{ "architectures": [ "DeiTModel" ], "attention_head_size": 88, "attention_probs_dropout_prob": 0.0, "encoder_stride": 14, "hidden_act": "gelu", "hidden_dropout_prob": 0.0, "hidden_size": 1408, "image_size": 224, "initializer_range": 0.02, "intermediate_size": 3072, "layer_norm_eps": 1e-12, "model_type": "deit", "num_attention_heads": 16, "num_channels": 3, "num_hidden_layers": 12, "patch_size": 16, "qkv_bias": true, "torch_dtype": "float32", "transformers_version": "4.39.0.dev0" }