vscode accelerate deepspeed配置

accelerate训练

bash 复制代码
{
    // Use IntelliSense to learn about possible attributes.
    // Hover to view descriptions of existing attributes.
    // For more information, visit: https://go.microsoft.com/fwlink/?linkid=830387
    "version": "0.2.0",
    "configurations": [
        {
            "name": "Python: Current File",
            "type": "python",
            "request": "launch",
            // "program": "${file}",
            "module": "accelerate.commands.launch",
            "console": "integratedTerminal",
            "env": {
                "CUDA_VISIBLE_DEVICES": "4,5,6,7"  // 与
            },
            "args": [
                 train sdxl lora
                "--num_processes=4",
                "examples/text_to_image/train_text_to_image_lora_sdxl.py",
                "--dataset_name","lambdalabs/naruto-blip-captions",
                "--output_dir","exp",
                "--resolution","1024", "--random_flip",
                "--train_batch_size","1",
                "--num_train_epochs","2", "--checkpointing_steps","500",
                "--learning_rate","1e-04", "--lr_scheduler","constant", "--lr_warmup_steps","0",
                "--mixed_precision","fp16",
                "--seed","42",
                "--output_dir","sd-naruto-model-lora-sdxl",
                "--validation_prompt","cute dragon creature",
            ]   
        }
    ]
}

accelerate+deepspeed训练

bash 复制代码
{
    // Use IntelliSense to learn about possible attributes.
    // Hover to view descriptions of existing attributes.
    // For more information, visit: https://go.microsoft.com/fwlink/?linkid=830387
    "version": "0.2.0",
    "configurations": [
        {
            "name": "Python: Current File",
            "type": "python",
            "request": "launch",
            // "program": "${file}",
            "module": "accelerate.commands.launch",
            "console": "integratedTerminal",
            "env": {
                "CUDA_VISIBLE_DEVICES": "4,5,6,7"
            },
            "args": [
                 train sdxl lora
                "--config_file","my/deepspeed_config/default_config.yaml",  //跟上面只有这个区别
                "--num_processes=4",
                "examples/text_to_image/train_text_to_image_lora_sdxl.py",
                "--dataset_name","lambdalabs/naruto-blip-captions",
                "--output_dir","exp",
                "--resolution","1024", "--random_flip",
                "--train_batch_size","1",
                "--num_train_epochs","2", "--checkpointing_steps","500",
                "--learning_rate","1e-04", "--lr_scheduler","constant", "--lr_warmup_steps","0",
                "--mixed_precision","fp16",
                "--seed","42",
                "--output_dir","sd-naruto-model-lora-sdxl",
                "--validation_prompt","cute dragon creature",
            ]   
        }
    ]
}
相关推荐
张紫娃5 小时前
idea 如何看一个类的 .class 文件
java·ide·intellij-idea
NoteDeep 深度笔记10 小时前
DOM 被改乱了怎么办?Notedeep 编辑器的 normalize 设计
前端·javascript·编辑器
顧棟14 小时前
组件生命周期设计:从硬编码初始化到状态机
java·ide
jianqiang.xue1 天前
Espressif-IDE进入维护模式:乐鑫把ESP32开发的宝押给VS Code
前端·css·ide
Solange__3 天前
Blender 3D模型优化:普通复制、共享网格、实例、合并怎么选?
人工智能·3d·ai·自然语言处理·编辑器·blender·材质
2601_967980114 天前
世界上最好的Python编辑器是什么?我投PyCharm一票
ide·python·pycharm·数据科学·jupyter notebook
FungLeo4 天前
成为全栈·Vue 管理后台篇·编辑器图片上传:从文件选择到文章附件关联
前端·vue.js·编辑器·markdown·文件上传·formdata·附件管理
拼图2096 天前
VSCode 正则表达式使用指南
ide·vscode·正则表达式