vLLM v0.33.0rc2 released with bugfix for gfx950
Key point
vLLM v0.33.0rc2 has been released. This version of the high-throughput and memory-efficient inference and serving engine for LLMs includes a bugfix to skip AITER tuned GEMM when M*N >= 2^31 on gfx950.
vLLM v0.33.0rc2 has been released. vLLM is a high-throughput and memory-efficient inference and serving engine for LLMs.
This release includes a bugfix for ROCm on gfx950 to skip AITER tuned GEMM when M*N >= 2^31. This change was signed-off by Micah Williamson.
This item is a summary. The full, checkable version is at the original source:
Read the original at vLLM ↗