deepspeedai / deepspeedai/DeepSpeed
can deepspeed fast-gen support int8 weightonly inference
Open
@HeyangQin is already working on this.
Since Dec 15, 2023.
enhancement
inference
- Dominant language
- Python
- Stars
- 43.1k
- Forks
- 5k
- Avg merge
- 4d 15h
- Merged PRs (30d)
- 112
Description
I test the fastgen on codellama-34b a100, the speed is as:
tensorrt llm 1a100 19token/s, fast-gen 21token/s
the speed of 2a100 on fast-gen is 36token/s
however the speed of tensorrt llm 1a100 is 46token/s on int4 quant, 30token/s on int8.
int4/int8 quantization is the best inference speedup strategy, so can fast-gen support int4/int8 weightonly.
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.