Dimension Error

[etetteh]

I am working on clinical time series forecasting with different architectures including wavenet. I came across your wavenet blog post, very informative, and reproduced the work.
However, I get an error using a codebase with the wavenet model. I have tried several times, but can't find a solution. Please advise. Thank you.

ValueError: Error when checking target: expected lambda_1 to have 3 dimensions, but got array with shape (64, 1)

[evanatyourservice]

Hi, post your code and data shape and we may be able to help you better. The data in the examples is shaped so each row is a sequence, and each column is a timestep in those sequences. There is 240,000 sequences or something like that, 550 timesteps each, so the resulting data shape is (240000, 550) it is then reshaped before the model to (240000, 550, 1). So try to format your data after this and it will work.

For example, if you have 1000 examples of heart beats, 300 seconds each, format it so it’s (1000, 300, 1) and go from there.

[etetteh]

Thank you for your response. I will check the data input format and get back to you as soon as possible.

…

On Sun, Oct 13, 2019, 10:46 Evan Walters ***@***.***> wrote: Hi, post your code and data shape and we may be able to help you better. The data in the examples is shaped so each row is a sequence, and each column is a timestep in that sequence. There is 240,000 sequences or something like that, 550 timesteps each, so the resulting data shape is (240000, 550) it is then reshaped before the model to (240000, 550, 1). So try to format your data after this and it will work. For example, you have 100 examples of heart beats, 300 seconds each. Format it so it’s (100, 300, 1) and go from there. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#7?email_source=notifications&email_token=AGZQ72HIE5USKYHWJNIGXB3QOLOAFA5CNFSM4I7W5C32YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEBCRUOA#issuecomment-541399608>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AGZQ72BDPX6EDONAFQGZ45LQOLOAFANCNFSM4I7W5C3Q> .

[etetteh]

To simply put it, the data I am using is being extracted from so many .cvs files into a single .cvs (linked .cvs if I may put it) and put Keras data generator.
I tried getting the shapes doing:

ret = next(train_data_gen)
x, y = ret["data"]
cur_names = ret["names"]
cur_ts = ret["ts"] 

since I can't infer the shapes directly from a keras generator. It turns out x.shape is (16, 215, 76) which is (batch_size, samples, features).
What shape do I pass to the Input Layer? I have tried a couple of them, but none is working. I can't also reshape x to 4ndim since Conv1D doesn't support 4 ndim. Please help. Thank you.

Thank you for your response. I will check the data input format and get back to you as soon as possible.
…
On Sun, Oct 13, 2019, 10:46 Evan Walters @.***> wrote: Hi, post your code and data shape and we may be able to help you better. The data in the examples is shaped so each row is a sequence, and each column is a timestep in that sequence. There is 240,000 sequences or something like that, 550 timesteps each, so the resulting data shape is (240000, 550) it is then reshaped before the model to (240000, 550, 1). So try to format your data after this and it will work. For example, you have 100 examples of heart beats, 300 seconds each. Format it so it’s (100, 300, 1) and go from there. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#7?email_source=notifications&email_token=AGZQ72HIE5USKYHWJNIGXB3QOLOAFA5CNFSM4I7W5C32YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEBCRUOA#issuecomment-541399608>, or unsubscribe https://github.com/notifications/unsubscribe-auth/AGZQ72BDPX6EDONAFQGZ45LQOLOAFANCNFSM4I7W5C3Q .