Release Notes

Each release has a corresponding branch in the NanoLLM GitHub repository and container images on DockerHub. For more info about running these, see the Installation Guide.

Container Versions

The latest builds following the main branch are dustynv/nano_llm:r35.4.1 for JetPack 5 and dustynv/nano_llm:r36.2.0 for JetPack 6. Check the tags on DockerHub for the versions below.

24.7

• Initial release of Agent Studio

24.6

• Added LLM backends for AWQ (--api=awq) and HuggingFace Transformers (--api=hf)

• Support for OpenAI-style tool calling and NousResearch/Hermes-2-Pro-Llama-3-8B

• Plugin for whisper_trt and VAD

24.5.1

• Added Video VILA on image sequences

• Added KVCache interface for KV cache manipulation

• Updated ChatHistory and ChatMessage with deletion/removal operators

24.5

• Added function calling with Llama 3

• Added VILA-1.5 with TensorRT for CLIP/SigLIP

24.4.1

• Added chat templates for Llama 3

• Added NanoLLM page to Jetson AI Lab

• Added simplified ‘Live Llava’ code example under nano_llm/vision/example.py

• Changed over some lingering references from local_llm

24.4

• Migration from local_llm in jetson-containers

• Generated docs site (dusty-nv.github.io/NanoLLM)

• Year/month/date versioning with stable releases being cut for each month