wd-llm-caption打标PYPI源V4

这是由DuKeG大佬做的pypi源，感谢DuKeG大佬！

首先先吹嘘一波，无脑吊打无脑吊打无脑吊打无脑吊打无脑吊打打标一键包！

这是由DuKeG大佬做的集合包，感谢DuKeG大佬！

推荐的理由是WD14+LLAMA3.2+ joy-caption+千问结合包，你可以组合使用和单独使用。

新增joy-two

优点是安装简单，很多人huggingface模型你无法下载，而且受限下载速度很慢，你可以通过阿里的modelscope下载，国内站速度很快。

缺点就是如果你想组合用，显存要大。

组合显存要求：

16G 可量化4bit

24G 可量化8bit

30G 非量化

如果你只用wd14 cpu就行

安装步骤：

你只要装了python就行

大部分人都装了stable diffusion，这个使用了torch==2.4.0和cu124，为避免冲突，创建虚拟环境

#创建虚拟环境和激活虚拟环境 window用户 右键powershell安装

python -m venv .venv

.\venv\Scripts\activate

#创建虚拟环境和激活虚拟环境 linux用户终端安装

python -m venv .venv

source .venv/bin/activate

安装提供的文件whl文件或者从下面的网址下载：

PIP源地址：https://pypi.org/manage/project/wd-llm-caption/releases/ 搜索wd-llm-caption

window用户右键powershell运行

!pip install /content/wd_llm_caption-0.1.1a0-py3-none-any.whl #路径换成自己的路径

!python wd-llm-caption-gui --share #gui的入口是wd-llm-caption-gui

#cli的入口命令是wd-llm-caption +训练集路径，其他参数参考--help说明

### Recommended by DuKeG

This comprehensive package is created by the esteemed DuKeG, and we extend our heartfelt gratitude to him!

#### Recommended Reasons:

- WD14 + LLAMA3.2 + joy-caption + Qianwen Combined Package: You can use these models in combination or individually.

- Simple Installation: Many people face difficulties downloading models from Hugging Face due to slow speeds and restrictions. This package allows you to download models from Alibaba's ModelScope, which offers much faster speeds within China.

- Memory Requirements: If you plan to use the models in combination, you will need a significant amount of GPU memory.

#### Memory Requirements for Combination:

- 16GB: Can be quantized to 4-bit.

- 24GB: Can be quantized to 4-bit.

- 30GB: Non-quantized.

If you only use WD14, a CPU is sufficient.

### Installation Steps:

You only need Python installed. Most people already have Stable Diffusion installed, which uses torch==2.4.0 and cu124. To avoid conflicts, create a virtual environment.

#### Create and Activate Virtual Environment (Windows Users):

Right-click on PowerShell and run:

```powershell

python -m venv .venv

.\venv\Scripts\activate

```

#### Create and Activate Virtual Environment (Linux Users):

Run in terminal:

```bash

python -m venv .venv

source .venv/bin/activate

```

#### Install the Provided WHL File or PIP源地址：https://pypi.org/manage/project/wd-llm-caption/releases/ 搜索wd-llm-caption:

For Windows users, right-click on PowerShell and run:

```powershell

!pip install /content/wd_llm_caption-0.1.1a0-py3-none-any.whl # Replace the path with your own path

!wd-llm-caption-gui --share # The GUI entry point is wd-llm-caption-gui

```

#### CLI Entry Command:

The CLI entry command is wd-llm-caption followed by the training set path. For other parameters, refer to the --help documentation.