{ "architectures": [ "InternVLModel" ], "downsample_ratio": 0.5, "dtype": "float32", "image_seq_length": 64, "image_token_id": 1, "model_type": "internvl", "projector_hidden_act": "gelu", "text_config": { "attention_dropout": 0.0, "bos_token_id": 3, "eos_token_id": 151645, "hidden_act": "silu", "hidden_size": 128, "initializer_range": 0.02, "intermediate_size": 37, "layer_types": [ "full_attention", "full_attention" ], "max_position_embeddings": 512, "max_window_layers": 28, "mlp_ratio": 4, "model_type": "qwen2", "num_attention_heads": 4, "num_hidden_layers": 2, "num_key_value_heads": 2, "output_channels": 64, "pad_token_id": 151643, "rms_norm_eps": 1e-06, "rope_parameters": { "rope_theta": 10000, "rope_type": "default" }, "sliding_window": null, "tie_word_embeddings": true, "use_cache": true, "use_sliding_window": false, "vocab_size": 151675 }, "tie_word_embeddings": true, "transformers_version": "5.16.0.dev0", "vision_config": { "attention_bias": false, "attention_dropout": 0.0, "hidden_act": "quick_gelu", "hidden_dropout_prob": 0.0, "hidden_size": 32, "image_size": [ 64, 64 ], "initializer_range": 0.02, "intermediate_size": 128, "layer_norm_eps": 1e-06, "layer_scale_init_value": 0.1, "model_type": "internvl_vision", "norm_type": "layer_norm", "num_attention_heads": 4, "num_channels": 3, "num_hidden_layers": 2, "patch_size": [ 4, 4 ], "projection_dropout": 0.0, "use_absolute_position_embeddings": true, "use_mask_token": false, "use_mean_pooling": true, "use_qk_norm": false }, "vision_feature_layer": -1, "vision_feature_select_strategy": "default" }