/yet-another-transformer

Transformer built from scratch in PyTorch for educational purposes.

Primary LanguagePythonMIT LicenseMIT

yet-another-transformer

Transformer build from scratch in PyTorch for educational purposes.

TODO List

  • Add trainer
  • Add decoding code
  • Train on wikitext dataset
  • Train seq2seq task on Friends dialog dataset (No significant results due to quality of the dataset)