Upload 22 files
Browse files- action_model/added_tokens.json +15 -0
- action_model/config.json +310 -0
- action_model/generation_config.json +14 -0
- action_model/inference.config.json +13 -0
- action_model/merges.txt +0 -0
- action_model/model.config.json +383 -0
- action_model/model.safetensors +3 -0
- action_model/preprocessor_config.json +24 -0
- action_model/special_tokens_map.json +49 -0
- action_model/tokenizer_config.json +129 -0
- action_model/vocab.json +0 -0
- progress_model/added_tokens.json +15 -0
- progress_model/config.json +310 -0
- progress_model/generation_config.json +14 -0
- progress_model/inference.config.json +13 -0
- progress_model/merges.txt +0 -0
- progress_model/model.config.json +383 -0
- progress_model/model.safetensors +3 -0
- progress_model/preprocessor_config.json +24 -0
- progress_model/special_tokens_map.json +49 -0
- progress_model/tokenizer_config.json +129 -0
- progress_model/vocab.json +0 -0
action_model/added_tokens.json
ADDED
|
@@ -0,0 +1,15 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"<act>": 151649,
|
| 3 |
+
"<generated_image>": 151652,
|
| 4 |
+
"<his>": 151650,
|
| 5 |
+
"<image>": 151654,
|
| 6 |
+
"<nav>": 151648,
|
| 7 |
+
"<obs>": 151651,
|
| 8 |
+
"<vila/sentinel>": 151653,
|
| 9 |
+
"<vila/video>": 151655,
|
| 10 |
+
"<|endoftext|>": 151643,
|
| 11 |
+
"<|im_end|>": 151645,
|
| 12 |
+
"<|im_start|>": 151644,
|
| 13 |
+
"[BOS]": 151646,
|
| 14 |
+
"[PAD]": 151647
|
| 15 |
+
}
|
action_model/config.json
ADDED
|
@@ -0,0 +1,310 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"Ubit": 100,
|
| 3 |
+
"_attn_implementation_autoset": true,
|
| 4 |
+
"architectures": [
|
| 5 |
+
"LlavaLlamaModel"
|
| 6 |
+
],
|
| 7 |
+
"babit": "E5M2",
|
| 8 |
+
"bobit": "E5M2",
|
| 9 |
+
"bwbit": "E5M2",
|
| 10 |
+
"chat_template": null,
|
| 11 |
+
"col_blocksize": -1,
|
| 12 |
+
"col_blocksize_optimizer": 128,
|
| 13 |
+
"draw_distribution_backward": false,
|
| 14 |
+
"draw_distribution_forward": false,
|
| 15 |
+
"drop_path_rate": 0.0,
|
| 16 |
+
"dynamic_s2": false,
|
| 17 |
+
"epsilon": 1e-10,
|
| 18 |
+
"epsilon_optimizer": 1e-15,
|
| 19 |
+
"fabit": "E4M3",
|
| 20 |
+
"first_order_bit": null,
|
| 21 |
+
"first_order_quant_type": null,
|
| 22 |
+
"fobit": "E4M3",
|
| 23 |
+
"fps": 0.0,
|
| 24 |
+
"fwbit": "E4M3",
|
| 25 |
+
"group_size": -1,
|
| 26 |
+
"hidden_size": 1536,
|
| 27 |
+
"image_aspect_ratio": "dynamic",
|
| 28 |
+
"interpolate_mode": "linear",
|
| 29 |
+
"llm_cfg": {
|
| 30 |
+
"_attn_implementation_autoset": false,
|
| 31 |
+
"add_cross_attention": false,
|
| 32 |
+
"architectures": [
|
| 33 |
+
"Qwen2ForCausalLM"
|
| 34 |
+
],
|
| 35 |
+
"attention_dropout": 0.0,
|
| 36 |
+
"bad_words_ids": null,
|
| 37 |
+
"begin_suppress_tokens": null,
|
| 38 |
+
"bos_token_id": 151643,
|
| 39 |
+
"chunk_size_feed_forward": 0,
|
| 40 |
+
"cross_attention_hidden_size": null,
|
| 41 |
+
"decoder_start_token_id": null,
|
| 42 |
+
"diversity_penalty": 0.0,
|
| 43 |
+
"do_sample": false,
|
| 44 |
+
"early_stopping": false,
|
| 45 |
+
"encoder_no_repeat_ngram_size": 0,
|
| 46 |
+
"eos_token_id": 151645,
|
| 47 |
+
"exponential_decay_length_penalty": null,
|
| 48 |
+
"finetuning_task": null,
|
| 49 |
+
"forced_bos_token_id": null,
|
| 50 |
+
"forced_eos_token_id": null,
|
| 51 |
+
"hidden_act": "silu",
|
| 52 |
+
"hidden_size": 1536,
|
| 53 |
+
"id2label": {
|
| 54 |
+
"0": "LABEL_0",
|
| 55 |
+
"1": "LABEL_1"
|
| 56 |
+
},
|
| 57 |
+
"initializer_range": 0.02,
|
| 58 |
+
"intermediate_size": 8960,
|
| 59 |
+
"is_decoder": false,
|
| 60 |
+
"is_encoder_decoder": false,
|
| 61 |
+
"label2id": {
|
| 62 |
+
"LABEL_0": 0,
|
| 63 |
+
"LABEL_1": 1
|
| 64 |
+
},
|
| 65 |
+
"length_penalty": 1.0,
|
| 66 |
+
"max_length": 20,
|
| 67 |
+
"max_position_embeddings": 32768,
|
| 68 |
+
"max_window_layers": 28,
|
| 69 |
+
"min_length": 0,
|
| 70 |
+
"model_max_length": 4096,
|
| 71 |
+
"model_type": "qwen2",
|
| 72 |
+
"no_repeat_ngram_size": 0,
|
| 73 |
+
"num_attention_heads": 12,
|
| 74 |
+
"num_beam_groups": 1,
|
| 75 |
+
"num_beams": 1,
|
| 76 |
+
"num_hidden_layers": 28,
|
| 77 |
+
"num_key_value_heads": 2,
|
| 78 |
+
"num_return_sequences": 1,
|
| 79 |
+
"output_attentions": false,
|
| 80 |
+
"output_hidden_states": false,
|
| 81 |
+
"output_scores": false,
|
| 82 |
+
"pad_token_id": null,
|
| 83 |
+
"prefix": null,
|
| 84 |
+
"problem_type": null,
|
| 85 |
+
"pruned_heads": {},
|
| 86 |
+
"remove_invalid_values": false,
|
| 87 |
+
"repetition_penalty": 1.0,
|
| 88 |
+
"return_dict": true,
|
| 89 |
+
"return_dict_in_generate": false,
|
| 90 |
+
"rms_norm_eps": 1e-06,
|
| 91 |
+
"rope_scaling": null,
|
| 92 |
+
"rope_theta": 1000000.0,
|
| 93 |
+
"sep_token_id": null,
|
| 94 |
+
"sliding_window": null,
|
| 95 |
+
"suppress_tokens": null,
|
| 96 |
+
"task_specific_params": null,
|
| 97 |
+
"temperature": 1.0,
|
| 98 |
+
"tf_legacy_loss": false,
|
| 99 |
+
"tie_encoder_decoder": false,
|
| 100 |
+
"tie_word_embeddings": true,
|
| 101 |
+
"tokenizer_class": null,
|
| 102 |
+
"tokenizer_model_max_length": 4096,
|
| 103 |
+
"tokenizer_padding_side": "right",
|
| 104 |
+
"top_k": 50,
|
| 105 |
+
"top_p": 1.0,
|
| 106 |
+
"torch_dtype": "bfloat16",
|
| 107 |
+
"torchscript": false,
|
| 108 |
+
"typical_p": 1.0,
|
| 109 |
+
"use_bfloat16": false,
|
| 110 |
+
"use_cache": false,
|
| 111 |
+
"use_sliding_window": false,
|
| 112 |
+
"vocab_size": 151651
|
| 113 |
+
},
|
| 114 |
+
"max_tiles": 12,
|
| 115 |
+
"min_blockunit_col": 4,
|
| 116 |
+
"min_blockunit_row": 4,
|
| 117 |
+
"min_tiles": 1,
|
| 118 |
+
"mlp_path": null,
|
| 119 |
+
"mm_hidden_size": 1152,
|
| 120 |
+
"mm_projector": "mlp_downsample_3x3_fix",
|
| 121 |
+
"mm_projector_cfg": {
|
| 122 |
+
"_attn_implementation_autoset": false,
|
| 123 |
+
"add_cross_attention": false,
|
| 124 |
+
"architectures": [
|
| 125 |
+
"MultimodalProjector"
|
| 126 |
+
],
|
| 127 |
+
"bad_words_ids": null,
|
| 128 |
+
"begin_suppress_tokens": null,
|
| 129 |
+
"bos_token_id": null,
|
| 130 |
+
"chunk_size_feed_forward": 0,
|
| 131 |
+
"cross_attention_hidden_size": null,
|
| 132 |
+
"decoder_start_token_id": null,
|
| 133 |
+
"diversity_penalty": 0.0,
|
| 134 |
+
"do_sample": false,
|
| 135 |
+
"early_stopping": false,
|
| 136 |
+
"encoder_no_repeat_ngram_size": 0,
|
| 137 |
+
"eos_token_id": null,
|
| 138 |
+
"exponential_decay_length_penalty": null,
|
| 139 |
+
"finetuning_task": null,
|
| 140 |
+
"forced_bos_token_id": null,
|
| 141 |
+
"forced_eos_token_id": null,
|
| 142 |
+
"id2label": {
|
| 143 |
+
"0": "LABEL_0",
|
| 144 |
+
"1": "LABEL_1"
|
| 145 |
+
},
|
| 146 |
+
"is_decoder": false,
|
| 147 |
+
"is_encoder_decoder": false,
|
| 148 |
+
"label2id": {
|
| 149 |
+
"LABEL_0": 0,
|
| 150 |
+
"LABEL_1": 1
|
| 151 |
+
},
|
| 152 |
+
"length_penalty": 1.0,
|
| 153 |
+
"max_length": 20,
|
| 154 |
+
"min_length": 0,
|
| 155 |
+
"mm_projector_type": "mlp_downsample_3x3_fix",
|
| 156 |
+
"model_type": "v2l_projector",
|
| 157 |
+
"no_repeat_ngram_size": 0,
|
| 158 |
+
"num_beam_groups": 1,
|
| 159 |
+
"num_beams": 1,
|
| 160 |
+
"num_return_sequences": 1,
|
| 161 |
+
"output_attentions": false,
|
| 162 |
+
"output_hidden_states": false,
|
| 163 |
+
"output_scores": false,
|
| 164 |
+
"pad_token_id": null,
|
| 165 |
+
"prefix": null,
|
| 166 |
+
"problem_type": null,
|
| 167 |
+
"pruned_heads": {},
|
| 168 |
+
"remove_invalid_values": false,
|
| 169 |
+
"repetition_penalty": 1.0,
|
| 170 |
+
"return_dict": true,
|
| 171 |
+
"return_dict_in_generate": false,
|
| 172 |
+
"sep_token_id": null,
|
| 173 |
+
"suppress_tokens": null,
|
| 174 |
+
"task_specific_params": null,
|
| 175 |
+
"temperature": 1.0,
|
| 176 |
+
"tf_legacy_loss": false,
|
| 177 |
+
"tie_encoder_decoder": false,
|
| 178 |
+
"tie_word_embeddings": true,
|
| 179 |
+
"tokenizer_class": null,
|
| 180 |
+
"top_k": 50,
|
| 181 |
+
"top_p": 1.0,
|
| 182 |
+
"torch_dtype": "bfloat16",
|
| 183 |
+
"torchscript": false,
|
| 184 |
+
"typical_p": 1.0,
|
| 185 |
+
"use_bfloat16": false
|
| 186 |
+
},
|
| 187 |
+
"mm_projector_lr": null,
|
| 188 |
+
"mm_use_im_patch_token": false,
|
| 189 |
+
"mm_use_im_start_end": false,
|
| 190 |
+
"mm_vision_select_feature": "cls_patch",
|
| 191 |
+
"mm_vision_select_layer": -2,
|
| 192 |
+
"model_dtype": "torch.bfloat16",
|
| 193 |
+
"model_type": "llava_llama",
|
| 194 |
+
"num_time_tokens": 0,
|
| 195 |
+
"num_video_frames": 8,
|
| 196 |
+
"pad_block": false,
|
| 197 |
+
"pad_to_multiple_of": 0,
|
| 198 |
+
"qchoice": "none",
|
| 199 |
+
"quantize_model": false,
|
| 200 |
+
"refine_attn_blocksize": false,
|
| 201 |
+
"refine_col_blocksize": 4,
|
| 202 |
+
"refine_ln_blocksize": false,
|
| 203 |
+
"refine_ln_blocksize_but_only_backward": false,
|
| 204 |
+
"refine_ln_blocksize_but_only_forward": false,
|
| 205 |
+
"refine_ln_pertoken": false,
|
| 206 |
+
"refine_mlp_blocksize": false,
|
| 207 |
+
"refine_residual_fp": false,
|
| 208 |
+
"refine_row_blocksize": 4,
|
| 209 |
+
"row_blocksize": -1,
|
| 210 |
+
"row_blocksize_optimizer": 1,
|
| 211 |
+
"s2": false,
|
| 212 |
+
"s2_max_split_size": 336,
|
| 213 |
+
"s2_resize_output_to_scale_idx": 0,
|
| 214 |
+
"s2_scales": "336,672,1008",
|
| 215 |
+
"second_order_bit": null,
|
| 216 |
+
"second_order_quant_type": null,
|
| 217 |
+
"soft_ce_std": 1.0,
|
| 218 |
+
"symm": true,
|
| 219 |
+
"time_token_format": "<t{t}>",
|
| 220 |
+
"time_token_ids": [],
|
| 221 |
+
"transformers_version": "4.46.0",
|
| 222 |
+
"tune_language_model": true,
|
| 223 |
+
"tune_mm_projector": true,
|
| 224 |
+
"tune_vision_tower": true,
|
| 225 |
+
"use_quantize_optimizer": false,
|
| 226 |
+
"version": "auto",
|
| 227 |
+
"vision_resolution": -1,
|
| 228 |
+
"vision_tower": "Efficient-Large-Model/paligemma-siglip-so400m-patch14-448",
|
| 229 |
+
"vision_tower_cfg": {
|
| 230 |
+
"_attn_implementation_autoset": false,
|
| 231 |
+
"add_cross_attention": false,
|
| 232 |
+
"architectures": [
|
| 233 |
+
"SiglipVisionModel"
|
| 234 |
+
],
|
| 235 |
+
"attention_dropout": 0.0,
|
| 236 |
+
"bad_words_ids": null,
|
| 237 |
+
"begin_suppress_tokens": null,
|
| 238 |
+
"bos_token_id": null,
|
| 239 |
+
"chunk_size_feed_forward": 0,
|
| 240 |
+
"cross_attention_hidden_size": null,
|
| 241 |
+
"decoder_start_token_id": null,
|
| 242 |
+
"diversity_penalty": 0.0,
|
| 243 |
+
"do_sample": false,
|
| 244 |
+
"early_stopping": false,
|
| 245 |
+
"encoder_no_repeat_ngram_size": 0,
|
| 246 |
+
"eos_token_id": null,
|
| 247 |
+
"exponential_decay_length_penalty": null,
|
| 248 |
+
"finetuning_task": null,
|
| 249 |
+
"forced_bos_token_id": null,
|
| 250 |
+
"forced_eos_token_id": null,
|
| 251 |
+
"hidden_act": "gelu_pytorch_tanh",
|
| 252 |
+
"hidden_size": 1152,
|
| 253 |
+
"id2label": {
|
| 254 |
+
"0": "LABEL_0",
|
| 255 |
+
"1": "LABEL_1"
|
| 256 |
+
},
|
| 257 |
+
"image_size": 448,
|
| 258 |
+
"intermediate_size": 4304,
|
| 259 |
+
"is_decoder": false,
|
| 260 |
+
"is_encoder_decoder": false,
|
| 261 |
+
"label2id": {
|
| 262 |
+
"LABEL_0": 0,
|
| 263 |
+
"LABEL_1": 1
|
| 264 |
+
},
|
| 265 |
+
"layer_norm_eps": 1e-06,
|
| 266 |
+
"length_penalty": 1.0,
|
| 267 |
+
"max_length": 20,
|
| 268 |
+
"min_length": 0,
|
| 269 |
+
"model_type": "siglip_vision_model",
|
| 270 |
+
"no_repeat_ngram_size": 0,
|
| 271 |
+
"num_attention_heads": 16,
|
| 272 |
+
"num_beam_groups": 1,
|
| 273 |
+
"num_beams": 1,
|
| 274 |
+
"num_channels": 3,
|
| 275 |
+
"num_hidden_layers": 27,
|
| 276 |
+
"num_image_tokens": 256,
|
| 277 |
+
"num_return_sequences": 1,
|
| 278 |
+
"output_attentions": false,
|
| 279 |
+
"output_hidden_states": false,
|
| 280 |
+
"output_scores": false,
|
| 281 |
+
"pad_token_id": null,
|
| 282 |
+
"patch_size": 14,
|
| 283 |
+
"prefix": null,
|
| 284 |
+
"problem_type": null,
|
| 285 |
+
"projection_dim": 2048,
|
| 286 |
+
"projector_hidden_act": "gelu_fast",
|
| 287 |
+
"pruned_heads": {},
|
| 288 |
+
"remove_invalid_values": false,
|
| 289 |
+
"repetition_penalty": 1.0,
|
| 290 |
+
"return_dict": true,
|
| 291 |
+
"return_dict_in_generate": false,
|
| 292 |
+
"sep_token_id": null,
|
| 293 |
+
"suppress_tokens": null,
|
| 294 |
+
"task_specific_params": null,
|
| 295 |
+
"temperature": 1.0,
|
| 296 |
+
"tf_legacy_loss": false,
|
| 297 |
+
"tie_encoder_decoder": false,
|
| 298 |
+
"tie_word_embeddings": true,
|
| 299 |
+
"tokenizer_class": null,
|
| 300 |
+
"top_k": 50,
|
| 301 |
+
"top_p": 1.0,
|
| 302 |
+
"torch_dtype": "bfloat16",
|
| 303 |
+
"torchscript": false,
|
| 304 |
+
"typical_p": 1.0,
|
| 305 |
+
"use_bfloat16": false,
|
| 306 |
+
"vision_use_head": false
|
| 307 |
+
},
|
| 308 |
+
"vision_tower_lr": null,
|
| 309 |
+
"weight_memory_efficient": true
|
| 310 |
+
}
|
action_model/generation_config.json
ADDED
|
@@ -0,0 +1,14 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"bos_token_id": 151643,
|
| 3 |
+
"do_sample": true,
|
| 4 |
+
"eos_token_id": [
|
| 5 |
+
151645,
|
| 6 |
+
151643
|
| 7 |
+
],
|
| 8 |
+
"pad_token_id": 151643,
|
| 9 |
+
"repetition_penalty": 1.1,
|
| 10 |
+
"temperature": 0.7,
|
| 11 |
+
"top_k": 20,
|
| 12 |
+
"top_p": 0.8,
|
| 13 |
+
"transformers_version": "4.46.0"
|
| 14 |
+
}
|
action_model/inference.config.json
ADDED
|
@@ -0,0 +1,13 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"__config_type__": "robo_orchard_lab.inference.basic:InferencePipelineCfg",
|
| 3 |
+
"class_type": "robo_orchard_lab.inference.basic:InferencePipeline",
|
| 4 |
+
"model_cfg": null,
|
| 5 |
+
"processor": {
|
| 6 |
+
"__config_type__": "robo_orchard_lab.models.progress_think.action_processor:ActionModelProcessorCfg",
|
| 7 |
+
"class_type": "robo_orchard_lab.models.progress_think.action_processor:ActionModelProcessor",
|
| 8 |
+
"load_image": true,
|
| 9 |
+
"transforms": null
|
| 10 |
+
},
|
| 11 |
+
"collate_fn": null,
|
| 12 |
+
"batch_size": 1
|
| 13 |
+
}
|
action_model/merges.txt
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
action_model/model.config.json
ADDED
|
@@ -0,0 +1,383 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"__config_type__": "robo_orchard_lab.models.progress_think.action_model:ActionModelConfig",
|
| 3 |
+
"class_type": "robo_orchard_lab.models.progress_think.action_model:ActionModel",
|
| 4 |
+
"model_dtype": "torch.float16",
|
| 5 |
+
"llava_llama_config": {
|
| 6 |
+
"Ubit": 100,
|
| 7 |
+
"_attn_implementation_autoset": true,
|
| 8 |
+
"architectures": [
|
| 9 |
+
"LlavaLlamaModel"
|
| 10 |
+
],
|
| 11 |
+
"babit": "E5M2",
|
| 12 |
+
"bobit": "E5M2",
|
| 13 |
+
"bwbit": "E5M2",
|
| 14 |
+
"chat_template": null,
|
| 15 |
+
"col_blocksize": -1,
|
| 16 |
+
"col_blocksize_optimizer": 128,
|
| 17 |
+
"draw_distribution_backward": false,
|
| 18 |
+
"draw_distribution_forward": false,
|
| 19 |
+
"drop_path_rate": 0.0,
|
| 20 |
+
"dynamic_s2": false,
|
| 21 |
+
"epsilon": 1e-10,
|
| 22 |
+
"epsilon_optimizer": 1e-15,
|
| 23 |
+
"fabit": "E4M3",
|
| 24 |
+
"first_order_bit": null,
|
| 25 |
+
"first_order_quant_type": null,
|
| 26 |
+
"fobit": "E4M3",
|
| 27 |
+
"fps": 0.0,
|
| 28 |
+
"fwbit": "E4M3",
|
| 29 |
+
"group_size": -1,
|
| 30 |
+
"hidden_size": 1536,
|
| 31 |
+
"image_aspect_ratio": "dynamic",
|
| 32 |
+
"image_encoder": {
|
| 33 |
+
"_target_": "robo_orchard_lab.models.monodream.multimodal_encoder.BasicImageEncoder"
|
| 34 |
+
},
|
| 35 |
+
"interpolate_mode": "linear",
|
| 36 |
+
"llm_cfg": {
|
| 37 |
+
"_attn_implementation_autoset": false,
|
| 38 |
+
"add_cross_attention": false,
|
| 39 |
+
"architectures": [
|
| 40 |
+
"Qwen2ForCausalLM"
|
| 41 |
+
],
|
| 42 |
+
"attention_dropout": 0.0,
|
| 43 |
+
"bad_words_ids": null,
|
| 44 |
+
"begin_suppress_tokens": null,
|
| 45 |
+
"bos_token_id": 151643,
|
| 46 |
+
"chunk_size_feed_forward": 0,
|
| 47 |
+
"cross_attention_hidden_size": null,
|
| 48 |
+
"decoder_start_token_id": null,
|
| 49 |
+
"diversity_penalty": 0.0,
|
| 50 |
+
"do_sample": false,
|
| 51 |
+
"early_stopping": false,
|
| 52 |
+
"encoder_no_repeat_ngram_size": 0,
|
| 53 |
+
"eos_token_id": 151645,
|
| 54 |
+
"exponential_decay_length_penalty": null,
|
| 55 |
+
"finetuning_task": null,
|
| 56 |
+
"forced_bos_token_id": null,
|
| 57 |
+
"forced_eos_token_id": null,
|
| 58 |
+
"hidden_act": "silu",
|
| 59 |
+
"hidden_size": 1536,
|
| 60 |
+
"id2label": {
|
| 61 |
+
"0": "LABEL_0",
|
| 62 |
+
"1": "LABEL_1"
|
| 63 |
+
},
|
| 64 |
+
"initializer_range": 0.02,
|
| 65 |
+
"intermediate_size": 8960,
|
| 66 |
+
"is_decoder": false,
|
| 67 |
+
"is_encoder_decoder": false,
|
| 68 |
+
"label2id": {
|
| 69 |
+
"LABEL_0": 0,
|
| 70 |
+
"LABEL_1": 1
|
| 71 |
+
},
|
| 72 |
+
"length_penalty": 1.0,
|
| 73 |
+
"max_length": 20,
|
| 74 |
+
"max_position_embeddings": 32768,
|
| 75 |
+
"max_window_layers": 28,
|
| 76 |
+
"min_length": 0,
|
| 77 |
+
"model_max_length": 4096,
|
| 78 |
+
"model_type": "qwen2",
|
| 79 |
+
"no_repeat_ngram_size": 0,
|
| 80 |
+
"num_attention_heads": 12,
|
| 81 |
+
"num_beam_groups": 1,
|
| 82 |
+
"num_beams": 1,
|
| 83 |
+
"num_hidden_layers": 28,
|
| 84 |
+
"num_key_value_heads": 2,
|
| 85 |
+
"num_return_sequences": 1,
|
| 86 |
+
"output_attentions": false,
|
| 87 |
+
"output_hidden_states": false,
|
| 88 |
+
"output_scores": false,
|
| 89 |
+
"pad_token_id": null,
|
| 90 |
+
"prefix": null,
|
| 91 |
+
"problem_type": null,
|
| 92 |
+
"pruned_heads": {},
|
| 93 |
+
"remove_invalid_values": false,
|
| 94 |
+
"repetition_penalty": 1.0,
|
| 95 |
+
"return_dict": true,
|
| 96 |
+
"return_dict_in_generate": false,
|
| 97 |
+
"rms_norm_eps": 1e-6,
|
| 98 |
+
"rope_scaling": null,
|
| 99 |
+
"rope_theta": 1000000.0,
|
| 100 |
+
"sep_token_id": null,
|
| 101 |
+
"sliding_window": null,
|
| 102 |
+
"suppress_tokens": null,
|
| 103 |
+
"task_specific_params": null,
|
| 104 |
+
"temperature": 1.0,
|
| 105 |
+
"tf_legacy_loss": false,
|
| 106 |
+
"tie_encoder_decoder": false,
|
| 107 |
+
"tie_word_embeddings": true,
|
| 108 |
+
"tokenizer_class": null,
|
| 109 |
+
"tokenizer_model_max_length": 4096,
|
| 110 |
+
"tokenizer_padding_side": "right",
|
| 111 |
+
"top_k": 50,
|
| 112 |
+
"top_p": 1.0,
|
| 113 |
+
"torch_dtype": "bfloat16",
|
| 114 |
+
"torchscript": false,
|
| 115 |
+
"typical_p": 1.0,
|
| 116 |
+
"use_bfloat16": false,
|
| 117 |
+
"use_cache": false,
|
| 118 |
+
"use_sliding_window": false,
|
| 119 |
+
"vocab_size": 151656
|
| 120 |
+
},
|
| 121 |
+
"max_tiles": 12,
|
| 122 |
+
"min_blockunit_col": 4,
|
| 123 |
+
"min_blockunit_row": 4,
|
| 124 |
+
"min_tiles": 1,
|
| 125 |
+
"mlp_path": null,
|
| 126 |
+
"mm_hidden_size": 1152,
|
| 127 |
+
"mm_projector": "mlp_downsample_3x3_fix",
|
| 128 |
+
"mm_projector_cfg": {
|
| 129 |
+
"_attn_implementation_autoset": false,
|
| 130 |
+
"add_cross_attention": false,
|
| 131 |
+
"architectures": [
|
| 132 |
+
"MultimodalProjector"
|
| 133 |
+
],
|
| 134 |
+
"bad_words_ids": null,
|
| 135 |
+
"begin_suppress_tokens": null,
|
| 136 |
+
"bos_token_id": null,
|
| 137 |
+
"chunk_size_feed_forward": 0,
|
| 138 |
+
"cross_attention_hidden_size": null,
|
| 139 |
+
"decoder_start_token_id": null,
|
| 140 |
+
"diversity_penalty": 0.0,
|
| 141 |
+
"do_sample": false,
|
| 142 |
+
"early_stopping": false,
|
| 143 |
+
"encoder_no_repeat_ngram_size": 0,
|
| 144 |
+
"eos_token_id": null,
|
| 145 |
+
"exponential_decay_length_penalty": null,
|
| 146 |
+
"finetuning_task": null,
|
| 147 |
+
"forced_bos_token_id": null,
|
| 148 |
+
"forced_eos_token_id": null,
|
| 149 |
+
"id2label": {
|
| 150 |
+
"0": "LABEL_0",
|
| 151 |
+
"1": "LABEL_1"
|
| 152 |
+
},
|
| 153 |
+
"is_decoder": false,
|
| 154 |
+
"is_encoder_decoder": false,
|
| 155 |
+
"label2id": {
|
| 156 |
+
"LABEL_0": 0,
|
| 157 |
+
"LABEL_1": 1
|
| 158 |
+
},
|
| 159 |
+
"length_penalty": 1.0,
|
| 160 |
+
"max_length": 20,
|
| 161 |
+
"min_length": 0,
|
| 162 |
+
"mm_projector_type": "mlp_downsample_3x3_fix",
|
| 163 |
+
"model_type": "v2l_projector",
|
| 164 |
+
"no_repeat_ngram_size": 0,
|
| 165 |
+
"num_beam_groups": 1,
|
| 166 |
+
"num_beams": 1,
|
| 167 |
+
"num_return_sequences": 1,
|
| 168 |
+
"output_attentions": false,
|
| 169 |
+
"output_hidden_states": false,
|
| 170 |
+
"output_scores": false,
|
| 171 |
+
"pad_token_id": null,
|
| 172 |
+
"prefix": null,
|
| 173 |
+
"problem_type": null,
|
| 174 |
+
"pruned_heads": {},
|
| 175 |
+
"remove_invalid_values": false,
|
| 176 |
+
"repetition_penalty": 1.0,
|
| 177 |
+
"return_dict": true,
|
| 178 |
+
"return_dict_in_generate": false,
|
| 179 |
+
"sep_token_id": null,
|
| 180 |
+
"suppress_tokens": null,
|
| 181 |
+
"task_specific_params": null,
|
| 182 |
+
"temperature": 1.0,
|
| 183 |
+
"tf_legacy_loss": false,
|
| 184 |
+
"tie_encoder_decoder": false,
|
| 185 |
+
"tie_word_embeddings": true,
|
| 186 |
+
"tokenizer_class": null,
|
| 187 |
+
"top_k": 50,
|
| 188 |
+
"top_p": 1.0,
|
| 189 |
+
"torch_dtype": "bfloat16",
|
| 190 |
+
"torchscript": false,
|
| 191 |
+
"typical_p": 1.0,
|
| 192 |
+
"use_bfloat16": false
|
| 193 |
+
},
|
| 194 |
+
"mm_projector_lr": null,
|
| 195 |
+
"mm_use_im_patch_token": false,
|
| 196 |
+
"mm_use_im_start_end": false,
|
| 197 |
+
"mm_vision_select_feature": "cls_patch",
|
| 198 |
+
"mm_vision_select_layer": -2,
|
| 199 |
+
"model_dtype": "torch.bfloat16",
|
| 200 |
+
"model_type": "llava_llama",
|
| 201 |
+
"num_time_tokens": 0,
|
| 202 |
+
"num_video_frames": 8,
|
| 203 |
+
"pad_block": false,
|
| 204 |
+
"pad_to_multiple_of": 0,
|
| 205 |
+
"qchoice": "none",
|
| 206 |
+
"quantize_model": false,
|
| 207 |
+
"refine_attn_blocksize": false,
|
| 208 |
+
"refine_col_blocksize": 4,
|
| 209 |
+
"refine_ln_blocksize": false,
|
| 210 |
+
"refine_ln_blocksize_but_only_backward": false,
|
| 211 |
+
"refine_ln_blocksize_but_only_forward": false,
|
| 212 |
+
"refine_ln_pertoken": false,
|
| 213 |
+
"refine_mlp_blocksize": false,
|
| 214 |
+
"refine_residual_fp": false,
|
| 215 |
+
"refine_row_blocksize": 4,
|
| 216 |
+
"row_blocksize": -1,
|
| 217 |
+
"row_blocksize_optimizer": 1,
|
| 218 |
+
"s2": false,
|
| 219 |
+
"s2_max_split_size": 336,
|
| 220 |
+
"s2_resize_output_to_scale_idx": 0,
|
| 221 |
+
"s2_scales": "336,672,1008",
|
| 222 |
+
"second_order_bit": null,
|
| 223 |
+
"second_order_quant_type": null,
|
| 224 |
+
"soft_ce_std": 1.0,
|
| 225 |
+
"symm": true,
|
| 226 |
+
"time_token_format": "<t{t}>",
|
| 227 |
+
"time_token_ids": [],
|
| 228 |
+
"transformers_version": "4.46.0",
|
| 229 |
+
"tune_language_model": true,
|
| 230 |
+
"tune_mm_projector": true,
|
| 231 |
+
"tune_vision_tower": true,
|
| 232 |
+
"use_quantize_optimizer": false,
|
| 233 |
+
"version": "auto",
|
| 234 |
+
"video_encoder": {
|
| 235 |
+
"_target_": "robo_orchard_lab.models.monodream.multimodal_encoder.BasicVideoEncoder"
|
| 236 |
+
},
|
| 237 |
+
"vision_resolution": -1,
|
| 238 |
+
"vision_tower": "Efficient-Large-Model/paligemma-siglip-so400m-patch14-448",
|
| 239 |
+
"vision_tower_cfg": {
|
| 240 |
+
"_attn_implementation_autoset": false,
|
| 241 |
+
"add_cross_attention": false,
|
| 242 |
+
"architectures": [
|
| 243 |
+
"SiglipVisionModel"
|
| 244 |
+
],
|
| 245 |
+
"attention_dropout": 0.0,
|
| 246 |
+
"bad_words_ids": null,
|
| 247 |
+
"begin_suppress_tokens": null,
|
| 248 |
+
"bos_token_id": null,
|
| 249 |
+
"chunk_size_feed_forward": 0,
|
| 250 |
+
"cross_attention_hidden_size": null,
|
| 251 |
+
"decoder_start_token_id": null,
|
| 252 |
+
"diversity_penalty": 0.0,
|
| 253 |
+
"do_sample": false,
|
| 254 |
+
"early_stopping": false,
|
| 255 |
+
"encoder_no_repeat_ngram_size": 0,
|
| 256 |
+
"eos_token_id": null,
|
| 257 |
+
"exponential_decay_length_penalty": null,
|
| 258 |
+
"finetuning_task": null,
|
| 259 |
+
"forced_bos_token_id": null,
|
| 260 |
+
"forced_eos_token_id": null,
|
| 261 |
+
"hidden_act": "gelu_pytorch_tanh",
|
| 262 |
+
"hidden_size": 1152,
|
| 263 |
+
"id2label": {
|
| 264 |
+
"0": "LABEL_0",
|
| 265 |
+
"1": "LABEL_1"
|
| 266 |
+
},
|
| 267 |
+
"image_size": 448,
|
| 268 |
+
"intermediate_size": 4304,
|
| 269 |
+
"is_decoder": false,
|
| 270 |
+
"is_encoder_decoder": false,
|
| 271 |
+
"label2id": {
|
| 272 |
+
"LABEL_0": 0,
|
| 273 |
+
"LABEL_1": 1
|
| 274 |
+
},
|
| 275 |
+
"layer_norm_eps": 1e-6,
|
| 276 |
+
"length_penalty": 1.0,
|
| 277 |
+
"max_length": 20,
|
| 278 |
+
"min_length": 0,
|
| 279 |
+
"model_type": "siglip_vision_model",
|
| 280 |
+
"no_repeat_ngram_size": 0,
|
| 281 |
+
"num_attention_heads": 16,
|
| 282 |
+
"num_beam_groups": 1,
|
| 283 |
+
"num_beams": 1,
|
| 284 |
+
"num_channels": 3,
|
| 285 |
+
"num_hidden_layers": 27,
|
| 286 |
+
"num_image_tokens": 256,
|
| 287 |
+
"num_return_sequences": 1,
|
| 288 |
+
"output_attentions": false,
|
| 289 |
+
"output_hidden_states": false,
|
| 290 |
+
"output_scores": false,
|
| 291 |
+
"pad_token_id": null,
|
| 292 |
+
"patch_size": 14,
|
| 293 |
+
"prefix": null,
|
| 294 |
+
"problem_type": null,
|
| 295 |
+
"projection_dim": 2048,
|
| 296 |
+
"projector_hidden_act": "gelu_fast",
|
| 297 |
+
"pruned_heads": {},
|
| 298 |
+
"remove_invalid_values": false,
|
| 299 |
+
"repetition_penalty": 1.0,
|
| 300 |
+
"return_dict": true,
|
| 301 |
+
"return_dict_in_generate": false,
|
| 302 |
+
"sep_token_id": null,
|
| 303 |
+
"suppress_tokens": null,
|
| 304 |
+
"task_specific_params": null,
|
| 305 |
+
"temperature": 1.0,
|
| 306 |
+
"tf_legacy_loss": false,
|
| 307 |
+
"tie_encoder_decoder": false,
|
| 308 |
+
"tie_word_embeddings": true,
|
| 309 |
+
"tokenizer_class": null,
|
| 310 |
+
"top_k": 50,
|
| 311 |
+
"top_p": 1.0,
|
| 312 |
+
"torch_dtype": "bfloat16",
|
| 313 |
+
"torchscript": false,
|
| 314 |
+
"typical_p": 1.0,
|
| 315 |
+
"use_bfloat16": false,
|
| 316 |
+
"vision_use_head": false
|
| 317 |
+
},
|
| 318 |
+
"vision_tower_lr": null,
|
| 319 |
+
"weight_memory_efficient": true
|
| 320 |
+
},
|
| 321 |
+
"projector_config": {
|
| 322 |
+
"architectures": [
|
| 323 |
+
"MultimodalProjector"
|
| 324 |
+
],
|
| 325 |
+
"mm_projector_type": "mlp_downsample_3x3_fix",
|
| 326 |
+
"model_type": "v2l_projector",
|
| 327 |
+
"torch_dtype": "bfloat16",
|
| 328 |
+
"transformers_version": "4.46.0"
|
| 329 |
+
},
|
| 330 |
+
"vision_config": {
|
| 331 |
+
"architectures": [
|
| 332 |
+
"SiglipVisionModel"
|
| 333 |
+
],
|
| 334 |
+
"attention_dropout": 0.0,
|
| 335 |
+
"hidden_act": "gelu_pytorch_tanh",
|
| 336 |
+
"hidden_size": 1152,
|
| 337 |
+
"image_size": 448,
|
| 338 |
+
"intermediate_size": 4304,
|
| 339 |
+
"layer_norm_eps": 1e-6,
|
| 340 |
+
"model_type": "siglip_vision_model",
|
| 341 |
+
"num_attention_heads": 16,
|
| 342 |
+
"num_channels": 3,
|
| 343 |
+
"num_hidden_layers": 27,
|
| 344 |
+
"num_image_tokens": 256,
|
| 345 |
+
"patch_size": 14,
|
| 346 |
+
"projection_dim": 2048,
|
| 347 |
+
"projector_hidden_act": "gelu_fast",
|
| 348 |
+
"torch_dtype": "bfloat16",
|
| 349 |
+
"transformers_version": "4.46.0",
|
| 350 |
+
"vision_use_head": false
|
| 351 |
+
},
|
| 352 |
+
"llm_config": {
|
| 353 |
+
"architectures": [
|
| 354 |
+
"Qwen2ForCausalLM"
|
| 355 |
+
],
|
| 356 |
+
"attention_dropout": 0.0,
|
| 357 |
+
"bos_token_id": 151643,
|
| 358 |
+
"eos_token_id": 151645,
|
| 359 |
+
"hidden_act": "silu",
|
| 360 |
+
"hidden_size": 1536,
|
| 361 |
+
"initializer_range": 0.02,
|
| 362 |
+
"intermediate_size": 8960,
|
| 363 |
+
"max_position_embeddings": 32768,
|
| 364 |
+
"max_window_layers": 28,
|
| 365 |
+
"model_max_length": 4096,
|
| 366 |
+
"model_type": "qwen2",
|
| 367 |
+
"num_attention_heads": 12,
|
| 368 |
+
"num_hidden_layers": 28,
|
| 369 |
+
"num_key_value_heads": 2,
|
| 370 |
+
"rms_norm_eps": 1e-6,
|
| 371 |
+
"rope_scaling": null,
|
| 372 |
+
"rope_theta": 1000000.0,
|
| 373 |
+
"sliding_window": null,
|
| 374 |
+
"tie_word_embeddings": true,
|
| 375 |
+
"tokenizer_model_max_length": 4096,
|
| 376 |
+
"tokenizer_padding_side": "right",
|
| 377 |
+
"torch_dtype": "bfloat16",
|
| 378 |
+
"transformers_version": "4.46.0",
|
| 379 |
+
"use_cache": false,
|
| 380 |
+
"use_sliding_window": false,
|
| 381 |
+
"vocab_size": 151656
|
| 382 |
+
}
|
| 383 |
+
}
|
action_model/model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:94e0bd1a46cc267556de401a2af410f79cabff38ff45b472bda532088e0837cb
|
| 3 |
+
size 4000400976
|
action_model/preprocessor_config.json
ADDED
|
@@ -0,0 +1,24 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"do_convert_rgb": null,
|
| 3 |
+
"do_normalize": true,
|
| 4 |
+
"do_rescale": true,
|
| 5 |
+
"do_resize": true,
|
| 6 |
+
"image_mean": [
|
| 7 |
+
0.5,
|
| 8 |
+
0.5,
|
| 9 |
+
0.5
|
| 10 |
+
],
|
| 11 |
+
"image_processor_type": "SiglipImageProcessor",
|
| 12 |
+
"image_std": [
|
| 13 |
+
0.5,
|
| 14 |
+
0.5,
|
| 15 |
+
0.5
|
| 16 |
+
],
|
| 17 |
+
"processor_class": "SiglipProcessor",
|
| 18 |
+
"resample": 3,
|
| 19 |
+
"rescale_factor": 0.00392156862745098,
|
| 20 |
+
"size": {
|
| 21 |
+
"height": 448,
|
| 22 |
+
"width": 448
|
| 23 |
+
}
|
| 24 |
+
}
|
action_model/special_tokens_map.json
ADDED
|
@@ -0,0 +1,49 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"additional_special_tokens": [
|
| 3 |
+
"<|im_start|>",
|
| 4 |
+
"<|im_end|>",
|
| 5 |
+
"<generated_image>",
|
| 6 |
+
{
|
| 7 |
+
"content": "<vila/sentinel>",
|
| 8 |
+
"lstrip": false,
|
| 9 |
+
"normalized": false,
|
| 10 |
+
"rstrip": false,
|
| 11 |
+
"single_word": false
|
| 12 |
+
},
|
| 13 |
+
{
|
| 14 |
+
"content": "<image>",
|
| 15 |
+
"lstrip": false,
|
| 16 |
+
"normalized": false,
|
| 17 |
+
"rstrip": false,
|
| 18 |
+
"single_word": false
|
| 19 |
+
},
|
| 20 |
+
{
|
| 21 |
+
"content": "<vila/video>",
|
| 22 |
+
"lstrip": false,
|
| 23 |
+
"normalized": false,
|
| 24 |
+
"rstrip": false,
|
| 25 |
+
"single_word": false
|
| 26 |
+
}
|
| 27 |
+
],
|
| 28 |
+
"bos_token": {
|
| 29 |
+
"content": "[BOS]",
|
| 30 |
+
"lstrip": false,
|
| 31 |
+
"normalized": false,
|
| 32 |
+
"rstrip": false,
|
| 33 |
+
"single_word": false
|
| 34 |
+
},
|
| 35 |
+
"eos_token": {
|
| 36 |
+
"content": "<|im_end|>",
|
| 37 |
+
"lstrip": false,
|
| 38 |
+
"normalized": false,
|
| 39 |
+
"rstrip": false,
|
| 40 |
+
"single_word": false
|
| 41 |
+
},
|
| 42 |
+
"pad_token": {
|
| 43 |
+
"content": "[PAD]",
|
| 44 |
+
"lstrip": false,
|
| 45 |
+
"normalized": false,
|
| 46 |
+
"rstrip": false,
|
| 47 |
+
"single_word": false
|
| 48 |
+
}
|
| 49 |
+
}
|
action_model/tokenizer_config.json
ADDED
|
@@ -0,0 +1,129 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"add_prefix_space": false,
|
| 3 |
+
"added_tokens_decoder": {
|
| 4 |
+
"151643": {
|
| 5 |
+
"content": "<|endoftext|>",
|
| 6 |
+
"lstrip": false,
|
| 7 |
+
"normalized": false,
|
| 8 |
+
"rstrip": false,
|
| 9 |
+
"single_word": false,
|
| 10 |
+
"special": true
|
| 11 |
+
},
|
| 12 |
+
"151644": {
|
| 13 |
+
"content": "<|im_start|>",
|
| 14 |
+
"lstrip": false,
|
| 15 |
+
"normalized": false,
|
| 16 |
+
"rstrip": false,
|
| 17 |
+
"single_word": false,
|
| 18 |
+
"special": true
|
| 19 |
+
},
|
| 20 |
+
"151645": {
|
| 21 |
+
"content": "<|im_end|>",
|
| 22 |
+
"lstrip": false,
|
| 23 |
+
"normalized": false,
|
| 24 |
+
"rstrip": false,
|
| 25 |
+
"single_word": false,
|
| 26 |
+
"special": true
|
| 27 |
+
},
|
| 28 |
+
"151646": {
|
| 29 |
+
"content": "[BOS]",
|
| 30 |
+
"lstrip": false,
|
| 31 |
+
"normalized": false,
|
| 32 |
+
"rstrip": false,
|
| 33 |
+
"single_word": false,
|
| 34 |
+
"special": true
|
| 35 |
+
},
|
| 36 |
+
"151647": {
|
| 37 |
+
"content": "[PAD]",
|
| 38 |
+
"lstrip": false,
|
| 39 |
+
"normalized": false,
|
| 40 |
+
"rstrip": false,
|
| 41 |
+
"single_word": false,
|
| 42 |
+
"special": true
|
| 43 |
+
},
|
| 44 |
+
"151648": {
|
| 45 |
+
"content": "<nav>",
|
| 46 |
+
"lstrip": false,
|
| 47 |
+
"normalized": false,
|
| 48 |
+
"rstrip": false,
|
| 49 |
+
"single_word": false,
|
| 50 |
+
"special": true
|
| 51 |
+
},
|
| 52 |
+
"151649": {
|
| 53 |
+
"content": "<act>",
|
| 54 |
+
"lstrip": false,
|
| 55 |
+
"normalized": false,
|
| 56 |
+
"rstrip": false,
|
| 57 |
+
"single_word": false,
|
| 58 |
+
"special": true
|
| 59 |
+
},
|
| 60 |
+
"151650": {
|
| 61 |
+
"content": "<his>",
|
| 62 |
+
"lstrip": false,
|
| 63 |
+
"normalized": false,
|
| 64 |
+
"rstrip": false,
|
| 65 |
+
"single_word": false,
|
| 66 |
+
"special": true
|
| 67 |
+
},
|
| 68 |
+
"151651": {
|
| 69 |
+
"content": "<obs>",
|
| 70 |
+
"lstrip": false,
|
| 71 |
+
"normalized": false,
|
| 72 |
+
"rstrip": false,
|
| 73 |
+
"single_word": false,
|
| 74 |
+
"special": true
|
| 75 |
+
},
|
| 76 |
+
"151652": {
|
| 77 |
+
"content": "<generated_image>",
|
| 78 |
+
"lstrip": false,
|
| 79 |
+
"normalized": false,
|
| 80 |
+
"rstrip": false,
|
| 81 |
+
"single_word": false,
|
| 82 |
+
"special": true
|
| 83 |
+
},
|
| 84 |
+
"151653": {
|
| 85 |
+
"content": "<vila/sentinel>",
|
| 86 |
+
"lstrip": false,
|
| 87 |
+
"normalized": false,
|
| 88 |
+
"rstrip": false,
|
| 89 |
+
"single_word": false,
|
| 90 |
+
"special": true
|
| 91 |
+
},
|
| 92 |
+
"151654": {
|
| 93 |
+
"content": "<image>",
|
| 94 |
+
"lstrip": false,
|
| 95 |
+
"normalized": false,
|
| 96 |
+
"rstrip": false,
|
| 97 |
+
"single_word": false,
|
| 98 |
+
"special": true
|
| 99 |
+
},
|
| 100 |
+
"151655": {
|
| 101 |
+
"content": "<vila/video>",
|
| 102 |
+
"lstrip": false,
|
| 103 |
+
"normalized": false,
|
| 104 |
+
"rstrip": false,
|
| 105 |
+
"single_word": false,
|
| 106 |
+
"special": true
|
| 107 |
+
}
|
| 108 |
+
},
|
| 109 |
+
"additional_special_tokens": [
|
| 110 |
+
"<|im_start|>",
|
| 111 |
+
"<|im_end|>",
|
| 112 |
+
"<generated_image>",
|
| 113 |
+
"<vila/sentinel>",
|
| 114 |
+
"<image>",
|
| 115 |
+
"<vila/video>"
|
| 116 |
+
],
|
| 117 |
+
"bos_token": "[BOS]",
|
| 118 |
+
"chat_template": "{% if messages[0]['role'] != 'system' %}{{ '<|im_start|>system\\nYou are a helpful assistant<|im_end|>\\n' }}{% endif %}{% for message in messages if message['content'] is not none %}{{ '<|im_start|>' + message['role'] + '\\n' + message['content'] + '<|im_end|>' + '\\n' }}{% endfor %}{% if add_generation_prompt %}{{ '<|im_start|>assistant\\n' }}{% endif %}",
|
| 119 |
+
"clean_up_tokenization_spaces": false,
|
| 120 |
+
"eos_token": "<|im_end|>",
|
| 121 |
+
"errors": "replace",
|
| 122 |
+
"legacy": false,
|
| 123 |
+
"model_max_length": 4096,
|
| 124 |
+
"pad_token": "[PAD]",
|
| 125 |
+
"padding_side": "right",
|
| 126 |
+
"split_special_tokens": false,
|
| 127 |
+
"tokenizer_class": "Qwen2Tokenizer",
|
| 128 |
+
"unk_token": null
|
| 129 |
+
}
|
action_model/vocab.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
progress_model/added_tokens.json
ADDED
|
@@ -0,0 +1,15 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"<act>": 151649,
|
| 3 |
+
"<generated_image>": 151652,
|
| 4 |
+
"<his>": 151650,
|
| 5 |
+
"<image>": 151654,
|
| 6 |
+
"<nav>": 151648,
|
| 7 |
+
"<obs>": 151651,
|
| 8 |
+
"<vila/sentinel>": 151653,
|
| 9 |
+
"<vila/video>": 151655,
|
| 10 |
+
"<|endoftext|>": 151643,
|
| 11 |
+
"<|im_end|>": 151645,
|
| 12 |
+
"<|im_start|>": 151644,
|
| 13 |
+
"[BOS]": 151646,
|
| 14 |
+
"[PAD]": 151647
|
| 15 |
+
}
|
progress_model/config.json
ADDED
|
@@ -0,0 +1,310 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"Ubit": 100,
|
| 3 |
+
"_attn_implementation_autoset": true,
|
| 4 |
+
"architectures": [
|
| 5 |
+
"LlavaLlamaModel"
|
| 6 |
+
],
|
| 7 |
+
"babit": "E5M2",
|
| 8 |
+
"bobit": "E5M2",
|
| 9 |
+
"bwbit": "E5M2",
|
| 10 |
+
"chat_template": null,
|
| 11 |
+
"col_blocksize": -1,
|
| 12 |
+
"col_blocksize_optimizer": 128,
|
| 13 |
+
"draw_distribution_backward": false,
|
| 14 |
+
"draw_distribution_forward": false,
|
| 15 |
+
"drop_path_rate": 0.0,
|
| 16 |
+
"dynamic_s2": false,
|
| 17 |
+
"epsilon": 1e-10,
|
| 18 |
+
"epsilon_optimizer": 1e-15,
|
| 19 |
+
"fabit": "E4M3",
|
| 20 |
+
"first_order_bit": null,
|
| 21 |
+
"first_order_quant_type": null,
|
| 22 |
+
"fobit": "E4M3",
|
| 23 |
+
"fps": 0.0,
|
| 24 |
+
"fwbit": "E4M3",
|
| 25 |
+
"group_size": -1,
|
| 26 |
+
"hidden_size": 1536,
|
| 27 |
+
"image_aspect_ratio": "dynamic",
|
| 28 |
+
"interpolate_mode": "linear",
|
| 29 |
+
"llm_cfg": {
|
| 30 |
+
"_attn_implementation_autoset": false,
|
| 31 |
+
"add_cross_attention": false,
|
| 32 |
+
"architectures": [
|
| 33 |
+
"Qwen2ForCausalLM"
|
| 34 |
+
],
|
| 35 |
+
"attention_dropout": 0.0,
|
| 36 |
+
"bad_words_ids": null,
|
| 37 |
+
"begin_suppress_tokens": null,
|
| 38 |
+
"bos_token_id": 151643,
|
| 39 |
+
"chunk_size_feed_forward": 0,
|
| 40 |
+
"cross_attention_hidden_size": null,
|
| 41 |
+
"decoder_start_token_id": null,
|
| 42 |
+
"diversity_penalty": 0.0,
|
| 43 |
+
"do_sample": false,
|
| 44 |
+
"early_stopping": false,
|
| 45 |
+
"encoder_no_repeat_ngram_size": 0,
|
| 46 |
+
"eos_token_id": 151645,
|
| 47 |
+
"exponential_decay_length_penalty": null,
|
| 48 |
+
"finetuning_task": null,
|
| 49 |
+
"forced_bos_token_id": null,
|
| 50 |
+
"forced_eos_token_id": null,
|
| 51 |
+
"hidden_act": "silu",
|
| 52 |
+
"hidden_size": 1536,
|
| 53 |
+
"id2label": {
|
| 54 |
+
"0": "LABEL_0",
|
| 55 |
+
"1": "LABEL_1"
|
| 56 |
+
},
|
| 57 |
+
"initializer_range": 0.02,
|
| 58 |
+
"intermediate_size": 8960,
|
| 59 |
+
"is_decoder": false,
|
| 60 |
+
"is_encoder_decoder": false,
|
| 61 |
+
"label2id": {
|
| 62 |
+
"LABEL_0": 0,
|
| 63 |
+
"LABEL_1": 1
|
| 64 |
+
},
|
| 65 |
+
"length_penalty": 1.0,
|
| 66 |
+
"max_length": 20,
|
| 67 |
+
"max_position_embeddings": 32768,
|
| 68 |
+
"max_window_layers": 28,
|
| 69 |
+
"min_length": 0,
|
| 70 |
+
"model_max_length": 4096,
|
| 71 |
+
"model_type": "qwen2",
|
| 72 |
+
"no_repeat_ngram_size": 0,
|
| 73 |
+
"num_attention_heads": 12,
|
| 74 |
+
"num_beam_groups": 1,
|
| 75 |
+
"num_beams": 1,
|
| 76 |
+
"num_hidden_layers": 28,
|
| 77 |
+
"num_key_value_heads": 2,
|
| 78 |
+
"num_return_sequences": 1,
|
| 79 |
+
"output_attentions": false,
|
| 80 |
+
"output_hidden_states": false,
|
| 81 |
+
"output_scores": false,
|
| 82 |
+
"pad_token_id": null,
|
| 83 |
+
"prefix": null,
|
| 84 |
+
"problem_type": null,
|
| 85 |
+
"pruned_heads": {},
|
| 86 |
+
"remove_invalid_values": false,
|
| 87 |
+
"repetition_penalty": 1.0,
|
| 88 |
+
"return_dict": true,
|
| 89 |
+
"return_dict_in_generate": false,
|
| 90 |
+
"rms_norm_eps": 1e-06,
|
| 91 |
+
"rope_scaling": null,
|
| 92 |
+
"rope_theta": 1000000.0,
|
| 93 |
+
"sep_token_id": null,
|
| 94 |
+
"sliding_window": null,
|
| 95 |
+
"suppress_tokens": null,
|
| 96 |
+
"task_specific_params": null,
|
| 97 |
+
"temperature": 1.0,
|
| 98 |
+
"tf_legacy_loss": false,
|
| 99 |
+
"tie_encoder_decoder": false,
|
| 100 |
+
"tie_word_embeddings": true,
|
| 101 |
+
"tokenizer_class": null,
|
| 102 |
+
"tokenizer_model_max_length": 4096,
|
| 103 |
+
"tokenizer_padding_side": "right",
|
| 104 |
+
"top_k": 50,
|
| 105 |
+
"top_p": 1.0,
|
| 106 |
+
"torch_dtype": "bfloat16",
|
| 107 |
+
"torchscript": false,
|
| 108 |
+
"typical_p": 1.0,
|
| 109 |
+
"use_bfloat16": false,
|
| 110 |
+
"use_cache": false,
|
| 111 |
+
"use_sliding_window": false,
|
| 112 |
+
"vocab_size": 151651
|
| 113 |
+
},
|
| 114 |
+
"max_tiles": 12,
|
| 115 |
+
"min_blockunit_col": 4,
|
| 116 |
+
"min_blockunit_row": 4,
|
| 117 |
+
"min_tiles": 1,
|
| 118 |
+
"mlp_path": null,
|
| 119 |
+
"mm_hidden_size": 1152,
|
| 120 |
+
"mm_projector": "mlp_downsample_3x3_fix",
|
| 121 |
+
"mm_projector_cfg": {
|
| 122 |
+
"_attn_implementation_autoset": false,
|
| 123 |
+
"add_cross_attention": false,
|
| 124 |
+
"architectures": [
|
| 125 |
+
"MultimodalProjector"
|
| 126 |
+
],
|
| 127 |
+
"bad_words_ids": null,
|
| 128 |
+
"begin_suppress_tokens": null,
|
| 129 |
+
"bos_token_id": null,
|
| 130 |
+
"chunk_size_feed_forward": 0,
|
| 131 |
+
"cross_attention_hidden_size": null,
|
| 132 |
+
"decoder_start_token_id": null,
|
| 133 |
+
"diversity_penalty": 0.0,
|
| 134 |
+
"do_sample": false,
|
| 135 |
+
"early_stopping": false,
|
| 136 |
+
"encoder_no_repeat_ngram_size": 0,
|
| 137 |
+
"eos_token_id": null,
|
| 138 |
+
"exponential_decay_length_penalty": null,
|
| 139 |
+
"finetuning_task": null,
|
| 140 |
+
"forced_bos_token_id": null,
|
| 141 |
+
"forced_eos_token_id": null,
|
| 142 |
+
"id2label": {
|
| 143 |
+
"0": "LABEL_0",
|
| 144 |
+
"1": "LABEL_1"
|
| 145 |
+
},
|
| 146 |
+
"is_decoder": false,
|
| 147 |
+
"is_encoder_decoder": false,
|
| 148 |
+
"label2id": {
|
| 149 |
+
"LABEL_0": 0,
|
| 150 |
+
"LABEL_1": 1
|
| 151 |
+
},
|
| 152 |
+
"length_penalty": 1.0,
|
| 153 |
+
"max_length": 20,
|
| 154 |
+
"min_length": 0,
|
| 155 |
+
"mm_projector_type": "mlp_downsample_3x3_fix",
|
| 156 |
+
"model_type": "v2l_projector",
|
| 157 |
+
"no_repeat_ngram_size": 0,
|
| 158 |
+
"num_beam_groups": 1,
|
| 159 |
+
"num_beams": 1,
|
| 160 |
+
"num_return_sequences": 1,
|
| 161 |
+
"output_attentions": false,
|
| 162 |
+
"output_hidden_states": false,
|
| 163 |
+
"output_scores": false,
|
| 164 |
+
"pad_token_id": null,
|
| 165 |
+
"prefix": null,
|
| 166 |
+
"problem_type": null,
|
| 167 |
+
"pruned_heads": {},
|
| 168 |
+
"remove_invalid_values": false,
|
| 169 |
+
"repetition_penalty": 1.0,
|
| 170 |
+
"return_dict": true,
|
| 171 |
+
"return_dict_in_generate": false,
|
| 172 |
+
"sep_token_id": null,
|
| 173 |
+
"suppress_tokens": null,
|
| 174 |
+
"task_specific_params": null,
|
| 175 |
+
"temperature": 1.0,
|
| 176 |
+
"tf_legacy_loss": false,
|
| 177 |
+
"tie_encoder_decoder": false,
|
| 178 |
+
"tie_word_embeddings": true,
|
| 179 |
+
"tokenizer_class": null,
|
| 180 |
+
"top_k": 50,
|
| 181 |
+
"top_p": 1.0,
|
| 182 |
+
"torch_dtype": "bfloat16",
|
| 183 |
+
"torchscript": false,
|
| 184 |
+
"typical_p": 1.0,
|
| 185 |
+
"use_bfloat16": false
|
| 186 |
+
},
|
| 187 |
+
"mm_projector_lr": null,
|
| 188 |
+
"mm_use_im_patch_token": false,
|
| 189 |
+
"mm_use_im_start_end": false,
|
| 190 |
+
"mm_vision_select_feature": "cls_patch",
|
| 191 |
+
"mm_vision_select_layer": -2,
|
| 192 |
+
"model_dtype": "torch.bfloat16",
|
| 193 |
+
"model_type": "llava_llama",
|
| 194 |
+
"num_time_tokens": 0,
|
| 195 |
+
"num_video_frames": 8,
|
| 196 |
+
"pad_block": false,
|
| 197 |
+
"pad_to_multiple_of": 0,
|
| 198 |
+
"qchoice": "none",
|
| 199 |
+
"quantize_model": false,
|
| 200 |
+
"refine_attn_blocksize": false,
|
| 201 |
+
"refine_col_blocksize": 4,
|
| 202 |
+
"refine_ln_blocksize": false,
|
| 203 |
+
"refine_ln_blocksize_but_only_backward": false,
|
| 204 |
+
"refine_ln_blocksize_but_only_forward": false,
|
| 205 |
+
"refine_ln_pertoken": false,
|
| 206 |
+
"refine_mlp_blocksize": false,
|
| 207 |
+
"refine_residual_fp": false,
|
| 208 |
+
"refine_row_blocksize": 4,
|
| 209 |
+
"row_blocksize": -1,
|
| 210 |
+
"row_blocksize_optimizer": 1,
|
| 211 |
+
"s2": false,
|
| 212 |
+
"s2_max_split_size": 336,
|
| 213 |
+
"s2_resize_output_to_scale_idx": 0,
|
| 214 |
+
"s2_scales": "336,672,1008",
|
| 215 |
+
"second_order_bit": null,
|
| 216 |
+
"second_order_quant_type": null,
|
| 217 |
+
"soft_ce_std": 1.0,
|
| 218 |
+
"symm": true,
|
| 219 |
+
"time_token_format": "<t{t}>",
|
| 220 |
+
"time_token_ids": [],
|
| 221 |
+
"transformers_version": "4.46.0",
|
| 222 |
+
"tune_language_model": true,
|
| 223 |
+
"tune_mm_projector": true,
|
| 224 |
+
"tune_vision_tower": true,
|
| 225 |
+
"use_quantize_optimizer": false,
|
| 226 |
+
"version": "auto",
|
| 227 |
+
"vision_resolution": -1,
|
| 228 |
+
"vision_tower": "Efficient-Large-Model/paligemma-siglip-so400m-patch14-448",
|
| 229 |
+
"vision_tower_cfg": {
|
| 230 |
+
"_attn_implementation_autoset": false,
|
| 231 |
+
"add_cross_attention": false,
|
| 232 |
+
"architectures": [
|
| 233 |
+
"SiglipVisionModel"
|
| 234 |
+
],
|
| 235 |
+
"attention_dropout": 0.0,
|
| 236 |
+
"bad_words_ids": null,
|
| 237 |
+
"begin_suppress_tokens": null,
|
| 238 |
+
"bos_token_id": null,
|
| 239 |
+
"chunk_size_feed_forward": 0,
|
| 240 |
+
"cross_attention_hidden_size": null,
|
| 241 |
+
"decoder_start_token_id": null,
|
| 242 |
+
"diversity_penalty": 0.0,
|
| 243 |
+
"do_sample": false,
|
| 244 |
+
"early_stopping": false,
|
| 245 |
+
"encoder_no_repeat_ngram_size": 0,
|
| 246 |
+
"eos_token_id": null,
|
| 247 |
+
"exponential_decay_length_penalty": null,
|
| 248 |
+
"finetuning_task": null,
|
| 249 |
+
"forced_bos_token_id": null,
|
| 250 |
+
"forced_eos_token_id": null,
|
| 251 |
+
"hidden_act": "gelu_pytorch_tanh",
|
| 252 |
+
"hidden_size": 1152,
|
| 253 |
+
"id2label": {
|
| 254 |
+
"0": "LABEL_0",
|
| 255 |
+
"1": "LABEL_1"
|
| 256 |
+
},
|
| 257 |
+
"image_size": 448,
|
| 258 |
+
"intermediate_size": 4304,
|
| 259 |
+
"is_decoder": false,
|
| 260 |
+
"is_encoder_decoder": false,
|
| 261 |
+
"label2id": {
|
| 262 |
+
"LABEL_0": 0,
|
| 263 |
+
"LABEL_1": 1
|
| 264 |
+
},
|
| 265 |
+
"layer_norm_eps": 1e-06,
|
| 266 |
+
"length_penalty": 1.0,
|
| 267 |
+
"max_length": 20,
|
| 268 |
+
"min_length": 0,
|
| 269 |
+
"model_type": "siglip_vision_model",
|
| 270 |
+
"no_repeat_ngram_size": 0,
|
| 271 |
+
"num_attention_heads": 16,
|
| 272 |
+
"num_beam_groups": 1,
|
| 273 |
+
"num_beams": 1,
|
| 274 |
+
"num_channels": 3,
|
| 275 |
+
"num_hidden_layers": 27,
|
| 276 |
+
"num_image_tokens": 256,
|
| 277 |
+
"num_return_sequences": 1,
|
| 278 |
+
"output_attentions": false,
|
| 279 |
+
"output_hidden_states": false,
|
| 280 |
+
"output_scores": false,
|
| 281 |
+
"pad_token_id": null,
|
| 282 |
+
"patch_size": 14,
|
| 283 |
+
"prefix": null,
|
| 284 |
+
"problem_type": null,
|
| 285 |
+
"projection_dim": 2048,
|
| 286 |
+
"projector_hidden_act": "gelu_fast",
|
| 287 |
+
"pruned_heads": {},
|
| 288 |
+
"remove_invalid_values": false,
|
| 289 |
+
"repetition_penalty": 1.0,
|
| 290 |
+
"return_dict": true,
|
| 291 |
+
"return_dict_in_generate": false,
|
| 292 |
+
"sep_token_id": null,
|
| 293 |
+
"suppress_tokens": null,
|
| 294 |
+
"task_specific_params": null,
|
| 295 |
+
"temperature": 1.0,
|
| 296 |
+
"tf_legacy_loss": false,
|
| 297 |
+
"tie_encoder_decoder": false,
|
| 298 |
+
"tie_word_embeddings": true,
|
| 299 |
+
"tokenizer_class": null,
|
| 300 |
+
"top_k": 50,
|
| 301 |
+
"top_p": 1.0,
|
| 302 |
+
"torch_dtype": "bfloat16",
|
| 303 |
+
"torchscript": false,
|
| 304 |
+
"typical_p": 1.0,
|
| 305 |
+
"use_bfloat16": false,
|
| 306 |
+
"vision_use_head": false
|
| 307 |
+
},
|
| 308 |
+
"vision_tower_lr": null,
|
| 309 |
+
"weight_memory_efficient": true
|
| 310 |
+
}
|
progress_model/generation_config.json
ADDED
|
@@ -0,0 +1,14 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"bos_token_id": 151643,
|
| 3 |
+
"do_sample": true,
|
| 4 |
+
"eos_token_id": [
|
| 5 |
+
151645,
|
| 6 |
+
151643
|
| 7 |
+
],
|
| 8 |
+
"pad_token_id": 151643,
|
| 9 |
+
"repetition_penalty": 1.1,
|
| 10 |
+
"temperature": 0.7,
|
| 11 |
+
"top_k": 20,
|
| 12 |
+
"top_p": 0.8,
|
| 13 |
+
"transformers_version": "4.46.0"
|
| 14 |
+
}
|
progress_model/inference.config.json
ADDED
|
@@ -0,0 +1,13 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"__config_type__": "robo_orchard_lab.inference.basic:InferencePipelineCfg",
|
| 3 |
+
"class_type": "robo_orchard_lab.inference.basic:InferencePipeline",
|
| 4 |
+
"model_cfg": null,
|
| 5 |
+
"processor": {
|
| 6 |
+
"__config_type__": "robo_orchard_lab.models.progress_think.progress_processor:ProgressModelProcessorCfg",
|
| 7 |
+
"class_type": "robo_orchard_lab.models.progress_think.progress_processor:ProgressModelProcessor",
|
| 8 |
+
"load_image": true,
|
| 9 |
+
"transforms": null
|
| 10 |
+
},
|
| 11 |
+
"collate_fn": null,
|
| 12 |
+
"batch_size": 1
|
| 13 |
+
}
|
progress_model/merges.txt
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
progress_model/model.config.json
ADDED
|
@@ -0,0 +1,383 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"__config_type__": "robo_orchard_lab.models.progress_think.progress_model:ProgressModelConfig",
|
| 3 |
+
"class_type": "robo_orchard_lab.models.progress_think.progress_model:ProgressModel",
|
| 4 |
+
"model_dtype": "torch.float16",
|
| 5 |
+
"llava_llama_config": {
|
| 6 |
+
"Ubit": 100,
|
| 7 |
+
"_attn_implementation_autoset": true,
|
| 8 |
+
"architectures": [
|
| 9 |
+
"LlavaLlamaModel"
|
| 10 |
+
],
|
| 11 |
+
"babit": "E5M2",
|
| 12 |
+
"bobit": "E5M2",
|
| 13 |
+
"bwbit": "E5M2",
|
| 14 |
+
"chat_template": null,
|
| 15 |
+
"col_blocksize": -1,
|
| 16 |
+
"col_blocksize_optimizer": 128,
|
| 17 |
+
"draw_distribution_backward": false,
|
| 18 |
+
"draw_distribution_forward": false,
|
| 19 |
+
"drop_path_rate": 0.0,
|
| 20 |
+
"dynamic_s2": false,
|
| 21 |
+
"epsilon": 1e-10,
|
| 22 |
+
"epsilon_optimizer": 1e-15,
|
| 23 |
+
"fabit": "E4M3",
|
| 24 |
+
"first_order_bit": null,
|
| 25 |
+
"first_order_quant_type": null,
|
| 26 |
+
"fobit": "E4M3",
|
| 27 |
+
"fps": 0.0,
|
| 28 |
+
"fwbit": "E4M3",
|
| 29 |
+
"group_size": -1,
|
| 30 |
+
"hidden_size": 1536,
|
| 31 |
+
"image_aspect_ratio": "dynamic",
|
| 32 |
+
"image_encoder": {
|
| 33 |
+
"_target_": "robo_orchard_lab.models.monodream.multimodal_encoder.BasicImageEncoder"
|
| 34 |
+
},
|
| 35 |
+
"interpolate_mode": "linear",
|
| 36 |
+
"llm_cfg": {
|
| 37 |
+
"_attn_implementation_autoset": false,
|
| 38 |
+
"add_cross_attention": false,
|
| 39 |
+
"architectures": [
|
| 40 |
+
"Qwen2ForCausalLM"
|
| 41 |
+
],
|
| 42 |
+
"attention_dropout": 0.0,
|
| 43 |
+
"bad_words_ids": null,
|
| 44 |
+
"begin_suppress_tokens": null,
|
| 45 |
+
"bos_token_id": 151643,
|
| 46 |
+
"chunk_size_feed_forward": 0,
|
| 47 |
+
"cross_attention_hidden_size": null,
|
| 48 |
+
"decoder_start_token_id": null,
|
| 49 |
+
"diversity_penalty": 0.0,
|
| 50 |
+
"do_sample": false,
|
| 51 |
+
"early_stopping": false,
|
| 52 |
+
"encoder_no_repeat_ngram_size": 0,
|
| 53 |
+
"eos_token_id": 151645,
|
| 54 |
+
"exponential_decay_length_penalty": null,
|
| 55 |
+
"finetuning_task": null,
|
| 56 |
+
"forced_bos_token_id": null,
|
| 57 |
+
"forced_eos_token_id": null,
|
| 58 |
+
"hidden_act": "silu",
|
| 59 |
+
"hidden_size": 1536,
|
| 60 |
+
"id2label": {
|
| 61 |
+
"0": "LABEL_0",
|
| 62 |
+
"1": "LABEL_1"
|
| 63 |
+
},
|
| 64 |
+
"initializer_range": 0.02,
|
| 65 |
+
"intermediate_size": 8960,
|
| 66 |
+
"is_decoder": false,
|
| 67 |
+
"is_encoder_decoder": false,
|
| 68 |
+
"label2id": {
|
| 69 |
+
"LABEL_0": 0,
|
| 70 |
+
"LABEL_1": 1
|
| 71 |
+
},
|
| 72 |
+
"length_penalty": 1.0,
|
| 73 |
+
"max_length": 20,
|
| 74 |
+
"max_position_embeddings": 32768,
|
| 75 |
+
"max_window_layers": 28,
|
| 76 |
+
"min_length": 0,
|
| 77 |
+
"model_max_length": 4096,
|
| 78 |
+
"model_type": "qwen2",
|
| 79 |
+
"no_repeat_ngram_size": 0,
|
| 80 |
+
"num_attention_heads": 12,
|
| 81 |
+
"num_beam_groups": 1,
|
| 82 |
+
"num_beams": 1,
|
| 83 |
+
"num_hidden_layers": 28,
|
| 84 |
+
"num_key_value_heads": 2,
|
| 85 |
+
"num_return_sequences": 1,
|
| 86 |
+
"output_attentions": false,
|
| 87 |
+
"output_hidden_states": false,
|
| 88 |
+
"output_scores": false,
|
| 89 |
+
"pad_token_id": null,
|
| 90 |
+
"prefix": null,
|
| 91 |
+
"problem_type": null,
|
| 92 |
+
"pruned_heads": {},
|
| 93 |
+
"remove_invalid_values": false,
|
| 94 |
+
"repetition_penalty": 1.0,
|
| 95 |
+
"return_dict": true,
|
| 96 |
+
"return_dict_in_generate": false,
|
| 97 |
+
"rms_norm_eps": 1e-6,
|
| 98 |
+
"rope_scaling": null,
|
| 99 |
+
"rope_theta": 1000000.0,
|
| 100 |
+
"sep_token_id": null,
|
| 101 |
+
"sliding_window": null,
|
| 102 |
+
"suppress_tokens": null,
|
| 103 |
+
"task_specific_params": null,
|
| 104 |
+
"temperature": 1.0,
|
| 105 |
+
"tf_legacy_loss": false,
|
| 106 |
+
"tie_encoder_decoder": false,
|
| 107 |
+
"tie_word_embeddings": true,
|
| 108 |
+
"tokenizer_class": null,
|
| 109 |
+
"tokenizer_model_max_length": 4096,
|
| 110 |
+
"tokenizer_padding_side": "right",
|
| 111 |
+
"top_k": 50,
|
| 112 |
+
"top_p": 1.0,
|
| 113 |
+
"torch_dtype": "bfloat16",
|
| 114 |
+
"torchscript": false,
|
| 115 |
+
"typical_p": 1.0,
|
| 116 |
+
"use_bfloat16": false,
|
| 117 |
+
"use_cache": false,
|
| 118 |
+
"use_sliding_window": false,
|
| 119 |
+
"vocab_size": 151656
|
| 120 |
+
},
|
| 121 |
+
"max_tiles": 12,
|
| 122 |
+
"min_blockunit_col": 4,
|
| 123 |
+
"min_blockunit_row": 4,
|
| 124 |
+
"min_tiles": 1,
|
| 125 |
+
"mlp_path": null,
|
| 126 |
+
"mm_hidden_size": 1152,
|
| 127 |
+
"mm_projector": "mlp_downsample_3x3_fix",
|
| 128 |
+
"mm_projector_cfg": {
|
| 129 |
+
"_attn_implementation_autoset": false,
|
| 130 |
+
"add_cross_attention": false,
|
| 131 |
+
"architectures": [
|
| 132 |
+
"MultimodalProjector"
|
| 133 |
+
],
|
| 134 |
+
"bad_words_ids": null,
|
| 135 |
+
"begin_suppress_tokens": null,
|
| 136 |
+
"bos_token_id": null,
|
| 137 |
+
"chunk_size_feed_forward": 0,
|
| 138 |
+
"cross_attention_hidden_size": null,
|
| 139 |
+
"decoder_start_token_id": null,
|
| 140 |
+
"diversity_penalty": 0.0,
|
| 141 |
+
"do_sample": false,
|
| 142 |
+
"early_stopping": false,
|
| 143 |
+
"encoder_no_repeat_ngram_size": 0,
|
| 144 |
+
"eos_token_id": null,
|
| 145 |
+
"exponential_decay_length_penalty": null,
|
| 146 |
+
"finetuning_task": null,
|
| 147 |
+
"forced_bos_token_id": null,
|
| 148 |
+
"forced_eos_token_id": null,
|
| 149 |
+
"id2label": {
|
| 150 |
+
"0": "LABEL_0",
|
| 151 |
+
"1": "LABEL_1"
|
| 152 |
+
},
|
| 153 |
+
"is_decoder": false,
|
| 154 |
+
"is_encoder_decoder": false,
|
| 155 |
+
"label2id": {
|
| 156 |
+
"LABEL_0": 0,
|
| 157 |
+
"LABEL_1": 1
|
| 158 |
+
},
|
| 159 |
+
"length_penalty": 1.0,
|
| 160 |
+
"max_length": 20,
|
| 161 |
+
"min_length": 0,
|
| 162 |
+
"mm_projector_type": "mlp_downsample_3x3_fix",
|
| 163 |
+
"model_type": "v2l_projector",
|
| 164 |
+
"no_repeat_ngram_size": 0,
|
| 165 |
+
"num_beam_groups": 1,
|
| 166 |
+
"num_beams": 1,
|
| 167 |
+
"num_return_sequences": 1,
|
| 168 |
+
"output_attentions": false,
|
| 169 |
+
"output_hidden_states": false,
|
| 170 |
+
"output_scores": false,
|
| 171 |
+
"pad_token_id": null,
|
| 172 |
+
"prefix": null,
|
| 173 |
+
"problem_type": null,
|
| 174 |
+
"pruned_heads": {},
|
| 175 |
+
"remove_invalid_values": false,
|
| 176 |
+
"repetition_penalty": 1.0,
|
| 177 |
+
"return_dict": true,
|
| 178 |
+
"return_dict_in_generate": false,
|
| 179 |
+
"sep_token_id": null,
|
| 180 |
+
"suppress_tokens": null,
|
| 181 |
+
"task_specific_params": null,
|
| 182 |
+
"temperature": 1.0,
|
| 183 |
+
"tf_legacy_loss": false,
|
| 184 |
+
"tie_encoder_decoder": false,
|
| 185 |
+
"tie_word_embeddings": true,
|
| 186 |
+
"tokenizer_class": null,
|
| 187 |
+
"top_k": 50,
|
| 188 |
+
"top_p": 1.0,
|
| 189 |
+
"torch_dtype": "bfloat16",
|
| 190 |
+
"torchscript": false,
|
| 191 |
+
"typical_p": 1.0,
|
| 192 |
+
"use_bfloat16": false
|
| 193 |
+
},
|
| 194 |
+
"mm_projector_lr": null,
|
| 195 |
+
"mm_use_im_patch_token": false,
|
| 196 |
+
"mm_use_im_start_end": false,
|
| 197 |
+
"mm_vision_select_feature": "cls_patch",
|
| 198 |
+
"mm_vision_select_layer": -2,
|
| 199 |
+
"model_dtype": "torch.bfloat16",
|
| 200 |
+
"model_type": "llava_llama",
|
| 201 |
+
"num_time_tokens": 0,
|
| 202 |
+
"num_video_frames": 8,
|
| 203 |
+
"pad_block": false,
|
| 204 |
+
"pad_to_multiple_of": 0,
|
| 205 |
+
"qchoice": "none",
|
| 206 |
+
"quantize_model": false,
|
| 207 |
+
"refine_attn_blocksize": false,
|
| 208 |
+
"refine_col_blocksize": 4,
|
| 209 |
+
"refine_ln_blocksize": false,
|
| 210 |
+
"refine_ln_blocksize_but_only_backward": false,
|
| 211 |
+
"refine_ln_blocksize_but_only_forward": false,
|
| 212 |
+
"refine_ln_pertoken": false,
|
| 213 |
+
"refine_mlp_blocksize": false,
|
| 214 |
+
"refine_residual_fp": false,
|
| 215 |
+
"refine_row_blocksize": 4,
|
| 216 |
+
"row_blocksize": -1,
|
| 217 |
+
"row_blocksize_optimizer": 1,
|
| 218 |
+
"s2": false,
|
| 219 |
+
"s2_max_split_size": 336,
|
| 220 |
+
"s2_resize_output_to_scale_idx": 0,
|
| 221 |
+
"s2_scales": "336,672,1008",
|
| 222 |
+
"second_order_bit": null,
|
| 223 |
+
"second_order_quant_type": null,
|
| 224 |
+
"soft_ce_std": 1.0,
|
| 225 |
+
"symm": true,
|
| 226 |
+
"time_token_format": "<t{t}>",
|
| 227 |
+
"time_token_ids": [],
|
| 228 |
+
"transformers_version": "4.46.0",
|
| 229 |
+
"tune_language_model": true,
|
| 230 |
+
"tune_mm_projector": true,
|
| 231 |
+
"tune_vision_tower": true,
|
| 232 |
+
"use_quantize_optimizer": false,
|
| 233 |
+
"version": "auto",
|
| 234 |
+
"video_encoder": {
|
| 235 |
+
"_target_": "robo_orchard_lab.models.monodream.multimodal_encoder.BasicVideoEncoder"
|
| 236 |
+
},
|
| 237 |
+
"vision_resolution": -1,
|
| 238 |
+
"vision_tower": "Efficient-Large-Model/paligemma-siglip-so400m-patch14-448",
|
| 239 |
+
"vision_tower_cfg": {
|
| 240 |
+
"_attn_implementation_autoset": false,
|
| 241 |
+
"add_cross_attention": false,
|
| 242 |
+
"architectures": [
|
| 243 |
+
"SiglipVisionModel"
|
| 244 |
+
],
|
| 245 |
+
"attention_dropout": 0.0,
|
| 246 |
+
"bad_words_ids": null,
|
| 247 |
+
"begin_suppress_tokens": null,
|
| 248 |
+
"bos_token_id": null,
|
| 249 |
+
"chunk_size_feed_forward": 0,
|
| 250 |
+
"cross_attention_hidden_size": null,
|
| 251 |
+
"decoder_start_token_id": null,
|
| 252 |
+
"diversity_penalty": 0.0,
|
| 253 |
+
"do_sample": false,
|
| 254 |
+
"early_stopping": false,
|
| 255 |
+
"encoder_no_repeat_ngram_size": 0,
|
| 256 |
+
"eos_token_id": null,
|
| 257 |
+
"exponential_decay_length_penalty": null,
|
| 258 |
+
"finetuning_task": null,
|
| 259 |
+
"forced_bos_token_id": null,
|
| 260 |
+
"forced_eos_token_id": null,
|
| 261 |
+
"hidden_act": "gelu_pytorch_tanh",
|
| 262 |
+
"hidden_size": 1152,
|
| 263 |
+
"id2label": {
|
| 264 |
+
"0": "LABEL_0",
|
| 265 |
+
"1": "LABEL_1"
|
| 266 |
+
},
|
| 267 |
+
"image_size": 448,
|
| 268 |
+
"intermediate_size": 4304,
|
| 269 |
+
"is_decoder": false,
|
| 270 |
+
"is_encoder_decoder": false,
|
| 271 |
+
"label2id": {
|
| 272 |
+
"LABEL_0": 0,
|
| 273 |
+
"LABEL_1": 1
|
| 274 |
+
},
|
| 275 |
+
"layer_norm_eps": 1e-6,
|
| 276 |
+
"length_penalty": 1.0,
|
| 277 |
+
"max_length": 20,
|
| 278 |
+
"min_length": 0,
|
| 279 |
+
"model_type": "siglip_vision_model",
|
| 280 |
+
"no_repeat_ngram_size": 0,
|
| 281 |
+
"num_attention_heads": 16,
|
| 282 |
+
"num_beam_groups": 1,
|
| 283 |
+
"num_beams": 1,
|
| 284 |
+
"num_channels": 3,
|
| 285 |
+
"num_hidden_layers": 27,
|
| 286 |
+
"num_image_tokens": 256,
|
| 287 |
+
"num_return_sequences": 1,
|
| 288 |
+
"output_attentions": false,
|
| 289 |
+
"output_hidden_states": false,
|
| 290 |
+
"output_scores": false,
|
| 291 |
+
"pad_token_id": null,
|
| 292 |
+
"patch_size": 14,
|
| 293 |
+
"prefix": null,
|
| 294 |
+
"problem_type": null,
|
| 295 |
+
"projection_dim": 2048,
|
| 296 |
+
"projector_hidden_act": "gelu_fast",
|
| 297 |
+
"pruned_heads": {},
|
| 298 |
+
"remove_invalid_values": false,
|
| 299 |
+
"repetition_penalty": 1.0,
|
| 300 |
+
"return_dict": true,
|
| 301 |
+
"return_dict_in_generate": false,
|
| 302 |
+
"sep_token_id": null,
|
| 303 |
+
"suppress_tokens": null,
|
| 304 |
+
"task_specific_params": null,
|
| 305 |
+
"temperature": 1.0,
|
| 306 |
+
"tf_legacy_loss": false,
|
| 307 |
+
"tie_encoder_decoder": false,
|
| 308 |
+
"tie_word_embeddings": true,
|
| 309 |
+
"tokenizer_class": null,
|
| 310 |
+
"top_k": 50,
|
| 311 |
+
"top_p": 1.0,
|
| 312 |
+
"torch_dtype": "bfloat16",
|
| 313 |
+
"torchscript": false,
|
| 314 |
+
"typical_p": 1.0,
|
| 315 |
+
"use_bfloat16": false,
|
| 316 |
+
"vision_use_head": false
|
| 317 |
+
},
|
| 318 |
+
"vision_tower_lr": null,
|
| 319 |
+
"weight_memory_efficient": true
|
| 320 |
+
},
|
| 321 |
+
"projector_config": {
|
| 322 |
+
"architectures": [
|
| 323 |
+
"MultimodalProjector"
|
| 324 |
+
],
|
| 325 |
+
"mm_projector_type": "mlp_downsample_3x3_fix",
|
| 326 |
+
"model_type": "v2l_projector",
|
| 327 |
+
"torch_dtype": "bfloat16",
|
| 328 |
+
"transformers_version": "4.46.0"
|
| 329 |
+
},
|
| 330 |
+
"vision_config": {
|
| 331 |
+
"architectures": [
|
| 332 |
+
"SiglipVisionModel"
|
| 333 |
+
],
|
| 334 |
+
"attention_dropout": 0.0,
|
| 335 |
+
"hidden_act": "gelu_pytorch_tanh",
|
| 336 |
+
"hidden_size": 1152,
|
| 337 |
+
"image_size": 448,
|
| 338 |
+
"intermediate_size": 4304,
|
| 339 |
+
"layer_norm_eps": 1e-6,
|
| 340 |
+
"model_type": "siglip_vision_model",
|
| 341 |
+
"num_attention_heads": 16,
|
| 342 |
+
"num_channels": 3,
|
| 343 |
+
"num_hidden_layers": 27,
|
| 344 |
+
"num_image_tokens": 256,
|
| 345 |
+
"patch_size": 14,
|
| 346 |
+
"projection_dim": 2048,
|
| 347 |
+
"projector_hidden_act": "gelu_fast",
|
| 348 |
+
"torch_dtype": "bfloat16",
|
| 349 |
+
"transformers_version": "4.46.0",
|
| 350 |
+
"vision_use_head": false
|
| 351 |
+
},
|
| 352 |
+
"llm_config": {
|
| 353 |
+
"architectures": [
|
| 354 |
+
"Qwen2ForCausalLM"
|
| 355 |
+
],
|
| 356 |
+
"attention_dropout": 0.0,
|
| 357 |
+
"bos_token_id": 151643,
|
| 358 |
+
"eos_token_id": 151645,
|
| 359 |
+
"hidden_act": "silu",
|
| 360 |
+
"hidden_size": 1536,
|
| 361 |
+
"initializer_range": 0.02,
|
| 362 |
+
"intermediate_size": 8960,
|
| 363 |
+
"max_position_embeddings": 32768,
|
| 364 |
+
"max_window_layers": 28,
|
| 365 |
+
"model_max_length": 4096,
|
| 366 |
+
"model_type": "qwen2",
|
| 367 |
+
"num_attention_heads": 12,
|
| 368 |
+
"num_hidden_layers": 28,
|
| 369 |
+
"num_key_value_heads": 2,
|
| 370 |
+
"rms_norm_eps": 1e-6,
|
| 371 |
+
"rope_scaling": null,
|
| 372 |
+
"rope_theta": 1000000.0,
|
| 373 |
+
"sliding_window": null,
|
| 374 |
+
"tie_word_embeddings": true,
|
| 375 |
+
"tokenizer_model_max_length": 4096,
|
| 376 |
+
"tokenizer_padding_side": "right",
|
| 377 |
+
"torch_dtype": "bfloat16",
|
| 378 |
+
"transformers_version": "4.46.0",
|
| 379 |
+
"use_cache": false,
|
| 380 |
+
"use_sliding_window": false,
|
| 381 |
+
"vocab_size": 151656
|
| 382 |
+
}
|
| 383 |
+
}
|
progress_model/model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:15582afcca8fe26b81a963a54bdcf01e0b513e56230f60da9b461d0b8f835885
|
| 3 |
+
size 4000400976
|
progress_model/preprocessor_config.json
ADDED
|
@@ -0,0 +1,24 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"do_convert_rgb": null,
|
| 3 |
+
"do_normalize": true,
|
| 4 |
+
"do_rescale": true,
|
| 5 |
+
"do_resize": true,
|
| 6 |
+
"image_mean": [
|
| 7 |
+
0.5,
|
| 8 |
+
0.5,
|
| 9 |
+
0.5
|
| 10 |
+
],
|
| 11 |
+
"image_processor_type": "SiglipImageProcessor",
|
| 12 |
+
"image_std": [
|
| 13 |
+
0.5,
|
| 14 |
+
0.5,
|
| 15 |
+
0.5
|
| 16 |
+
],
|
| 17 |
+
"processor_class": "SiglipProcessor",
|
| 18 |
+
"resample": 3,
|
| 19 |
+
"rescale_factor": 0.00392156862745098,
|
| 20 |
+
"size": {
|
| 21 |
+
"height": 448,
|
| 22 |
+
"width": 448
|
| 23 |
+
}
|
| 24 |
+
}
|
progress_model/special_tokens_map.json
ADDED
|
@@ -0,0 +1,49 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"additional_special_tokens": [
|
| 3 |
+
"<|im_start|>",
|
| 4 |
+
"<|im_end|>",
|
| 5 |
+
"<generated_image>",
|
| 6 |
+
{
|
| 7 |
+
"content": "<vila/sentinel>",
|
| 8 |
+
"lstrip": false,
|
| 9 |
+
"normalized": false,
|
| 10 |
+
"rstrip": false,
|
| 11 |
+
"single_word": false
|
| 12 |
+
},
|
| 13 |
+
{
|
| 14 |
+
"content": "<image>",
|
| 15 |
+
"lstrip": false,
|
| 16 |
+
"normalized": false,
|
| 17 |
+
"rstrip": false,
|
| 18 |
+
"single_word": false
|
| 19 |
+
},
|
| 20 |
+
{
|
| 21 |
+
"content": "<vila/video>",
|
| 22 |
+
"lstrip": false,
|
| 23 |
+
"normalized": false,
|
| 24 |
+
"rstrip": false,
|
| 25 |
+
"single_word": false
|
| 26 |
+
}
|
| 27 |
+
],
|
| 28 |
+
"bos_token": {
|
| 29 |
+
"content": "[BOS]",
|
| 30 |
+
"lstrip": false,
|
| 31 |
+
"normalized": false,
|
| 32 |
+
"rstrip": false,
|
| 33 |
+
"single_word": false
|
| 34 |
+
},
|
| 35 |
+
"eos_token": {
|
| 36 |
+
"content": "<|im_end|>",
|
| 37 |
+
"lstrip": false,
|
| 38 |
+
"normalized": false,
|
| 39 |
+
"rstrip": false,
|
| 40 |
+
"single_word": false
|
| 41 |
+
},
|
| 42 |
+
"pad_token": {
|
| 43 |
+
"content": "[PAD]",
|
| 44 |
+
"lstrip": false,
|
| 45 |
+
"normalized": false,
|
| 46 |
+
"rstrip": false,
|
| 47 |
+
"single_word": false
|
| 48 |
+
}
|
| 49 |
+
}
|
progress_model/tokenizer_config.json
ADDED
|
@@ -0,0 +1,129 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"add_prefix_space": false,
|
| 3 |
+
"added_tokens_decoder": {
|
| 4 |
+
"151643": {
|
| 5 |
+
"content": "<|endoftext|>",
|
| 6 |
+
"lstrip": false,
|
| 7 |
+
"normalized": false,
|
| 8 |
+
"rstrip": false,
|
| 9 |
+
"single_word": false,
|
| 10 |
+
"special": true
|
| 11 |
+
},
|
| 12 |
+
"151644": {
|
| 13 |
+
"content": "<|im_start|>",
|
| 14 |
+
"lstrip": false,
|
| 15 |
+
"normalized": false,
|
| 16 |
+
"rstrip": false,
|
| 17 |
+
"single_word": false,
|
| 18 |
+
"special": true
|
| 19 |
+
},
|
| 20 |
+
"151645": {
|
| 21 |
+
"content": "<|im_end|>",
|
| 22 |
+
"lstrip": false,
|
| 23 |
+
"normalized": false,
|
| 24 |
+
"rstrip": false,
|
| 25 |
+
"single_word": false,
|
| 26 |
+
"special": true
|
| 27 |
+
},
|
| 28 |
+
"151646": {
|
| 29 |
+
"content": "[BOS]",
|
| 30 |
+
"lstrip": false,
|
| 31 |
+
"normalized": false,
|
| 32 |
+
"rstrip": false,
|
| 33 |
+
"single_word": false,
|
| 34 |
+
"special": true
|
| 35 |
+
},
|
| 36 |
+
"151647": {
|
| 37 |
+
"content": "[PAD]",
|
| 38 |
+
"lstrip": false,
|
| 39 |
+
"normalized": false,
|
| 40 |
+
"rstrip": false,
|
| 41 |
+
"single_word": false,
|
| 42 |
+
"special": true
|
| 43 |
+
},
|
| 44 |
+
"151648": {
|
| 45 |
+
"content": "<nav>",
|
| 46 |
+
"lstrip": false,
|
| 47 |
+
"normalized": false,
|
| 48 |
+
"rstrip": false,
|
| 49 |
+
"single_word": false,
|
| 50 |
+
"special": true
|
| 51 |
+
},
|
| 52 |
+
"151649": {
|
| 53 |
+
"content": "<act>",
|
| 54 |
+
"lstrip": false,
|
| 55 |
+
"normalized": false,
|
| 56 |
+
"rstrip": false,
|
| 57 |
+
"single_word": false,
|
| 58 |
+
"special": true
|
| 59 |
+
},
|
| 60 |
+
"151650": {
|
| 61 |
+
"content": "<his>",
|
| 62 |
+
"lstrip": false,
|
| 63 |
+
"normalized": false,
|
| 64 |
+
"rstrip": false,
|
| 65 |
+
"single_word": false,
|
| 66 |
+
"special": true
|
| 67 |
+
},
|
| 68 |
+
"151651": {
|
| 69 |
+
"content": "<obs>",
|
| 70 |
+
"lstrip": false,
|
| 71 |
+
"normalized": false,
|
| 72 |
+
"rstrip": false,
|
| 73 |
+
"single_word": false,
|
| 74 |
+
"special": true
|
| 75 |
+
},
|
| 76 |
+
"151652": {
|
| 77 |
+
"content": "<generated_image>",
|
| 78 |
+
"lstrip": false,
|
| 79 |
+
"normalized": false,
|
| 80 |
+
"rstrip": false,
|
| 81 |
+
"single_word": false,
|
| 82 |
+
"special": true
|
| 83 |
+
},
|
| 84 |
+
"151653": {
|
| 85 |
+
"content": "<vila/sentinel>",
|
| 86 |
+
"lstrip": false,
|
| 87 |
+
"normalized": false,
|
| 88 |
+
"rstrip": false,
|
| 89 |
+
"single_word": false,
|
| 90 |
+
"special": true
|
| 91 |
+
},
|
| 92 |
+
"151654": {
|
| 93 |
+
"content": "<image>",
|
| 94 |
+
"lstrip": false,
|
| 95 |
+
"normalized": false,
|
| 96 |
+
"rstrip": false,
|
| 97 |
+
"single_word": false,
|
| 98 |
+
"special": true
|
| 99 |
+
},
|
| 100 |
+
"151655": {
|
| 101 |
+
"content": "<vila/video>",
|
| 102 |
+
"lstrip": false,
|
| 103 |
+
"normalized": false,
|
| 104 |
+
"rstrip": false,
|
| 105 |
+
"single_word": false,
|
| 106 |
+
"special": true
|
| 107 |
+
}
|
| 108 |
+
},
|
| 109 |
+
"additional_special_tokens": [
|
| 110 |
+
"<|im_start|>",
|
| 111 |
+
"<|im_end|>",
|
| 112 |
+
"<generated_image>",
|
| 113 |
+
"<vila/sentinel>",
|
| 114 |
+
"<image>",
|
| 115 |
+
"<vila/video>"
|
| 116 |
+
],
|
| 117 |
+
"bos_token": "[BOS]",
|
| 118 |
+
"chat_template": "{% if messages[0]['role'] != 'system' %}{{ '<|im_start|>system\\nYou are a helpful assistant<|im_end|>\\n' }}{% endif %}{% for message in messages if message['content'] is not none %}{{ '<|im_start|>' + message['role'] + '\\n' + message['content'] + '<|im_end|>' + '\\n' }}{% endfor %}{% if add_generation_prompt %}{{ '<|im_start|>assistant\\n' }}{% endif %}",
|
| 119 |
+
"clean_up_tokenization_spaces": false,
|
| 120 |
+
"eos_token": "<|im_end|>",
|
| 121 |
+
"errors": "replace",
|
| 122 |
+
"legacy": false,
|
| 123 |
+
"model_max_length": 4096,
|
| 124 |
+
"pad_token": "[PAD]",
|
| 125 |
+
"padding_side": "right",
|
| 126 |
+
"split_special_tokens": false,
|
| 127 |
+
"tokenizer_class": "Qwen2Tokenizer",
|
| 128 |
+
"unk_token": null
|
| 129 |
+
}
|
progress_model/vocab.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|