[Feature] Support JIT set kv cache (#16273)
This commit is contained in:
@@ -53,6 +53,7 @@ def flashinfer_qknorm(
|
||||
q_weight: torch.Tensor,
|
||||
k_weight: torch.Tensor,
|
||||
) -> None:
|
||||
from flashinfer import rmsnorm
|
||||
|
||||
rmsnorm(q, q_weight, out=q)
|
||||
rmsnorm(k, k_weight, out=k)
|
||||
|
||||
Reference in New Issue
Block a user