fix: correct gqa_decode_attn reduction + add gqa_prefill_attn
- gqa_decode_attn: rewrite to per-KV-head, K in smem - gqa_prefill_attn: new kernel for Q_len > 1 with GQA
This commit is contained in:
@@ -27,3 +27,4 @@ def register(name: str, sources: list[str] | None = None, **kwargs):
|
||||
|
||||
|
||||
register("gqa_decode_attn")
|
||||
register("gqa_prefill_attn")
|
||||
|
||||
Reference in New Issue
Block a user