add eo1phys_none_img6default_b8_s46600/checkpoints/011650/pretrained_model/config.json
Browse files
eo1phys_none_img6default_b8_s46600/checkpoints/011650/pretrained_model/config.json
ADDED
|
@@ -0,0 +1,273 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"type": "eo1_phys",
|
| 3 |
+
"n_obs_steps": 1,
|
| 4 |
+
"input_features": {
|
| 5 |
+
"observation.images.third": {
|
| 6 |
+
"type": "VISUAL",
|
| 7 |
+
"shape": [
|
| 8 |
+
3,
|
| 9 |
+
360,
|
| 10 |
+
480
|
| 11 |
+
]
|
| 12 |
+
},
|
| 13 |
+
"observation.images.wrist": {
|
| 14 |
+
"type": "VISUAL",
|
| 15 |
+
"shape": [
|
| 16 |
+
3,
|
| 17 |
+
360,
|
| 18 |
+
480
|
| 19 |
+
]
|
| 20 |
+
},
|
| 21 |
+
"observation.state": {
|
| 22 |
+
"type": "STATE",
|
| 23 |
+
"shape": [
|
| 24 |
+
8
|
| 25 |
+
]
|
| 26 |
+
}
|
| 27 |
+
},
|
| 28 |
+
"output_features": {
|
| 29 |
+
"action": {
|
| 30 |
+
"type": "ACTION",
|
| 31 |
+
"shape": [
|
| 32 |
+
80
|
| 33 |
+
]
|
| 34 |
+
}
|
| 35 |
+
},
|
| 36 |
+
"device": "cuda",
|
| 37 |
+
"dtype": "bfloat16",
|
| 38 |
+
"use_amp": false,
|
| 39 |
+
"use_peft": false,
|
| 40 |
+
"push_to_hub": false,
|
| 41 |
+
"repo_id": null,
|
| 42 |
+
"private": null,
|
| 43 |
+
"tags": null,
|
| 44 |
+
"license": null,
|
| 45 |
+
"pretrained_path": "lerobot/eo1-base",
|
| 46 |
+
"pretrained_revision": null,
|
| 47 |
+
"vlm_base": "lerobot/eo1-base",
|
| 48 |
+
"vlm_config": {
|
| 49 |
+
"_name_or_path": "",
|
| 50 |
+
"action_pass_id": 151672,
|
| 51 |
+
"action_token_id": 151666,
|
| 52 |
+
"architectures": [
|
| 53 |
+
"Qwen2_5_VLForConditionalGeneration"
|
| 54 |
+
],
|
| 55 |
+
"chunk_size_feed_forward": 0,
|
| 56 |
+
"dtype": "bfloat16",
|
| 57 |
+
"eos_token_id": 151645,
|
| 58 |
+
"id2label": {
|
| 59 |
+
"0": "LABEL_0",
|
| 60 |
+
"1": "LABEL_1"
|
| 61 |
+
},
|
| 62 |
+
"image_token_id": 151655,
|
| 63 |
+
"is_encoder_decoder": false,
|
| 64 |
+
"label2id": {
|
| 65 |
+
"LABEL_0": 0,
|
| 66 |
+
"LABEL_1": 1
|
| 67 |
+
},
|
| 68 |
+
"model_type": "qwen2_5_vl",
|
| 69 |
+
"output_attentions": false,
|
| 70 |
+
"output_hidden_states": false,
|
| 71 |
+
"pad_token_id": 151643,
|
| 72 |
+
"problem_type": null,
|
| 73 |
+
"return_dict": true,
|
| 74 |
+
"state_token_id": 151669,
|
| 75 |
+
"text_config": {
|
| 76 |
+
"_name_or_path": "",
|
| 77 |
+
"architectures": null,
|
| 78 |
+
"attention_dropout": 0.0,
|
| 79 |
+
"bos_token_id": 151643,
|
| 80 |
+
"chunk_size_feed_forward": 0,
|
| 81 |
+
"dtype": "bfloat16",
|
| 82 |
+
"eos_token_id": 151645,
|
| 83 |
+
"hidden_act": "silu",
|
| 84 |
+
"hidden_size": 2048,
|
| 85 |
+
"id2label": {
|
| 86 |
+
"0": "LABEL_0",
|
| 87 |
+
"1": "LABEL_1"
|
| 88 |
+
},
|
| 89 |
+
"initializer_range": 0.02,
|
| 90 |
+
"intermediate_size": 11008,
|
| 91 |
+
"is_encoder_decoder": false,
|
| 92 |
+
"label2id": {
|
| 93 |
+
"LABEL_0": 0,
|
| 94 |
+
"LABEL_1": 1
|
| 95 |
+
},
|
| 96 |
+
"layer_types": [
|
| 97 |
+
"full_attention",
|
| 98 |
+
"full_attention",
|
| 99 |
+
"full_attention",
|
| 100 |
+
"full_attention",
|
| 101 |
+
"full_attention",
|
| 102 |
+
"full_attention",
|
| 103 |
+
"full_attention",
|
| 104 |
+
"full_attention",
|
| 105 |
+
"full_attention",
|
| 106 |
+
"full_attention",
|
| 107 |
+
"full_attention",
|
| 108 |
+
"full_attention",
|
| 109 |
+
"full_attention",
|
| 110 |
+
"full_attention",
|
| 111 |
+
"full_attention",
|
| 112 |
+
"full_attention",
|
| 113 |
+
"full_attention",
|
| 114 |
+
"full_attention",
|
| 115 |
+
"full_attention",
|
| 116 |
+
"full_attention",
|
| 117 |
+
"full_attention",
|
| 118 |
+
"full_attention",
|
| 119 |
+
"full_attention",
|
| 120 |
+
"full_attention",
|
| 121 |
+
"full_attention",
|
| 122 |
+
"full_attention",
|
| 123 |
+
"full_attention",
|
| 124 |
+
"full_attention",
|
| 125 |
+
"full_attention",
|
| 126 |
+
"full_attention",
|
| 127 |
+
"full_attention",
|
| 128 |
+
"full_attention",
|
| 129 |
+
"full_attention",
|
| 130 |
+
"full_attention",
|
| 131 |
+
"full_attention",
|
| 132 |
+
"full_attention"
|
| 133 |
+
],
|
| 134 |
+
"max_position_embeddings": 128000,
|
| 135 |
+
"max_window_layers": 70,
|
| 136 |
+
"model_type": "qwen2_5_vl_text",
|
| 137 |
+
"num_attention_heads": 16,
|
| 138 |
+
"num_hidden_layers": 36,
|
| 139 |
+
"num_key_value_heads": 2,
|
| 140 |
+
"output_attentions": false,
|
| 141 |
+
"output_hidden_states": false,
|
| 142 |
+
"pad_token_id": null,
|
| 143 |
+
"problem_type": null,
|
| 144 |
+
"return_dict": true,
|
| 145 |
+
"rms_norm_eps": 1e-06,
|
| 146 |
+
"rope_parameters": {
|
| 147 |
+
"mrope_section": [
|
| 148 |
+
16,
|
| 149 |
+
24,
|
| 150 |
+
24
|
| 151 |
+
],
|
| 152 |
+
"rope_theta": 1000000.0,
|
| 153 |
+
"rope_type": "default",
|
| 154 |
+
"type": "mrope"
|
| 155 |
+
},
|
| 156 |
+
"sliding_window": null,
|
| 157 |
+
"use_cache": true,
|
| 158 |
+
"use_sliding_window": false,
|
| 159 |
+
"vocab_size": 151936
|
| 160 |
+
},
|
| 161 |
+
"tie_word_embeddings": true,
|
| 162 |
+
"transformers_version": "5.5.4",
|
| 163 |
+
"use_cache": false,
|
| 164 |
+
"video_token_id": 151656,
|
| 165 |
+
"vision_config": {
|
| 166 |
+
"_name_or_path": "",
|
| 167 |
+
"architectures": null,
|
| 168 |
+
"chunk_size_feed_forward": 0,
|
| 169 |
+
"depth": 32,
|
| 170 |
+
"dtype": null,
|
| 171 |
+
"fullatt_block_indexes": [
|
| 172 |
+
7,
|
| 173 |
+
15,
|
| 174 |
+
23,
|
| 175 |
+
31
|
| 176 |
+
],
|
| 177 |
+
"hidden_act": "silu",
|
| 178 |
+
"hidden_size": 1280,
|
| 179 |
+
"id2label": {
|
| 180 |
+
"0": "LABEL_0",
|
| 181 |
+
"1": "LABEL_1"
|
| 182 |
+
},
|
| 183 |
+
"in_channels": 3,
|
| 184 |
+
"in_chans": 3,
|
| 185 |
+
"initializer_range": 0.02,
|
| 186 |
+
"intermediate_size": 3420,
|
| 187 |
+
"is_encoder_decoder": false,
|
| 188 |
+
"label2id": {
|
| 189 |
+
"LABEL_0": 0,
|
| 190 |
+
"LABEL_1": 1
|
| 191 |
+
},
|
| 192 |
+
"model_type": "qwen2_5_vl",
|
| 193 |
+
"num_heads": 16,
|
| 194 |
+
"out_hidden_size": 2048,
|
| 195 |
+
"output_attentions": false,
|
| 196 |
+
"output_hidden_states": false,
|
| 197 |
+
"patch_size": 14,
|
| 198 |
+
"problem_type": null,
|
| 199 |
+
"return_dict": true,
|
| 200 |
+
"spatial_merge_size": 2,
|
| 201 |
+
"spatial_patch_size": 14,
|
| 202 |
+
"temporal_patch_size": 2,
|
| 203 |
+
"tokens_per_second": 2,
|
| 204 |
+
"window_size": 112
|
| 205 |
+
},
|
| 206 |
+
"vision_end_token_id": 151653,
|
| 207 |
+
"vision_start_token_id": 151652,
|
| 208 |
+
"vision_token_id": 151654,
|
| 209 |
+
"vocab_size": 151936
|
| 210 |
+
},
|
| 211 |
+
"image_min_pixels": 50176,
|
| 212 |
+
"image_max_pixels": 100352,
|
| 213 |
+
"use_fast_processor": true,
|
| 214 |
+
"chunk_size": 10,
|
| 215 |
+
"n_action_steps": 10,
|
| 216 |
+
"max_state_dim": 32,
|
| 217 |
+
"max_action_dim": 32,
|
| 218 |
+
"num_denoise_steps": 10,
|
| 219 |
+
"num_action_layers": 2,
|
| 220 |
+
"action_act": "linear",
|
| 221 |
+
"time_sampling_beta_alpha": 1.5,
|
| 222 |
+
"time_sampling_beta_beta": 1.0,
|
| 223 |
+
"time_sampling_scale": 0.999,
|
| 224 |
+
"time_sampling_offset": 0.001,
|
| 225 |
+
"min_period": 0.004,
|
| 226 |
+
"max_period": 4.0,
|
| 227 |
+
"supervise_padding_action_dims": true,
|
| 228 |
+
"supervise_padding_actions": true,
|
| 229 |
+
"force_fp32_autocast": true,
|
| 230 |
+
"attn_implementation": "sdpa",
|
| 231 |
+
"gradient_checkpointing": false,
|
| 232 |
+
"recipe_path": null,
|
| 233 |
+
"recipe": null,
|
| 234 |
+
"tokenizer_max_length": 2560,
|
| 235 |
+
"text_temperature": 0.0,
|
| 236 |
+
"text_top_p": 1.0,
|
| 237 |
+
"flow_loss_weight": 1.0,
|
| 238 |
+
"text_loss_weight": 1.0,
|
| 239 |
+
"normalization_mapping": {
|
| 240 |
+
"VISUAL": "IDENTITY",
|
| 241 |
+
"STATE": "MEAN_STD",
|
| 242 |
+
"ACTION": "MEAN_STD"
|
| 243 |
+
},
|
| 244 |
+
"optimizer_lr": 0.0001,
|
| 245 |
+
"optimizer_betas": [
|
| 246 |
+
0.9,
|
| 247 |
+
0.999
|
| 248 |
+
],
|
| 249 |
+
"optimizer_eps": 1e-08,
|
| 250 |
+
"optimizer_weight_decay": 0.1,
|
| 251 |
+
"optimizer_grad_clip_norm": 1.0,
|
| 252 |
+
"scheduler_warmup_steps": 900,
|
| 253 |
+
"scheduler_decay_steps": 46600,
|
| 254 |
+
"scheduler_decay_lr": 0.0,
|
| 255 |
+
"n_history": 3,
|
| 256 |
+
"history_cameras": null,
|
| 257 |
+
"camera_max_pixels": null,
|
| 258 |
+
"command_steps": 10,
|
| 259 |
+
"command_dim": 8,
|
| 260 |
+
"quaternion_identity_norm": true,
|
| 261 |
+
"batched_vision_attention": true,
|
| 262 |
+
"freeze_vision_tower": false,
|
| 263 |
+
"reasoning_form": null,
|
| 264 |
+
"memory_blank_weight": 0.1,
|
| 265 |
+
"memory_outdated_weight": 0.1,
|
| 266 |
+
"max_new_text_tokens": 256,
|
| 267 |
+
"tokens_path": null,
|
| 268 |
+
"extra_tokens": null,
|
| 269 |
+
"extra_tokens_ready": false,
|
| 270 |
+
"overlay_path": null,
|
| 271 |
+
"overlay_constants": null,
|
| 272 |
+
"overlay_camera": "observation.images.third"
|
| 273 |
+
}
|