NeilKim/diffusion-so100-policy-2
012
1{2 "type": "diffusion",3 "n_obs_steps": 2,4 "normalization_mapping": {5 "VISUAL": "MEAN_STD",6 "STATE": "MIN_MAX",7 "ACTION": "MIN_MAX"8 },9 "input_features": {10 "observation.state": {11 "type": "STATE",12 "shape": [13 614 ]15 },16 "observation.images.hand": {17 "type": "VISUAL",18 "shape": [19 3,20 480,21 64022 ]23 },24 "observation.images.top": {25 "type": "VISUAL",26 "shape": [27 3,28 480,29 64030 ]31 }32 },33 "output_features": {34 "action": {35 "type": "ACTION",36 "shape": [37 638 ]39 }40 },41 "device": "cuda",42 "use_amp": false,43 "push_to_hub": true,44 "repo_id": "NeilKim/diffusion-so100-policy-2",45 "private": null,46 "tags": null,47 "license": null,48 "horizon": 16,49 "n_action_steps": 8,50 "drop_n_last_frames": 7,51 "vision_backbone": "resnet18",52 "crop_shape": [53 84,54 8455 ],56 "crop_is_random": true,57 "pretrained_backbone_weights": null,58 "use_group_norm": true,59 "spatial_softmax_num_keypoints": 32,60 "use_separate_rgb_encoder_per_camera": false,61 "down_dims": [62 512,63 1024,64 204865 ],66 "kernel_size": 5,67 "n_groups": 8,68 "diffusion_step_embed_dim": 128,69 "use_film_scale_modulation": true,70 "noise_scheduler_type": "DDPM",71 "num_train_timesteps": 100,72 "beta_schedule": "squaredcos_cap_v2",73 "beta_start": 0.0001,74 "beta_end": 0.02,75 "prediction_type": "epsilon",76 "clip_sample": true,77 "clip_sample_range": 1.0,78 "num_inference_steps": null,79 "do_mask_loss_for_padding": false,80 "optimizer_lr": 0.0001,81 "optimizer_betas": [82 0.95,83 0.99984 ],85 "optimizer_eps": 1e-08,86 "optimizer_weight_decay": 1e-06,87 "scheduler_name": "cosine",88 "scheduler_warmup_steps": 50089}