Go to file
Meng Zhang a0b438da06
Add python transformer backend for tabby (mainly used for local dev / test in non-cuda environment) (#6)
* Add python backend

* Split docker-compose.triton.yml

* update makefile
2023-03-23 14:14:33 +08:00
.github/workflows Update docker.yml 2023-03-22 23:59:40 +08:00
admin Add admin panel (w/ streamlit) and logging (w/ vectordev) (#4) 2023-03-22 23:18:12 +08:00
converter Add docker compose (#3) 2023-03-22 02:42:47 +08:00
deployment Add python transformer backend for tabby (mainly used for local dev / test in non-cuda environment) (#6) 2023-03-23 14:14:33 +08:00
preprocess Add basic filters 2023-03-16 18:11:24 +08:00
server Add python transformer backend for tabby (mainly used for local dev / test in non-cuda environment) (#6) 2023-03-23 14:14:33 +08:00
testdata/tiny-70M Reduce repo size (#5) 2023-03-23 13:10:59 +08:00
.dockerignore Update Dockerfile 2023-03-22 02:52:41 +08:00
.gitattributes Add docker compose (#3) 2023-03-22 02:42:47 +08:00
.gitignore Add admin panel (w/ streamlit) and logging (w/ vectordev) (#4) 2023-03-22 23:18:12 +08:00
.pre-commit-config.yaml Switch pre-commit to commit 2023-03-16 18:11:54 +08:00
LICENSE Create LICENSE 2023-03-16 17:28:10 +08:00
Makefile Add python transformer backend for tabby (mainly used for local dev / test in non-cuda environment) (#6) 2023-03-23 14:14:33 +08:00
poetry.lock Add admin panel (w/ streamlit) and logging (w/ vectordev) (#4) 2023-03-22 23:18:12 +08:00
pyproject.toml Add admin panel (w/ streamlit) and logging (w/ vectordev) (#4) 2023-03-22 23:18:12 +08:00
README.md Add python transformer backend for tabby (mainly used for local dev / test in non-cuda environment) (#6) 2023-03-23 14:14:33 +08:00

🐾 Tabby

License Code style: black Docker build status

Warning

This repository is undering heavy construction, everything changes fast.

An opensource / on-prem alternative to GitHub Copilot

Contents

  • admin: Admin panel for monitoring / settings purpose.
  • server: API server for completion requests. It also logs users' selections (as feedback to model's quality).
  • deployment: Container related deployment configs.
  • converter: Converts a transformers causal LM model into TensorRT / FasterTransformer serving formats.
  • preprocess: Preprocess files into datasets
  • tabformer: NOT RELEASED Trainer(PEFT w/RLHF) for tabby models.

Development

Assuming Linux workstation with:

  1. docker
  2. docker w/ gpu driver
  3. python 3.10

Use make setup-development-environment to setup basic dev environment, and make dev to start local development server.