diff options
| author | justheuristic <justheuristic@gmail.com> | 2022-09-17 23:38:09 +0300 |
|---|---|---|
| committer | justheuristic <justheuristic@gmail.com> | 2022-09-17 23:38:09 +0300 |
| commit | eac9aca460ee7afb6d0cbc61ae43a95120d34f29 (patch) | |
| tree | ec0dd8c9865adbf14769b50fc85e2cd8bf26fc16 /bitsandbytes/autograd | |
| parent | a9fe0ff98c3293d972eb7a638b9887df0bc0d30d (diff) | |
cast bias too
Diffstat (limited to 'bitsandbytes/autograd')
| -rw-r--r-- | bitsandbytes/autograd/_functions.py | 2 |
1 files changed, 2 insertions, 0 deletions
diff --git a/bitsandbytes/autograd/_functions.py b/bitsandbytes/autograd/_functions.py index dc79bb1..6d9229b 100644 --- a/bitsandbytes/autograd/_functions.py +++ b/bitsandbytes/autograd/_functions.py @@ -234,6 +234,8 @@ class MatMul8bitLt(torch.autograd.Function): if A_dtype != torch.float16: warnings.warn(f"MatMul8bitLt: input matrix will be converted from {A_dtype} to float16") A = A.to(torch.float16) + if bias is not None: + bias = bias.to(torch.float16) # 1. Quantize A if len(A.shape) == 3: |