- Attaching nvidia-bug-report.log.gz would definitely increase chances of people looking into this.
- Where have you obtained driver version 572.16 from? I couldn’t find any downloads for it…
- have you verified that it works in a simpler setup? for example directly on Linux without Win+WSL? Also, would it be feasible to recreate the environment of the container directly in Linux to omit the docker layer? (it’s an honest question, I have no idea TBH).
[WSL2 + Docker + TensorFlow] CUDA failed to initialize in TensorFlow container (RTX 5090, CUDA 12.8)
1 Like