danielsanjosepro/dit_meanflow_stack_multi_v1

1{2    "type": "dit_meanflow",3    "n_obs_steps": 2,4    "normalization_mapping": {5        "VISUAL": "MEAN_STD",6        "STATE": "MIN_MAX",7        "ACTION": "MEAN_STD"8    },9    "input_features": {10        "observation.images.primary": {11            "type": "VISUAL",12            "shape": [13                3,14                256,15                25616            ]17        },18        "observation.images.wrist": {19            "type": "VISUAL",20            "shape": [21                3,22                256,23                25624            ]25        },26        "observation.state.cartesian": {27            "type": "STATE",28            "shape": [29                630            ]31        },32        "observation.state.gripper": {33            "type": "STATE",34            "shape": [35                136            ]37        },38        "observation.state.joints": {39            "type": "STATE",40            "shape": [41                742            ]43        },44        "observation.state.sensors_ft_sensor": {45            "type": "STATE",46            "shape": [47                648            ]49        },50        "observation.state.target": {51            "type": "STATE",52            "shape": [53                654            ]55        },56        "observation.state": {57            "type": "STATE",58            "shape": [59                2660            ]61        }62    },63    "output_features": {64        "action": {65            "type": "ACTION",66            "shape": [67                768            ]69        }70    },71    "device": "cuda",72    "use_amp": false,73    "push_to_hub": true,74    "repo_id": "danielsanjosepro/dit_meanflow_stack_multi_v1",75    "private": null,76    "tags": null,77    "license": null,78    "horizon": 16,79    "n_action_steps": 8,80    "use_proprioceptive": true,81    "drop_n_last_frames": 7,82    "vision_backbone": "resnet18",83    "crop_shape": null,84    "crop_is_random": false,85    "pretrained_backbone_weights": null,86    "use_group_norm": true,87    "spatial_softmax_num_keypoints": 32,88    "use_separate_rgb_encoder_per_camera": true,89    "frequency_embedding_dim": 256,90    "hidden_dim": 512,91    "num_blocks": 6,92    "num_heads": 16,93    "dropout": 0.1,94    "dim_feedforward": 4096,95    "activation": "gelu",96    "training_noise_sampling": "uniform",97    "clip_sample": true,98    "clip_sample_range": 1.0,99    "num_inference_steps": 100,100    "do_mask_loss_for_padding": false,101    "optimizer_lr": 0.0001,102    "optimizer_betas": [103        0.95,104        0.999105    ],106    "optimizer_eps": 1e-08,107    "optimizer_weight_decay": 1e-06,108    "scheduler_name": "cosine",109    "scheduler_warmup_steps": 500,110    "use_meanflow": true,111    "do_multi_step_sampling": true,112    "inference_timesteps": 100,113    "time_distribution": "uniform",114    "log_norm_mu": -0.4,115    "log_norm_sigma": 1.0,116    "use_autograd_functional_jvp": false,117    "use_adaptative_loss": true,118    "flow_ratio": 0.5,119    "cfg_prob": 0.1,120    "cfg_omega": 2.0,121    "cfg_w": 2.0,122    "cfg_kappa": 0.0,123    "optimizer_grad_clip_norm": 0.5124}