| { | |
| "common": { | |
| "action_dim": 14, | |
| "state_dim": 14, | |
| "num_video_frames": 8, | |
| "video_height": 384, | |
| "video_width": 320, | |
| "global_downsample_rate": 3, | |
| "video_action_freq_ratio": 2, | |
| "action_chunk_size": 16 | |
| }, | |
| "action_expert": { | |
| "hidden_size": 1024, | |
| "ffn_dim_multiplier": 4, | |
| "norm_eps": 1e-05 | |
| }, | |
| "und_expert": { | |
| "hidden_size": 512, | |
| "ffn_dim_multiplier": 4, | |
| "norm_eps": 1e-05, | |
| "vlm": { | |
| "input_dim": 2048, | |
| "projector_type": "mlp3x_silu" | |
| } | |
| }, | |
| "time_distribution": { | |
| "timestep_sample_method": "logit_normal", | |
| "sigmoid_scale": 1.0, | |
| "min_t": 0.0, | |
| "max_t": 1.0 | |
| }, | |
| "ema": { | |
| "enabled": false, | |
| "update_after_step": 0, | |
| "inv_gamma": 1.0, | |
| "power": 0.75, | |
| "min_value": 0.0, | |
| "max_value": 0.9999 | |
| }, | |
| "geometry_jepa": { | |
| "enable": true, | |
| "lambda_geo": 0.01, | |
| "cache_dir": "/mnt/workspace/Tao/cache/vggt_teacher_robotwin_joint_future_slot_v1", | |
| "teacher_protocol_id": "vggt_tlayout_3view_percam_q32_d768_centered_joint_future_slot_v1", | |
| "horizon_choices": [ | |
| 1, | |
| 2, | |
| 4, | |
| 8 | |
| ], | |
| "max_horizon": 8, | |
| "num_queries": 96, | |
| "hidden_dim": 768, | |
| "adapter_num_heads": 8, | |
| "adapter_depth": 1, | |
| "adapter_dropout": 0.0, | |
| "predictor_depth": 2, | |
| "predictor_num_heads": 8, | |
| "predictor_dropout": 0.0, | |
| "action_hidden_dim": 768, | |
| "target_mode": "joint_future_slot", | |
| "alpha_cos": 1.0, | |
| "beta_mse": 0.05, | |
| "gamma_token_cos": 1.0, | |
| "use_pool_cosine": true, | |
| "use_token_mse": true, | |
| "detach_target": true, | |
| "freeze_wam_for_geo": false, | |
| "lr_scale": 1.0, | |
| "log_grad_norm": true | |
| } | |
| } |