Skip to content

Latest commit

 

History

History
37 lines (28 loc) · 1.87 KB

UPCOMING.md

File metadata and controls

37 lines (28 loc) · 1.87 KB

Upcoming Features

Here is a list of features, in no particular order, of what I hope to add to aitextgen, given that it's feasible.

Training Features

  • Sparse Transformers?
  • Include keyword conditioning as well
    • For vocab prefix, use heuristics on tokens, since very speedy and can scale high with strong tokenizers
  • Context support, provided as a named dict to a generation function e sequence stronger than later text (e.g. recent tweets than older tweets!)

Generation Features

  • Allow a user-curated mode
  • Allow returning probabilities of next token
    • Console BG coloring of token prediction confidence?
  • Allow excluding of tokens by index (e.g. allow model generation without the letter e)
  • For postfix prediction, allow returning best guess or probabilities of all classes.
  • Calculate text Dimensionality by using activation weights of all tokens prior to EOS token.
    • Use dimensionality to calculate a similarity score from generated texts to real texts; scores below a threshold may be considered incoherent and can be discarded.
  • Dedupe existing texts when generating using token id matching.
  • Allow cycling context tokens when generating
  • Unlimited text generation via sliding context window (need to include a warning if doing so)

Deployment Features

  • Use ray async actors for async generation. May need a custom async generation function.
    • Use websockets in starlette so output can be returned token by token.
  • Export function: PyTorch trace, TensorFlow Serving, TensorFlow.js, CoreML

Quality-of-Life Features

  • Have a super minimal version that can be distributed with the PyPi package. (< 3MB)
    • Include a small model trainable on a CPU, trained from a distilled larger model
  • Support any CLM model, not just GPT2.
  • Support CPU-based XLA for faster PyTorch CPU Train/Predict performance