| { | |
| "_commit_hash": null, | |
| "_name_or_path": "recruit-jp/japanese-clip-vit-b-32-roberta-base", | |
| "architectures": [ | |
| "JapaneseCLIPModel" | |
| ], | |
| "auto_map": { | |
| "AutoModel": "modeling_japanese_clip.JapaneseCLIPModel", | |
| "AutoConfig": "configuration_japanese_clip.JapaneseCLIPConfig" | |
| }, | |
| "torch_dtype": "float32", | |
| "transformers_version": "4.36.2", | |
| "model_type": "japanese_clip", | |
| "text_config": { | |
| "_name_or_path": "", | |
| "architectures": [ | |
| "RobertaModel" | |
| ], | |
| "attention_probs_dropout_prob": 0.1, | |
| "bos_token_id": 1, | |
| "eos_token_id": 2, | |
| "gradient_checkpointing": false, | |
| "hidden_act": "gelu", | |
| "hidden_dropout_prob": 0.1, | |
| "hidden_size": 768, | |
| "initializer_range": 0.02, | |
| "intermediate_size": 3072, | |
| "layer_norm_eps": 1e-05, | |
| "max_position_embeddings": 514, | |
| "model_type": "roberta", | |
| "num_attention_heads": 12, | |
| "num_hidden_layers": 12, | |
| "pad_token_id": 3, | |
| "position_embedding_type": "absolute", | |
| "transformers_version": "4.6.1", | |
| "type_vocab_size": 2, | |
| "use_cache": true, | |
| "vocab_size": 32000 | |
| }, | |
| "vision_config": { | |
| "_name_or_path": "", | |
| "image_size": 224, | |
| "patch_size": 32, | |
| "width": 768, | |
| "layers": 12, | |
| "mlp_ratio": 4.0, | |
| "ls_init_value": null, | |
| "attentional_pool": false, | |
| "attn_pooler_queries": 256, | |
| "attn_pooler_heads": 8, | |
| "output_dim": 512, | |
| "patch_dropout": 0.0, | |
| "no_ln_pre": false, | |
| "pool_type": "tok", | |
| "final_ln_after_pool": false, | |
| "output_tokens": false | |
| } | |
| } |