'RandomResizedCrop' raises exception

[SerezD]

I would like to train on Imagenet dataset, and I have set up two different pipelines for train/test:

img_train_pipeline = [CenterCropRGBImageDecoder(output_size=(self.image_size, self.image_size), ratio=1.),
                      RandomResizedCrop(scale=(0.8, 1.), ratio=(0.9, 1.1), size=self.image_size),
                      RandomHorizontalFlip(),
                      ToTensor(),
                      ToTorchImage(),
                      Convert(torch.float16),
                      Normalize(IMAGENET_MEAN, IMAGENET_STD)
                      ]

img_test_pipeline = [CenterCropRGBImageDecoder(output_size=(self.image_size, self.image_size), ratio=1.),
                     ToTensor(), 
                     ToTorchImage(),
                     Convert(torch.float16),
                     Normalize(IMAGENET_MEAN, IMAGENET_STD),
                     ]

The test pipeline works fine, but the train pipeline raises an error. In particular:

Traceback (most recent call last):
  File "/home/dserez/anaconda3/envs/ffcv/lib/python3.10/threading.py", line 1016, in _bootstrap_inner
    self.run()
  File "/home/dserez/anaconda3/envs/ffcv/lib/python3.10/site-packages/ffcv/loader/epoch_iterator.py", line 79, in run
    result = self.run_pipeline(b_ix, ixes, slot, events[slot])
  File "/home/dserez/anaconda3/envs/ffcv/lib/python3.10/site-packages/ffcv/loader/epoch_iterator.py", line 133, in run_pipeline
    result = code(*args)
  File "/home/dserez/anaconda3/envs/ffcv/lib/python3.10/site-packages/numba/core/dispatcher.py", line 468, in _compile_for_args
    error_rewrite(e, 'typing')
  File "/home/dserez/anaconda3/envs/ffcv/lib/python3.10/site-packages/numba/core/dispatcher.py", line 409, in error_rewrite
    raise e.with_traceback(None)
numba.core.errors.TypingError: Failed in nopython mode pipeline (step: nopython frontend)
Failed in nopython mode pipeline (step: nopython frontend)
No implementation of function Function(<ufunc 'log'>) found for signature:

 >>> log(reflected list(float64)<iv=None>)

There are 2 candidate implementations:
  - Of which 2 did not match due to:
  Overload in function 'Numpy_rules_ufunc.generic': File: numba/core/typing/npydecl.py: Line 98.
    With argument(s): '(reflected list(float64)<iv=None>)':
   Rejected as the implementation raised a specific error:
     TypingError: can't resolve ufunc log for types [reflected list(float64)<iv=None>]
  raised from /home/dserez/anaconda3/envs/ffcv/lib/python3.10/site-packages/numba/core/typing/npydecl.py:107
During: resolving callee type: Function(<ufunc 'log'>)
During: typing of call at /home/dserez/anaconda3/envs/ffcv/lib/python3.10/site-packages/ffcv/transforms/utils/fast_crop.py (19)
File "../../../../../anaconda3/envs/ffcv/lib/python3.10/site-packages/ffcv/transforms/utils/fast_crop.py", line 19:
def get_random_crop(height, width, scale, ratio):
    <source elided>
    area = height * width
    log_ratio = np.log(ratio)
    ^
During: resolving callee type: type(CPUDispatcher(<function RandomResizedCrop.generate_code.<locals>.random_resized_crop at 0x7f7d53ce9750>))
During: typing of call at  (2)
During: resolving callee type: type(CPUDispatcher(<function RandomResizedCrop.generate_code.<locals>.random_resized_crop at 0x7f7d53ce9750>))
During: typing of call at  (2)
File "/home/dserez/.../src", line 2:
<source missing, REPL/exec in use?>
Exception in thread Thread-8:
Traceback (most recent call last):
  File "/home/dserez/anaconda3/envs/ffcv/lib/python3.10/threading.py", line 1016, in _bootstrap_inner
    self.run()
  File "/home/dserez/anaconda3/envs/ffcv/lib/python3.10/site-packages/ffcv/loader/epoch_iterator.py", line 79, in run
    result = self.run_pipeline(b_ix, ixes, slot, events[slot])
  File "/home/dserez/anaconda3/envs/ffcv/lib/python3.10/site-packages/ffcv/loader/epoch_iterator.py", line 133, in run_pipeline
    result = code(*args)
  File "/home/dserez/anaconda3/envs/ffcv/lib/python3.10/site-packages/numba/core/dispatcher.py", line 468, in _compile_for_args
    error_rewrite(e, 'typing')
  File "/home/dserez/anaconda3/envs/ffcv/lib/python3.10/site-packages/numba/core/dispatcher.py", line 409, in error_rewrite
    raise e.with_traceback(None)
numba.core.errors.TypingError: Failed in nopython mode pipeline (step: nopython frontend)
Failed in nopython mode pipeline (step: nopython frontend)
No implementation of function Function(<ufunc 'log'>) found for signature:

 >>> log(reflected list(float64)<iv=None>)

There are 2 candidate implementations:
  - Of which 2 did not match due to:
  Overload in function 'Numpy_rules_ufunc.generic': File: numba/core/typing/npydecl.py: Line 98.
    With argument(s): '(reflected list(float64)<iv=None>)':
   Rejected as the implementation raised a specific error:
     TypingError: can't resolve ufunc log for types [reflected list(float64)<iv=None>]
  raised from /home/dserez/anaconda3/envs/ffcv/lib/python3.10/site-packages/numba/core/typing/npydecl.py:107
During: resolving callee type: Function(<ufunc 'log'>)
During: typing of call at /home/dserez/anaconda3/envs/ffcv/lib/python3.10/site-packages/ffcv/transforms/utils/fast_crop.py (19)
File "../../../../../anaconda3/envs/ffcv/lib/python3.10/site-packages/ffcv/transforms/utils/fast_crop.py", line 19:
def get_random_crop(height, width, scale, ratio):
    <source elided>
    area = height * width
    log_ratio = np.log(ratio)
    ^
During: resolving callee type: type(CPUDispatcher(<function RandomResizedCrop.generate_code.<locals>.random_resized_crop at 0x7fc05cea12d0>))
During: typing of call at  (2)
During: resolving callee type: type(CPUDispatcher(<function RandomResizedCrop.generate_code.<locals>.random_resized_crop at 0x7fc05cea12d0>))
During: typing of call at  (2)
File "/home/dserez/.../src", line 2:
<source missing, REPL/exec in use?>

[andrewilyas]

I recently merged in a fix to RandomResizedCrop (#288) so this should be fixed in v1.0 coming out this week!

[andrewilyas]

This should now be fixed and uploaded on PyPI!