Sequentialrecsys

[Leavingseason]

Description

We provide 4 sequential models in this update for deeprec, namely ASVD (non-sequential, just to compare with), GRU4Rec (RNN based sequential model), Caser (CNN based sequential model), and SLi-Rec (time-aware RNN base sequential model, published in IJCAI'19 by MSRA).
We provide a jupyter notebook in quick_start.
We use a public dataset, Amazon review dataset, for demonstration. In the quick start notebook, the script will automatically download the dataset, so there is no need for us to host the dataset.
We provide unit test and smoke test for sequential models.

Sequential recommenders is a type of recommender models with increasing importance. In this update, we aim to enable the repo with sequential recommender models.

Related Issues

Checklist:

• I have followed the contribution guidelines and code style for this project.
• I have added tests covering my contributions.
• I have updated the documentation accordingly.

[review-notebook-app]

Check out this pull request on 

You'll be able to see Jupyter notebook diff and discuss changes. Powered by ReviewNB.

[miguelgfierro]

there is an error in the gpu unit test:

tests/unit/test_notebooks_gpu.py ....F.                                  [100%]

=================================== FAILURES ===================================
_________________________________ test_xdeepfm _________________________________

notebooks = {'als_deep_dive': '/data/home/recocat/agent/_work/5/s/notebooks/02_model/als_deep_dive.ipynb', 'als_pyspark': '/data/h...pynb', 'cornac_bpr_deep_dive': '/data/home/recocat/agent/_work/5/s/notebooks/02_model/cornac_bpr_deep_dive.ipynb', ...}

    @pytest.mark.notebooks
    @pytest.mark.gpu
    def test_xdeepfm(notebooks):
        notebook_path = notebooks["xdeepfm_quickstart"]
        pm.execute_notebook(
            notebook_path,
            OUTPUT_NOTEBOOK,
            kernel_name=KERNEL_NAME,
            parameters=dict(
                EPOCHS_FOR_SYNTHETIC_RUN=1,
                EPOCHS_FOR_CRITEO_RUN=1,
                BATCH_SIZE_SYNTHETIC=128,
>               BATCH_SIZE_CRITEO=512,
            ),
        )

tests/unit/test_notebooks_gpu.py:69: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 
/data/anaconda/envs/reco_gpu/lib/python3.6/site-packages/papermill/execute.py:100: in execute_notebook
    raise_for_execution_errors(nb, output_path)
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 

nb = {'cells': [{'cell_type': 'code', 'metadata': {'inputHidden': True, 'hide_input': True}, 'execution_count': None, 'sour...nd_time': '2019-12-13T03:56:00.431270', 'duration': 12.523554, 'exception': True}}, 'nbformat': 4, 'nbformat_minor': 2}
output_path = 'output.ipynb'

    def raise_for_execution_errors(nb, output_path):
        """Assigned parameters into the appropriate place in the input notebook
    
        Parameters
        ----------
        nb : NotebookNode
           Executable notebook object
        output_path : str
           Path to write executed notebook
        """
        error = None
        for cell in nb.cells:
            if cell.get("outputs") is None:
                continue
    
            for output in cell.outputs:
                if output.output_type == "error":
                    error = PapermillExecutionError(
                        exec_count=cell.execution_count,
                        source=cell.source,
                        ename=output.ename,
                        evalue=output.evalue,
                        traceback=output.traceback,
                    )
                    break
    
        if error:
            # Write notebook back out with the Error Message at the top of the Notebook.
            error_msg = ERROR_MESSAGE_TEMPLATE % str(error.exec_count)
            error_msg_cell = nbformat.v4.new_code_cell(
                source="%%html\n" + error_msg,
                outputs=[
                    nbformat.v4.new_output(output_type="display_data", data={"text/html": error_msg})
                ],
                metadata={"inputHidden": True, "hide_input": True},
            )
            nb.cells = [error_msg_cell] + nb.cells
            write_ipynb(nb, output_path)
>           raise error
E           papermill.exceptions.PapermillExecutionError: 
E           ---------------------------------------------------------------------------
E           Exception encountered at "In [9]":
E           ---------------------------------------------------------------------------
E           ValueError                                Traceback (most recent call last)
E           <ipython-input-9-1a477a30c87f> in <module>
E           ----> 1 model.fit(train_file, valid_file)
E           
E           /data/home/recocat/agent/_work/5/s/reco_utils/recommender/deeprec/models/base_model.py in fit(self, train_file, valid_file, test_file)
E               420             for batch_data_input in self.iterator.load_data_from_file(train_file):
E               421                 step_result = self.train(train_sess, batch_data_input)
E           --> 422                 (_, step_loss, step_data_loss, summary) = step_result
E               423                 if self.hparams.write_tfevents:
E               424                     self.writer.add_summary(summary, step)
E           
E           ValueError: too many values to unpack (expected 4)

@Leavingseason, one question, can you access this https://dev.azure.com/best-practices/recommenders/_build/results?buildId=18584 and see all the logs, etc?

[miguelgfierro]

this is absolutely awesome

[miguelgfierro]

@Leavingseason, feel free to merge when you think it is finished. After this is merged, I'll start working on the 4 deep dives #1013

[Leavingseason]

@Leavingseason, feel free to merge when you think it is finished. After this is merged, I'll start working on the 4 deep dives #1013

Unfortunately I an not authorized to merge this pull request... @yueguoguo @anargyri @gramhagen do you have any comments?