W3 programming assignment1 (NMT)

When I load the trained model “model.load_weights(‘models/model.h5’)”,
it shows wrong outputs for test cases as below:

There may be an error in your code, probably for the modelf() function, or perhaps in one_step_attention().

Or, maybe you need to re-run the cells in the notebook that train the model.
It’s always wise to re-run all of the cells every time you open the notebook.

hum, I got 100, so I think model is correct.

I run the code several times, still the same.

This issue is reported occasionally, I’ll see if I can find any of the threads that discuss how to fix it.

1 Like

@balaji.ambresh! Isn’t that output similar to the one for which you opened the git issue?

That is correct. I first saw the issue in this topic and filed this ticket. After the following message, I decided not to visit that topic again.

[snippet removed by self]

1 Like

Okay, Balaji. ~~~~~~

@Lena_Tannaz_Trigg, check your code in modelf() where you use the post_activation_LSTM_cell() function. It’s “Step 2.B”. You need to pass the correct parameters for the inputs and initial_state.

See the instructions and hints in the notebook.

The unit tests in the notebook do not catch all possible errors in your code.

1 Like

Thanks for the guidance