{ "architectures": [ "StableDiffusionSafetyChecker" ], "initializer_factor": 1.0, "logit_scale_init_value": 2.6592, "model_type": "clip", "projection_dim": 64, "text_config": { "attention_dropout": 0.1, "bos_token_id": 0, "dropout": 0.1, "eos_token_id": 2, "gradient_checkpointing": false, "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": "clip_text_model", "num_attention_heads": 4, "num_hidden_layers": 5, "projection_dim": 512, "torch_dtype": "float32", "vocab_size": 99 }, "torch_dtype": "float32", "transformers_version": "4.53.3", "vision_config": { "attention_dropout": 0.1, "dropout": 0.1, "gradient_checkpointing": false, "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, "model_type": "clip_vision_model", "num_attention_heads": 4, "num_channels": 3, "num_hidden_layers": 5, "patch_size": 2, "projection_dim": 512, "torch_dtype": "float32" }, "vocab_size": 1000 }