Modify a deep learning repo to generate instead of classify

I am looking for a freelancer who can modify a PyTorch deep learning repo to generate sequences instead of classifying them. The code currently works with text data and audio, and the desired output format is plain text or audio. Ideal skills and experience for this project include:
- Proficiency in PyTorch
- Experience with deep learning algorithms
- Familiarity with working with text data
- Ability to modify existing code to meet project requirements.
-Experience with Seq2Seq, Encoder/Decoder, and Transformer architectures