Music Generation Unexpected Indices

I’ve finished the music generation exercise. I’ve passed all the test. However, predict_and_sample isn’t producing the expected argmax indices. I’ve been quite thorough in terms of coding up the function. I’ve tried a few test: checking the expected dims etc. I still can’t see why it would not produce the expected indices… The function is working, but not producing the right argmax indices…
Please can you assist.

Please read this note in the python notebook:

  • Your results may likely differ because Keras’ results are not completely predictable.

Thanks Balaji. So I can submit (and hopefully get 100%)?

Please go ahead & submit.

Thanks Balaji. I got 100% :+1: