Open francisduval opened 8 months ago
I can reproduce this on macOS.
kedro new --name=test-custom-resolver-viz --starter=spaceflights-pandas
everything works ✔️ pip install torch
everything still works ✔️ src/test_custom_resolver_viz/settings.py
as explained in https://github.com/kedro-org/kedro-viz/issues/1712#issue-2090023920 everything still works ✔️ module__act
from https://github.com/kedro-org/kedro-viz/issues/1712#issue-2090023920 to conf/base/parameters_data_science.yml
now kedro viz run
fails 💥 I don't think this is specific to PyTorch though, maybe Kedro Viz is trying to serialise the getattr
in the lambda
as JSON and failing.
Thanks! One option that works is to use the getattr function directly in nodes.py.
Description
When using a custom resolver, the pipeline runs normally, but it becomes impossible to use the
kedro viz run
command.Context
I want to be able to use a custom resolver while still be able to visualize my pipelines.
Steps to Reproduce
Define a custom resolver, i.e. put those lines in settings.py (this resolver transforms an activation function given as a string like 'ReLU' into a class object):
In parameters.yml, use your resolver:
Then, run
kedro viz run
Expected Result
Should render the kedro visualization.
Actual Result
The Kedro visualization does not render, and this error is thrown:
Your Environment
conda environment kedro version 0.19.1 Python 3.10.13 Windows 10, Version 21H2 (version du système d'exploitation 19044.3803)