good first issue
Repository metrics
- Stars
- (6,010 stars)
- PR merge metrics
- (平均マージ 1d 11h) (30d で 31 merged PRs)
説明
Currently the context parallelism's training time is like 2 times when we use cp=2 than cp=1, we should try to optimize this for long sequence, maybe reduce the all gather entropy and logp and calculate the local loss to reduce the numbers of communication