def static_functions_pipeline(input_string):
# Creates a pipeline with a list of functions
pipe = Pipeline([remove_spaces, remove_special_chars, lowercase])
# Invokes pipeline
output = pipe(input_string)
print(f"""output ==> {output}""")
def dynamic_functions_pipeline(input_string, pipe_funcs):
# Creates a pipeline with a list of functions using using globals()
pipe = Pipeline([globals()[func] for func in pipe_funcs])
# Invokes pipeline
output = pipe(input_string)
print(f"""output ==> {output}""")
def make_cifar_item_tfm(th_img_tfms=None):
img_tfms = [cfnp2img_tfm]
if th_img_tfms is not None:
# assumes th_img_tfms incl ToTensor (cnvt2 PIL.Image -> tensor + div by 255)
img_tfms += [th_img_tfms]
else:
img_tfms += [cfimg_tfm, cfimg2float_tfm]
return CifarTupleTransform(x_tfm=Pipeline(img_tfms), y_tfm=i2t_tfm)
def __init__(self, **kwargs):
self.tfms = Pipeline(audio_item_tfms(**kwargs))
def after_batch(self:th_data.DataLoader):
'return empty pipeline when fastai learner looks for after_batch'
return Pipeline()
def test_pipeline():
pipe = Pipeline([f, g])
assert 2.0 == pipe(3)