auto-round
fb2fca6a
- perf: cut W4A8 prefill traffic with int8-in and fused top-k reduce
Go
Login via GitHub
Home
Pricing
FAQ
Install
Login
via GitHub
Commit
View On
GitHub
Commit
4 days ago
perf: cut W4A8 prefill traffic with int8-in and fused top-k reduce Co-authored-by: a32543254 <53296245+a32543254@users.noreply.github.com>
References
#2143 - feat: W4A8 ARK XPU MoE kernel (int4 weight / int8 compute) with prefill + decode
Author
Copilot
Parents
4318efaa
Loading