tensorflow / tensorflow/tensorboard
Trace enhancements - idle time details, cublas, cudnn, etc ops display.
Open
@qiuminxu is already working on this.
Since Apr 8, 2020.
core:backend
stat:awaiting tensorflower
type:feature
- Dominant language
- TypeScript
- Stars
- 7.2k
- Forks
- 1.7k
- Avg merge
- 4d 22h
- Merged PRs (30d)
- 1
Description
Current Tensorflow traces show idle times, which may or may not indicate opportunity to optimize. Is there a way to show what is the cause of the idle time ? Most TF ops are usually implemented with platform specific libraries (e.g. cublas, cudnn, etc.) . Most platforms have there own profiling tools (e.g. nvvp) that traces executions at platform specific levels. It is virtually impossible to map between the two, but that mapping is invaluable in improving performance.
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Assessment
This issue has not been assessed yet.