AutoDL云平台deepfacelab_linux教程（远程桌面）

无双爱吃糖呀

(deepfacelab) root@autodl-container-1d3449b00a-57f9985d:~# conda install cudnn==8.0.4
Collecting package metadata (current_repodata.json): done
Solving environment: failed with initial frozen solve. Retrying with flexible solve.
Collecting package metadata (repodata.json): done
Solving environment: failed with initial frozen solve. Retrying with flexible solve.

PackagesNotFoundError: The following packages are not available from current channels:

  - cudnn==8.0.4

Current channels:

  - https://mirrors.ustc.edu.cn/anaconda/pkgs/main/linux-64
  - https://mirrors.ustc.edu.cn/anaconda/pkgs/main/noarch
  - https://mirrors.ustc.edu.cn/anaconda/pkgs/free/linux-64
  - https://mirrors.ustc.edu.cn/anaconda/pkgs/free/noarch
  - https://repo.anaconda.com/pkgs/main/linux-64
  - https://repo.anaconda.com/pkgs/main/noarch
  - https://repo.anaconda.com/pkgs/r/linux-64
  - https://repo.anaconda.com/pkgs/r/noarch

To search for alternate channels that may provide the conda package you're
looking for, navigate to

    https://anaconda.org

and use the search bar at the top of the page.

没有8.0.4啊

xiaobaidamowang

遇到几个小问题，不过搞定了，30、40系列的弄到最后打开训练不知道为什么没发成功调动线程。。还有一个问题：20系列能训练模型了，但是为什么除了train SAEHD以外scripts里面的程序打不开额？

大赤焰

我测试提示找不到文件是咋回事。。

朱雀陵光

运行解压人脸素材包出错，当然试运行train也是这样报错，无法运行，请指教，谢谢。已经是在有卡模式，显卡是2080ti 22G


解决了，是缺少numexpr模块

用pip install 安装一下就好了

又出现了新的问题，可以解包人脸，但是仍然无法跑训练，大篇的报错。真的想放弃了，太麻烦了

baka9

原来教程在这里

读书人荣哥

试了3080和3090以及4090都会报错Check failed: ret == 0 (11 vs. 0)Thread creation via pthread_create() failed.实际再往上看能够看到真正的报错原因OpenBLAS blas_thread_init: pthread_create failed for thread 1 of 12: Resource temporarily unavailable OpenBLAS blas_thread_init: RLIMIT_NPROC -1 current, -1 max,应该是平台容器配置的问题,解决方法:在启动脚本前面加export OPENBLAS_NUM_THREADS=1