccaa002d717 +size 7298653