From 52c70b210ce9d66e9afb9d18e086c3d05408c492 Mon Sep 17 00:00:00 2001 From: zofia <110436990+zufangzhu@users.noreply.github.com> Date: Tue, 11 Aug 2026 16:43:33 +0800 Subject: [PATCH] [XPU] [Linear] enable torch linear backend for blockwise gemm on xpu (#50826) Signed-off-by: Zhu, Zufang Co-authored-by: Kunshang Ji --- .buildkite/intel_jobs/test-intel.yaml | 1 + vllm/model_executor/kernels/linear/__init__.py | 1 + 2 files changed, 2 insertions(+) diff --git a/.buildkite/intel_jobs/test-intel.yaml b/.buildkite/intel_jobs/test-intel.yaml index d2afbc7eef..7106404b8a 100644 --- a/.buildkite/intel_jobs/test-intel.yaml +++ b/.buildkite/intel_jobs/test-intel.yaml @@ -72,6 +72,7 @@ steps: 'python3 examples/basic/offline_inference/generate.py --linear-backend xpu --model RedHatAI/Meta-Llama-3.1-8B-Instruct-FP8 --enforce-eager --max-model-len 4096 && python3 examples/basic/offline_inference/generate.py --linear-backend torch --model RedHatAI/Meta-Llama-3.1-8B-Instruct-FP8 --enforce-eager --max-model-len 4096 && python3 examples/basic/offline_inference/generate.py --linear-backend xpu --model neuralmagic/Llama-3.2-1B-Instruct-FP8-dynamic --enforce-eager --max-model-len 4096 && + python3 examples/basic/offline_inference/generate.py --linear-backend torch --model Qwen/Qwen3-4B-Instruct-2507-FP8 --enforce-eager --max-model-len 4096 && python3 examples/basic/offline_inference/generate.py --linear-backend xpu --model meta-llama/Llama-3.2-1B-Instruct --quantization fp8 --enforce-eager --max-model-len 4096 && python3 examples/basic/offline_inference/generate.py --linear-backend torch --model meta-llama/Llama-3.2-1B-Instruct --quantization fp8 --enforce-eager --max-model-len 4096 ' diff --git a/vllm/model_executor/kernels/linear/__init__.py b/vllm/model_executor/kernels/linear/__init__.py index 90e983653e..9915255a84 100644 --- a/vllm/model_executor/kernels/linear/__init__.py +++ b/vllm/model_executor/kernels/linear/__init__.py @@ -426,6 +426,7 @@ _POSSIBLE_FP8_BLOCK_KERNELS: dict[ PlatformEnum.XPU: [ XPUFp8BlockScaledMMKernel, TritonFp8BlockScaledMMKernel, + BlockWiseTorchFP8ScaledMMLinearKernel, ], }