Hey! This is a sample transformer app I wrote to test deployment to platforms.
To setup the dependencies:
pip install -r requirements.txt
To run the server:
python app.py
Quick documentation on some endpoints:
/api/test : (GET) tests generation
/api/generate : (POST) tests user input to generate long sequence.
Parameters of /api/generate:
input: the initial context / phrase / passage that the model will base its generation off. Have fun here!
iterations: number of samples the model generates. These samples are eventually combined in the end.
min_sample_len: minimum length of chunks from the sample.
max_sample_len: maximum length of a sample. Make sure this stays below ~1000
past_context_len: number of tokens from the previous iteration to use as context for next model iteration