Fix GPU OOM (#6564)

Co-authored-by: michael <michael.zhang@amd.com>
2025-05-25 07:38:39 +08:00
parent 24c035f2e3
commit 7a5e6ce1cb
8 changed files with 15 additions and 16 deletions
--- a/test/srt/test_bench_one_batch.py
+++ b/test/srt/test_bench_one_batch.py
@@ -62,7 +62,10 @@ class TestBenchOneBatch(CustomTestCase):
                f"### test_torch_compile_tp2_bs1 (Mixtral-8x7B)\n"
                f"output_throughput: {output_throughput:.2f} token/s\n"
            )
-            self.assertGreater(output_throughput, 220)
+            if os.getenv("SGLANG_AMD_CI") == "1":
+                self.assertGreater(output_throughput, 200)
+            else:
+                self.assertGreater(output_throughput, 220)


 if __name__ == "__main__":