ragavsachdeva/magiv3

1{2  "_name_or_path": "magiv3",3  "architectures": [4    "Florence2ForConditionalGeneration"5  ],6  "auto_map": {7    "AutoConfig": "configuration_florence2.Florence2Config",8    "AutoModel": "modeling_florence2.Florence2ForConditionalGeneration",9    "AutoModelForCausalLM": "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": 1024,35    "decoder_attention_heads": 16,36    "decoder_ffn_dim": 4096,37    "decoder_layerdrop": 0.0,38    "decoder_layers": 12,39    "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": 16,45    "encoder_ffn_dim": 4096,46    "encoder_layerdrop": 0.0,47    "encoder_layers": 12,48    "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": 12,78    "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    "pruned_heads": {},86    "remove_invalid_values": false,87    "repetition_penalty": 1.0,88    "return_dict": true,89    "return_dict_in_generate": false,90    "scale_embedding": false,91    "sep_token_id": null,92    "suppress_tokens": null,93    "task_specific_params": null,94    "temperature": 1.0,95    "tf_legacy_loss": false,96    "tie_encoder_decoder": false,97    "tie_word_embeddings": true,98    "tokenizer_class": null,99    "top_k": 50,100    "top_p": 1.0,101    "torch_dtype": null,102    "torchscript": false,103    "typical_p": 1.0,104    "use_bfloat16": false,105    "use_cache": true,106    "vocab_size": 51293107  },108  "torch_dtype": "float16",109  "transformers_version": "4.45.2",110  "vision_config": {111    "_name_or_path": "",112    "add_cross_attention": false,113    "architectures": null,114    "bad_words_ids": null,115    "begin_suppress_tokens": null,116    "bos_token_id": null,117    "chunk_size_feed_forward": 0,118    "cross_attention_hidden_size": null,119    "decoder_start_token_id": null,120    "depths": [121      1,122      1,123      9,124      1125    ],126    "dim_embed": [127      256,128      512,129      1024,130      2048131    ],132    "diversity_penalty": 0.0,133    "do_sample": false,134    "drop_path_rate": 0.1,135    "early_stopping": false,136    "enable_checkpoint": false,137    "encoder_no_repeat_ngram_size": 0,138    "eos_token_id": null,139    "exponential_decay_length_penalty": null,140    "finetuning_task": null,141    "forced_bos_token_id": null,142    "forced_eos_token_id": null,143    "id2label": {144      "0": "LABEL_0",145      "1": "LABEL_1"146    },147    "image_feature_source": [148      "spatial_avg_pool",149      "temporal_avg_pool"150    ],151    "image_pos_embed": {152      "max_pos_embeddings": 50,153      "type": "learned_abs_2d"154    },155    "is_decoder": false,156    "is_encoder_decoder": false,157    "label2id": {158      "LABEL_0": 0,159      "LABEL_1": 1160    },161    "length_penalty": 1.0,162    "max_length": 20,163    "min_length": 0,164    "model_type": "",165    "no_repeat_ngram_size": 0,166    "num_beam_groups": 1,167    "num_beams": 1,168    "num_groups": [169      8,170      16,171      32,172      64173    ],174    "num_heads": [175      8,176      16,177      32,178      64179    ],180    "num_return_sequences": 1,181    "output_attentions": false,182    "output_hidden_states": false,183    "output_scores": false,184    "pad_token_id": null,185    "patch_padding": [186      3,187      1,188      1,189      1190    ],191    "patch_prenorm": [192      false,193      true,194      true,195      true196    ],197    "patch_size": [198      7,199      3,200      3,201      3202    ],203    "patch_stride": [204      4,205      2,206      2,207      2208    ],209    "prefix": null,210    "problem_type": null,211    "projection_dim": 1024,212    "pruned_heads": {},213    "remove_invalid_values": false,214    "repetition_penalty": 1.0,215    "return_dict": true,216    "return_dict_in_generate": false,217    "sep_token_id": null,218    "suppress_tokens": null,219    "task_specific_params": null,220    "temperature": 1.0,221    "tf_legacy_loss": false,222    "tie_encoder_decoder": false,223    "tie_word_embeddings": true,224    "tokenizer_class": null,225    "top_k": 50,226    "top_p": 1.0,227    "torch_dtype": null,228    "torchscript": false,229    "typical_p": 1.0,230    "use_bfloat16": false,231    "visual_temporal_embedding": {232      "max_temporal_embeddings": 100,233      "type": "COSINE"234    },235    "window_size": 12236  },237  "vocab_size": 51293238}239