基于llama.cpp的一些和提示词相关的节点，目前包括美化提示词和类似clip-interrogator的图片反推

Use llama.cpp to assist in generating some nodes related to prompt words, including beautifying prompt words and image recognition similar to clip-interrogator

• [2024-06-22] 新增Florence-2-large图片反推模型节点 (Added Florence-2-large image interrogation model node)
• [2024-06-20] 新增选择本机ollama模型的节点 (Added nodes to select local ollama models)
• [2024-06-05] 新增千问2.0预设模型 (Added Qianwen 2.0 preset model)
• [2024-06-05] 可选chat_format,图片反推后处理 (Optional chat_format, post-processing after image interrogation)
• [2024-06-04] 新增了一些预设模型 (Added some preset models)
• [2024-06-04] 新增通用节点,支持手动选择模型 (Add universal node, support manual selection of models)
• [2024-05-30] 添加ImageCaptionerConfig节点来支持批量生成提示词 (Add ImageCaptionerConfig node to support batch generation of prompt words)
• [2024-05-24] 运行后在当前节点显示生成的提示词 (Display the generated prompt words in the current node after running)
• [2024-05-24] 兼容清华智谱API (Compatible with Zhipu API)
• [2024-05-24] 使用A1111权重缩放,感谢ComfyUI_ADV_CLIP_emb (Use A1111 weight scaling, thanks to ComfyUI_ADV_CLIP_emb)
• [2024-05-13] 新增OpenAI API节点 (add OpenAI API node)
• [2024-04-30] 支持自定义指令 (Support for custom instructions)
• [2024-04-30] 添加llava-v1.6-vicuna-13b (add llava-v1.6-vicuna-13b)
• [2024-04-30] 添加翻译
• [2024-04-28] 新增Phi-3-mini节点 (add Phi-3-mini node)

1. Clone this repo into custom_nodes folder.
2. Restart ComfyUI.

• MZ_Florence2CLIPTextEncode

• ModelConfigManualSelect (Ollama)

• CLIPTextEncode (LLamaCPP Universal)

• ModelConfigManualSelect(LLamaCPP)

• ModelConfigDownloaderSelect(LLamaCPP)

• CLIPTextEncode (ImageInterrogator)

• ModelConfigManualSelect(ImageInterrogator)

• ModelConfigDownloaderSelect(ImageInterrogator)

• CLIPTextEncode (OpenAI API)

  

• CLIPTextEncode (Phi-3)

  

• CLIPTextEncode (LLama3)

  

• ImageInterrogator (LLava)

  Enable parameter sd_format

• ImageCaptionerConfig

• LLamaCPPOptions

  

• CustomizeInstruct

  

• BaseLLamaCPPCLIPTextEncode (可以手动传入模型路径/You can directly pass in the model path)

• BaseLLavaImageInterrogator (可以手动传入模型路径/You can directly pass in the model path)

升级llama_cpp_python的版本到最新版本,前往 https://github.com/abetlen/llama-cpp-python/releases 下载安装

CUDA版本切换到12.1,如果你使用秋叶启动器,高级设置->环境维护->安装PyTorch->选择版本中选择CUDA 12.1的版本

保持网络畅通,该上魔法上魔法,或者手动安装llama_cpp_python

• https://github.com/comfyanonymous/ComfyUI
• https://github.com/ggerganov/llama.cpp
• https://github.com/BlenderNeko/ComfyUI_ADV_CLIP_emb

• 绿泡泡: minrszone
• Bilibili: minus_zone
• 小红书: MinusZoneAI
• 爱发电: MinusZoneAI