deepseek-ai / deepseek-ai/profile-data
Why there is such huge difference between dispatch (F) and dispatch (B) execution time in training profile data
Open
- Dominant language
- No language data
- Stars
- 1.2k
- Forks
- 152
- PR merge metrics
- No merged PRs in 30d
Description
In the profile data, dispatch (F) takes about 21-23 ms to finish while dispatch (B) finished within 0.16 ms. I expect they should be finished nearly the same time, or not too far. Is there any reason you can share with us? :-)
Contributor guide
No contributing guide indexed for this repository
Assessment
This issue has not been assessed yet.