{ "architectures": [ "HunYuanVLForConditionalGeneration" ], "dtype": "float32", "im_end_id": 120119, "im_newline_id": 120121, "im_start_id": 120118, "image_token_id": 5, "model_type": "hunyuan_vl", "text_config": { "attention_bias": false, "attention_dropout": 0.0, "bos_token_id": 120000, "eod_token_id": 3, "eos_token_id": 120007, "head_dim": 16, "hidden_act": "silu", "hidden_size": 64, "initializer_range": 0.02, "intermediate_size": 128, "max_position_embeddings": 200, "model_type": "hunyuan_vl_text", "num_attention_heads": 4, "num_hidden_layers": 2, "num_key_value_heads": 4, "pad_token_id": 120002, "pretraining_tp": 1, "rms_norm_eps": 1e-05, "rope_parameters": { "mrope_section": [ 2, 2, 2, 2 ], "rope_theta": 10000.0, "rope_type": "default" }, "sep_token_id": 4, "tie_word_embeddings": false, "use_cache": true, "vocab_size": 120818 }, "tie_word_embeddings": false, "transformers_version": "5.16.0.dev0", "vision_config": { "attention_dropout": 0.0, "hidden_act": "silu", "hidden_size": 64, "img_max_token_num": 4096, "intermediate_size": 128, "interpolate_mode": "bilinear", "max_image_size": 64, "max_vit_seq_len": 16, "min_image_size": 64, "model_type": "hunyuan_vl_vision", "num_attention_heads": 4, "num_channels": 3, "num_hidden_layers": 2, "num_key_value_heads": 4, "out_hidden_size": 64, "patch_size": 16, "rms_norm_eps": 1e-05, "spatial_merge_size": 1, "temporal_patch_size": 1, "text_hidden_size": 64 } }