LLM Chronicles #4.6: Building an Encoder/Decoder RNN in PyTorch to Translate from English to Italian
HTML-код
- Опубликовано: 9 сен 2024
- In this lab we'll use PyTorch to build a encoder/decoder RNN for language translation, similar to the model described in the 2014 paper "Sequence to Sequence Learning with Neural Networks” by OpenAI's chief scientist Ilya Sutskever. Our RNN layers will use LSTM (Long Short-Term Memory) cells.
🖹 Lab Notebook: colab.research...
🕤 Timestamps:
00:10 - Sutskever's Paper on Language Translation with RNNs
00:27 - Loading the Dataset, Tokenizer and Vocabulary
02:40 - PyTorch Dataset and DataLoader Objects
05:20 - Encoder / Decoder RNN in PyTorch
08:15 - Inference / Forward Pass for Translation
10:10 - Training Loop, Teacher Forcing
15:54 - Using/Evaluating the Trained Model
19:15 - Bi-Directional RNN for the Encoder
22:10 - Using/Evaluating the Trained Bi-Directional RNN
23:29 - Comparing our model to Sutskever's Paper
References:
- "Sequence to Sequence Learning with Neural Networks", arxiv.org/abs/...
Don't forget to like, share, and subscribe for more neural network adventures!
🚨 Clarifications & Corrections 🚨
1. At 25:04 - The size of the model is in millions of parameters, not Megabytes!
Thanks for starting with basics❤
Thanks for the support! Next episode is on Attention, we'll add that to this model getting a pretty good outcome.