issues
search
fhswf
/
tagflip-autonlp
Automate NLP tasks
https://autonlp.informatik.fh-swf.de
MIT License
3
stars
2
forks
source link
fix(training): use GPU for training
#29
Closed
cgawron
closed
2 years ago
cgawron
commented
2 years ago
Remove explicit setting of CUDA_AVAILABLE_DEVICES in training script. In trainings started via mlflow in docker, this conflicts with the external settings. This results in training running on CPU only.