search

tensorflow / ranking

Learning to Rank in TensorFlow
Apache License 2.0
2.74k stars 477 forks source link

Mismatch between latest PyPi version and git tagged version 0.1.4 #118

Closed hopeemac closed 5 years ago

hopeemac commented 5 years ago

A pip install of tensorflow-rank==0.1.4 seems to install a prior version. "0.1.3dev" is returned as the version when I run print(tfr.__version__). Also the input_size is still a required parameter for the encode_listwise_features function even though it is marked as deprecated the function on master.

ramakumar1729 commented 5 years ago

@hopeemac , thanks for pointing this out. The version number was updated to 0.1.4 in a commit after the package was uploaded to PyPI, this should have been done before. Since we cannot overwrite the package on PyPI without creating a new version, we will take care of this in subsequent releases.

The change deprecating input_size was added to master after the release of tfr-0.1.4. To use the master version of the code, please compile the package from source, as outlined in README.

karsures commented 5 years ago

@ramakumar1729 The sparse features demo on colab doesn't work after the commit. This is the error I get:

encode_listwise_features() takes at least 4 arguments (5 given)

Is there a quick fix rather than building from source? Thank you.

ramakumar1729 commented 5 years ago

@karsures Thanks for pointing this out. I've just pushed a new release, the colab should work now.

@hopeemac : The release for TFR 0.1.5 should address the two concerns you pointed out.

lorenzospataro commented 5 years ago

@ramakumar1729 I can see the new release on github but not in Pypi: https://pypi.org/project/tensorflow-ranking/#history When I "pip install --upgrade" I don't get the latest version (0.1.5)... Is there any way to fix this? Thanks

ramakumar1729 commented 5 years ago

@lorenzospataro : Thanks for pointing this out. This is fixed.

lorenzospataro commented 5 years ago

@ramakumar1729 Great, thank you!