diff --git a/.buildkite/intel_jobs/misc_intel.yaml b/.buildkite/intel_jobs/misc_intel.yaml index 5d38f7f329..20a9f5bd17 100644 --- a/.buildkite/intel_jobs/misc_intel.yaml +++ b/.buildkite/intel_jobs/misc_intel.yaml @@ -62,7 +62,8 @@ steps: pytest -v -s v1/logits_processors --ignore=v1/logits_processors/test_custom_online.py --ignore=v1/logits_processors/test_custom_offline.py && pytest -v -s v1/test_oracle.py && pytest -v -s v1/test_request.py && - pytest -v -s v1/test_outputs.py' + pytest -v -s v1/test_outputs.py && + pytest -v -s v1/sample/test_topk_topp_sampler.py' - label: XPU CPU Offload timeout_in_minutes: 60 diff --git a/vllm/v1/sample/ops/topk_topp_triton.py b/vllm/v1/sample/ops/topk_topp_triton.py old mode 100644 new mode 100755 index bfe6fd6ae5..d20cac37fc --- a/vllm/v1/sample/ops/topk_topp_triton.py +++ b/vllm/v1/sample/ops/topk_topp_triton.py @@ -929,8 +929,12 @@ def apply_top_k_top_p_triton( normal_cdf_to_sigma_table, percentile_to_std_table = tables # Smaller tiles compile and run faster on CPU; GPU benefits from larger tiles. + # On XPU, large BLOCK_SIZE causes precision loss in the single-pass pivot + # approximation; use smaller tiles for accurate top-p results. if logits.device.type == "cpu": block_size, block_size_trunc = 256, 128 + elif logits.device.type == "xpu": + block_size, block_size_trunc = 4096, 2048 else: block_size, block_size_trunc = 8192, 4096