Skip to content

clip模型safetensors,会大幅提升速度 #2

Description

@xy9144

原8.55g大小gguf模型,反量化为bf16,再w4a8量化为safetensors模型大小10.8g
测试相同默认参数gguf模型用时42秒,safetensors模型仅需24秒

先安装原版节点和模型正常跑通gguf工作流,超分节点如果找不节点安装可以去掉

用本修改的节点包,替换原版节点2个脚本文件
llada_gguf_adapter.py
llada/pipeline_llada_image.py

然后用参考脚本反量化gguf模型为bf16
再用参考脚本w4a8量化为safetensors

LLaDa-Image_ComfyUI.zip

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions