File size: 9,112 Bytes
e03488e | 1 2 | {"global_step": 2895, "inference_config": {"schema_version": 1, "values": {"model.model_id": "Wan-AI/Wan2.2-TI2V-5B", "model.tokenizer_model_id": "Wan-AI/Wan2.2-TI2V-5B", "model.tokenizer_max_len": 128, "model.redirect_common_files": true, "model.arch": "rgb_dep_sem", "model.action_conditioning": "predicted_depth_head_semantic", "model.attention_layout": "rgb_auxiliary_causal", "model.depth_projection": "independent", "model.semantic_projection": "independent", "model.input_paradigm": "t_concat", "model.view_layout": "robotwin_t", "model.proprio_dim": 20, "model.video_dit_config": {"in_dim": 48, "out_dim": 48, "text_dim": 4096, "freq_dim": 256, "patch_size": [1, 2, 2], "num_layers": 30, "has_image_input": false, "has_image_pos_emb": false, "has_ref_conv": false, "add_control_adapter": false, "in_dim_control_adapter": 24, "seperated_timestep": true, "require_vae_embedding": false, "require_clip_embedding": false, "fuse_vae_embedding_in_latents": true, "action_conditioned": false, "action_dim": 20, "action_group_causal_mask_mode": "group_diagonal", "video_attention_mask_mode": "first_frame_causal", "moe": {"enabled": true, "routing": "per_sample", "num_experts": 4, "top_k": 0, "use_shared_expert": true, "moe_layers": [21, 24, 27], "router_cond": "modality", "token_modality_bias": true, "token_dispatch": "dense_masked", "loss_lambda_aux": 0.01, "loss_lambda_entropy": 0.0, "aux_warmup_steps": 144, "aux_warmup_ratio": 0.05, "aux_warmup_unit": "optimizer_step", "init_noise_std": 0.02}, "attn_head_dim": 128, "eps": 1e-06, "ffn_dim": 14336, "hidden_dim": 3072, "num_heads": 24}, "model.action_dit_config": {"action_dim": 20, "text_dim": 4096, "freq_dim": 256, "num_layers": 30, "moe": null, "attn_head_dim": 128, "eps": 1e-06, "ffn_dim": 4096, "hidden_dim": 1024, "num_heads": 24}, "model.video_scheduler": {"train_shift": 5.0, "infer_shift": 5.0, "num_train_timesteps": 1000}, "model.action_scheduler": {"train_shift": 5.0, "infer_shift": 5.0, "num_train_timesteps": 1000}, "model.wrist": {"view_size": [128, 160], "frame": "first", "attends_video": false, "rope_offset": null, "cameras": ["left", "right"]}, "model.loss": {"lambda_video": 1.0, "lambda_action": 1.0, "lambda_semantic": 1.0, "lambda_depth": 1.0}, "model.semantic.compression": {"type": "svae", "latent_dim": 48, "output_normalization": "layernorm", "model_config": {"input_dim": 768, "latent_dim": 48, "num_heads": 16, "num_layers": 3, "intermediate_size": 2048, "dropout": 0.0}, "artifact_identity": {"schema_version": 1, "encoder": {"backbone": "dinov3_vitb16", "feat_dim": 768, "weights_sha256": "73cec8be7427c8655ceced13ce62f6e20a1fa90d1b4d4a550df17a1144081a7c", "source_sha256": "c5bb90cf2c71dd5700762b97da6a6d58404fdbc2d1303e7b70b40b43b13951fb", "wrapper_sha256": "3d83792f6472f88feff63b15eb69bdcb43532ee2e277b29c064476c229369c93", "pool_sha256": "d9ee695c8c9b48e3afc6f76e6a4e4f42880dfcee3d57c73fc991b339c35c40f1", "layout_sha256": "399ec692097a95e559406ddaebd5e5ca4bd39693ad915047e527655c23140204", "feature_space": "semwam_dino_normalized_patch_features_post_temporal_pool"}, "features": {"geometry": {"num_frames": 33, "stride": 1, "action_video_freq_ratio": 4, "layout": "robotwin_t", "input_paradigm": "t_concat", "video_size": [384, 320], "wrist_view_size": [128, 160], "wrist_frame": "first"}}, "fingerprint": "d3383726600435ddc50852e7e12db67c393b1a2f54568803214753980bf3597d"}, "state_sha256": "e8760f44c242bb79232f2d3740dda71956cac2001ba8c00ceaec77ea80e36961"}, "model.semantic.backbone": "dinov3_vitb16", "model.semantic.feat_dim": 768, "model.semantic.fusion_layers": [7, 23], "model.semantic.temporal_pool": "mean", "model.semantic.parametrization": "v_pred", "model.semantic.velocity_loss_form": "direct_error", "model.semantic.loss_diagnostics": true, "model.semantic.loss_lambda_semantic": 1.0, "model.depth.condition_source": "estimated", "model.depth.target_source": "dataset", "model.depth.condition_temporal_context": "current_frame", "model.depth.repr": "dvae", "model.depth.estimator": "vda", "model.depth.intrinsics": {"source": "auto", "sim": {"high": {"fovy_deg": 37.0, "ref_h": 480, "ref_w": 640}, "left": {"fovy_deg": 37.0, "ref_h": 480, "ref_w": 640}, "right": {"fovy_deg": 37.0, "ref_h": 480, "ref_w": 640}, "image": {"fovy_deg": 45.0, "ref_h": 224, "ref_w": 224}, "wrist_image": {"fovy_deg": 75.0, "ref_h": 224, "ref_w": 224}}}, "model.depth.pointmap": {"norm_bounds": {"x": [-0.6, 0.6], "y": [-0.6, 0.6], "z": [0.0, 1.5]}, "max_depth_m": 2.0}, "model.depth.colormap": {"name": "turbo", "near_m": 0.1, "far_m": 2.0}, "model.depth.vda": {"backbone": "vitb", "input_size": 518, "fp32": false, "expected_sha256": "f6f58576b9680a112f2428d4f39aff92656d3ae85745b0164675ace8d5b1fade"}, "model.depth.unidepth": {"version": "v2", "backbone": "vitb14", "resolution_level": null}, "model.depth.temporal_pool": "mean", "model.depth.loss_lambda_depth": 1.0, "model.depth.vae": {"expected_sha256": "c869ccb52c79b398e0f173f3054cfa229f620ef108ceeec82ad0670ca8bb1957", "identity": {"type": "prism_wan22_depth", "weights_sha256": "c869ccb52c79b398e0f173f3054cfa229f620ef108ceeec82ad0670ca8bb1957", "latent_channels": 48, "spatial_factor": 16, "temporal_factor": 4, "input_channels": 1, "latent_mean": [0.4887385368347168, -2.0717055797576904, -0.09361350536346436, -1.9305212497711182, 0.6804881691932678, -0.18310534954071045, -0.017370181158185005, -0.05728627368807793, -0.7831002473831177, -1.8836477994918823, -0.2327776700258255, 1.9035433530807495, 0.46046552062034607, -0.6453059911727905, -1.4464999437332153, 0.21215873956680298, 1.3156590461730957, 0.007310659624636173, 0.745159924030304, -0.7342119216918945, -0.3538833260536194, 2.9265501499176025, 1.9496617317199707, 0.3119705617427826, -1.0393187999725342, -0.45994818210601807, -0.13466505706310272, -1.2324262857437134, 0.3329582214355469, -0.20562906563282013, 1.6063544750213623, -0.37199077010154724, 0.10520733147859573, -0.005950210150331259, -1.021066427230835, -0.16209781169891357, -0.8151541948318481, 2.7302606105804443, -0.5384726524353027, -3.4191009998321533, -0.4929030239582062, 2.472595691680908, 0.09065636992454529, 0.1566804200410843, 0.6077839732170105, 0.16436097025871277, 0.574763834476471, -1.018074631690979], "latent_std": [0.8759883046150208, 4.3842878341674805, 0.600670576095581, 4.464785575866699, 1.4707190990447998, 0.7163799405097961, 0.39080148935317993, 0.521407425403595, 2.459089994430542, 4.432198524475098, 0.9659329056739807, 4.131554126739502, 1.7189385890960693, 1.1850491762161255, 3.289055347442627, 0.6653669476509094, 2.821338176727295, 0.44802847504615784, 1.4034730195999146, 1.3087146282196045, 0.9007586240768433, 6.145054340362549, 3.761077404022217, 1.2898753881454468, 2.543177366256714, 1.234912633895874, 0.765889585018158, 2.0544753074645996, 0.9371235966682434, 0.6772623062133789, 3.1669251918792725, 0.5857013463973999, 0.5159927606582642, 0.5009284615516663, 2.351271152496338, 0.6470572352409363, 1.8693331480026245, 5.487253189086914, 1.4262751340866089, 7.51733922958374, 0.8003208637237549, 4.9757843017578125, 0.6112902760505676, 0.5305010676383972, 1.6170129776000977, 0.4477066993713379, 1.232177972793579, 2.2924752235412598], "preprocess": {"unit": "metres", "near_m": 0.0, "far_m": 2.5, "resize": "bilinear", "align_corners": false, "layout": "robotwin_t", "normalization": "linear_minus_one_one", "posterior": "mean"}}, "preprocess": {"unit": "metres", "near_m": 0.0, "far_m": 2.5, "resize": "bilinear", "align_corners": false, "layout": "robotwin_t", "normalization": "linear_minus_one_one", "posterior": "mean"}}, "model.depth.defm.backbone": "defm_vit_s14", "model.depth.defm.feat_dim": 384, "model.depth.defm.patch_size": 14, "data.train.num_frames": 33, "data.train.action_video_freq_ratio": 4, "data.normalization_statistics_sha256": "93a1d2c296037547209132c79d5966c44e80d147db2359ca952574bd5795bef1", "data.train.processor.shape_meta": {"images": [{"key": "cam_high", "raw_shape": [3, 240, 320], "shape": [3, 240, 320]}, {"key": "cam_left_wrist", "raw_shape": [3, 240, 320], "shape": [3, 240, 320]}, {"key": "cam_right_wrist", "raw_shape": [3, 240, 320], "shape": [3, 240, 320]}], "action": [{"key": "default", "raw_shape": 16, "shape": 20}], "state": [{"key": "default", "raw_shape": 16, "shape": 20}]}, "data.train.processor.num_obs_steps": 33, "data.train.processor.num_output_cameras": 3, "data.train.processor.action_output_dim": 20, "data.train.processor.proprio_output_dim": 20, "data.train.processor.action_state_transforms": [{"_target_": "wam.datasets.xtrainer2.Xtrainer2RelativePose"}], "data.train.processor.use_stepwise_action_norm": false, "data.train.processor.norm_default_mode": "z-score", "data.train.processor.action_state_merger": {"_target_": "wam.datasets.lerobot.transforms.action_state_merger.ConcatLeftAlign"}, "data.train.processor.train_transforms": [{"_target_": "wam.datasets.lerobot.transforms.image.ToTensor"}], "data.train.processor.val_transforms": [{"_target_": "wam.datasets.lerobot.transforms.image.ToTensor"}], "data.train.processor.norm_clip_enabled": false, "model.model_family": "wam", "model.depth.fusion_layers": null}}}
|