ml-collections

This project publishes no release notes. Between v1.0.0 and v1.1.0 there were 10 commits, 9 of them substantive:

• remove usage of Six Py2+3 compatibility layer
• Better error message when flag parsing fail
• Support overwriting ConfigDict through flags
• Support accept_new_attributes in DEFINE_config_dict.
• Pass allow_dotted_keys to recursive call of ConfigDict.
• Set allow_dotted_keys equal to original class during copy operations
• Fix error when assigning a reference field to a dict field
• Reraise error keep original message
• Release ml_collections 1.1.0

This project publishes no release notes. Between v0.1.1 and v1.0.0 there were 105 commits, 96 of them substantive:

• Create python-publish to publish a release
• Update python-publish to use Twine to push to pypi
• Internal change only.
• Improve readability of a typing error message to make it clear what is the original type and what is it the type of new variable where overriding failed.
• Fix DEFINE_config_dataclass to have the correct type when accessed via flags and to belong to the correct module.
• Use ast module to split string. Remove broken Python2 support.
• Config_dict: provide a little more detail in a certain error message.
• Adding file to keep only in the github repo.
• Typo in README.md
• Make the examples use the public API.
• Update recommended usage of absl.flags.
• Silence some pytype errors.
• Use preferred import pattern from ml_collections import config_flags.
• Add an ability to provide custom parser function for DEFINE_config_dataclass, so that an entire dataclass can be specified.
• Make ml_collection imports more targeted (config_dict or config_flags).
• Remove unused comments related to Python 2 compatibility.
• Remove unused comments related to Python 2 compatibility.
• Remove unused comments related to Python 2 compatibility.
• Remove unused comments related to Python 2 compatibility.
• Support overriding enum values with flags.

…and 76 more.