Hacker News: NVIDIA introduces TensorRT-LLM for accelerating LLM inference on H100/A100 GPUs

Story dated 2023-09-08 by the Algolia API; 69 points and 33 comments at access 2026-10-04. Included because the thread's top comment is itself a sentiment record: at launch the friction was gated access ('must be registered with the NVIDIA Developer Program'), and the read of the vendor's 'ready-to-run, optimized versions of many LLMs' claim was contested in the comments as a bring-your-own-weights claim. Points are a popularity proxy, not a measurement.