Fix for issue #970

[elogicaadith]

Description

existing code, in base_model.py, under the fit function -

if self.hparams.save_model:

        if epoch % self.hparams.save_epoch == 0:
            checkpoint_path = self.saver.save(
                sess=train_sess,
                save_path= self.hparams.MODEL_DIR + "epoch_" + str(epoch),
            )

Modified code:
from os.path import join
.
.
.

if self.hparams.save_model:
if epoch % self.hparams.save_epoch == 0:
save_path_str = join(self.hparams.MODEL_DIR, "epoch_" + str(epoch))
checkpoint_path = self.saver.save(
sess=train_sess,
save_path= save_path_str,
)

Azure Machine Learning service tests :-

Code in AML -

AML outputs folder -

Local machine tests (using xdeepfm_criteo notebook in 00_quick_start) :-

prepare_hparams() -

local directory -

When running the xDeepFM script, the model was not getting saved to the ./outputs directory in the Azure Machine Learning Service. This was due to the way in which the string concatenation for save directory was set up - the model directory & the file names were concatenated without a "/" in between them. As a result, Azure Machine Learning Service was not able to save models.

Related Issues

This is a fix for Issue #970

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.

[elogicaadith]

@miguelgfierro - Hello Miguel! I'm now using os.path.join to concatenate the save path string. :)

[miguelgfierro]

awesome, thanks!