fix bug
This commit is contained in:
parent
234004ceb5
commit
d28ee06202
|
|
@ -108,9 +108,9 @@ def matmul4bit(x, qweight, scales, zeros, groupsize=-1):
|
|||
if np.prod(x.shape[:-1]) > auto_switch_thd:
|
||||
output = _matmul4bit_v1_recons(x, qweight, scales, zeros)
|
||||
else:
|
||||
output = _matmul4bit_v1(x, qweight, scales.float(), zeros)
|
||||
output = _matmul4bit_v1(x, qweight, scales.float(), zeros.float())
|
||||
else:
|
||||
output = _matmul4bit_v1(x, qweight, scales.float(), zeros)
|
||||
output = _matmul4bit_v1(x, qweight, scales.float(), zeros.float())
|
||||
else:
|
||||
# use v2
|
||||
if use_new:
|
||||
|
|
|
|||
Loading…
Reference in New Issue