Source82/florence-2-large-4bit
07
1{2 "_name_or_path": "microsoft/Florence-2-large",3 "architectures": [4 "Florence2ForConditionalGeneration"5 ],6 "auto_map": {7 "AutoConfig": "microsoft/Florence-2-large--configuration_florence2.Florence2Config",8 "AutoModelForCausalLM": "microsoft/Florence-2-large--modeling_florence2.Florence2ForConditionalGeneration"9 },10 "bos_token_id": 0,11 "eos_token_id": 2,12 "ignore_index": -100,13 "is_encoder_decoder": true,14 "model_type": "florence2",15 "pad_token_id": 1,16 "projection_dim": 1024,17 "quantization_config": {18 "_load_in_4bit": true,19 "_load_in_8bit": false,20 "bnb_4bit_compute_dtype": "bfloat16",21 "bnb_4bit_quant_storage": "uint8",22 "bnb_4bit_quant_type": "fp4",23 "bnb_4bit_use_double_quant": true,24 "llm_int8_enable_fp32_cpu_offload": false,25 "llm_int8_has_fp16_weight": false,26 "llm_int8_skip_modules": null,27 "llm_int8_threshold": 6.0,28 "load_in_4bit": true,29 "load_in_8bit": false,30 "quant_method": "bitsandbytes"31 },32 "text_config": {33 "_attn_implementation_autoset": true,34 "_name_or_path": "",35 "activation_dropout": 0.1,36 "activation_function": "gelu",37 "add_bias_logits": false,38 "add_cross_attention": false,39 "add_final_layer_norm": false,40 "architectures": null,41 "attention_dropout": 0.1,42 "bad_words_ids": null,43 "begin_suppress_tokens": null,44 "bos_token_id": 0,45 "chunk_size_feed_forward": 0,46 "classif_dropout": 0.1,47 "classifier_dropout": 0.0,48 "cross_attention_hidden_size": null,49 "d_model": 1024,50 "decoder_attention_heads": 16,51 "decoder_ffn_dim": 4096,52 "decoder_layerdrop": 0.0,53 "decoder_layers": 12,54 "decoder_start_token_id": 2,55 "diversity_penalty": 0.0,56 "do_sample": false,57 "dropout": 0.1,58 "early_stopping": true,59 "encoder_attention_heads": 16,60 "encoder_ffn_dim": 4096,61 "encoder_layerdrop": 0.0,62 "encoder_layers": 12,63 "encoder_no_repeat_ngram_size": 0,64 "eos_token_id": 2,65 "exponential_decay_length_penalty": null,66 "finetuning_task": null,67 "forced_bos_token_id": 0,68 "forced_eos_token_id": 2,69 "gradient_checkpointing": false,70 "id2label": {71 "0": "LABEL_0",72 "1": "LABEL_1",73 "2": "LABEL_2"74 },75 "init_std": 0.02,76 "is_decoder": false,77 "is_encoder_decoder": true,78 "label2id": {79 "LABEL_0": 0,80 "LABEL_1": 1,81 "LABEL_2": 282 },83 "length_penalty": 1.0,84 "max_length": 20,85 "max_position_embeddings": 4096,86 "min_length": 0,87 "model_type": "florence2_language",88 "no_repeat_ngram_size": 3,89 "normalize_before": false,90 "num_beam_groups": 1,91 "num_beams": 3,92 "num_hidden_layers": 12,93 "num_return_sequences": 1,94 "output_attentions": false,95 "output_hidden_states": false,96 "output_scores": false,97 "pad_token_id": 1,98 "prefix": null,99 "problem_type": null,100 "pruned_heads": {},101 "remove_invalid_values": false,102 "repetition_penalty": 1.0,103 "return_dict": true,104 "return_dict_in_generate": false,105 "scale_embedding": false,106 "sep_token_id": null,107 "suppress_tokens": null,108 "task_specific_params": null,109 "temperature": 1.0,110 "tf_legacy_loss": false,111 "tie_encoder_decoder": false,112 "tie_word_embeddings": true,113 "tokenizer_class": null,114 "top_k": 50,115 "top_p": 1.0,116 "torch_dtype": null,117 "torchscript": false,118 "typical_p": 1.0,119 "use_bfloat16": false,120 "use_cache": true,121 "vocab_size": 51289122 },123 "torch_dtype": "float16",124 "transformers_version": "4.48.3",125 "vision_config": {126 "_attn_implementation_autoset": false,127 "_name_or_path": "",128 "add_cross_attention": false,129 "architectures": null,130 "bad_words_ids": null,131 "begin_suppress_tokens": null,132 "bos_token_id": null,133 "chunk_size_feed_forward": 0,134 "cross_attention_hidden_size": null,135 "decoder_start_token_id": null,136 "depths": [137 1,138 1,139 9,140 1141 ],142 "dim_embed": [143 256,144 512,145 1024,146 2048147 ],148 "diversity_penalty": 0.0,149 "do_sample": false,150 "drop_path_rate": 0.1,151 "early_stopping": false,152 "enable_checkpoint": false,153 "encoder_no_repeat_ngram_size": 0,154 "eos_token_id": null,155 "exponential_decay_length_penalty": null,156 "finetuning_task": null,157 "forced_bos_token_id": null,158 "forced_eos_token_id": null,159 "id2label": {160 "0": "LABEL_0",161 "1": "LABEL_1"162 },163 "image_feature_source": [164 "spatial_avg_pool",165 "temporal_avg_pool"166 ],167 "image_pos_embed": {168 "max_pos_embeddings": 50,169 "type": "learned_abs_2d"170 },171 "is_decoder": false,172 "is_encoder_decoder": false,173 "label2id": {174 "LABEL_0": 0,175 "LABEL_1": 1176 },177 "length_penalty": 1.0,178 "max_length": 20,179 "min_length": 0,180 "model_type": "davit",181 "no_repeat_ngram_size": 0,182 "num_beam_groups": 1,183 "num_beams": 1,184 "num_groups": [185 8,186 16,187 32,188 64189 ],190 "num_heads": [191 8,192 16,193 32,194 64195 ],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_padding": [202 3,203 1,204 1,205 1206 ],207 "patch_prenorm": [208 false,209 true,210 true,211 true212 ],213 "patch_size": [214 7,215 3,216 3,217 3218 ],219 "patch_stride": [220 4,221 2,222 2,223 2224 ],225 "prefix": null,226 "problem_type": null,227 "projection_dim": 1024,228 "pruned_heads": {},229 "remove_invalid_values": false,230 "repetition_penalty": 1.0,231 "return_dict": true,232 "return_dict_in_generate": false,233 "sep_token_id": null,234 "suppress_tokens": null,235 "task_specific_params": null,236 "temperature": 1.0,237 "tf_legacy_loss": false,238 "tie_encoder_decoder": false,239 "tie_word_embeddings": true,240 "tokenizer_class": null,241 "top_k": 50,242 "top_p": 1.0,243 "torch_dtype": null,244 "torchscript": false,245 "typical_p": 1.0,246 "use_bfloat16": false,247 "visual_temporal_embedding": {248 "max_temporal_embeddings": 100,249 "type": "COSINE"250 },251 "window_size": 12252 },253 "vocab_size": 51289254}255