{ "exp": "Wan+JEPA", "seed": 1039, "auto_tag_run": false, "eval": false, "debug": false, "timestamp": "\"2606150056\"", "log": { "logging_dir": "logs", "report_to": "wandb", "log_freq": 100 }, "wandb": { "project": "psi", "entity": "keycharon", "group": "finetune", "id": "wtg4c4ql", "name": "Wan+JEPA.wholeb.acvjepa.nullstate.cosine.lr1.0e-04.b256.gpus8.\"2606150056\"", "resume": "allow" }, "train": { "num_workers": 8, "overfit_single_batch": false, "name": "finetune", "resume_from_checkpoint": null, "skip_resumed_steps": false, "hf_token": ".hf_token", "lora": false, "output_dir": "/mnt/data0/runs", "gradient_accumulation_steps": 1, "mixed_precision": "bf16", "max_grad_norm": 1.0, "optimizer_foreach": null, "train_batch_size": 32, "val_batch_size": 16, "val_num_batches": 20, "checkpointing_steps": 100, "max_checkpoints_to_keep": 5, "validation_steps": 0, "learning_rate": 0.0001, "lr_scheduler_type": "cosine", "lr_scheduler_kwargs": { "betas": [ 0.95, 0.999 ], "weight_decay": 1e-6, "eps": 1e-8 }, "scheduler_specific_kwargs": {}, "data_parallel": "deepspeed", "sharding_strategy": "full-shard", "deepspeed_config": "/mnt/data0/Psi0/scripts/deepspeed/zero3.json", "enable_gradient_checkpointing": true, "enable_mixed_precision_training": true, "reduce_in_full_precision": true, "max_training_steps": 100000, "num_train_epochs": null, "warmup_steps": 1000, "warmup_ratio": null }, "data": { "transform": { "repack": { "dataset_name": "wholebody", "stage": "postpre", "action_chunk_size": 25, "pad_action_dim": 66, "pad_state_dim": 45, "num_past_frames": 1, "use_delta_actions": false }, "model": { "resize": { "size": [ 270, 480 ] }, "center_crop": { "size": [ 270, 480 ] }, "jepa_resize": { "size": [ 256, 256 ] }, "jepa_center_crop": { "size": [ 256, 256 ] }, "color_jitter": { "brightness": 0.2, "contrast": [ 0.8, 1.2 ], "saturation": [ 0.8, 1.2 ], "hue": 0.05 }, "gaussian_noise": { "mean": 0.0, "std": 3.0, "prob_skip": 0.1 }, "img_aug": false, "adaptive_resize": false, "img_sizes": { "egodex": [ 270, 480 ], "he": [ 240, 320 ] } }, "field": { "stat_path": "/home/baai/wam_ckpt/stats/wam_stat_no_human.json", "state_stat_path": "/home/baai/wam_ckpt/stats/wam_stat_state_no_human.json", "stat_action_key": "wam_smpl", "stat_state_key": "wam_smpl", "action_norm_type": "std", "normalize_state": true, "state_norm_type": "bounds_q99", "clip_value": null, "eps": 1e-6, "low": null, "high": null } }, "root_dir": "/mnt/data0/posttrain_data_new", "annotation_dir": null, "ego_video_dir": null, "exo_video_dir": null, "require_exo_video": false, "upsample_rate": 1, "chunk_size": 25, "num_past_frames": 0, "img_history_size": 1, "use_delta_actions": false, "data_downsample": 1, "force_rewrite_cache": false }, "model": { "resnet_store_path": null, "pretrained_action_header_path": null, "rtc": true, "max_delay": 8, "action_dim": 66, "action_chunk_size": 25, "action_exec_horizon": 25, "observation_horizon": 1, "img_chunk": 1, "n_cams": 1, "use_obs": "add_token", "dropout": 0.1, "noise_scheduler": "flow", "train_diffusion_steps": 1000, "eval_diffusion_steps": 10, "share_cam_features": false, "early_fusion": false, "odim": 66, "n_conditions": 1, "token_fusion": "concat", "loss_w": [ 0.1, 0.2, 0.1 ], "time_dim": 256, "hidden_dim": 1536, "num_blocks": 6, "dim_feedforward": 2048, "nhead": 24, "activation": "gelu", "view_feature_dim": 1920, "use_film": false, "combined_temb": false, "use_dit": false, "weight_decay": 0.01, "model_name_or_path": "/home/baai/wam_ckpt/omni_cross_38900", "vlm_ckpt_step": null, "tune_vlm": false, "tune_mm_llm": false, "tune_mm_vision": false, "tune_mm_mlp": false, "gradient_checkpointing": true, "lang_backbone_lr": 0.00001, "mm_projector_lr": 0.00001, "vision_tower_lr": 1e-6, "optim": "adamw_torch", "model_max_length": 4096, "data_flatten": true, "data_packing": true, "max_pixels": 451584, "min_pixels": 12544, "posttrain_mode": "ac_vjepa", "ac_vjepa_jepa_loss_weight": 1.0, "ac_vjepa_action_loss_weight": 1.0, "ac_vjepa_abs_action_loss_weight": 0.0, "ac_vjepa_use_action_tokenizer": false, "ac_vjepa_action_tokenizer_type": "fast", "ac_vjepa_action_tokenizer_ckpt": null, "ac_vjepa_action_tokenizer_bins": 2048, "ac_vjepa_action_token_loss_weight": 0.0, "ac_vjepa_action_token_use_absolute": true, "ac_vjepa_hidden_dim": 768, "ac_vjepa_depth": 12, "ac_vjepa_num_heads": 16, "ac_vjepa_latent_dim": 768, "ac_vjepa_state_dim": null, "ac_vjepa_use_state": false, "ac_vjepa_force_null_state": true, "ac_vjepa_state_dropout_prob": 0.1, "ac_vjepa_state_dropout_prob_end": 0.8, "ac_vjepa_state_dropout_anneal_steps": 30000, "ac_vjepa_image_size": 256, "ac_vjepa_patch_size": 16, "ac_vjepa_encoder_hidden_dim": 1408, "ac_vjepa_encoder_depth": 40, "ac_vjepa_encoder_heads": 16, "ac_vjepa_encoder_mlp_ratio": 4.363636363636363, "ac_vjepa_encoder_use_rope": true, "ac_vjepa_encoder_ckpt_key": "target_encoder", "ac_vjepa_predictor_ckpt_key": "predictor", "ac_vjepa_load_predictor_from_ckpt": false, "ac_vjepa_freeze_frame_encoder": true, "ac_vjepa_force_predictor_trainable": true } }