Embedding for categorical variables about pysddr HOT 4 CLOSED

Dear xwz-123,

thank you for you feedback. We fixed the bug in commit 16140e3

We also tested it with the code and data you provided. The code as is, does not run, as the output shape of the neural networks is a 3D Tensor (due to the embedding layer).

In order for it to work you can use e.g. a nn.Flatten before output layer.
This deep_models_dict should work:

deep_models_dict = {
'd1': {
'model': nn.Sequential(nn.Embedding(num_emb[0], emb_dim[0]), nn.ReLU(),nn.Flatten(start_dim=1), nn.Linear(emb_dim[0], 3)),
'output_shape': 3
},
'd2': {
'model': nn.Sequential(nn.Embedding(num_emb[1], emb_dim[1]), nn.ReLU(),nn.Flatten(start_dim=1), nn.Linear(emb_dim[1], 3)),
'output_shape': 3
}
}

Please check if it works now.

from pysddr.

Dear @DoTha,

thanks for your reply. I've used the code provided by your, but I still get the same error as before, i.e.
Expected tensor for argument #1 'indices' to have one of the following scalar types: Long, Int; but got torch.FloatTensor instead (while checking arguments for embedding)

from pysddr.

Dear @xwz-123

just to be sure: when you say you use the code provided
do you mean that you pulled the latest commit from our git repo,
or do you mean you use the deep_models_dict that I provided?

Because the error that you get should be fixed in our latest commit (the deep_model_dict I provided does NOT fix the bug).

What might also be a problem:
if you installed with pip without the -e flag, then you might (additionally to pulling the latest version of the code) need to uninstall the package and reinstall it again

please tell me if the problem prevails

from pysddr.

@DoTha I've indeed forgotten to pull the latest commit, sorry for my mistake! Now the codes run smoothly. Thanks a lot for your help!

from pysddr.