diff --git a/tests/metrics/test_metrics.py b/tests/metrics/test_metrics.py
index 8b301f5ba0c49..5694061e17e02 100644
--- a/tests/metrics/test_metrics.py
+++ b/tests/metrics/test_metrics.py
@@ -169,7 +169,7 @@ def test_engine_log_metrics_regression(
 
 
 @pytest.mark.parametrize("model", MODELS)
-@pytest.mark.parametrize("dtype", ["float"])
+@pytest.mark.parametrize("dtype", ["half"])
 @pytest.mark.parametrize("max_tokens", [10])
 def test_metric_spec_decode(
     vllm_runner,
@@ -188,6 +188,10 @@ def test_metric_spec_decode(
                      num_speculative_tokens=k,
                      use_v2_block_manager=True) as vllm_model:
 
+        # Force log interval to be 0 to catch all metrics.
+        stat_logger = vllm_model.model.llm_engine.stat_loggers['prometheus']
+        stat_logger.local_interval = 0
+
         # Note that the purpose of this test is to verify spec decode
         # metrics instead of functional correctness, so the expected values
         # are intended to be loose.
@@ -204,7 +208,6 @@ def test_metric_spec_decode(
         prompts = example_prompts[:1]
 
         _ = vllm_model.generate_greedy(prompts, max_tokens)
-        stat_logger = vllm_model.model.llm_engine.stat_loggers['prometheus']
         for metric_name, is_expected in metric_name_to_expected_fn.items():
             metric_val = getattr(
                 stat_logger.metrics,