有N卡,且内存在6G以上,并安装了对应的CUDA环境 可以登录外网 一、下载so-vits-svc 1. 创建新的conda环境 首先你需要在电脑上安装Anaconda,生成一个新的环境,方便不同Python包的版本管理: conda create --name so-vits-svc 激活该环境: conda activate so-vits-svc 2. Git clone项目 首先把该项目下载到本地...
一:前期准备 1.安装CUDA驱动 首先我们要检查显卡的CUDA Toolkit版本,命令行界面输入:nvidia-smi,回车。通过表格右上角的CUDA Version来确定你应该下载的CUDA Toolkit驱动版本 CUDA驱动地址:developer.nvidia.com/cu 选择“自定义”,然后全选,点击下一步 一直下一步就行了 2.安装ffmpeg 下载地址:gyan.dev/ffmpeg/bui...
答:数据集切片切太长了,5-10秒差不多。 报错:CUDA error: CUBLAS_STATUS_NOT_INITIALIZED when calling 'cublasCreate(handle)' 答:爆显存了,基本上跟CUDA有关的报错大都是爆显存…… 报错:torch.multiprocessing.spawn.ProcessExitedException: process 0 terminated with exit code 3221225477 答:调大虚拟内存,管...
打开“软件和更新” 点击列表第五个的“附加驱动”,选择“Nvidia driver metapackage 来自 515 (专有)”驱动,并应用(这个515自带的CUDA版本应该是11.7,正好适用于我们安装so-vits-svc的环境) 4. 安装好重启,打开终端,输入nvidia-smi,如果驱动正常安装,这里会出现Driver版本和CUDA版本还有显卡的使用情况(如果想要一直...
前往NVIDIA-CUDA 官网下载与系统对应的Cuda 版本 以Cuda-11.7 版本为例,根据自己的系统和需求选择安装(一般本地 Windows 用户请依次选择Windows, x86_64, 系统版本, exe(local)) 安装成功之后在 cmd 控制台中输入nvcc -V, 出现类似以下内容则安装成功:nvcc...
# 系统环境 Ubuntu 22.04 LTS # 显卡环境 NVIDIA-SMI 470.182.03 Driver Version: 470.182.03 CUDA Version: 11.4 # python环境 Python 3.10 # GPU环境 Tesla T4 16G * 1 # CPU环境 8核32GB 365404354399d97e8333b.png 87053539fd2c945364ffb.png 音频处理 为了训练,我们需要将音频文件分离成人声和伴奏两个音...
load WARNING:xformers:WARNING[XFORMERS]: xFormers can't load C++/CUDA extensions. xFormers was built for: PyTorch 1.13.1+cu117 with CUDA 1107 (you have 2.0.1+cu117) Python 3.10.9 (you have 3.10.6) Please reinstall xformers (see https://github.com/facebookresearch/xformers#in...
However, excessively long clips may lead to issues such as torch.cuda.OutOfMemoryError. To facilitate the slicing process, you can use audio-slicer-GUI or audio-slicer-CLI In general, only the Minimum Interval needs to be adjusted. For spoken audio, the default value usually suffices, while...
Too long may lead to torch.cuda.OutOfMemoryError during training or even pre-processing. By using audio-slicer-GUI or audio-slicer-CLI In general, only the Minimum Interval needs to be adjusted. For statement audio it usually remains default. For singing audio it can be adjusted to 100 or...
torch.cuda.set_device(rank) collate_fn = TextAudioCollate() all_in_mem = hps.train.all_in_mem # If you have enough memory, turn on this option to avoid disk IO and speed up training. train_dataset = TextAudioSpeakerLoader(hps.data.training_files, hps, all_in_mem=all_in_mem) ...