[versioning] Save ML-Agents version in checkpoints and check on load #4035
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
chriselion
reviewed
May 28, 2020
|
|
||
|
|
||
| logger = get_logger(__name__) | ||
|
|
||
|
|
||
| # This is the version number of the inputs and outputs of the model, and | ||
| # determines compatibility with inference in Barracuda. | ||
| API_VERSION_NUMBER = 2 |
There was a problem hiding this comment.
MODEL_FORMAT_VERSION_NUMBER? "API" doesn't really feel right here.
chriselion
reviewed
May 28, 2020
| :param version_string: The semantic-versioned version string (X.Y.Z). | ||
| :return: A Tuple containing (major_ver, minor_ver, patch_ver). | ||
| """ | ||
| split_ver = version_string.split(".")[0:3] # Remove dev tag |
There was a problem hiding this comment.
I think you can use distutils.version.LooseVersion to simplify this:
>>> from distutils.version import LooseVersion
>>> v = LooseVersion("1.2.3.dev4")
>>> v
LooseVersion ('1.2.3.dev4')
>>> v.version
[1, 2, 3, 'dev', 4]
>>> v.version[0:3]
[1, 2, 3]
There was a problem hiding this comment.
Oh, that is very handy! Updated to use the LooseVersion class.
chriselion
approved these changes
May 28, 2020
added 8 commits
May 28, 2020 13:56
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Proposed change(s)
This PR saves the semantic version of the trainer package (
major,minor, andpatch) as 3 variables in the TF graph. It is also exported into the .nn file. This lets us check whether a checkpoint is being loaded from the same version of ML-Agents, and (in the future) check which version of ML-Agents created a particular NN file.Note that this is different than the existing version number in the NN, which is checked by C#. That number corresponds to the input and output tensors. It is possible, when upgrading the Trainer code, for an NN file to remain compatible with C# but not be loadable into Python for training (e.g. if the network architecture changes).
Currently, we throw a warning if the versions don't match when a user tries to load a checkpoint.
Types of change(s)
Checklist