add fp16 kv_cache test

vllm-project · Aug 30, 2024 · b601bb6 · b601bb6
1 parent 01d5dd6
commit b601bb6
Showing 1 changed file with 2 additions and 2 deletions.
diff --git a/tests/basic_correctness/test_chunked_prefill.py b/tests/basic_correctness/test_chunked_prefill.py
@@ -79,8 +79,8 @@ def test_models(
 
 @pytest.mark.parametrize("kv_cache_dtype,model",
                          [("fp8_e5m2", m)
-                          for m in E5M2_KV_MODELS] + [("fp8_e4m3", m)
-                                                      for m in E4M3_KV_MODELS]
+                          for m in E5M2_KV_MODELS]
+                         + [("fp8_e4m3", m) for m in E4M3_KV_MODELS]
                          + [("auto", "Qwen/Qwen2-1.5B-Instruct")])
 # Due to low-precision numerical divergence, we only test logprob of 4 tokens
 @pytest.mark.parametrize("max_tokens", [4])