| { | |
| "architectures": [ | |
| "SiglipForImageClassification" | |
| ], | |
| "dtype": "float32", | |
| "initializer_factor": 1.0, | |
| "model_type": "siglip", | |
| "text_config": { | |
| "attention_dropout": 0.1, | |
| "bos_token_id": 49406, | |
| "dropout": 0.1, | |
| "eos_token_id": 1, | |
| "hidden_act": "gelu_pytorch_tanh", | |
| "hidden_size": 64, | |
| "initializer_range": 0.02, | |
| "intermediate_size": 37, | |
| "layer_norm_eps": 1e-06, | |
| "max_position_embeddings": 512, | |
| "model_type": "siglip_text_model", | |
| "num_attention_heads": 2, | |
| "num_hidden_layers": 2, | |
| "pad_token_id": 1, | |
| "projection_size": 64, | |
| "vocab_size": 32000 | |
| }, | |
| "transformers_version": "5.16.0.dev0", | |
| "vision_config": { | |
| "attention_dropout": 0.1, | |
| "dropout": 0.1, | |
| "hidden_act": "gelu_pytorch_tanh", | |
| "hidden_size": 64, | |
| "image_size": 4, | |
| "initializer_range": 0.02, | |
| "intermediate_size": 37, | |
| "layer_norm_eps": 1e-06, | |
| "model_type": "siglip_vision_model", | |
| "num_attention_heads": 2, | |
| "num_channels": 3, | |
| "num_hidden_layers": 2, | |
| "patch_size": 2 | |
| } | |
| } | |