giobin/idefics3_random_connector

1{2  "architectures": [3    "Idefics3ForConditionalGeneration"4  ],5  "hidden_size": 4096,6  "image_token_id": 128257,7  "model_type": "idefics3",8  "scale_factor": 2,9  "text_config": {10    "_flash_attn_2_enabled": true,11    "_name_or_path": "None",12    "add_cross_attention": false,13    "architectures": [14      "Idefics3ForVisionText2Text"15    ],16    "attention_bias": false,17    "attention_dropout": 0.0,18    "auto_map": {19      "AutoConfig": "configuration_idefics3.Idefics3Config",20      "AutoModelForCausalLM": "modeling_idefics3.Idefics3ForVisionText2Text"21    },22    "bad_words_ids": null,23    "begin_suppress_tokens": null,24    "bos_token_id": 128000,25    "chunk_size_feed_forward": 0,26    "cross_attention_hidden_size": null,27    "decoder_start_token_id": null,28    "diversity_penalty": 0.0,29    "do_sample": false,30    "early_stopping": false,31    "encoder_no_repeat_ngram_size": 0,32    "eos_token_id": [33      128001,34      128008,35      12800936    ],37    "exponential_decay_length_penalty": null,38    "finetuning_task": null,39    "forced_bos_token_id": null,40    "forced_eos_token_id": null,41    "hidden_act": "silu",42    "hidden_size": 4096,43    "id2label": {44      "0": "LABEL_0",45      "1": "LABEL_1"46    },47    "initializer_range": 0.02,48    "intermediate_size": 14336,49    "is_decoder": false,50    "is_encoder_decoder": false,51    "label2id": {52      "LABEL_0": 0,53      "LABEL_1": 154    },55    "length_penalty": 1.0,56    "max_length": 20,57    "max_position_embeddings": 131072,58    "min_length": 0,59    "mlp_bias": false,60    "model_type": "llama",61    "neftune_noise_alpha": 0.0,62    "no_repeat_ngram_size": 0,63    "num_attention_heads": 32,64    "num_beam_groups": 1,65    "num_beams": 1,66    "num_hidden_layers": 32,67    "num_key_value_heads": 8,68    "num_return_sequences": 1,69    "output_attentions": false,70    "output_hidden_states": false,71    "output_scores": false,72    "pad_token_id": 128002,73    "perceiver_config": {74      "_name_or_path": "",75      "add_cross_attention": false,76      "architectures": null,77      "attention_dropout": 0.0,78      "bad_words_ids": null,79      "begin_suppress_tokens": null,80      "bos_token_id": null,81      "chunk_size_feed_forward": 0,82      "cross_attention_hidden_size": null,83      "decoder_start_token_id": null,84      "diversity_penalty": 0.0,85      "do_sample": false,86      "early_stopping": false,87      "encoder_no_repeat_ngram_size": 0,88      "eos_token_id": null,89      "exponential_decay_length_penalty": null,90      "finetuning_task": null,91      "forced_bos_token_id": null,92      "forced_eos_token_id": null,93      "hidden_act": "silu",94      "id2label": {95        "0": "LABEL_0",96        "1": "LABEL_1"97      },98      "is_decoder": false,99      "is_encoder_decoder": false,100      "label2id": {101        "LABEL_0": 0,102        "LABEL_1": 1103      },104      "length_penalty": 1.0,105      "max_length": 20,106      "min_length": 0,107      "model_type": "idefics3",108      "no_repeat_ngram_size": 0,109      "num_beam_groups": 1,110      "num_beams": 1,111      "num_key_value_heads": 1,112      "num_return_sequences": 1,113      "output_attentions": false,114      "output_hidden_states": false,115      "output_scores": false,116      "pad_token_id": 128002,117      "prefix": null,118      "problem_type": null,119      "pruned_heads": {},120      "qk_layer_norms_perceiver": false,121      "remove_invalid_values": false,122      "repetition_penalty": 1.0,123      "resampler_depth": 6,124      "resampler_head_dim": 96,125      "resampler_n_heads": 16,126      "resampler_n_latents": 64,127      "return_dict": true,128      "return_dict_in_generate": false,129      "sep_token_id": null,130      "suppress_tokens": null,131      "task_specific_params": null,132      "temperature": 1.0,133      "tf_legacy_loss": false,134      "tie_encoder_decoder": false,135      "tie_word_embeddings": true,136      "tokenizer_class": null,137      "top_k": 50,138      "top_p": 1.0,139      "torch_dtype": null,140      "torchscript": false,141      "transformers_version": "4.43.2",142      "typical_p": 1.0,143      "use_bfloat16": false144    },145    "prefix": null,146    "pretraining_tp": 1,147    "problem_type": null,148    "pruned_heads": {},149    "qk_layer_norms": false,150    "remove_invalid_values": false,151    "repetition_penalty": 1.0,152    "return_dict": true,153    "return_dict_in_generate": false,154    "rms_norm_eps": 1e-05,155    "rope_scaling": {156      "factor": 8.0,157      "high_freq_factor": 4.0,158      "low_freq_factor": 1.0,159      "original_max_position_embeddings": 8192,160      "rope_type": "llama3"161    },162    "rope_theta": 500000.0,163    "sep_token_id": null,164    "suppress_tokens": null,165    "task_specific_params": null,166    "temperature": 1.0,167    "tf_legacy_loss": false,168    "tie_encoder_decoder": false,169    "tie_word_embeddings": false,170    "tokenizer_class": null,171    "top_k": 50,172    "top_p": 1.0,173    "torch_dtype": "bfloat16",174    "torchscript": false,175    "typical_p": 1.0,176    "use_bfloat16": false,177    "use_cache": true,178    "use_resampler": false,179    "vocab_size": 128259180  },181  "tie_word_embeddings": false,182  "torch_dtype": "float32",183  "transformers_version": "4.44.0.dev0",184  "use_cache": true,185  "vision_config": {186    "_name_or_path": "",187    "add_cross_attention": false,188    "architectures": null,189    "attention_dropout": 0.0,190    "bad_words_ids": null,191    "begin_suppress_tokens": null,192    "bos_token_id": null,193    "chunk_size_feed_forward": 0,194    "cross_attention_hidden_size": null,195    "decoder_start_token_id": null,196    "diversity_penalty": 0.0,197    "do_sample": false,198    "early_stopping": false,199    "encoder_no_repeat_ngram_size": 0,200    "eos_token_id": null,201    "exponential_decay_length_penalty": null,202    "finetuning_task": null,203    "forced_bos_token_id": null,204    "forced_eos_token_id": null,205    "hidden_act": "gelu_pytorch_tanh",206    "hidden_size": 1152,207    "id2label": {208      "0": "LABEL_0",209      "1": "LABEL_1"210    },211    "image_size": 364,212    "initializer_range": 0.02,213    "intermediate_size": 4304,214    "is_decoder": false,215    "is_encoder_decoder": false,216    "label2id": {217      "LABEL_0": 0,218      "LABEL_1": 1219    },220    "layer_norm_eps": 1e-06,221    "length_penalty": 1.0,222    "max_length": 20,223    "min_length": 0,224    "model_type": "idefics3",225    "no_repeat_ngram_size": 0,226    "num_attention_heads": 16,227    "num_beam_groups": 1,228    "num_beams": 1,229    "num_channels": 3,230    "num_hidden_layers": 27,231    "num_return_sequences": 1,232    "old_vision_model_name": "/fsx/hugo/siglip-so400m-14-364-flash-attn2-navit",233    "output_attentions": false,234    "output_hidden_states": false,235    "output_scores": false,236    "pad_token_id": 128002,237    "patch_size": 14,238    "prefix": null,239    "problem_type": null,240    "pruned_heads": {},241    "remove_invalid_values": false,242    "repetition_penalty": 1.0,243    "return_dict": true,244    "return_dict_in_generate": false,245    "sep_token_id": null,246    "suppress_tokens": null,247    "task_specific_params": null,248    "temperature": 1.0,249    "tf_legacy_loss": false,250    "tie_encoder_decoder": false,251    "tie_word_embeddings": true,252    "tokenizer_class": null,253    "top_k": 50,254    "top_p": 1.0,255    "torch_dtype": null,256    "torchscript": false,257    "typical_p": 1.0,258    "use_bfloat16": false259  },260  "vocab_size": 128259261}262