naavox/multitask-dit-2

1{2    "type": "multi_task_dit",3    "n_obs_steps": 2,4    "input_features": {5        "observation.state": {6            "type": "STATE",7            "shape": [8                179            ]10        },11        "observation.images.gripper_camera": {12            "type": "VISUAL",13            "shape": [14                3,15                384,16                38417            ]18        }19    },20    "output_features": {21        "action": {22            "type": "ACTION",23            "shape": [24                525            ]26        }27    },28    "device": "cuda",29    "use_amp": false,30    "use_peft": false,31    "push_to_hub": true,32    "repo_id": "naavox/multitask-dit-2",33    "private": null,34    "tags": null,35    "license": null,36    "pretrained_path": null,37    "horizon": 64,38    "n_action_steps": 32,39    "objective": "flow_matching",40    "noise_scheduler_type": "DDPM",41    "num_train_timesteps": 100,42    "beta_schedule": "squaredcos_cap_v2",43    "beta_start": 0.0001,44    "beta_end": 0.02,45    "prediction_type": "epsilon",46    "clip_sample": true,47    "clip_sample_range": 1.0,48    "num_inference_steps": null,49    "sigma_min": 0.0,50    "num_integration_steps": 100,51    "integration_method": "euler",52    "timestep_sampling_strategy": "beta",53    "timestep_sampling_s": 0.999,54    "timestep_sampling_alpha": 1.5,55    "timestep_sampling_beta": 1.0,56    "hidden_dim": 512,57    "num_layers": 6,58    "num_heads": 8,59    "dropout": 0.1,60    "use_positional_encoding": false,61    "timestep_embed_dim": 256,62    "use_rope": true,63    "rope_base": 10000.0,64    "vision_encoder_name": "openai/clip-vit-base-patch16",65    "use_separate_rgb_encoder_per_camera": false,66    "vision_encoder_lr_multiplier": 0.1,67    "image_resize_shape": [68        256,69        25670    ],71    "image_crop_shape": [72        224,73        22474    ],75    "image_crop_is_random": true,76    "text_encoder_name": "openai/clip-vit-base-patch16",77    "tokenizer_max_length": 77,78    "tokenizer_padding": "max_length",79    "tokenizer_padding_side": "right",80    "tokenizer_truncation": true,81    "normalization_mapping": {82        "VISUAL": "MEAN_STD",83        "STATE": "MIN_MAX",84        "ACTION": "MIN_MAX"85    },86    "optimizer_lr": 2e-05,87    "optimizer_betas": [88        0.95,89        0.99990    ],91    "optimizer_eps": 1e-08,92    "optimizer_weight_decay": 0.0,93    "scheduler_name": "cosine",94    "scheduler_warmup_steps": 0,95    "do_mask_loss_for_padding": false,96    "drop_n_last_frames": 3197}