vpj/python_autocomplete
A simple neural network for python autocompletion
Employs an LSTM model with token-level encoding and beam search to predict up to ~10 characters ahead, achieving 30-50% keystroke savings on Python code. Trains on preprocessed source code with comments, strings, and blank lines removed, using tokenization rather than character-level or byte-pair encoding. Includes a pre-trained checkpoint on TensorFlow's model repository, though beam search overhead currently limits practical editor integration.
821 stars. No commits in the last 6 months.
Stars
821
Forks
127
Language
Python
License
MIT
Category
Last pushed
Aug 09, 2020
Commits (30d)
0
Get this data via API
curl "https://pt-edge.onrender.com/api/v1/quality/ml-frameworks/vpj/python_autocomplete"
Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.
Higher-rated alternatives
minimaxir/textgenrnn
Easily train your own text-generating neural network of any size and complexity on any text...
jackcook/predictive-spy
Spying on Apple’s new predictive text model
yuazi/keyboard-AI
Self-learning keyboard layout optimizer using n-gram statistics and an ergonomic cost model
realAltamashAli/bilingual-swiftkey-nlp
A high-performance bilingual (EN/DE) text prediction engine using Katz Back-off N-gram modeling,...
subhadipsinha722133/Spelling-Checker-Web-App
📝 A simple and intuitive web application built with Streamlit and pyspellchecker to help you...