bloom model still not support In-flight batching and paged kv cache

NVIDIA / TensorRT-LLM

TensorRT-LLM provides users with an easy-to-use Python API to define Large Language Models (LLMs) and build TensorRT engines that contain state-of-the-art optimizations to perform inference efficiently on NVIDIA GPUs. TensorRT-LLM also contains components to create Python and C++ runtimes that execute those TensorRT engines.

https://nvidia.github.io/TensorRT-LLM

Apache License 2.0

8.03k stars 883 forks source link

bloom model still not support In-flight batching and paged kv cache #134

Open luxiushu2023 opened 10 months ago

luxiushu2023 commented 10 months ago

### Tasks

ncomly-nvidia commented 9 months ago

ACK - yep that is true! We are working on some improvements that should make adding these features for a model much more straight forward (if any work at all)!