Closed rodluger closed 3 years ago
This is an aspect of the FunctionGraph
and optimization framework that is in the process of being fixed. After refactoring the core optimization logic that relies on this silent failure behavior, this warning will become an error.
It should be possible to filter these warnings using warnings.filterwarnings
, but, if that doesn't work well enough, we could change it to a debug message emitted via logging
—at least until we've fixed Aesara's own code.
Ok, awesome -- thanks for the quick response. The filterwarnings approach should work fine. I really appreciate all the hard work you guys are putting into this!
Thanks!
Description of your problem or feature request
As of #298 I am getting hundreds of
UserWarnings
on code that used to run seamlessly. The warnings are issued here. Previously,theano
andtheano-pymc
simply failed silently at this step. Perhaps there's some config setting I can change to suppress these, but I couldn't find anything in the docs.Please provide a minimal, self-contained, and reproducible example.
Below is just one of many ways to trigger the
UserWarning
:Please provide the full traceback of any errors.
The output is correct, but a warning is raised complaining about how a variable cannot be replaced as it is not in the
FunctionGraph
.Please provide any additional information below.
Versions and main components
Aesara config:
``` floatX ({'float64', 'float32', 'float16'}) Doc: Default floating-point precision for python casts. Note: float16 support is experimental, use at your own risk. Value: float64 warn_float64 ({'raise', 'pdb', 'ignore', 'warn'}) Doc: Do an action when a tensor variable with float64 dtype is created. They can't be run on the GPU with the current(old) gpu back-end and are slow with gamer GPUs. Value: ignore pickle_test_value (