auto-round
b1f816bb - perf: vectorize Qwen4-Exp QSA indexer to speed up full-attention blocks (#2579)

Commit

2 days ago

perf: vectorize Qwen4-Exp QSA indexer to speed up full-attention blocks (#2579) Signed-off-by: lvliang-intel <liang1.lv@intel.com>

References

Author

lvliang-intel

Parents