llama.cpp/CHANGELOG.md

100 lines
1.7 KiB
Markdown
Raw Normal View History

2023-05-26 21:32:34 +00:00
# Changelog
All notable changes to this project will be documented in this file.
The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
## [Unreleased]
2023-06-26 12:53:54 +00:00
## [0.1.66]
## Added
- (llama.cpp) New model API
## Fixed
- Performance issue during eval caused by looped np.concatenate call
- State pickling issue when saving cache to disk
2023-06-20 15:25:44 +00:00
## [0.1.65]
2023-06-20 15:25:10 +00:00
### Added
- (llama.cpp) Fix struct misalignment bug
2023-06-18 13:48:43 +00:00
## [0.1.64]
2023-06-15 02:15:44 +00:00
### Added
2023-06-17 17:37:14 +00:00
- (llama.cpp) Update llama.cpp
2023-06-17 17:39:48 +00:00
- Fix docs for seed. Set -1 for random.
2023-06-17 17:37:14 +00:00
## [0.1.63]
### Added
2023-06-15 02:15:44 +00:00
- (llama.cpp) Add full gpu utilisation in CUDA
- (llama.cpp) Add get_vocab
- (llama.cpp) Add low_vram parameter
- (server) Add logit_bias parameter
2023-06-10 22:19:48 +00:00
## [0.1.62]
2023-06-10 16:22:39 +00:00
### Fixed
- Metal support working
- Cache re-enabled
2023-06-10 03:25:38 +00:00
## [0.1.61]
### Fixed
- Fix broken pip installation
2023-06-09 15:52:07 +00:00
## [0.1.60]
### NOTE
2023-06-09 15:01:42 +00:00
- This release was deleted due to a bug with the packaging system that caused pip installations to fail.
2023-06-09 01:53:38 +00:00
2023-06-09 15:01:42 +00:00
### Fixed
2023-06-09 15:01:42 +00:00
- Truncate max_tokens in create_completion so requested tokens doesn't exceed context size.
- Temporarily disable cache for completion requests
2023-06-09 15:01:42 +00:00
2023-06-08 07:26:49 +00:00
## [v0.1.59]
2023-06-06 21:01:10 +00:00
### Added
- (llama.cpp) k-quants support
- (server) mirostat sampling parameters to server
2023-06-05 22:17:29 +00:00
2023-06-08 07:26:49 +00:00
### Fixed
- Support both `.so` and `.dylib` for `libllama` on MacOS
2023-06-05 03:31:51 +00:00
## [v0.1.58]
2023-06-06 21:01:10 +00:00
### Added
- (llama.cpp) Metal Silicon support
2023-06-05 03:30:42 +00:00
2023-06-05 03:30:10 +00:00
## [v0.1.57]
2023-06-06 21:01:10 +00:00
### Added
- (llama.cpp) OpenLlama 3B support
2023-06-05 03:30:10 +00:00
2023-05-30 07:07:36 +00:00
## [v0.1.56]
2023-05-26 21:32:34 +00:00
### Added
2023-06-06 21:01:10 +00:00
- (misc) Added first version of the changelog
- (server) Use async routes
- (python-api) Use numpy for internal buffers to reduce memory usage and improve performance.
2023-05-27 00:23:49 +00:00
### Fixed
2023-06-06 21:01:10 +00:00
- (python-api) Performance bug in stop sequence check slowing down streaming.