A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 14, 05:30 UTC
[CI Perf] Only test bfloat16 for tests/compile/test_fusion_all_reduce.py (#23132)
Signed-off-by: mgoin <mgoin64@gmail.com>
Michael Goin committed
Aug 19, 2025 at 22:18 UTC
d46d417b5897d7eddb002b61b19e8cba029c3dda
1 file changed
+1
-1
tests/compile/test_fusion_all_reduce.py
+1
-1
@@ -148,7 +148,7 @@ class TestAllReduceFusedAddRMSNormStaticQuantFP4Model(torch.nn.Module):
148
@pytest.mark.parametrize("batch_size", [8])
149
@pytest.mark.parametrize("seq_len", [8])
150
@pytest.mark.parametrize("hidden_size", [16])
151
-@pytest.mark.parametrize("dtype", [torch.float16, torch.bfloat16])
151
+@pytest.mark.parametrize("dtype", [torch.bfloat16])
152
@pytest.mark.skipif(envs.VLLM_TARGET_DEVICE not in ["cuda"],
153
reason="Only test on CUDA")
154
@pytest.mark.skipif(