Using RNN to make english to french translations. Developed as coursework in the "Deep Learning Fundamentals" Nanodegree @ Udacity. This project passed all specifications defined by the program, as shown in the Project Information section.
- The function text_to_ids is implemented correctly.
- The parameters are set to reasonable numbers.
- The number of epochs is set adequately, as the training loss is low and not improving much with more training.
- Batch size is large enough to train efficiently, and it is obviously small enough to fit in memory.
- RNN size seems adequate.
- The number of layers seems good.
- Encoding and decoding embedding sizes are set okay.
- Learning rate looks fine. Not too large, not too small.
- Keep probability is in adequate range.
- The project ended with a validation and test accuracy that is at least 90.00%
Your NN translated most of the words correctly. I encourage you to try translating other sentences from the dataset and see how your NN performs on them. Unfortunately, in this project you will rarely see perfect translation.