How to pull upstream changes into the vendored cloudpickle
Right now, we will need to update the vendored cloudpickle manually. Here are the steps to do so:
- Overwrite the
cloudpickle.pyandcloudpickle_fast.pyfrom the upstream - Update the version in python/ray/cloudpickle/__init__.py
- Run performance benchmarks to validate no regression. (e.g. microbenchmark, scalability tests, etc.)