Upload config.json with huggingface_hub
Browse files- config.json +88 -0
config.json
ADDED
|
@@ -0,0 +1,88 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"type": "smolvla",
|
| 3 |
+
"n_obs_steps": 1,
|
| 4 |
+
"normalization_mapping": {
|
| 5 |
+
"VISUAL": "IDENTITY",
|
| 6 |
+
"STATE": "MEAN_STD",
|
| 7 |
+
"ACTION": "MEAN_STD"
|
| 8 |
+
},
|
| 9 |
+
"input_features": {
|
| 10 |
+
"observation.images.wrist_image": {
|
| 11 |
+
"type": "VISUAL",
|
| 12 |
+
"shape": [
|
| 13 |
+
256,
|
| 14 |
+
256,
|
| 15 |
+
3
|
| 16 |
+
]
|
| 17 |
+
},
|
| 18 |
+
"observation.images.image": {
|
| 19 |
+
"type": "VISUAL",
|
| 20 |
+
"shape": [
|
| 21 |
+
256,
|
| 22 |
+
256,
|
| 23 |
+
3
|
| 24 |
+
]
|
| 25 |
+
},
|
| 26 |
+
"observation.state": {
|
| 27 |
+
"type": "STATE",
|
| 28 |
+
"shape": [
|
| 29 |
+
8
|
| 30 |
+
]
|
| 31 |
+
}
|
| 32 |
+
},
|
| 33 |
+
"output_features": {
|
| 34 |
+
"action": {
|
| 35 |
+
"type": "ACTION",
|
| 36 |
+
"shape": [
|
| 37 |
+
7
|
| 38 |
+
]
|
| 39 |
+
}
|
| 40 |
+
},
|
| 41 |
+
"device": "cuda",
|
| 42 |
+
"use_amp": false,
|
| 43 |
+
"push_to_hub": false,
|
| 44 |
+
"repo_id": null,
|
| 45 |
+
"private": null,
|
| 46 |
+
"tags": null,
|
| 47 |
+
"license": null,
|
| 48 |
+
"chunk_size": 50,
|
| 49 |
+
"n_action_steps": 50,
|
| 50 |
+
"max_state_dim": 32,
|
| 51 |
+
"max_action_dim": 32,
|
| 52 |
+
"resize_imgs_with_padding": [
|
| 53 |
+
512,
|
| 54 |
+
512
|
| 55 |
+
],
|
| 56 |
+
"empty_cameras": 0,
|
| 57 |
+
"adapt_to_pi_aloha": false,
|
| 58 |
+
"use_delta_joint_actions_aloha": false,
|
| 59 |
+
"tokenizer_max_length": 48,
|
| 60 |
+
"num_steps": 10,
|
| 61 |
+
"use_cache": true,
|
| 62 |
+
"freeze_vision_encoder": true,
|
| 63 |
+
"train_expert_only": true,
|
| 64 |
+
"train_state_proj": true,
|
| 65 |
+
"optimizer_lr": 0.0001,
|
| 66 |
+
"optimizer_betas": [
|
| 67 |
+
0.9,
|
| 68 |
+
0.95
|
| 69 |
+
],
|
| 70 |
+
"optimizer_eps": 1e-08,
|
| 71 |
+
"optimizer_weight_decay": 1e-10,
|
| 72 |
+
"optimizer_grad_clip_norm": 10,
|
| 73 |
+
"scheduler_warmup_steps": 1000,
|
| 74 |
+
"scheduler_decay_steps": 30000,
|
| 75 |
+
"scheduler_decay_lr": 2.5e-06,
|
| 76 |
+
"vlm_model_name": "HuggingFaceTB/SmolVLM2-500M-Video-Instruct",
|
| 77 |
+
"load_vlm_weights": true,
|
| 78 |
+
"add_image_special_tokens": false,
|
| 79 |
+
"attention_mode": "cross_attn",
|
| 80 |
+
"prefix_length": -1,
|
| 81 |
+
"pad_language_to": "longest",
|
| 82 |
+
"num_expert_layers": -1,
|
| 83 |
+
"num_vlm_layers": 16,
|
| 84 |
+
"self_attn_every_n_layers": 2,
|
| 85 |
+
"expert_width_multiplier": 0.75,
|
| 86 |
+
"min_period": 0.004,
|
| 87 |
+
"max_period": 4.0
|
| 88 |
+
}
|