Skip to content

Commit 96f4ac8

Browse files
committed
fix shape
Signed-off-by: jiqing-feng <[email protected]>
1 parent 1f30627 commit 96f4ac8

File tree

1 file changed

+1
-1
lines changed

1 file changed

+1
-1
lines changed

bitsandbytes/backends/cpu_xpu_common.py

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -401,7 +401,7 @@ def quantize_4bit_impl(
401401
def dequant_8bit(A, offset, quant_state):
402402
assert A.dtype == torch.uint8
403403
absmax = quant_state.code[A.reshape(-1).int()]
404-
absmax = (absmax.view(-1, 256) * quant_state.absmax.view(-1, 1)).reshape(quant_state.shape).to(quant_state.dtype)
404+
absmax = (absmax.view(-1, 256) * quant_state.absmax.view(-1, 1)).to(quant_state.dtype).reshape(A.shape)
405405
absmax += offset
406406
return absmax
407407

0 commit comments

Comments
 (0)