Common Issues¶

CUDA Error: no kernel image is available¶

You might receive this error when using an old version of the container or if you installed pytorch in a virtual environment.

To resolve this, please either:

• Re-install pytorch following these directions.
• Try re-running your job with a GPU type of lower compute capability, such as gtx_1080_ti.

NVIDIA A100 is not compatible with the current PyTorch¶

You might receive this error when using an old version of the container or if you installed pytorch in a virtual environment.

To resolve this, please either:

• Re-install pytorch following these directions.
• Try re-running your job with a GPU type of lower compute capability, such as gtx_1080_ti.

Job on hold - AssocGrpGRES¶

Due to high demand for GPUs, labs are limited in the number of concurrent GPU jobs they can have running. This limit is presently the number of GPUs the lab funded plus two.