{ "run_id": "dit4dit_streaming", "run_root_dir": "runs", "seed": 42, "trackers": [ "jsonl" ], "is_debug": false, "provenance": { "official_revision": "66a6f3a12e2c8740157c1e478795952c040d31dd", "official_checkpoint_sha256": "10cdc70029d84519942157ff741fe570a78f161be3fda6ff1e87286b4c7faa29", "official_checkpoint_bytes": 21283256413 }, "framework": { "name": "DiT4DiT", "cosmos25": { "base_model": "${oc.env:COSMOS_MODEL_PATH}", "revision": null, "local_files_only": true, "torch_dtype": "bfloat16", "device": "cpu", "trainable": true, "training": "joint", "future_num_inference_steps": 1, "future_loss_type": "flow_matching", "extract_layer": 17, "attn_implementation": "flash_attention_2", "vl_hidden_dim": 2048, "conditional_frame_timestep": 0.0001, "flow_matching": { "time_distribution": "logit_normal", "high_sigma_ratio": 0.05, "high_sigma_min": 0.98 } }, "action_model": { "action_model_type": "DiT-B", "hidden_size": 2560, "add_pos_embed": true, "max_seq_len": 1024, "action_dim": 8, "state_dim": 16, "past_action_window_size": 8, "future_action_window_size": 7, "action_horizon": 8, "repeated_diffusion_steps": 8, "noise_beta_alpha": 1.5, "noise_beta_beta": 1.0, "noise_s": 0.999, "num_timestep_buckets": 1000, "num_inference_timesteps": 4, "num_target_vision_tokens": 32, "streaming": true, "streaming_prefix_steps": 8, "streaming_execution_steps": 4, "streaming_noise_levels": 32, "streaming_warm_start_steps": 4, "streaming_ladder_probability": 0.8, "diffusion_model_cfg": { "cross_attention_dim": 2048, "dropout": 0.2, "final_dropout": true, "interleave_self_attention": true, "norm_type": "ada_norm", "num_layers": 16, "output_dim": 2560, "positional_embeddings": null } } }, "datasets": { "vla_data": { "dataset_py": "lerobot_datasets", "data_root_dir": "unused-for-evaluation", "lerobot_version": "v3.0", "data_mix": "libero_all", "include_state": true, "action_type": "joint_position", "default_image_resolution": [ 3, 224, 224 ], "per_device_batch_size": 16, "load_all_data_for_training": true, "obs": [ "image_0" ], "image_size": [ 224, 224 ], "delete_pause_frame": false, "streaming": true, "video_delta_indices": [ -4, -3, -2, -1, 0, 2, 4, 6, 8 ], "action_delta_indices": [ -8, -7, -6, -5, -4, -3, -2, -1, 0, 1, 2, 3, 4, 5, 6, 7 ], "max_state_dim": 16, "max_action_dim": 8, "action_video_freq_ratio": 1, "video_backend": "torchvision_av" } }, "trainer": { "max_train_steps": 20000, "num_warmup_steps": 1000, "save_interval": 1000, "learning_rate": { "base": 1e-05, "backbone_interface": 1e-05, "action_model": 0.0001 }, "lr_preset": "official", "min_lr": 5e-07, "repeated_diffusion_steps": 4, "gradient_clipping": 1.0, "gradient_accumulation_steps": 1, "num_workers": 4, "logging_frequency": 10, "enable_gradient_checkpointing": true, "cache_frozen_text": true, "gradient_validation_interval": 20, "prefetch_factor": 2, "state_save_interval": 1000, "first_state_save_step": 50, "keep_last_training_states": 2, "keep_last_weights": 3, "landmark_steps": [ 5000, 10000, 20000 ], "mixed_precision": "bf16", "parameter_dtype": "bfloat16", "loss_scale": { "future_video": 1.0 }, "optimizer": { "name": "AdamW", "betas": [ 0.9, 0.95 ], "eps": 1e-08, "weight_decay": 1e-08 }, "pretrained_checkpoint": null }, "output_dir": "runs/streaming" }