{ "auto_tag_run": false, "data": { "decision_labels_path": "/training/EmbodiedDyna-runs/sg2t37-v3varlen-20261002/sam3v2/decision_labels_sg2t37.json", "decision_labels_sha256": "82beae2fc8f1609da06e9605652fc3a22b2adc9804a751770cbcf13e6ec2ba79", "framesamp_dataset_revision": "799b00f3a1e08a0328d54bff13f1b8d8a7489205", "framesamp_feature_manifest_sha256": "1d809e0cbf5d1018551ca01f46ff197ff32f654a9f79179b95c4623190764c23", "framesamp_feature_root": "/training/EmbodiedDyna-psi0-framesamp-data/features-sg2t37v3varlen/qwen-vision-merger-raw1/4efc80fed43af5e3503f7fe3d803fc32a880c6db/SimFactoryShellGame2T-v0", "framesamp_feature_scope": "full", "native_keyframe_manifest_sha256": "67f7023b19d47bea3461aa97ec0fd983cc33acce374798ac8f17ce1f1c42a545", "native_keyframe_root": "/training/keyframe-native7x77-sg2t37v3-20261002", "native_keyframe_selector": "sam3", "object_memory_manifest_sha256": "875398dcb3e9e2e21d7c60689c5584fc0b3825a03fd26c39d434be51e12a2eee", "object_memory_root": "/training/EmbodiedDyna-runs/sg2t37-v3varlen-20261002/sam3v2/objects-flexle7c3", "root_dir": "/training/EmbodiedDyna-psi0-framesamp-data/psi-data-sg2t37v3varlen/799b00f3a1e08a0328d54bff13f1b8d8a7489205/extracted", "symbolic_memory_mode": "disabled", "symbolic_memory_rows_path": null, "symbolic_memory_seed": 292285, "symbolic_use_phase_action_mask": true, "train_repo_ids": [ "SimFactoryShellGame2T-v0" ], "transform": { "field": { "action_max": [ 0.20344078540802002, 0.17177540063858032, 0.5726423859596252, 0.060586556792259216, 0.030544286593794823, 0.03791704773902893, 0.03118515945971012, 1.0442190170288086, 0.3619808554649353, 4.94015648655477e-06, 1.567848563194275, 1.737970232963562, 1.5620214939117432, 1.5995635986328125, 0.026498660445213318, 0.43013057112693787, 0.04635608568787575, 0.06244596838951111, 0.04676114767789841, 0.012783452868461609, 0.41602256894111633, 0.003717988496646285, 0.2647508382797241, 0.5406146049499512, 0.9597413539886475, 1.1283698081970215, 1.4887322187423706, 0.7542539238929749, 0.14426511526107788, 0.111517533659935, 0.0343802385032177, 0.75, 0.25, 0.20000000298023224, 0.0, 0.08828353881835938 ], "action_min": [ -0.2876609265804291, -0.14852772653102875, -0.04031404107809067, -0.2837502062320709, -0.3911275565624237, -0.4982183575630188, -0.4393625855445862, -1.0449014902114868, -0.25999966263771057, -0.9059439301490784, -7.455036393366754e-05, -1.7042308172676712e-05, 0.0, 0.0, -0.05483657866716385, -0.04589128494262695, -0.4559757709503174, 0.0, -0.4417652189731598, -0.12400467693805695, -0.04442235454916954, -1.1999903917312622, -0.7976195812225342, -0.3771253526210785, -0.704856812953949, -0.27635878324508667, 0.0, -1.557715654373169, -0.05882887542247772, -0.16772495210170746, -0.14286065101623535, 0.75, -0.18000000715255737, -0.20000000298023224, 0.0, -0.053018126636743546 ], "action_norm_masks": [ true, true, true, true, true, true, false ], "action_norm_type": "bounds", "normalize_state": true, "pad_action_dim": 36, "pad_state_dim": 36, "stat_action_key": "action", "stat_path": "meta/stats_psi0.json", "stat_state_key": "states", "state_max": [ 0.20360559225082397, 0.1720379739999771, 0.5726733803749084, 0.0044308616779744625, 0.0022593322210013866, 0.004013810306787491, 0.002306972397491336, 1.0320099592208862, 0.7252905964851379, 0.017051663249731064, 1.1951128244400024, 1.696426272392273, 1.0122944116592407, 1.5571244955062866, 0.03242148831486702, 0.42924612760543823, 0.0469878651201725, 0.06706680357456207, 0.04684685170650482, 0.0148002402856946, 0.41609084606170654, 0.010240839794278145, 0.264823317527771, 0.542106568813324, 0.9625779986381531, 1.1291245222091675, 1.4903392791748047, 0.7545167207717896, 0.0, 0.0, 0.0, 0.75, 0.0, 0.0, 0.0, 0.0 ], "state_min": [ -0.2878865599632263, -0.1484350711107254, -0.0029822050128132105, -0.2838318347930908, -0.39114177227020264, -0.4984891712665558, -0.43937596678733826, -1.0394816398620605, -0.051454901695251465, -0.8826939463615417, -0.024508588016033173, -0.019203325733542442, -0.017816338688135147, -0.019350627437233925, -0.05183826759457588, -0.047514408826828, -0.45884960889816284, 0.00024363555712625384, -0.4422004818916321, -0.12305071204900742, -0.04441813379526138, -1.1937716007232666, -0.7942893505096436, -0.37647077441215515, -0.7034225463867188, -0.2765703499317169, -6.341539119603112e-05, -1.5577093362808228, 0.0, -0.15000000596046448, 0.0, 0.75, 0.0, 0.0, 0.0, 0.0 ], "use_norm_mask": false }, "model": { "adaptive_resize": false, "center_crop": { "size": [ 180, 320 ] }, "color_jitter": { "brightness": 0.2, "contrast": [ 0.8, 1.2 ], "hue": 0.05, "saturation": [ 0.8, 1.2 ] }, "gaussian_noise": { "mean": 0.0, "prob_skip": 0.1, "std": 3.0 }, "img_aug": true, "img_sizes": { "egodex": [ 270, 480 ], "he": [ 240, 320 ] }, "resize": { "size": [ 180, 320 ] } }, "repack": { "action_chunk_size": 30, "dataset_name": "simple", "num_past_frames": 0, "pad_action_dim": 36, "pad_state_dim": 36 } }, "val_repo_ids": [ "SimFactoryShellGame2T-v0" ], "video_backend": "decord" }, "debug": false, "eval": false, "exp": "r-50e6d5fadaaa", "log": { "log_freq": 100, "logging_dir": "logs", "report_to": "wandb" }, "model": { "action_chunk_size": 30, "action_dim": 36, "action_exec_horizon": 30, "activation": "gelu", "combined_temb": false, "data_flatten": true, "data_packing": true, "decision_head_enabled": true, "decision_head_inputs": "both", "decision_loss_weight": 0.5, "decision_num_classes": 3, "decision_teacher_forcing": 0.7, "dim_feedforward": 2048, "dropout": 0.1, "early_fusion": false, "eval_diffusion_steps": 10, "framesamp_dataset_revision": "799b00f3a1e08a0328d54bff13f1b8d8a7489205", "framesamp_enabled": true, "framesamp_entry_probe_only": false, "framesamp_feature_dim": 2048, "framesamp_feature_manifest": "/training/EmbodiedDyna-psi0-framesamp-data/features-sg2t37v3varlen/qwen-vision-merger-raw1/4efc80fed43af5e3503f7fe3d803fc32a880c6db/SimFactoryShellGame2T-v0/manifest.json", "framesamp_feature_manifest_sha256": "1d809e0cbf5d1018551ca01f46ff197ff32f654a9f79179b95c4623190764c23", "framesamp_feature_scope": "full", "framesamp_include_current": true, "framesamp_memory_budget": 539, "framesamp_memory_clock": "execution_frame", "framesamp_memory_head_dim": 256, "framesamp_memory_heads": 6, "framesamp_memory_kv_heads": 1, "framesamp_modulation_init": "robomme_normal_std_0_002", "framesamp_off_clock_mapping": "all_raw_frames_le_current", "framesamp_online_memory_view": "rgb_head_stereo_left", "framesamp_padding_side": "right", "framesamp_pool_type": "none", "framesamp_position_dim": 768, "framesamp_query_stride": 1, "framesamp_temporal_position_source": "raw_frame_index", "framesamp_tokens_per_frame": 77, "framesamp_training_origin": "common_psi0_warm_start_step0", "gradient_checkpointing": true, "hidden_dim": 1536, "img_chunk": 1, "lang_backbone_lr": 1e-05, "loss_w": [ 0.1, 0.2, 0.1 ], "max_delay": 8, "max_pixels": 451584, "max_world_tokens": 256, "min_pixels": 12544, "mm_projector_lr": 1e-05, "model_max_length": 4096, "model_name_or_path": "/training/EmbodiedDyna/.psi/cache/checkpoints/psi0/pre.fast.1by1.2601091803.ckpt.ego200k.he30k", "n_cams": 1, "n_conditions": 0, "native_keyframe_contract_version": 1, "native_keyframe_enabled": true, "native_keyframe_grid_hw": [ 7, 11 ], "native_keyframe_manifest_sha256": "67f7023b19d47bea3461aa97ec0fd983cc33acce374798ac8f17ce1f1c42a545", "native_keyframe_root": "/training/keyframe-native7x77-sg2t37v3-20261002", "native_keyframe_selector": "sam3", "native_keyframe_spatial_coordinates": "cell_centers_normalized_to_16", "nhead": 24, "noise_scheduler": "flow", "num_blocks": 6, "object_memory_contract_version": 1, "object_memory_enabled": true, "object_memory_gate_init": "zero_out_proj", "object_memory_geometry_dim": 8, "object_memory_manifest_sha256": "875398dcb3e9e2e21d7c60689c5584fc0b3825a03fd26c39d434be51e12a2eee", "object_memory_root": "/training/EmbodiedDyna-runs/sg2t37-v3varlen-20261002/sam3v2/objects-flexle7c3", "object_memory_slots": 2, "observation_horizon": 1, "odim": 36, "optim": "adamw_torch", "pretrained_action_header_path": "/training/EmbodiedDyna/.psi/cache/checkpoints/psi0/postpre.1by1.pad36.2601131206.ckpt.he30k", "resnet_store_path": null, "rtc": true, "share_cam_features": false, "time_dim": 256, "token_fusion": "concat", "train_diffusion_steps": 1000, "tune_mm_llm": false, "tune_mm_mlp": false, "tune_mm_vision": false, "tune_vlm": false, "use_dit": false, "use_film": false, "use_obs": "add_token", "video_height": 160, "video_latent_channels": 48, "video_latent_frames": 3, "video_num_frames": 9, "video_vae_filename": "Wan2.2_VAE.pth", "video_vae_path": null, "video_vae_repo_id": "Wan-AI/Wan2.2-TI2V-5B", "video_vae_tiled": false, "video_width": 320, "view_feature_dim": 2048, "vision_tower_lr": 1e-06, "vlm_ckpt_step": null, "warm_start_checkpoint": "/training/EmbodiedDyna-runs/sg2t37-v3varlen-20261002/sam3v2/training-40k-dectf-bind/runs/finetune_keyframe_native_object/r-985fae5fc791/checkpoints/ckpt_40000", "weight_decay": 0.01, "world_learning_rate": null, "world_loss_weight": 1.0, "world_patch_size": 2 }, "seed": 292285, "timestamp": "50e6d5fadaaa", "train": { "checkpointing_steps": 2500, "data_parallel": "ddp", "deepspeed_config": "/training/EmbodiedDyna-litlamp-fs-20260926/src/InternVLA/config/deepseeds/zero3.json", "enable_gradient_checkpointing": true, "enable_mixed_precision_training": true, "gradient_accumulation_steps": 2, "hf_token": ".hf_token", "learning_rate": 5e-05, "lora": false, "lr_scheduler_kwargs": { "betas": [ 0.95, 0.999 ], "eps": 1e-08, "weight_decay": 1e-06 }, "lr_scheduler_type": "cosine", "max_checkpoints_to_keep": 4, "max_grad_norm": 1.0, "max_training_steps": 10000, "mixed_precision": "bf16", "name": "finetune_keyframe_native_object", "num_train_epochs": null, "num_workers": 8, "optimizer_foreach": null, "output_dir": "/training/EmbodiedDyna-runs/sg2t37-v3varlen-20261002/sam3v2/training-10k-flexkf/runs", "overfit_single_batch": false, "reduce_in_full_precision": true, "resume_from_checkpoint": null, "scheduler_specific_kwargs": {}, "sharding_strategy": "full-shard", "skip_resumed_steps": false, "train_batch_size": 8, "val_batch_size": 8, "val_num_batches": 20, "validation_steps": 2000, "warmup_ratio": null, "warmup_steps": 200 }, "wandb": { "entity": "wholebodymem", "group": "r-522a7df20da1", "id": "r-50e6d5fadaaa", "name": "r-50e6d5fadaaa", "project": "wholebodymem", "resume": "never", "tags": [ "Psi0", "native-keyframe", "7x77", "sam3", "sam3-object-memory", "gated-objects", "gate-zero_out_proj", "10k", "ShellGame2tV3cf_flexkf10k", "shellgame2tv3cf-flexkf10k", "full-dataset", "training-set-diagnostic", "cpu-bf16-bank", "decision-head-both", "decision-teacher-forcing", "warm-start" ] } }