v2.2.0
What's Changed
This release brings updates to the backends and includes a fix for recompilation of LocalAI with go-rwkv (#1473). To note, it also tries to reduce the image size by allowing some backends (transformers-based) to share the same environment.
With this release inline templates and models as URLs are supported, for example:
name: mixtral
parameters:
model: https://huggingface.co/TheBloke/Mixtral-8x7B-v0.1-GGUF/resolve/main/mixtral-8x7b-v0.1.Q2_K.gguf
# or huggingface://TheBloke/Mixtral-8x7B-v0.1-GGUF/mixtral-8x7b-v0.1.Q2_K.gguf@main
template:
completion: |
Complete the following: {{.Input}}
Exciting New Features 🎉
- docs: automatically track latest versions by @mudler in #1451
- feat: inline templates and accept URLs in models by @mudler in #1452
- feat(conda): share envs with transformer-based backends by @mudler in #1465
👒 Dependencies
- ⬆️ Update ggerganov/llama.cpp by @localai-bot in #1450
- ⬆️ Update ggerganov/llama.cpp by @localai-bot in #1455
- ⬆️ Update ggerganov/llama.cpp by @localai-bot in #1457
- ⬆️ Update ggerganov/whisper.cpp by @localai-bot in #1462
- ⬆️ Update ggerganov/llama.cpp by @localai-bot in #1461
- ⬆️ Update donomii/go-rwkv.cpp by @localai-bot in #1470
Other Changes
- How To (Updates and Fixes) by @lunamidori5 in #1456
- Revert ":arrow_up: Update donomii/go-rwkv.cpp" by @mudler in #1474
Full Changelog: v2.1.0...v2.2.0