[Misc] Bump up test_fused_moe tolerance (vllm-project#10364)

Signed-off-by: ElizaWszola <[email protected]>
opendatahub-io · Nov 15, 2024 · 79ee45b · 79ee45b
1 parent 691a3ec
commit 79ee45b
Showing 1 changed file with 1 addition and 1 deletion.
diff --git a/tests/kernels/test_moe.py b/tests/kernels/test_moe.py
@@ -45,7 +45,7 @@ def test_fused_moe(
     score = torch.randn((m, e), device="cuda", dtype=dtype)
     triton_output = fused_moe(a, w1, w2, score, topk, renormalize=False)
     torch_output = torch_moe(a, w1, w2, score, topk)
-    torch.testing.assert_close(triton_output, torch_output, atol=1e-2, rtol=0)
+    torch.testing.assert_close(triton_output, torch_output, atol=2e-2, rtol=0)
 
 
 @pytest.mark.parametrize("dtype",