alibaba / alibaba/EasyTransfer
8卡的线性度很低
Open
- Dominant language
- Python
- Stars
- 861
- Forks
- 162
- PR merge metrics
- No merged PRs in 30d
Description
跑fashionbert的多卡，发现8卡的性能跟4卡的性能差不多。
请问是用BundleCSVReader读数据部分的代码的问题吗？
请问这份代码在多卡的情况下验证过吗？
Contributor guide
No contributing guide indexed for this repository
Research direction
Start by reproducing the FashionBERT multi-GPU run at four and eight cards, then inspect the BundleCSVReader data-loading path and the multi-card training entry point. Compare throughput and scaling between both runs; the issue is done when the cause of the similar performance is identified and the validated behavior is documented or fixed.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python
- Domain
- distributed-systems, machine-learning, performance
- Issue type
- Bug
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Stale
- Clarity
- Needs clarification
- Newbie friendliness
- 25/100