Anyone see the TensorRT-LLM announcement from NVIDIA?

Would be amazing if the end result of this is a near doubling of the available message cap!


Looks like they can quantize to FP8 pretty easily:

NVIDIA H100 GPUs with TensorRT-LLM give users the ability to convert their model weights into a new FP8 format easily and compile their models to take advantage of optimized FP8 kernels automatically. This is made possible through Hopper Transformer Engine technology and done without having to change any model code.

Also found this line interesting:

TensorRT-LLM improves ease of use and extensibility through an open-source modular Python API for defining, optimizing, and executing new architectures and enhancements as LLMs evolve, and can be customized easily.

1 Like

NVIDIA has been working closely with leading companies, including Meta, Anyscale, Cohere, Deci, Grammarly, Mistral AI, MosaicML, now a part of Databricks, OctoML, Tabnine, and Together AI, to accelerate and optimize LLM inference. (coming open-source software)

Who’s not mentiioned? The one that is likely already crushing quants to the limit.