[ROCm] Enable bitsandbytes quantization support on ROCm (#34688)

Signed-off-by: badaoui <abdennacerbadaoui0@gmail.com>
This commit is contained in:
BADAOUI Abdennacer
2026-02-21 00:34:55 -08:00
committed by GitHub
parent 2aab2bb543
commit 8dc8a99b56
8 changed files with 29 additions and 44 deletions
+1 -3
View File
@@ -244,10 +244,8 @@ class RocmPlatform(Platform):
"mxfp4",
"petit_nvfp4",
"torchao",
"bitsandbytes",
]
# bitsandbytes not supported on gfx9 (warp size 64 limitation)
if not on_gfx9():
supported_quantization += ["bitsandbytes"]
@classmethod
def import_kernels(cls) -> None: