[Docs] Fix stale W4A16 NVFP4 default kernel in ModelOpt docs (#59135)
Signed-off-by: Rajeev Warrier <warrier.rajeev.95@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
d648847fe4
commit
30e956fad3
@@ -2314,9 +2314,9 @@ def select_linear_kernel(
|
||||
w = spec.weight
|
||||
assert isinstance(w, QuantKey), f"resolve() must supply a weight key, got {w!r}"
|
||||
if w.dtype == FP4_DTYPE:
|
||||
# W4A16 (activation is None) → use_a16=True defaults to Marlin *and*
|
||||
# honors --linear-backend (matches upstream ModelOptNvFp4W4A16LinearMethod
|
||||
# after #50273); W4A4 → use_a16=False.
|
||||
# W4A16 (activation is None) → use_a16=True, auto prefers FlashInfer
|
||||
# CuTe-DSL backend on SM100/103 when available and Marlin otherwise,
|
||||
# and honors --linear-backend (#50273); W4A4 → use_a16=False.
|
||||
return init_nvfp4_linear_kernel(use_a16=spec.activation is None)
|
||||
if w.scale.dtype == MXFP8_SCALE_DTYPE:
|
||||
return init_mxfp8_linear_kernel(
|
||||
|
||||
Reference in New Issue
Block a user