What is X in forward function of DeepFM?

[Jeriousman]

Describe the question(问题描述)

The DeepFM model has forward function which takes X as an input (It is not only for DeepFM but also other models). But What does X look like? As it is an input for forward function, I wanted to check the shape of X and what it is but DeepCTR takes keras fit, predict so the codes dont really show what you put as the input for DeepFM model. Can you tell me the input and pseudo-generate the input data? And where can I check X if Im missing how to check it?

[zanshuxun]

This is DeepCTR-Torch repository, and there is no keras fit/predict here. Do you mean DeepCTR repository?

[Jeriousman]

No it is clearly DeepCTR-Torch. It might not be Keras one but predict and fit are there. If you go to class BaseModel in models/basemodel.py, you can find

 def fit(self, x=None, y=None, batch_size=None, epochs=1, verbose=1, initial_epoch=0, validation_split=0.,
            validation_data=None, shuffle=True, callbacks=None):

and

 def predict(self, x, batch_size=256):

But these are 'small x' not 'large X'. If you see class Linear(nn.Module) in /models/basemodel.py, you can see large X def forward(self, X, sparse_feat_refine_weight=None):. And I believe this is the same input as X in def forward(self, X): in models/deepfm.py.

What I want to know is what the large X should be like. I am stuck in here for the moment. Can you please specify what the X is? the shape and what it is in an example in pseudo generated data with torch at least?

[zanshuxun]

It's not keras fit. It's DeepCTR-Torch fit, which is designed in https://github.com/shenweichen/DeepCTR-Torch/blob/master/deepctr_torch/models/basemodel.py#L136

The input X could be found here: https://github.com/shenweichen/DeepCTR-Torch/blob/master/deepctr_torch/models/basemodel.py#L244 , which is derived from train input.

[Jeriousman]

I will look it up and tell you about it! Thank you.