google-deepmind / google-deepmind/kinetics-i3d
out of memory
Open
- Dominant language
- Python
- Stars
- 1.8k
- Forks
- 467
- PR merge metrics
- No merged PRs in 30d
Description
When I ran evaluate_sample, I got "CUDA_ERROR_OUT_OF_MEMORY", but the results is correct. My GPU has 8GB memory.
Is it correct that the demo will cost that amount of memory? If so, does that mean I can not train by using this single GPU?
Contributor guide
Assessment
This issue has not been assessed yet.