modelscope/Nexus-Gen

1{2  "_name_or_path": "DiffSynth-Studio/Nexus-Gen",3  "architectures": [4    "Qwen2_5_VLForConditionalGeneration"5  ],6  "attention_dropout": 0.0,7  "auto_map": {8    "AutoConfig": "configuration_qwen2_5_vl.Qwen2_5_VLConfig",9    "AutoModel": "modeling_qwen2_5_vl.Qwen2_5_VLModel",10    "AutoModelForCausalLM": "modeling_qwen2_5_vl.Qwen2_5_VLForConditionalGeneration"11  },12  "bos_token_id": 151643,13  "eos_token_id": 151645,14  "hidden_act": "silu",15  "hidden_size": 3584,16  "image_token_id": 151655,17  "initializer_range": 0.02,18  "intermediate_size": 18944,19  "max_position_embeddings": 128000,20  "max_window_layers": 28,21  "model_type": "qwen2_5_vl",22  "num_attention_heads": 28,23  "num_hidden_layers": 28,24  "num_key_value_heads": 4,25  "pad_token_id": 151643,26  "rms_norm_eps": 1e-06,27  "rope_scaling": {28    "mrope_section": [29      16,30      24,31      2432    ],33    "rope_type": "default",34    "type": "default"35  },36  "rope_theta": 1000000.0,37  "sliding_window": 32768,38  "tie_word_embeddings": false,39  "torch_dtype": "bfloat16",40  "transformers_version": "4.49.0",41  "use_cache": false,42  "use_sliding_window": false,43  "video_token_id": 151656,44  "vision_config": {45    "hidden_size": 1280,46    "in_chans": 3,47    "model_type": "qwen2_5_vl",48    "spatial_patch_size": 14,49    "tokens_per_second": 2,50    "torch_dtype": "bfloat16"51  },52  "vision_end_token_id": 151653,53  "vision_start_token_id": 151652,54  "vision_token_id": 151654,55  "vocab_size": 15206456}57