Skip to content

PTQ-Level Efficiency, QAT-Level Accuracy | 训练后量化的效率,量化感知训练的精度

Enable Torch compilation for throughput tests (#2181)

Xin Hee55beddc5bf5
diff --git a/test/e2e/test_cuda/test_sglang_throughput.py b/test/e2e/test_cuda/test_sglang_throughput.py
index 5cb291343..dcdaf05a2 100644
--- a/test/e2e/test_cuda/test_sglang_throughput.py
+++ b/test/e2e/test_cuda/test_sglang_throughput.py
@@ -46,6 +46,8 @@ from typing import List
import pytest
import torch
+pytestmark = pytest.mark.enable_torch_compile
+
# ---------------------------------------------------------------------------
# Output sink
# ---------------------------------------------------------------------------
diff --git a/test/e2e/test_cuda/test_vllm_throughput.py b/test/e2e/test_cuda/test_vllm_throughput.py
index e2c0006ae..a13a339fe 100644
--- a/test/e2e/test_cuda/test_vllm_throughput.py
+++ b/test/e2e/test_cuda/test_vllm_throughput.py
@@ -55,6 +55,8 @@ from typing import List
import pytest
import torch
+pytestmark = pytest.mark.enable_torch_compile
+
# ---------------------------------------------------------------------------
# Output sink
# ---------------------------------------------------------------------------