PTQ-Level Efficiency, QAT-Level Accuracy | 训练后量化的效率,量化感知训练的精度
fix: update field extraction for ServerArgs in _run_sglang_worker (#2474)
Xin Hea065121579ca
diff --git a/test/integration/test_cuda/test_sglang.py b/test/integration/test_cuda/test_sglang.pyindex a89c0fd6f..f90e64ffd 100644--- a/test/integration/test_cuda/test_sglang.py+++ b/test/integration/test_cuda/test_sglang.py@@ -32,7 +32,7 @@ def _run_sglang_worker(model_path, connection):import sglang as sglfrom sglang.srt.server_args import ServerArgs- fields = {field.name for field in dataclasses.fields(ServerArgs)}+ fields = set(ServerArgs.__annotations__.keys())engine_kwargs = {"model_path": model_path, "mem_fraction_static": 0.5}if "disable_piecewise_cuda_graph" in fields:engine_kwargs["disable_piecewise_cuda_graph"] = True