FasterTransformer

The 'TensorRT-LLM-based inference server' in the Databricks best-practices post, and the baseline engine in the vLLM PagedAttention paper and in the SmoothQuant paper (which integrates SmoothQuant into FasterTransformer for its INT8 speed claims). Recorded so benchmark records can point at an engine id instead of free-text engine names. SmoothQuant's 1.56x speedup over FP16 is a FasterTransformer INT8 result, NOT a vLLM or TensorRT-LLM result. ||| MAINTENANCE (dated fact, GitHub API, 2026-10-04): last commit 2023-10-19 (35.5 months before this record), latest release release/v5.3_tag published 2023-01-23, 40 contributors (distinct committers including anonymous), open_issues_count 288 (issues PLUS pull requests - see [[ev-github-open-issues-count-includes-pull-requests]]), not archived, 6457 stars. See [[ev-maintenance-status-is-four-dated-facts]] for how to read this fact set. ||| DEPLOYMENT: NO EVIDENCE OF PRODUCTION DEPLOYMENT FOUND as of 2026-10-04 - repositories, READMEs, docs and issue trackers were searched. See [[ev-production-evidence-is-scarce-and-nearly-always-single-site]] for the grading scheme; note that for an artifact-evaluation or paper-code release, non-deployment is documented intent rather than a negative finding.