flakey test, see: vllm-project#7874 vllm-project#8051

noooop · Aug 31, 2024 · 57dc722 · 57dc722
1 parent 5e8eda1
commit 57dc722
Showing 1 changed file with 6 additions and 0 deletions.
diff --git a/tests/basic_correctness/test_chunked_prefill.py b/tests/basic_correctness/test_chunked_prefill.py
@@ -117,6 +117,12 @@ def test_models_with_fp8_kv_cache(
             "#7378: CUDA illegal memory access (undiagnosed) facebook/opt-125m"
         )
 
+    if ((model, kv_cache_dtype, chunked_prefill_token_size) ==
+            ("nm-testing/Qwen2-1.5B-Instruct-FP8-K-V", "fp8_e4m3", 4)):
+        pytest.skip(
+            "flakey test, see: #7874 #8051"
+        )
+
     max_num_seqs = chunked_prefill_token_size
     max_num_batched_tokens = chunked_prefill_token_size