bytedance / flux

A fast communication-overlapping library for tensor parallelism on GPUs.
Apache License 2.0
223 stars 17 forks source link

[QUESTION]is there a plan to support int8? #31

Closed Rainlin007 closed 2 months ago

Rainlin007 commented 3 months ago

There are only fp16 and bfp16 in the benchmark, is there a plan to support int8?

wenlei-bao commented 3 months ago

@Rainlin007 Thanks for your interests. We do have int8 support internally. So far we don't have a plan to open source that part yet.