Modelstore

**🆕 New functionality**

With thanks to [Tom in the MLOps Community](https://mlops-community.slack.com/archives/C0227QJCDS8/p1687777412934619) for the feedback:

* List all the states that have been set for a specific model with `get_model_states(self, domain: str, model_id: str)` [255](https://github.com/operatorai/modelstore/pull/255).
* Delete model states that are unused with `delete_model_state(self, state_name: str, skip_prompt: bool = False)` [251](https://github.com/operatorai/modelstore/pull/251).

**🐛 Bug fixes & general updates**

With thanks to [Tom in the MLOps Community](https://mlops-community.slack.com/archives/C0227QJCDS8/p1687777412934619) for the feedback:

* Model store cleans up temporary files that it generates, even if downloads or uploads are cancelled mid flight [258](https://github.com/operatorai/modelstore/pull/258).

And thanks to [Michael in the MLOps Community](https://mlops-community.slack.com/archives/C0227QJCDS8/p1695397710224629) for the feedback:

* Model store no longer creates its `artifacts.tar.gz` in the `os.getcwd()`, which was preventing modelstore from being used for parallel uploads [266](https://github.com/operatorai/modelstore/pull/266).

**🆕 New functionality**

Added support for any HDFS storage that you can access via [pydoop](https://crs4.github.io/pydoop/tutorial/hdfs_api.html#hdfs-api-tutorial): `ModelStore.from_hdfs()` [241](https://github.com/operatorai/modelstore/pull/241), thanks [sayandigital](https://github.com/sayandigital).

Updated the transformers manager: it no longer requires a `tokenizer` argument, so it can now be used to (for example) save/load SAM models [238](https://github.com/operatorai/modelstore/pull/238) or DPT models [#239](https://github.com/operatorai/modelstore/pull/239). Thank you, [Cate in the MLOps Community](https://mlops-community.slack.com/archives/C0227QJCDS8/p1683293544101389)

**🐛 Bug fixes & general updates**

Fixed [issues with saving & loading GPT-2 models](https://github.com/operatorai/modelstore/issues/233) in [#234](https://github.com/operatorai/modelstore/pull/234), thank you [sayandigital](https://github.com/sayandigital).

**🆕 New functionality**

Added support for Python 3.9, and an example where a yolov5 model is loaded from the torch hub, saved to modelstore, and then loaded back from the model store [225](https://github.com/operatorai/modelstore/pull/225), thanks to h4mmer in the MLOps Community.

**🐛 Bug fixes & general updates**

You can now use `modelstore` with a local MinIO server that doesn't have SSL enabled [223](workflows/investigate-minio-error), thanks to [sspillard](https://github.com/sspillard).

`modelstore` will now raise an exception if the user is attempting to upload a file from their current working directory that is called `artifacts.tar.gz` -- because that's the file that `modelstore` wants to create ([222](https://github.com/operatorai/modelstore/pull/222)).

**🆕 New functionality**

Added support for any s3-compatible object storage that you can access via [MinIO](https://min.io/): `ModelStore.from_minio()`.

Added support for [Spark ML Lib](https://spark.apache.org/) models.

You can move model stores that you created in one file system directory to another one (e.g., making it a mounted volume in a container) and `modelstore` will continue to work ([209](https://github.com/operatorai/modelstore/pull/209), thanks [hauks96](https://github.com/operatorai/modelstore/issues/189)).

**🐛 Bug fixes & general updates**

Fixes an issue where model archives would be overwritten if models are uploaded too quickly ([208](https://github.com/operatorai/modelstore/pull/208), thanks [shsnyder](https://github.com/operatorai/modelstore/issues/197)).

**🐛 Bug fixes & general updates**

A workaround for a security issue in the Python `tarfile` library was added ([203](https://github.com/operatorai/modelstore/pull/203), thanks [TrellixVulnTeam](https://github.com/TrellixVulnTeam)).

A security upgrade to the `protobuf` was merged ([202](https://github.com/operatorai/modelstore/pull/202), thanks dependabot) - this impacts developers of this library only.

**🆕 New functionality**

`modelstore` will issue a warning if you `load()` a model with a different version of Python than the version that was used to train the model ([192](https://github.com/operatorai/modelstore/pull/192)).

You can now add any extra metadata to your model when uploading it, using `upload(domain, model, extra_metadata={ ... })` ([185](https://github.com/operatorai/modelstore/pull/185)); if you want to upload extra _files_ with your model, then you should now use `extra_files=` instead of `extras=` ([#187](https://github.com/operatorai/modelstore/pull/187)).

**🐛 Bug fixes & general updates**

Fixed a bug when creating a GCP model store instance that reads from a public bucket ([176](https://github.com/operatorai/modelstore/pull/176)).

Added support for storing xgboost Booster models ([170](https://github.com/operatorai/modelstore/pull/170), [#177](https://github.com/operatorai/modelstore/pull/177)).

Refactored the library to use data classes for all model meta data ([178](https://github.com/operatorai/modelstore/pull/178))