[Perf] Reduce redundant Triton sampler warmup specializations (#58605)
Signed-off-by: mgoin <mgoin64@gmail.com> Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
This commit is contained in:
co-authored by
mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
parent
d882bddbea
commit
1ee7f78e80
@@ -103,7 +103,7 @@ def _update_min_larger_stats(data, above_mask, min_larger, num_min_larger, senti
|
||||
return min_larger, num_min_larger
|
||||
|
||||
|
||||
@triton.jit
|
||||
@triton.jit(do_not_specialize_on_alignment=["BATCH_SIZE"])
|
||||
def _topk_topp_kernel(
|
||||
LOGITS,
|
||||
LOGITS_STRIDE_0,
|
||||
@@ -1225,7 +1225,7 @@ def _topp_sb_combine(
|
||||
)
|
||||
|
||||
|
||||
@triton.jit
|
||||
@triton.jit(do_not_specialize_on_alignment=["ROUND"])
|
||||
def _topp_sb_step_kernel(
|
||||
LOGITS,
|
||||
LOGITS_STRIDE_0,
|
||||
|
||||
Reference in New Issue
Block a user