[torch.compile] Compile CustomOp.forward_native for SiluAndMul and QuantFP8 to avoid raw torch ops inside opaque custom ops (#32806)

Signed-off-by: Luka Govedič <lgovedic@redhat.com>
Signed-off-by: Luka Govedič <ProExpertProg@users.noreply.github.com>
Co-authored-by: Michael Goin <mgoin64@gmail.com>
This commit is contained in:
Luka Govedič
2026-01-22 19:52:26 -08:00
committed by GitHub
co-authored by Michael Goin
parent f61c9da711
commit 5e4e0e51f4
7 changed files with 52 additions and 13 deletions
+1
View File
@@ -336,6 +336,7 @@ class MatcherQuantFP8(MatcherCustomOp):
quant_key.scale.group_shape,
column_major_scales=has_col_major_scales,
use_ue8m0=is_e8m0,
compile_native=False,
)
def forward_rocm_aiter(