{ "architectures": [ "XCLIPModel" ], "dtype": "float32", "initializer_factor": 1.0, "logit_scale_init_value": 2.6592, "model_type": "xclip", "projection_dim": 64, "prompt_alpha": 0.1, "prompt_attention_dropout": 0.0, "prompt_hidden_act": "quick_gelu", "prompt_layers": 2, "prompt_num_attention_heads": 4, "prompt_projection_dropout": 0.0, "text_config": { "attention_dropout": 0.1, "bos_token_id": 49406, "dropout": 0.1, "eos_token_id": 49407, "hidden_act": "quick_gelu", "hidden_size": 32, "initializer_factor": 1.0, "initializer_range": 0.02, "intermediate_size": 37, "layer_norm_eps": 1e-05, "max_position_embeddings": 512, "model_type": "xclip_text_model", "num_attention_heads": 4, "num_hidden_layers": 2, "pad_token_id": 49407, "vocab_size": 49408 }, "transformers_version": "5.16.0.dev0", "vision_config": { "attention_dropout": 0.1, "drop_path_rate": 0.0, "dropout": 0.1, "hidden_act": "quick_gelu", "hidden_size": 32, "image_size": 30, "initializer_factor": 1.0, "initializer_range": 0.02, "intermediate_size": 37, "layer_norm_eps": 1e-05, "mit_hidden_size": 64, "mit_intermediate_size": 2048, "mit_num_attention_heads": 8, "mit_num_hidden_layers": 1, "model_type": "xclip_vision_model", "num_attention_heads": 4, "num_channels": 3, "num_frames": 8, "num_hidden_layers": 2, "patch_size": 2 } }