diff options
author | justheuristic <justheuristic@gmail.com> | 2022-09-17 20:46:04 +0300 |
---|---|---|
committer | justheuristic <justheuristic@gmail.com> | 2022-09-17 20:46:04 +0300 |
commit | cc4858c2fd48ef17a888b9d45bb35bb00e373eb8 (patch) | |
tree | e452b1e1ccbebd58e77cd4b0f6429366c670cd44 | |
parent | 3634fc738bc20e4041c75544d3f678f61ce2348c (diff) |
some kind of warning or something when this is first executed to make people aware that a cast happens and the operation quantization is performed in fp16.
-rw-r--r-- | bitsandbytes/autograd/_functions.py | 4 |
1 files changed, 4 insertions, 0 deletions
diff --git a/bitsandbytes/autograd/_functions.py b/bitsandbytes/autograd/_functions.py index bdcbec5..6d473e9 100644 --- a/bitsandbytes/autograd/_functions.py +++ b/bitsandbytes/autograd/_functions.py @@ -1,4 +1,6 @@ import operator +import warnings + import torch import bitsandbytes.functional as F @@ -229,6 +231,8 @@ class MatMul8bitLt(torch.autograd.Function): # Cast A to fp16 A_dtype = A.dtype + if A_dtype != torch.float16: + warnings.warn(f"MatMul8bitLt: temporarily casting input matrix from {A_dtype} to float16") A = A.to(torch.float16) # 1. Quantize A |