RVC-Project/Retrieval-based-Voice-Conversion-WebUI

Training is freezing the computer when training for more than 20 epochs.

Open

#1,711 创建于 2024年1月14日

在 GitHub 查看
 (0 评论) (0 反应) (0 负责人)Python (18,427 star) (2,849 fork)batch import
AMD/Intelhelp wanted

描述

Are there any options I can add to the invoking command to limit the VRAM usage. Like the way Automatic1111 for Stable Diffusion has --lowvram option.

Because when I'm trying to train a model for more than 20 epochs, it is freezing the entire computer. I have RX 6500 XT with 4GB, Main RAM is 16GB.

贡献者指南

技术栈
pythonpytorch
领域
machine learningperformance
议题类型
feature
难度面向新贡献者的预计实现难度，1 表示很小改动，5 表示专家级工作。
4
预计时间有经验贡献者完成调查、实现、测试并准备 pull request 的粗略时间范围。
3-5 days
活动状态议题当前的可参与程度：新鲜、活跃、陈旧、阻塞或等待维护者输入。
stale
清晰度议题是否清楚说明期望改动、验收标准和下一步。
mostly clear
前置要求
PythonPyTorchmemory managementCUDA programming
新手友好度1-100 的估计分数，表示该议题对首次贡献者的友好程度。
25
研究方向
The issue requests a lowvram option to limit VRAM usage during training, similar to Automatic1111's option. To implement, investigate the training script (likely in `train.py` or `infer.py`) to identify where model and data batches are loaded. Consider adding a command line argument using argparse that enables memory saving techniques like gradient checkpointing, mixed precision, or reducing batch size dynamically. Check the repository's existing argument parsing for guidance.