Pix2StructImageProcessor requires torch>=1.11.0
#24270
Conversation
| mel_shrink = torch.nn.functional.interpolate( | ||
| mel, size=[chunk_frames, 64], mode="bilinear", align_corners=False, antialias=False | ||
| ) | ||
| mel_shrink = torch.nn.functional.interpolate(mel, size=[chunk_frames, 64], mode="bilinear", align_corners=False) |
There was a problem hiding this comment.
the default value for antialias is False, no need to specify. Remove this so we don't need to require torch >= 1.11
|
The documentation is not available anymore as the PR was closed or merged. |
| @@ -192,7 +192,7 @@ def _random_mel_fusion(self, mel, total_frames, chunk_frames): | |||
|
|
|||
| mel = torch.tensor(mel[None, None, :]) | |||
| mel_shrink = torch.nn.functional.interpolate( | |||
| mel, size=[chunk_frames, 64], mode="bilinear", align_corners=False, antialias=False | |||
| mel, size=[chunk_frames, 64], mode="bilinear", align_corners=False | |||
There was a problem hiding this comment.
I know that it's False by default - but have we confirmed that it had this default behaviour before being added? (I'm assuming yes, otherwise a lot of things would have broken)
There was a problem hiding this comment.
antialias=False is there since the commit (main) where clap is added, so nothing earlier to compare against. As the default value is False, the change in this PR would just keep what we have on main, so I think we are good.
|
|
||
| logger = logging.get_logger(__name__) | ||
| DEFAULT_FONT_PATH = "ybelkada/fonts" | ||
|
|
||
|
|
||
| if is_torch_available() and not is_torch_greater_or_equal_than_1_11: |
There was a problem hiding this comment.
Perhaps we'll want a way to specify versions using or with something similar to requires_xxx or requires_backends(...).
Generally, users don't read warnings, and so have an error closer to when the object is used - instantiation or method calling - could be useful.
There was a problem hiding this comment.
I agree, but partially. In some modeling file (tapas, vilt), we have something
if not is_torch_greater_or_equal_than_xxx:
logger.warning(
f"You are using torch=={torch.__version__}, but torch>=xxx is required to use "
"TapasModel. Please upgrade torch."
)However, I understand it's not really feasible to do what you describe in a modeling file, but it could be done in a processor file.
Happy to apply the suggestion to this single file image_processing_pix2struct.py.
| if is_torch_available() and not is_torch_greater_or_equal_than_1_11: | ||
| raise ImportError( | ||
| f"You are using torch=={torch.__version__}, but torch>=1.11.0 is required to use " | ||
| "Pix2StructImageProcessor. Please upgrade torch." | ||
| ) |
There was a problem hiding this comment.
@amyeroberts Could you take a final look? 🙏
ImportError is taken from sth we have before
class PytorchGELUTanh(nn.Module):
def __init__(self):
super().__init__()
if version.parse(torch.__version__) < version.parse("1.12.0"):
raise ImportError(
f"You are using torch=={torch.__version__}, but torch>=1.12.0 is required to use "
"PytorchGELUTanh. Please upgrade torch."
)
* fix * fix * fix --------- Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>
What does this PR do?
So let's be nice to past CI ❤️ !
It's the argument
antialiasininterpolateonly supported intorch>=1.11.0: