update the code - SemanticDiff

Commit

19 days ago

update the code Signed-off-by: Dong, Bo1 <bo1.dong@intel.com>

References

#2090 - Opt the moe performance

#2108 - ark.moe: switch auto dispatch to total-token threshold + env override

#2111 - Optimize INT4 MoE prefill/decode (sym) XPU kernels with dedicated w4a16 tiling

#2143 - feat: W4A8 ARK XPU MoE kernel (int4 weight / int8 compute) with prefill + decode

Author

a32543254

a32543254

Committer

a32543254

a32543254

Parents

Loading