Seq2Seq

German to English Translator using pytorch Multi30k dataset

seq2seq.py: Used Encoder-Decoder Seq2Seq model

Metric used for evaluation: Bleu Score

Ways to improving model:-

• Using Attention
• Use Transformers
• Tuning Hyperparameters
• Increasing No. of Epochs
• Using Pre-trained Transformer model(Transfer Learning)

seq2seq_attn.py: Applied Attention to previous Seq2Seq model

Achieved Blue score of 25% with 50 epochs.

transformers.py: Used Transformer model for machine translation

Achieved Blue score of 28.4% with 50 epochs

(*see code for hyperparameters used)