Introducing common troubles using WSL (Windows Subsystem for Linux) and how to deal with them. When new troubles occur, I would like to add them as needed.
After starting WSL from PowerShell, you may not be able to log in without displaying an error message, or you may suddenly be forcibly logged out. The cause is unknown, but the problem can be solved by starting LxssManager according to the following procedure.
> net start LxssManager
> net stop LxssManager > net start LxssManager
When executing the run command of nvidia-docker from WSL, the following message may be displayed and the Docker container may not be started.
docker: Error response from daemon: OCI runtime create failed: container_linux.go ～
When connecting to GPU from WSL with nvidia-docker, it is necessary to install CUDA on Windows Subsystem for Linux (WSL) --Public Preview instead of the standard driver, but WSL is automatically updated by the driver. It seems that it may be updated to an unsupported driver.
The problem has been solved by downloading and updating the driver from CUDA on Windows Subsystem for Linux (WSL) --Public Preview.
We have summarized the troubles and countermeasures encountered while using WSL as a deep learning environment. Since the Insider Preview frequently updates Windows, new troubles may occur, but I have the impression that there are fewer troubles than I had expected.
Recommended Posts