Releases: groovybits/rsllm
Releases · groovybits/rsllm
v0.4.0 - Full GPU Saturation: fixes threading so as to allow concurrency
Concurrency is now worked out, fixes issues with using more ram than reasonable which could crash a system. Now it is well regulated and can use multiple threads concurrently yet not eat more memory than necessary and not have out of order NDI output.
What's Changed
- Threading - Consolidate image and speech generation functions into a concurrent async pipeline to NDI by @groovybits in #17
Full Changelog: v0.3.0...v0.4.0
v0.3.0
very alpha, basic functionality without good threading behavior and lack of concurrency of generation. WIP
What's Changed
- AI Network Analyzer: Integrate Pcap/DPDK logic from rscap as a separate library in rsllm by @groovybits in #1
- AI Network Analyzer - use packet capture analysis for LLM input by @groovybits in #2
- Control history and context size by @groovybits in #3
- Allow network capture to run in a thread and supply us with stream stats, not just the packets. by @groovybits in #4
- Stable diffusion by @groovybits in #5
- Candle mistral basic functionality by @groovybits in #6
- Gemma support! by @groovybits in #7
- Stable diffusion with candle by @groovybits in #8
- NDI subtitle burnin by @groovybits in #9
- Openai tts by @groovybits in #10
- Twitch chat client - Initial code WIP by @groovybits in #11
- metavoice, working! by @groovybits in #12
- Mimic3 TTS added for free alternative with good quality for now by @groovybits in #13
- break lines at 80 for terminal output by @groovybits in #14
- use function for output line breaking by @groovybits in #15
New Contributors
- @groovybits made their first contribution in #1
Full Changelog: https://github.com/groovybits/rsllm/commits/v0.3.0