| { |
| "beta_end": 0.02, |
| "beta_schedule": "squaredcos_cap_v2", |
| "beta_start": 0.0001, |
| "chunk_size": 100, |
| "clip_sample": true, |
| "clip_sample_range": 1.0, |
| "diffusion_step_embed_dim": 128, |
| "dim_feedforward": 3200, |
| "dim_model": 512, |
| "do_mask_loss_for_padding": true, |
| "down_dims": [ |
| 256, |
| 512 |
| ], |
| "dropout": 0.1, |
| "feedforward_activation": "relu", |
| "input_normalization_modes": { |
| "observation.images.cam_high": "mean_std", |
| "observation.images.cam_left_wrist": "mean_std", |
| "observation.state": "mean_std" |
| }, |
| "input_shapes": { |
| "observation.images.cam_high": [ |
| 3, |
| 480, |
| 640 |
| ], |
| "observation.images.cam_left_wrist": [ |
| 3, |
| 480, |
| 640 |
| ], |
| "observation.state": [ |
| 9 |
| ] |
| }, |
| "kernel_size": 5, |
| "n_action_steps": 100, |
| "n_encoder_layers": 4, |
| "n_groups": 8, |
| "n_heads": 8, |
| "n_obs_steps": 1, |
| "noise_scheduler_type": "DDPM", |
| "num_inference_steps": 10, |
| "num_train_timesteps": 100, |
| "output_normalization_modes": { |
| "action": "mean_std" |
| }, |
| "output_shapes": { |
| "action": [ |
| 9 |
| ] |
| }, |
| "pre_norm": false, |
| "prediction_type": "epsilon", |
| "pretrained_backbone_weights": "ResNet18_Weights.IMAGENET1K_V1", |
| "replace_final_stride_with_dilation": false, |
| "use_film_scale_modulation": true, |
| "vision_backbone": "resnet18" |
| } |
