Upload folder using huggingface_hub
Browse files- config.json +319 -0
- dataset_statistics.json +318 -0
- model.safetensors +3 -0
- singlevla_config/config.json +230 -0
config.json
ADDED
|
@@ -0,0 +1,319 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"action_dim": 10,
|
| 3 |
+
"action_head": "DiT",
|
| 4 |
+
"action_len": 20,
|
| 5 |
+
"architectures": [
|
| 6 |
+
"Eagle2_1BTwinVLA"
|
| 7 |
+
],
|
| 8 |
+
"attn_reweighting": true,
|
| 9 |
+
"denoiser": "FM",
|
| 10 |
+
"dit_scratch": false,
|
| 11 |
+
"enable_joint_attn": true,
|
| 12 |
+
"enable_moe": true,
|
| 13 |
+
"global_normalization": false,
|
| 14 |
+
"hz_interpolate": null,
|
| 15 |
+
"interpolate_gripper": false,
|
| 16 |
+
"knowledge_insulation": false,
|
| 17 |
+
"model_path": null,
|
| 18 |
+
"model_type": "Eagle2_1BTwinVLA",
|
| 19 |
+
"modeling": "denoising",
|
| 20 |
+
"normalization": "quantile",
|
| 21 |
+
"num_readouts": 1,
|
| 22 |
+
"readout_token_as_eos": false,
|
| 23 |
+
"share_decoder": true,
|
| 24 |
+
"share_embed_tokens": true,
|
| 25 |
+
"share_vision": true,
|
| 26 |
+
"singlevla_config": {
|
| 27 |
+
"_attn_implementation_autoset": false,
|
| 28 |
+
"_attn_implementation_internal": null,
|
| 29 |
+
"_commit_hash": "428b4d21376ff21d70b8b8830db6f6ab3907bfd8",
|
| 30 |
+
"_name_or_path": "jellyho/TwinVLA",
|
| 31 |
+
"action_dim": 10,
|
| 32 |
+
"action_head": "DiT",
|
| 33 |
+
"action_head_hidden_dim": 1024,
|
| 34 |
+
"action_len": 20,
|
| 35 |
+
"add_cross_attention": false,
|
| 36 |
+
"aggregation": "false",
|
| 37 |
+
"architectures": [
|
| 38 |
+
"Eagle2_1BVLA"
|
| 39 |
+
],
|
| 40 |
+
"auto_map": {},
|
| 41 |
+
"bad_words_ids": null,
|
| 42 |
+
"begin_suppress_tokens": null,
|
| 43 |
+
"bos_token_id": null,
|
| 44 |
+
"chunk_size_feed_forward": 0,
|
| 45 |
+
"cross_attention_hidden_size": null,
|
| 46 |
+
"dataset_statistics_path": null,
|
| 47 |
+
"decoder_start_token_id": null,
|
| 48 |
+
"denoiser": "FM",
|
| 49 |
+
"diffusion_batch": 32,
|
| 50 |
+
"dit_size": "DiT-B",
|
| 51 |
+
"diversity_penalty": 0.0,
|
| 52 |
+
"do_sample": false,
|
| 53 |
+
"downsample_ratio": 0.5,
|
| 54 |
+
"dynamic_image_size": true,
|
| 55 |
+
"early_stopping": false,
|
| 56 |
+
"efficient_loss": true,
|
| 57 |
+
"enable_cfg": true,
|
| 58 |
+
"encoder_no_repeat_ngram_size": 0,
|
| 59 |
+
"eos_token_id": null,
|
| 60 |
+
"exponential_decay_length_penalty": null,
|
| 61 |
+
"finetuning_task": null,
|
| 62 |
+
"force_image_size": 448,
|
| 63 |
+
"forced_bos_token_id": null,
|
| 64 |
+
"forced_eos_token_id": null,
|
| 65 |
+
"global_normalization": true,
|
| 66 |
+
"hz_interpolate": 20,
|
| 67 |
+
"id2label": {
|
| 68 |
+
"0": "LABEL_0",
|
| 69 |
+
"1": "LABEL_1"
|
| 70 |
+
},
|
| 71 |
+
"image_size": 224,
|
| 72 |
+
"interpolate_gripper": false,
|
| 73 |
+
"is_decoder": false,
|
| 74 |
+
"is_encoder_decoder": false,
|
| 75 |
+
"keep_aspect_ratio": false,
|
| 76 |
+
"knowledge_insulation": false,
|
| 77 |
+
"label2id": {
|
| 78 |
+
"LABEL_0": 0,
|
| 79 |
+
"LABEL_1": 1
|
| 80 |
+
},
|
| 81 |
+
"length_penalty": 1.0,
|
| 82 |
+
"llm_config": {
|
| 83 |
+
"_attn_implementation_autoset": true,
|
| 84 |
+
"_name_or_path": "./pretrained/Qwen2_5-0_5B-Instruct",
|
| 85 |
+
"add_cross_attention": false,
|
| 86 |
+
"architectures": [
|
| 87 |
+
"Qwen2ForCausalLM"
|
| 88 |
+
],
|
| 89 |
+
"attention_dropout": 0.0,
|
| 90 |
+
"auto_map": {
|
| 91 |
+
"AutoConfig": "configuration_qwen2.Qwen2Config",
|
| 92 |
+
"AutoModel": "modeling_qwen2.Qwen2Model",
|
| 93 |
+
"AutoModelForCausalLM": "modeling_qwen2.Qwen2ForCausalLM"
|
| 94 |
+
},
|
| 95 |
+
"bad_words_ids": null,
|
| 96 |
+
"begin_suppress_tokens": null,
|
| 97 |
+
"bos_token_id": 151643,
|
| 98 |
+
"chunk_size_feed_forward": 0,
|
| 99 |
+
"cross_attention_hidden_size": null,
|
| 100 |
+
"decoder_start_token_id": null,
|
| 101 |
+
"diversity_penalty": 0.0,
|
| 102 |
+
"do_sample": false,
|
| 103 |
+
"early_stopping": false,
|
| 104 |
+
"encoder_no_repeat_ngram_size": 0,
|
| 105 |
+
"eos_token_id": 151645,
|
| 106 |
+
"exponential_decay_length_penalty": null,
|
| 107 |
+
"finetuning_task": null,
|
| 108 |
+
"forced_bos_token_id": null,
|
| 109 |
+
"forced_eos_token_id": null,
|
| 110 |
+
"hidden_act": "silu",
|
| 111 |
+
"hidden_size": 896,
|
| 112 |
+
"id2label": {
|
| 113 |
+
"0": "LABEL_0",
|
| 114 |
+
"1": "LABEL_1"
|
| 115 |
+
},
|
| 116 |
+
"initializer_range": 0.02,
|
| 117 |
+
"intermediate_size": 4864,
|
| 118 |
+
"is_decoder": false,
|
| 119 |
+
"is_encoder_decoder": false,
|
| 120 |
+
"label2id": {
|
| 121 |
+
"LABEL_0": 0,
|
| 122 |
+
"LABEL_1": 1
|
| 123 |
+
},
|
| 124 |
+
"length_penalty": 1.0,
|
| 125 |
+
"max_length": 20,
|
| 126 |
+
"max_position_embeddings": 32768,
|
| 127 |
+
"max_window_layers": 21,
|
| 128 |
+
"min_length": 0,
|
| 129 |
+
"model_type": "qwen2",
|
| 130 |
+
"no_repeat_ngram_size": 0,
|
| 131 |
+
"num_attention_heads": 14,
|
| 132 |
+
"num_beam_groups": 1,
|
| 133 |
+
"num_beams": 1,
|
| 134 |
+
"num_hidden_layers": 24,
|
| 135 |
+
"num_key_value_heads": 2,
|
| 136 |
+
"num_return_sequences": 1,
|
| 137 |
+
"output_attentions": false,
|
| 138 |
+
"output_hidden_states": false,
|
| 139 |
+
"output_scores": false,
|
| 140 |
+
"pad_token_id": null,
|
| 141 |
+
"prefix": null,
|
| 142 |
+
"problem_type": null,
|
| 143 |
+
"pruned_heads": {},
|
| 144 |
+
"remove_invalid_values": false,
|
| 145 |
+
"repetition_penalty": 1.0,
|
| 146 |
+
"return_dict": true,
|
| 147 |
+
"return_dict_in_generate": false,
|
| 148 |
+
"rms_norm_eps": 1e-06,
|
| 149 |
+
"rope_scaling": null,
|
| 150 |
+
"rope_theta": 1000000.0,
|
| 151 |
+
"sep_token_id": null,
|
| 152 |
+
"sliding_window": 32768,
|
| 153 |
+
"suppress_tokens": null,
|
| 154 |
+
"task_specific_params": null,
|
| 155 |
+
"temperature": 1.0,
|
| 156 |
+
"tf_legacy_loss": false,
|
| 157 |
+
"tie_encoder_decoder": false,
|
| 158 |
+
"tie_word_embeddings": true,
|
| 159 |
+
"tokenizer_class": null,
|
| 160 |
+
"top_k": 50,
|
| 161 |
+
"top_p": 1.0,
|
| 162 |
+
"torch_dtype": "bfloat16",
|
| 163 |
+
"torchscript": false,
|
| 164 |
+
"transformers_version": "4.50.0.dev0",
|
| 165 |
+
"typical_p": 1.0,
|
| 166 |
+
"use_bfloat16": false,
|
| 167 |
+
"use_cache": false,
|
| 168 |
+
"use_sliding_window": false,
|
| 169 |
+
"vocab_size": 151674
|
| 170 |
+
},
|
| 171 |
+
"loss_version": "v4",
|
| 172 |
+
"max_dynamic_patch": 12,
|
| 173 |
+
"max_length": 20,
|
| 174 |
+
"min_dynamic_patch": 1,
|
| 175 |
+
"min_length": 0,
|
| 176 |
+
"mlp_checkpoint": true,
|
| 177 |
+
"model_path": "nvidia/Eagle2-1B",
|
| 178 |
+
"model_type": "Eagle2_1BVLA",
|
| 179 |
+
"modeling": "denoising",
|
| 180 |
+
"no_repeat_ngram_size": 0,
|
| 181 |
+
"normalization": "quantile",
|
| 182 |
+
"num_beam_groups": 1,
|
| 183 |
+
"num_beams": 1,
|
| 184 |
+
"num_readouts": 1,
|
| 185 |
+
"num_return_sequences": 1,
|
| 186 |
+
"output_attentions": false,
|
| 187 |
+
"output_hidden_states": false,
|
| 188 |
+
"output_scores": false,
|
| 189 |
+
"pad2square": false,
|
| 190 |
+
"pad_token_id": null,
|
| 191 |
+
"pre_feature_reduction": false,
|
| 192 |
+
"prefix": null,
|
| 193 |
+
"problem_type": null,
|
| 194 |
+
"pruned_heads": {},
|
| 195 |
+
"ps_version": "v2",
|
| 196 |
+
"readout_token_as_eos": false,
|
| 197 |
+
"remove_invalid_values": false,
|
| 198 |
+
"repetition_penalty": 1.0,
|
| 199 |
+
"return_dict": true,
|
| 200 |
+
"return_dict_in_generate": false,
|
| 201 |
+
"return_text": null,
|
| 202 |
+
"select_layer": -1,
|
| 203 |
+
"sep_token_id": null,
|
| 204 |
+
"state_dim": 10,
|
| 205 |
+
"stopping_token": "|",
|
| 206 |
+
"suppress_tokens": null,
|
| 207 |
+
"task_specific_params": null,
|
| 208 |
+
"temperature": 1.0,
|
| 209 |
+
"template": "qwen2-chat",
|
| 210 |
+
"test_denoising_steps": 10,
|
| 211 |
+
"tf_legacy_loss": false,
|
| 212 |
+
"tie_encoder_decoder": false,
|
| 213 |
+
"tie_word_embeddings": true,
|
| 214 |
+
"tokenizer_class": null,
|
| 215 |
+
"top_k": 50,
|
| 216 |
+
"top_p": 1.0,
|
| 217 |
+
"torch_dtype": "bfloat16",
|
| 218 |
+
"torchscript": false,
|
| 219 |
+
"train_denoising_steps": 100,
|
| 220 |
+
"typical_p": 1.0,
|
| 221 |
+
"use_backbone_lora": 0,
|
| 222 |
+
"use_bfloat16": false,
|
| 223 |
+
"use_llm_lora": 0,
|
| 224 |
+
"use_thumbnail": true,
|
| 225 |
+
"vision_config": {
|
| 226 |
+
"_attn_implementation_autoset": true,
|
| 227 |
+
"_name_or_path": "",
|
| 228 |
+
"add_cross_attention": false,
|
| 229 |
+
"architectures": [
|
| 230 |
+
"SiglipVisionModel"
|
| 231 |
+
],
|
| 232 |
+
"attention_dropout": 0.0,
|
| 233 |
+
"auto_map": {
|
| 234 |
+
"AutoConfig": "configuration_siglip.SiglipVisionConfig",
|
| 235 |
+
"AutoModel": "modeling_siglip.SiglipVisionModel"
|
| 236 |
+
},
|
| 237 |
+
"bad_words_ids": null,
|
| 238 |
+
"begin_suppress_tokens": null,
|
| 239 |
+
"bos_token_id": null,
|
| 240 |
+
"chunk_size_feed_forward": 0,
|
| 241 |
+
"cross_attention_hidden_size": null,
|
| 242 |
+
"decoder_start_token_id": null,
|
| 243 |
+
"diversity_penalty": 0.0,
|
| 244 |
+
"do_sample": false,
|
| 245 |
+
"drop_path_rate": 0.1,
|
| 246 |
+
"early_stopping": false,
|
| 247 |
+
"encoder_no_repeat_ngram_size": 0,
|
| 248 |
+
"eos_token_id": null,
|
| 249 |
+
"exponential_decay_length_penalty": null,
|
| 250 |
+
"finetuning_task": null,
|
| 251 |
+
"forced_bos_token_id": null,
|
| 252 |
+
"forced_eos_token_id": null,
|
| 253 |
+
"hidden_act": "gelu_pytorch_tanh",
|
| 254 |
+
"hidden_size": 1152,
|
| 255 |
+
"id2label": {
|
| 256 |
+
"0": "LABEL_0",
|
| 257 |
+
"1": "LABEL_1"
|
| 258 |
+
},
|
| 259 |
+
"image_size": 448,
|
| 260 |
+
"intermediate_size": 4304,
|
| 261 |
+
"is_decoder": false,
|
| 262 |
+
"is_encoder_decoder": false,
|
| 263 |
+
"label2id": {
|
| 264 |
+
"LABEL_0": 0,
|
| 265 |
+
"LABEL_1": 1
|
| 266 |
+
},
|
| 267 |
+
"layer_norm_eps": 1e-06,
|
| 268 |
+
"length_penalty": 1.0,
|
| 269 |
+
"max_length": 20,
|
| 270 |
+
"min_length": 0,
|
| 271 |
+
"model_type": "siglip_vision_model",
|
| 272 |
+
"no_repeat_ngram_size": 0,
|
| 273 |
+
"num_attention_heads": 16,
|
| 274 |
+
"num_beam_groups": 1,
|
| 275 |
+
"num_beams": 1,
|
| 276 |
+
"num_channels": 3,
|
| 277 |
+
"num_hidden_layers": 27,
|
| 278 |
+
"num_image_tokens": 1024,
|
| 279 |
+
"num_return_sequences": 1,
|
| 280 |
+
"output_attentions": false,
|
| 281 |
+
"output_hidden_states": false,
|
| 282 |
+
"output_scores": false,
|
| 283 |
+
"pad_token_id": null,
|
| 284 |
+
"patch_size": 14,
|
| 285 |
+
"prefix": null,
|
| 286 |
+
"problem_type": null,
|
| 287 |
+
"projection_dim": 2048,
|
| 288 |
+
"projector_hidden_act": "gelu_fast",
|
| 289 |
+
"pruned_heads": {},
|
| 290 |
+
"remove_invalid_values": false,
|
| 291 |
+
"repetition_penalty": 1.0,
|
| 292 |
+
"return_dict": true,
|
| 293 |
+
"return_dict_in_generate": false,
|
| 294 |
+
"sep_token_id": null,
|
| 295 |
+
"suppress_tokens": null,
|
| 296 |
+
"task_specific_params": null,
|
| 297 |
+
"temperature": 1.0,
|
| 298 |
+
"tf_legacy_loss": false,
|
| 299 |
+
"tie_encoder_decoder": false,
|
| 300 |
+
"tie_word_embeddings": true,
|
| 301 |
+
"tokenizer_class": null,
|
| 302 |
+
"top_k": 50,
|
| 303 |
+
"top_p": 1.0,
|
| 304 |
+
"torch_dtype": "bfloat16",
|
| 305 |
+
"torchscript": false,
|
| 306 |
+
"transformers_version": "4.50.0.dev0",
|
| 307 |
+
"typical_p": 1.0,
|
| 308 |
+
"use_bfloat16": false,
|
| 309 |
+
"vision_use_head": false
|
| 310 |
+
},
|
| 311 |
+
"vocab_size": 151674,
|
| 312 |
+
"vocab_start": null
|
| 313 |
+
},
|
| 314 |
+
"singlevla_config_path": "jellyho/TwinVLA",
|
| 315 |
+
"singlevla_pretrained_path": null,
|
| 316 |
+
"state_dim": 10,
|
| 317 |
+
"torch_dtype": "bfloat16",
|
| 318 |
+
"transformers_version": "4.50.0.dev0"
|
| 319 |
+
}
|
dataset_statistics.json
ADDED
|
@@ -0,0 +1,318 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"anubis_put_into_pot": {
|
| 3 |
+
"action": {
|
| 4 |
+
"mean": [
|
| 5 |
+
0.3237164318561554,
|
| 6 |
+
0.19323241710662842,
|
| 7 |
+
0.17074844241142273,
|
| 8 |
+
-0.40655800700187683,
|
| 9 |
+
-0.8302565813064575,
|
| 10 |
+
0.24439944326877594,
|
| 11 |
+
-0.6347360610961914,
|
| 12 |
+
0.08541738986968994,
|
| 13 |
+
-0.6577622890472412,
|
| 14 |
+
-0.14733541011810303,
|
| 15 |
+
0.3448340892791748,
|
| 16 |
+
-0.08713279664516449,
|
| 17 |
+
0.16664977371692657,
|
| 18 |
+
0.7354577779769897,
|
| 19 |
+
-0.4887841045856476,
|
| 20 |
+
0.21303653717041016,
|
| 21 |
+
-0.09639851748943329,
|
| 22 |
+
-0.553676187992096,
|
| 23 |
+
-0.6703258156776428,
|
| 24 |
+
0.6337327361106873
|
| 25 |
+
],
|
| 26 |
+
"std": [
|
| 27 |
+
0.047731053084135056,
|
| 28 |
+
0.06773686408996582,
|
| 29 |
+
0.03273946791887283,
|
| 30 |
+
0.1042996421456337,
|
| 31 |
+
0.11049824208021164,
|
| 32 |
+
0.25012287497520447,
|
| 33 |
+
0.21405042707920074,
|
| 34 |
+
0.27174288034439087,
|
| 35 |
+
0.19364885985851288,
|
| 36 |
+
0.9816686511039734,
|
| 37 |
+
0.07926628738641739,
|
| 38 |
+
0.06496380269527435,
|
| 39 |
+
0.05034809187054634,
|
| 40 |
+
0.17572304606437683,
|
| 41 |
+
0.33217287063598633,
|
| 42 |
+
0.18327680230140686,
|
| 43 |
+
0.19470512866973877,
|
| 44 |
+
0.3387109041213989,
|
| 45 |
+
0.2866652011871338,
|
| 46 |
+
0.7686191201210022
|
| 47 |
+
],
|
| 48 |
+
"max": [
|
| 49 |
+
0.3899821639060974,
|
| 50 |
+
0.40939435362815857,
|
| 51 |
+
0.26672402024269104,
|
| 52 |
+
-0.08145993202924728,
|
| 53 |
+
-0.36822038888931274,
|
| 54 |
+
0.8622857928276062,
|
| 55 |
+
0.13092871010303497,
|
| 56 |
+
0.7102192640304565,
|
| 57 |
+
-0.20073336362838745,
|
| 58 |
+
1.0,
|
| 59 |
+
0.549041748046875,
|
| 60 |
+
0.13987885415554047,
|
| 61 |
+
0.26468726992607117,
|
| 62 |
+
0.9996138215065002,
|
| 63 |
+
0.7353013753890991,
|
| 64 |
+
0.7054949402809143,
|
| 65 |
+
0.8297774195671082,
|
| 66 |
+
0.3165386915206909,
|
| 67 |
+
0.03685024008154869,
|
| 68 |
+
1.0
|
| 69 |
+
],
|
| 70 |
+
"min": [
|
| 71 |
+
0.14642871916294098,
|
| 72 |
+
0.1112271100282669,
|
| 73 |
+
0.10428614169359207,
|
| 74 |
+
-0.7812724709510803,
|
| 75 |
+
-0.9828734397888184,
|
| 76 |
+
-0.26845887303352356,
|
| 77 |
+
-0.9568612575531006,
|
| 78 |
+
-0.5920525789260864,
|
| 79 |
+
-0.9996806383132935,
|
| 80 |
+
-1.0,
|
| 81 |
+
0.1487334966659546,
|
| 82 |
+
-0.16685600578784943,
|
| 83 |
+
0.057327114045619965,
|
| 84 |
+
0.2669093906879425,
|
| 85 |
+
-0.9613909125328064,
|
| 86 |
+
-0.4603269398212433,
|
| 87 |
+
-0.6394798159599304,
|
| 88 |
+
-0.999998152256012,
|
| 89 |
+
-0.9998581409454346,
|
| 90 |
+
-1.0
|
| 91 |
+
],
|
| 92 |
+
"q01": [
|
| 93 |
+
0.18516795575618744,
|
| 94 |
+
0.12451772257685662,
|
| 95 |
+
0.11509101435542107,
|
| 96 |
+
-0.6908658498525619,
|
| 97 |
+
-0.9580670577287674,
|
| 98 |
+
-0.145108138024807,
|
| 99 |
+
-0.9188619738817215,
|
| 100 |
+
-0.5034985786676407,
|
| 101 |
+
-0.9934796422719956,
|
| 102 |
+
-1.0,
|
| 103 |
+
0.19577301174402237,
|
| 104 |
+
-0.16102446094155312,
|
| 105 |
+
0.06668414495885372,
|
| 106 |
+
0.3671954298019409,
|
| 107 |
+
-0.9094176268577576,
|
| 108 |
+
-0.2629955995082855,
|
| 109 |
+
-0.56613784968853,
|
| 110 |
+
-0.9951579666137695,
|
| 111 |
+
-0.9977369052171707,
|
| 112 |
+
-1.0
|
| 113 |
+
],
|
| 114 |
+
"q99": [
|
| 115 |
+
0.38379789978265755,
|
| 116 |
+
0.3623714971542357,
|
| 117 |
+
0.24623780891299243,
|
| 118 |
+
-0.13757481276989036,
|
| 119 |
+
-0.495397853255272,
|
| 120 |
+
0.8110066300630565,
|
| 121 |
+
-0.08928897663950969,
|
| 122 |
+
0.6101795786619182,
|
| 123 |
+
-0.2821595719456673,
|
| 124 |
+
1.0,
|
| 125 |
+
0.5267843568325042,
|
| 126 |
+
0.08202718742191788,
|
| 127 |
+
0.2494549164175987,
|
| 128 |
+
0.9867255806922912,
|
| 129 |
+
0.4620909711718558,
|
| 130 |
+
0.5912377506494515,
|
| 131 |
+
0.5964654320478437,
|
| 132 |
+
0.16175421819090668,
|
| 133 |
+
-0.044016047082841944,
|
| 134 |
+
1.0
|
| 135 |
+
],
|
| 136 |
+
"mask": [
|
| 137 |
+
true,
|
| 138 |
+
true,
|
| 139 |
+
true,
|
| 140 |
+
false,
|
| 141 |
+
false,
|
| 142 |
+
false,
|
| 143 |
+
false,
|
| 144 |
+
false,
|
| 145 |
+
false,
|
| 146 |
+
false,
|
| 147 |
+
true,
|
| 148 |
+
true,
|
| 149 |
+
true,
|
| 150 |
+
false,
|
| 151 |
+
false,
|
| 152 |
+
false,
|
| 153 |
+
false,
|
| 154 |
+
false,
|
| 155 |
+
false,
|
| 156 |
+
false
|
| 157 |
+
]
|
| 158 |
+
},
|
| 159 |
+
"proprio": {
|
| 160 |
+
"mean": [
|
| 161 |
+
0.32505786418914795,
|
| 162 |
+
0.19214771687984467,
|
| 163 |
+
0.1698455959558487,
|
| 164 |
+
-0.4101601243019104,
|
| 165 |
+
-0.8288484215736389,
|
| 166 |
+
0.24628858268260956,
|
| 167 |
+
-0.639862060546875,
|
| 168 |
+
0.08716083317995071,
|
| 169 |
+
-0.6574246287345886,
|
| 170 |
+
0.1716156154870987,
|
| 171 |
+
0.34529370069503784,
|
| 172 |
+
-0.08818725496530533,
|
| 173 |
+
0.16610461473464966,
|
| 174 |
+
0.7380656003952026,
|
| 175 |
+
-0.4896639585494995,
|
| 176 |
+
0.21104076504707336,
|
| 177 |
+
-0.09965088963508606,
|
| 178 |
+
-0.5533883571624756,
|
| 179 |
+
-0.6711149215698242,
|
| 180 |
+
0.7709687948226929
|
| 181 |
+
],
|
| 182 |
+
"std": [
|
| 183 |
+
0.04416600242257118,
|
| 184 |
+
0.06409642845392227,
|
| 185 |
+
0.030034804716706276,
|
| 186 |
+
0.10197774320840836,
|
| 187 |
+
0.11434831470251083,
|
| 188 |
+
0.2462652176618576,
|
| 189 |
+
0.2088947892189026,
|
| 190 |
+
0.26520225405693054,
|
| 191 |
+
0.19183579087257385,
|
| 192 |
+
0.6926007866859436,
|
| 193 |
+
0.07798658311367035,
|
| 194 |
+
0.06319615244865417,
|
| 195 |
+
0.049226533621549606,
|
| 196 |
+
0.17331941425800323,
|
| 197 |
+
0.3265608549118042,
|
| 198 |
+
0.1851140856742859,
|
| 199 |
+
0.1893804967403412,
|
| 200 |
+
0.3403646647930145,
|
| 201 |
+
0.28587019443511963,
|
| 202 |
+
0.4660526514053345
|
| 203 |
+
],
|
| 204 |
+
"max": [
|
| 205 |
+
0.38955727219581604,
|
| 206 |
+
0.3466160297393799,
|
| 207 |
+
0.24623136222362518,
|
| 208 |
+
-0.08630989491939545,
|
| 209 |
+
-0.3043893277645111,
|
| 210 |
+
0.8533033132553101,
|
| 211 |
+
0.0015313735930249095,
|
| 212 |
+
0.7063275575637817,
|
| 213 |
+
-0.20640863478183746,
|
| 214 |
+
1.003541111946106,
|
| 215 |
+
0.5462295413017273,
|
| 216 |
+
0.07705596089363098,
|
| 217 |
+
0.262810081243515,
|
| 218 |
+
0.9995301365852356,
|
| 219 |
+
0.5431351661682129,
|
| 220 |
+
0.7067950367927551,
|
| 221 |
+
0.6764119267463684,
|
| 222 |
+
0.30139848589897156,
|
| 223 |
+
0.025077033787965775,
|
| 224 |
+
1.014369249343872
|
| 225 |
+
],
|
| 226 |
+
"min": [
|
| 227 |
+
0.1910482496023178,
|
| 228 |
+
0.11194601655006409,
|
| 229 |
+
0.10973816365003586,
|
| 230 |
+
-0.7810142636299133,
|
| 231 |
+
-0.9814901351928711,
|
| 232 |
+
-0.21161042153835297,
|
| 233 |
+
-0.9549692869186401,
|
| 234 |
+
-0.5408061146736145,
|
| 235 |
+
-0.999822199344635,
|
| 236 |
+
-0.45464301109313965,
|
| 237 |
+
0.1878221333026886,
|
| 238 |
+
-0.16589593887329102,
|
| 239 |
+
0.06183049455285072,
|
| 240 |
+
0.2736285626888275,
|
| 241 |
+
-0.960009753704071,
|
| 242 |
+
-0.45437952876091003,
|
| 243 |
+
-0.6288766860961914,
|
| 244 |
+
-0.999998152256012,
|
| 245 |
+
-0.9999145269393921,
|
| 246 |
+
-0.49795544147491455
|
| 247 |
+
],
|
| 248 |
+
"q01": [
|
| 249 |
+
0.21300651252269745,
|
| 250 |
+
0.12324244387447834,
|
| 251 |
+
0.11845191389322281,
|
| 252 |
+
-0.6922157597541809,
|
| 253 |
+
-0.9584890645742417,
|
| 254 |
+
-0.1259118865430355,
|
| 255 |
+
-0.9182252979278565,
|
| 256 |
+
-0.46052598804235456,
|
| 257 |
+
-0.9951841980218887,
|
| 258 |
+
-0.44922900199890137,
|
| 259 |
+
0.2136411526799202,
|
| 260 |
+
-0.16165925472974776,
|
| 261 |
+
0.06899189889431,
|
| 262 |
+
0.3749441310763359,
|
| 263 |
+
-0.9038005238771438,
|
| 264 |
+
-0.2544199496507645,
|
| 265 |
+
-0.5673457723855972,
|
| 266 |
+
-0.9951095831394196,
|
| 267 |
+
-0.9976086699962616,
|
| 268 |
+
-0.3860650062561035
|
| 269 |
+
],
|
| 270 |
+
"q99": [
|
| 271 |
+
0.38241307526826857,
|
| 272 |
+
0.3363050788640975,
|
| 273 |
+
0.23329319134354592,
|
| 274 |
+
-0.14138195231556894,
|
| 275 |
+
-0.46486153453588586,
|
| 276 |
+
0.8059070360660553,
|
| 277 |
+
-0.0950157642364502,
|
| 278 |
+
0.6086657553911209,
|
| 279 |
+
-0.28258277803659465,
|
| 280 |
+
1.0017364025115967,
|
| 281 |
+
0.5247526270151138,
|
| 282 |
+
0.06677914083003997,
|
| 283 |
+
0.24800570905208577,
|
| 284 |
+
0.9867493206262589,
|
| 285 |
+
0.43939185500144884,
|
| 286 |
+
0.5892856574058531,
|
| 287 |
+
0.5718294656276701,
|
| 288 |
+
0.15633342444896692,
|
| 289 |
+
-0.04927438985556463,
|
| 290 |
+
1.0017364025115967
|
| 291 |
+
],
|
| 292 |
+
"mask": [
|
| 293 |
+
true,
|
| 294 |
+
true,
|
| 295 |
+
true,
|
| 296 |
+
false,
|
| 297 |
+
false,
|
| 298 |
+
false,
|
| 299 |
+
false,
|
| 300 |
+
false,
|
| 301 |
+
false,
|
| 302 |
+
false,
|
| 303 |
+
true,
|
| 304 |
+
true,
|
| 305 |
+
true,
|
| 306 |
+
false,
|
| 307 |
+
false,
|
| 308 |
+
false,
|
| 309 |
+
false,
|
| 310 |
+
false,
|
| 311 |
+
false,
|
| 312 |
+
false
|
| 313 |
+
]
|
| 314 |
+
},
|
| 315 |
+
"num_transitions": 12330,
|
| 316 |
+
"num_trajectories": 54
|
| 317 |
+
}
|
| 318 |
+
}
|
model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6c5793a679c97b60e4b4aa8ed772e5dc799e136aa661b6b66af2e3621134c96c
|
| 3 |
+
size 2889539864
|
singlevla_config/config.json
ADDED
|
@@ -0,0 +1,230 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_commit_hash": "428b4d21376ff21d70b8b8830db6f6ab3907bfd8",
|
| 3 |
+
"_name_or_path": "jellyho/TwinVLA",
|
| 4 |
+
"action_dim": 10,
|
| 5 |
+
"action_head": "DiT",
|
| 6 |
+
"action_head_hidden_dim": 1024,
|
| 7 |
+
"action_len": 20,
|
| 8 |
+
"aggregation": "false",
|
| 9 |
+
"architectures": [
|
| 10 |
+
"Eagle2_1BVLA"
|
| 11 |
+
],
|
| 12 |
+
"auto_map": {},
|
| 13 |
+
"dataset_statistics_path": null,
|
| 14 |
+
"denoiser": "FM",
|
| 15 |
+
"diffusion_batch": 32,
|
| 16 |
+
"dit_size": "DiT-B",
|
| 17 |
+
"downsample_ratio": 0.5,
|
| 18 |
+
"dynamic_image_size": true,
|
| 19 |
+
"efficient_loss": true,
|
| 20 |
+
"enable_cfg": true,
|
| 21 |
+
"force_image_size": 448,
|
| 22 |
+
"global_normalization": true,
|
| 23 |
+
"hz_interpolate": 20,
|
| 24 |
+
"image_size": 224,
|
| 25 |
+
"interpolate_gripper": false,
|
| 26 |
+
"keep_aspect_ratio": false,
|
| 27 |
+
"knowledge_insulation": false,
|
| 28 |
+
"llm_config": {
|
| 29 |
+
"_attn_implementation_autoset": true,
|
| 30 |
+
"_name_or_path": "./pretrained/Qwen2_5-0_5B-Instruct",
|
| 31 |
+
"add_cross_attention": false,
|
| 32 |
+
"architectures": [
|
| 33 |
+
"Qwen2ForCausalLM"
|
| 34 |
+
],
|
| 35 |
+
"attention_dropout": 0.0,
|
| 36 |
+
"auto_map": {
|
| 37 |
+
"AutoConfig": "configuration_qwen2.Qwen2Config",
|
| 38 |
+
"AutoModel": "modeling_qwen2.Qwen2Model",
|
| 39 |
+
"AutoModelForCausalLM": "modeling_qwen2.Qwen2ForCausalLM"
|
| 40 |
+
},
|
| 41 |
+
"bad_words_ids": null,
|
| 42 |
+
"begin_suppress_tokens": null,
|
| 43 |
+
"bos_token_id": 151643,
|
| 44 |
+
"chunk_size_feed_forward": 0,
|
| 45 |
+
"cross_attention_hidden_size": null,
|
| 46 |
+
"decoder_start_token_id": null,
|
| 47 |
+
"diversity_penalty": 0.0,
|
| 48 |
+
"do_sample": false,
|
| 49 |
+
"early_stopping": false,
|
| 50 |
+
"encoder_no_repeat_ngram_size": 0,
|
| 51 |
+
"eos_token_id": 151645,
|
| 52 |
+
"exponential_decay_length_penalty": null,
|
| 53 |
+
"finetuning_task": null,
|
| 54 |
+
"forced_bos_token_id": null,
|
| 55 |
+
"forced_eos_token_id": null,
|
| 56 |
+
"hidden_act": "silu",
|
| 57 |
+
"hidden_size": 896,
|
| 58 |
+
"id2label": {
|
| 59 |
+
"0": "LABEL_0",
|
| 60 |
+
"1": "LABEL_1"
|
| 61 |
+
},
|
| 62 |
+
"initializer_range": 0.02,
|
| 63 |
+
"intermediate_size": 4864,
|
| 64 |
+
"is_decoder": false,
|
| 65 |
+
"is_encoder_decoder": false,
|
| 66 |
+
"label2id": {
|
| 67 |
+
"LABEL_0": 0,
|
| 68 |
+
"LABEL_1": 1
|
| 69 |
+
},
|
| 70 |
+
"length_penalty": 1.0,
|
| 71 |
+
"max_length": 20,
|
| 72 |
+
"max_position_embeddings": 32768,
|
| 73 |
+
"max_window_layers": 21,
|
| 74 |
+
"min_length": 0,
|
| 75 |
+
"model_type": "qwen2",
|
| 76 |
+
"no_repeat_ngram_size": 0,
|
| 77 |
+
"num_attention_heads": 14,
|
| 78 |
+
"num_beam_groups": 1,
|
| 79 |
+
"num_beams": 1,
|
| 80 |
+
"num_hidden_layers": 24,
|
| 81 |
+
"num_key_value_heads": 2,
|
| 82 |
+
"num_return_sequences": 1,
|
| 83 |
+
"output_attentions": false,
|
| 84 |
+
"output_hidden_states": false,
|
| 85 |
+
"output_scores": false,
|
| 86 |
+
"pad_token_id": null,
|
| 87 |
+
"prefix": null,
|
| 88 |
+
"problem_type": null,
|
| 89 |
+
"pruned_heads": {},
|
| 90 |
+
"remove_invalid_values": false,
|
| 91 |
+
"repetition_penalty": 1.0,
|
| 92 |
+
"return_dict": true,
|
| 93 |
+
"return_dict_in_generate": false,
|
| 94 |
+
"rms_norm_eps": 1e-06,
|
| 95 |
+
"rope_scaling": null,
|
| 96 |
+
"rope_theta": 1000000.0,
|
| 97 |
+
"sep_token_id": null,
|
| 98 |
+
"sliding_window": 32768,
|
| 99 |
+
"suppress_tokens": null,
|
| 100 |
+
"task_specific_params": null,
|
| 101 |
+
"temperature": 1.0,
|
| 102 |
+
"tf_legacy_loss": false,
|
| 103 |
+
"tie_encoder_decoder": false,
|
| 104 |
+
"tie_word_embeddings": true,
|
| 105 |
+
"tokenizer_class": null,
|
| 106 |
+
"top_k": 50,
|
| 107 |
+
"top_p": 1.0,
|
| 108 |
+
"torch_dtype": "bfloat16",
|
| 109 |
+
"torchscript": false,
|
| 110 |
+
"transformers_version": "4.50.0.dev0",
|
| 111 |
+
"typical_p": 1.0,
|
| 112 |
+
"use_bfloat16": false,
|
| 113 |
+
"use_cache": false,
|
| 114 |
+
"use_sliding_window": false,
|
| 115 |
+
"vocab_size": 151674
|
| 116 |
+
},
|
| 117 |
+
"loss_version": "v4",
|
| 118 |
+
"max_dynamic_patch": 12,
|
| 119 |
+
"min_dynamic_patch": 1,
|
| 120 |
+
"mlp_checkpoint": true,
|
| 121 |
+
"model_path": "nvidia/Eagle2-1B",
|
| 122 |
+
"model_type": "Eagle2_1BVLA",
|
| 123 |
+
"modeling": "denoising",
|
| 124 |
+
"normalization": "quantile",
|
| 125 |
+
"num_readouts": 1,
|
| 126 |
+
"pad2square": false,
|
| 127 |
+
"pre_feature_reduction": false,
|
| 128 |
+
"ps_version": "v2",
|
| 129 |
+
"readout_token_as_eos": false,
|
| 130 |
+
"return_text": null,
|
| 131 |
+
"select_layer": -1,
|
| 132 |
+
"state_dim": 10,
|
| 133 |
+
"stopping_token": "|",
|
| 134 |
+
"template": "qwen2-chat",
|
| 135 |
+
"test_denoising_steps": 10,
|
| 136 |
+
"torch_dtype": "bfloat16",
|
| 137 |
+
"train_denoising_steps": 100,
|
| 138 |
+
"transformers_version": null,
|
| 139 |
+
"use_backbone_lora": 0,
|
| 140 |
+
"use_llm_lora": 0,
|
| 141 |
+
"use_thumbnail": true,
|
| 142 |
+
"vision_config": {
|
| 143 |
+
"_attn_implementation_autoset": true,
|
| 144 |
+
"_name_or_path": "",
|
| 145 |
+
"add_cross_attention": false,
|
| 146 |
+
"architectures": [
|
| 147 |
+
"SiglipVisionModel"
|
| 148 |
+
],
|
| 149 |
+
"attention_dropout": 0.0,
|
| 150 |
+
"auto_map": {
|
| 151 |
+
"AutoConfig": "configuration_siglip.SiglipVisionConfig",
|
| 152 |
+
"AutoModel": "modeling_siglip.SiglipVisionModel"
|
| 153 |
+
},
|
| 154 |
+
"bad_words_ids": null,
|
| 155 |
+
"begin_suppress_tokens": null,
|
| 156 |
+
"bos_token_id": null,
|
| 157 |
+
"chunk_size_feed_forward": 0,
|
| 158 |
+
"cross_attention_hidden_size": null,
|
| 159 |
+
"decoder_start_token_id": null,
|
| 160 |
+
"diversity_penalty": 0.0,
|
| 161 |
+
"do_sample": false,
|
| 162 |
+
"drop_path_rate": 0.1,
|
| 163 |
+
"early_stopping": false,
|
| 164 |
+
"encoder_no_repeat_ngram_size": 0,
|
| 165 |
+
"eos_token_id": null,
|
| 166 |
+
"exponential_decay_length_penalty": null,
|
| 167 |
+
"finetuning_task": null,
|
| 168 |
+
"forced_bos_token_id": null,
|
| 169 |
+
"forced_eos_token_id": null,
|
| 170 |
+
"hidden_act": "gelu_pytorch_tanh",
|
| 171 |
+
"hidden_size": 1152,
|
| 172 |
+
"id2label": {
|
| 173 |
+
"0": "LABEL_0",
|
| 174 |
+
"1": "LABEL_1"
|
| 175 |
+
},
|
| 176 |
+
"image_size": 448,
|
| 177 |
+
"intermediate_size": 4304,
|
| 178 |
+
"is_decoder": false,
|
| 179 |
+
"is_encoder_decoder": false,
|
| 180 |
+
"label2id": {
|
| 181 |
+
"LABEL_0": 0,
|
| 182 |
+
"LABEL_1": 1
|
| 183 |
+
},
|
| 184 |
+
"layer_norm_eps": 1e-06,
|
| 185 |
+
"length_penalty": 1.0,
|
| 186 |
+
"max_length": 20,
|
| 187 |
+
"min_length": 0,
|
| 188 |
+
"model_type": "siglip_vision_model",
|
| 189 |
+
"no_repeat_ngram_size": 0,
|
| 190 |
+
"num_attention_heads": 16,
|
| 191 |
+
"num_beam_groups": 1,
|
| 192 |
+
"num_beams": 1,
|
| 193 |
+
"num_channels": 3,
|
| 194 |
+
"num_hidden_layers": 27,
|
| 195 |
+
"num_image_tokens": 1024,
|
| 196 |
+
"num_return_sequences": 1,
|
| 197 |
+
"output_attentions": false,
|
| 198 |
+
"output_hidden_states": false,
|
| 199 |
+
"output_scores": false,
|
| 200 |
+
"pad_token_id": null,
|
| 201 |
+
"patch_size": 14,
|
| 202 |
+
"prefix": null,
|
| 203 |
+
"problem_type": null,
|
| 204 |
+
"projection_dim": 2048,
|
| 205 |
+
"projector_hidden_act": "gelu_fast",
|
| 206 |
+
"pruned_heads": {},
|
| 207 |
+
"remove_invalid_values": false,
|
| 208 |
+
"repetition_penalty": 1.0,
|
| 209 |
+
"return_dict": true,
|
| 210 |
+
"return_dict_in_generate": false,
|
| 211 |
+
"sep_token_id": null,
|
| 212 |
+
"suppress_tokens": null,
|
| 213 |
+
"task_specific_params": null,
|
| 214 |
+
"temperature": 1.0,
|
| 215 |
+
"tf_legacy_loss": false,
|
| 216 |
+
"tie_encoder_decoder": false,
|
| 217 |
+
"tie_word_embeddings": true,
|
| 218 |
+
"tokenizer_class": null,
|
| 219 |
+
"top_k": 50,
|
| 220 |
+
"top_p": 1.0,
|
| 221 |
+
"torch_dtype": "bfloat16",
|
| 222 |
+
"torchscript": false,
|
| 223 |
+
"transformers_version": "4.50.0.dev0",
|
| 224 |
+
"typical_p": 1.0,
|
| 225 |
+
"use_bfloat16": false,
|
| 226 |
+
"vision_use_head": false
|
| 227 |
+
},
|
| 228 |
+
"vocab_size": 151674,
|
| 229 |
+
"vocab_start": null
|
| 230 |
+
}
|