Xenova/tiny-random-Florence2ForConditionalGeneration
61.3k
1{2 "architectures": [3 "Florence2ForConditionalGeneration"4 ],5 "auto_map": {6 "AutoConfig": "configuration_florence2.Florence2Config",7 "AutoModelForCausalLM": "modeling_florence2.Florence2ForConditionalGeneration",8 "AutoModelForVision2Seq": "modeling_florence2.Florence2ForConditionalGeneration",9 "AutoModelForSeq2SeqLM": "modeling_florence2.Florence2ForConditionalGeneration"10 },11 "bos_token_id": 0,12 "eos_token_id": 2,13 "ignore_index": -100,14 "is_encoder_decoder": true,15 "model_type": "florence2",16 "pad_token_id": 1,17 "projection_dim": 1024,18 "text_config": {19 "_name_or_path": "",20 "activation_dropout": 0.1,21 "activation_function": "gelu",22 "add_bias_logits": false,23 "add_cross_attention": false,24 "add_final_layer_norm": false,25 "architectures": null,26 "attention_dropout": 0.1,27 "bad_words_ids": null,28 "begin_suppress_tokens": null,29 "bos_token_id": 0,30 "chunk_size_feed_forward": 0,31 "classif_dropout": 0.1,32 "classifier_dropout": 0.0,33 "cross_attention_hidden_size": null,34 "d_model": 32,35 "decoder_attention_heads": 2,36 "decoder_ffn_dim": 3072,37 "decoder_layerdrop": 0.0,38 "decoder_layers": 1,39 "decoder_start_token_id": 2,40 "diversity_penalty": 0.0,41 "do_sample": false,42 "dropout": 0.1,43 "early_stopping": true,44 "encoder_attention_heads": 2,45 "encoder_ffn_dim": 3072,46 "encoder_layerdrop": 0.0,47 "encoder_layers": 6,48 "encoder_no_repeat_ngram_size": 0,49 "eos_token_id": 2,50 "exponential_decay_length_penalty": null,51 "finetuning_task": null,52 "forced_bos_token_id": 0,53 "forced_eos_token_id": 2,54 "gradient_checkpointing": false,55 "id2label": {56 "0": "LABEL_0",57 "1": "LABEL_1",58 "2": "LABEL_2"59 },60 "init_std": 0.02,61 "is_decoder": false,62 "is_encoder_decoder": true,63 "label2id": {64 "LABEL_0": 0,65 "LABEL_1": 1,66 "LABEL_2": 267 },68 "length_penalty": 1.0,69 "max_length": 20,70 "max_position_embeddings": 1024,71 "min_length": 0,72 "model_type": "florence2_language",73 "no_repeat_ngram_size": 3,74 "normalize_before": false,75 "num_beam_groups": 1,76 "num_beams": 3,77 "num_hidden_layers": 1,78 "num_return_sequences": 1,79 "output_attentions": false,80 "output_hidden_states": false,81 "output_scores": false,82 "pad_token_id": 1,83 "prefix": null,84 "problem_type": null,85 "projection_dim": 32,86 "pruned_heads": {},87 "remove_invalid_values": false,88 "repetition_penalty": 1.0,89 "return_dict": true,90 "return_dict_in_generate": false,91 "scale_embedding": false,92 "sep_token_id": null,93 "suppress_tokens": null,94 "task_specific_params": null,95 "temperature": 1.0,96 "tf_legacy_loss": false,97 "tie_encoder_decoder": false,98 "tie_word_embeddings": true,99 "tokenizer_class": null,100 "top_k": 50,101 "top_p": 1.0,102 "torch_dtype": null,103 "torchscript": false,104 "typical_p": 1.0,105 "use_bfloat16": false,106 "use_cache": true,107 "vocab_size": 51289108 },109 "torch_dtype": "float32",110 "transformers_version": "4.41.2",111 "vision_config": {112 "_name_or_path": "",113 "add_cross_attention": false,114 "architectures": null,115 "bad_words_ids": null,116 "begin_suppress_tokens": null,117 "bos_token_id": null,118 "chunk_size_feed_forward": 0,119 "cross_attention_hidden_size": null,120 "decoder_start_token_id": null,121 "depths": [122 1,123 1,124 9,125 1126 ],127 "dim_embed": [128 32,129 32,130 32,131 32132 ],133 "diversity_penalty": 0.0,134 "do_sample": false,135 "drop_path_rate": 0.1,136 "early_stopping": false,137 "enable_checkpoint": false,138 "encoder_no_repeat_ngram_size": 0,139 "eos_token_id": null,140 "exponential_decay_length_penalty": null,141 "finetuning_task": null,142 "forced_bos_token_id": null,143 "forced_eos_token_id": null,144 "id2label": {145 "0": "LABEL_0",146 "1": "LABEL_1"147 },148 "image_feature_source": [149 "spatial_avg_pool",150 "temporal_avg_pool"151 ],152 "image_pos_embed": {153 "max_pos_embeddings": 50,154 "type": "learned_abs_2d"155 },156 "is_decoder": false,157 "is_encoder_decoder": false,158 "label2id": {159 "LABEL_0": 0,160 "LABEL_1": 1161 },162 "length_penalty": 1.0,163 "max_length": 20,164 "min_length": 0,165 "model_type": "davit",166 "no_repeat_ngram_size": 0,167 "num_beam_groups": 1,168 "num_beams": 1,169 "num_groups": [170 4,171 8,172 16,173 32174 ],175 "num_heads": [176 4,177 8,178 16,179 32180 ],181 "num_return_sequences": 1,182 "output_attentions": false,183 "output_hidden_states": false,184 "output_scores": false,185 "pad_token_id": null,186 "patch_padding": [187 3,188 1,189 1,190 1191 ],192 "patch_prenorm": [193 false,194 true,195 true,196 true197 ],198 "patch_size": [199 7,200 3,201 3,202 3203 ],204 "patch_stride": [205 4,206 2,207 2,208 2209 ],210 "prefix": null,211 "problem_type": null,212 "projection_dim": 32,213 "pruned_heads": {},214 "remove_invalid_values": false,215 "repetition_penalty": 1.0,216 "return_dict": true,217 "return_dict_in_generate": false,218 "sep_token_id": null,219 "suppress_tokens": null,220 "task_specific_params": null,221 "temperature": 1.0,222 "tf_legacy_loss": false,223 "tie_encoder_decoder": false,224 "tie_word_embeddings": true,225 "tokenizer_class": null,226 "top_k": 50,227 "top_p": 1.0,228 "torch_dtype": null,229 "torchscript": false,230 "typical_p": 1.0,231 "use_bfloat16": false,232 "visual_temporal_embedding": {233 "max_temporal_embeddings": 100,234 "type": "COSINE"235 },236 "window_size": 12237 },238 "vocab_size": 51289239}240 