Turbopilot Versions Save

Turbopilot is an open source large-language-model based code completion engine that runs locally on CPU

8 months ago

What's Changed

Temporary fix for stablecode and starcoder on mac by @ravenscroftj in https://github.com/ravenscroftj/turbopilot/pull/67

Full Changelog: https://github.com/ravenscroftj/turbopilot/compare/v0.2.0...v0.2.1

8 months ago

New: Turbopilot now supports full GPU offloading via CUDA and OpenCL by @ravenscroftj in https://github.com/ravenscroftj/turbopilot/pull/55
New: Turbopilot now supports StableCode models by @ravenscroftj in https://github.com/ravenscroftj/turbopilot/pull/43
New: Turbopilot is now interoperable with the Huggingface VSCode plugin by @ravenscroftj in https://github.com/ravenscroftj/turbopilot/pull/44
New: End-user control of generation temperature and top-p via command line arguments by @ravenscroftj in https://github.com/ravenscroftj/turbopilot/pull/52
New: Exposed batch size flag via cli by @ravenscroftj in https://github.com/ravenscroftj/turbopilot/pull/59
New: Added debug log level and added timings to model outputs by @ravenscroftj in https://github.com/ravenscroftj/turbopilot/pull/54
Fix: incorrect instructions in model docs by @aperullo in https://github.com/ravenscroftj/turbopilot/pull/57
Fix: prevent crashing of model due to multiple parallel requests by using locking @ravenscroftj in https://github.com/ravenscroftj/turbopilot/pull/58
Change: moved from alpine to ubuntu in Dockerfile.default by @nvtienanh in https://github.com/ravenscroftj/turbopilot/pull/60
Fix: download link on MODELS.md by @c01o in https://github.com/ravenscroftj/turbopilot/pull/61
Change: Implement better docker builds by @ravenscroftj in https://github.com/ravenscroftj/turbopilot/pull/62

Full Changelog: https://github.com/ravenscroftj/turbopilot/compare/v0.1.0...v0.2.0

9 months ago

Refactored codebase - now a single unified turbopilot binary that provides support for codegen and starcoder style models.
Support for starcoder, wizardcoder and santacoder models
Support for CUDA 11 and 12

Full Changelog: https://github.com/ravenscroftj/turbopilot/compare/v0.0.5...v0.1.0

11 months ago

Turbopilot now supports CUDA which significantly accelerates inference for long prompts on machines with NVIDIA cards builds by @ravenscroftj in https://github.com/ravenscroftj/turbopilot/pull/27
Turbopilot now comes with prebuilt windows binaries by @ravenscroftj in https://github.com/ravenscroftj/turbopilot/pull/29
Added links to download pre-converted and pre-quantized CodeGen mono models. by @virtualramblas and @CRD716 in https://github.com/ravenscroftj/turbopilot/pull/18 and https://github.com/ravenscroftj/turbopilot/pull/21

Full Changelog: https://github.com/ravenscroftj/turbopilot/compare/v0.0.4...v0.0.5

1 year ago

Added multi-threaded server support which should prevent health checks aimed at GET / from failing during prediction.
Separated autocomplete lambda into a separate C++ function so that it can be bound to /v1/completions, /v1/engines/copilot-codex/completions and /v1/engines/codegen/completions
Removed model from completion input as required param which stops the official copilot plugin from freaking out
Integrate latest changes from upstream ggml including some fixes for ARM NEON processor
Added Mac "universal binary" builds as part of CI
~~Support for fork of vscode-fauxpilot with a progress indicator is now available (PR is open upstream, please react/vote for it)~~ vscode-fauxcode now supports progress indication

1 year ago

Added 350M parameter codegen model to Google Drive folder
Added multi-arch docker images so that users can now directly run on Apple silicon and even raspberry pi 4
Now support pre-tokenized inputs passed into the API from a Python tokenizer (thanks to @thakkarparth007 for their PR - https://github.com/ravenscroftj/ggml/pull/2)

1 year ago

Project now builds on Mac OS (Thanks to @Dimitrije-V for their PR https://github.com/ravenscroftj/ggml/pull/1 and @dabdine for contributing some clearer Mac build instructions)
Fix inability to load vocab.json on converting the 16B model due to encoding of the file not being set by @swanserquack in #5
Improve performance of model by incorporating changes to GGML library from @ggerganov

Full Changelog: https://github.com/ravenscroftj/turbopilot/compare/0.0.1...0.0.2

1 year ago