[bug] group_size = 64 has bug
Open
backlog
- Dominant language
- Python
- Stars
- 8.1k
- Forks
- 748
- Avg merge
- 6d 2h
- Merged PRs (30d)
- 54
Description
只修改这两行代码,LLaMa-2-7B 模型无法得到正确的输出。
` Generate<64, Ops>(kernels_); group_sizes_.push_back(64);`
FP16 的输出:
@Input:
The first time I saw the movie, I was like, 'Oh my God,
_Output:
this is so cool.' I was like,
INT4 GroupSize 64 的输出:
@Input:
The first time I saw the movie, I was like, 'Oh my God,
_Output:
I was a little, I was a little
Contributor guide
Assessment
This issue has not been assessed yet.