Releases: software-mansion/react-native-executorch
Releases · software-mansion/react-native-executorch
v0.6.0
Announcing React Native ExecuTorch v0.6.0 🫶🏻
We are excited to release v0.6.0! This version brings some major performance improvements and introduces powerful new generative capabilities including Stable Diffusion, VAD and improved Whisper support.
What’s new?
- Major performance improvements: see benchmarks for more details
- Text to Image: Support for Stable Diffusion is here! docs
- Voice Activity Detection: Detect speech segments with FSMN-based VAD implementation. docs
- Quantized Whisper: We now support quantized Whisper models, resulting in 4x smaller file sizes without significant accuracy tradeoffs. docs
- LLM Control: You can now configure your generation with configurable
temperatureandtoppparameters. docs - ExecuTorch Runtime: Bumped to v1.0.0 to match upstream stability and features.
- Android Support: Added support for 16kB memory page sizes.
- General bug fixes
Breaking Changes ⚠️
Expo SDK Upgrade Required
React Native ExecuTorch now depends on the latest expo-file-system API changes. You must upgrade your project to Expo SDK 54 to use this version.
0.5.15
What's Changed
Fixed bug on android where number of threads in threadpools could be set to 0 in effect freezing the library. (issue should only impact simulators)
Full Changelog: v0.5.14...v0.5.15
0.5.13
What's Changed
- Remove CocoaPods cache step from iOS build workflow by @msluszniak in #654
- Fixed bug impacting quality of llms responses and sometimes causing models to keep generating until max number of tokens is reached.
Full Changelog: v0.5.12...v0.5.13
v0.4.10
What's Changed
- Fixed bug impacting quality of llms responses and sometimes causing models to keep generating until max number of tokens is reached.
Full Changelog: v0.4.9...v0.4.10
0.5.12
0.5.11
0.4.9
v0.5.10
v0.5.9
v0.5.8
What's Changed
- Added missing getGeneratedTokenCount for LLMModule.ts
Full Changelog: v0.5.7...v0.5.8