This is the pytorch model parameters and associated data used for training a small transformer model from scratch.
The transformer model is used to train for translation from hindi_latin to english.
Among the files, training dataset used to create the model is also there. Data used for training is semi-synthetic.
Steps for creating datasets:
Obtain actualuser questions in hindi and human translations thereof in english.
Prompt GPT to create variations of key words taking phonetics in account and giving a user persona.
