Run specific batch_size and n_features for jagged_mean

Author: yf225

benchmarks/run.py

@@ -99,7 +99,7 @@
         "jagged_mean_tritonbench",
         {"B": 32, "M": 8, "seqlen": 64}
         if os.environ.get("HELION_DEV_LOW_VRAM", "0") == "1"
-        else {},
+        else {"B": 512, "M": 64},
     ),
     "vector_add": ("tritonbench.operators.vector_add.operator", "examples.add", "add"),
     "embedding": (