-
Notifications
You must be signed in to change notification settings - Fork 5
Open
Description
When trying to debug compute_sharded_comparison_test.py I get the following error:
"RuntimeError: CUDA error: invalid device ordinal"
After a minute the following message shows up and the process hangs:
"GPU 0 loaded model."
I also tried following the run instructions in the readme, but in both cases, the process hangs after the gpus load the model.
The problem persists when running on a single gpu vs multiple gpus.
Metadata
Metadata
Assignees
Labels
No labels