v0.0.3

What's Changed

• Enable native CPU optimizations on Linux too by @alvarobartt in #9
• Update to CUDA 13.0, add missing 12.1 compute capability, and multi-target builds by @alvarobartt in #10
• Add apt-get upgrade to try to prevent some CVEs by @juanjucm in #11
• Add tests for {modernbert,laya}.rs with insta by @alvarobartt in #12
• Add --max-model-len & add Laya multi-lingual and typed-decisions by @alvarobartt in #13
• Add support for candle-flash-attn and candle-flash-attn-v3 in CUDA by @alvarobartt in #14
• Add safety note around unsafe for mmap2 by @alvarobartt in #18
• Update version to 0.0.3 & cargo update by @alvarobartt in #17

New Contributors

• @juanjucm made their first contribution in #11

Full Changelog: v0.0.2...v0.0.3

v0.0.2

What's Changed

• Update .github/workflows/build.yml by @alvarobartt in #1
• Add utoipa with /docs and openapi.json by @alvarobartt in #2
• Fix Dockerfile.cuda & fix Turing support in candle by @alvarobartt in #3
• Build CUDA compute capability based containers by @alvarobartt in #5
• Update version to 0.0.2 & cargo update by @alvarobartt in #4
• Add /metrics, --dtype, better ModernBert + Laya (CPU, Metal, and CUDA), and more by @alvarobartt in #7

Full Changelog: v0.0.1...v0.0.2

v0.0.1

Full Changelog: https://github.com/alvarobartt/sys1/commits/v0.0.1