|
{ |
|
"_name_or_path": "laion/CLIP-ViT-H-14-laion2B-s32B-b79K", |
|
"architectures": [ |
|
"CLIPModel" |
|
], |
|
"initializer_factor": 1.0, |
|
"logit_scale_init_value": 2.6592, |
|
"model_type": "clip", |
|
"projection_dim": 1024, |
|
"text_config": { |
|
"_name_or_path": "/work/dnai_explainability/unl4gen/safe-clip/vision-encoder-ft-checkpoints/1709914296/text_encoder/best_text_encoder_ft_recall_merged_lorafixed", |
|
"architectures": [ |
|
"CLIPTextModelWithProjection" |
|
], |
|
"bos_token_id": 0, |
|
"dropout": 0.0, |
|
"eos_token_id": 2, |
|
"hidden_act": "gelu", |
|
"hidden_size": 1024, |
|
"intermediate_size": 4096, |
|
"model_type": "clip_text_model", |
|
"num_attention_heads": 16, |
|
"num_hidden_layers": 23, |
|
"projection_dim": 1024, |
|
"torch_dtype": "float32" |
|
}, |
|
"torch_dtype": "float32", |
|
"transformers_version": "4.37.2", |
|
"vision_config": { |
|
"dropout": 0.0, |
|
"hidden_act": "gelu", |
|
"hidden_size": 1280, |
|
"intermediate_size": 5120, |
|
"model_type": "clip_vision_model", |
|
"num_attention_heads": 16, |
|
"num_hidden_layers": 32, |
|
"patch_size": 14, |
|
"projection_dim": 1024 |
|
} |
|
} |
|
|