[Bugfix][ROCm] Fix a few int4/int8 quantization errors (#52112)
Signed-off-by: Qiang Li <qiang.li2@amd.com>
This commit is contained in:
@@ -222,7 +222,6 @@ class MoeWNA16Method(FusedMoEMethodBase):
|
||||
else:
|
||||
scale = kInt4StaticGroupScale
|
||||
elif num_bits == 8:
|
||||
assert group_size == -1
|
||||
quant_type = INT8_DTYPE
|
||||
scale = kInt8StaticGroupScale
|
||||
else:
|
||||
|
||||
Reference in New Issue
Block a user