Add cmf_10x_batch_fusion test to Sandcastle and AIBench

Summary: As the title says. We want to add a test to cover batch fusion feature on cmf.

Reviewed By: jackiexu1992

Differential Revision: D48622074

fbshipit-source-id: 73c6cb52206805b9c6c3795789228bfacc35f80a

torchbenchmark/util/backends/torchdynamo.py

@@ -58,6 +58,11 @@ def parse_torchdynamo_args(model: 'torchbenchmark.util.model.BenchmarkModel', dy
         action='store_true',
         help="enable batch fusion in Inductor"
     )
+    parser.add_argument(
+        "--torchinductor_enable_split_cat_fx_pass",
+        action='store_true',
+        help="enable split_cat_fx_pass in Inductor"
+    )
     parser.add_argument(
         "--dynamo_disable_optimizer_step",
         type=distutils.util.strtobool,
@@ -92,7 +97,10 @@ def apply_torchdynamo_args(model: 'torchbenchmark.util.model.BenchmarkModel', ar
         if args.torchinductor_enable_group_fusion:
             torchinductor.config.group_fusion = True
         if args.torchinductor_enable_batch_fusion:
+            torchinductor.config.pattern_matcher = True
             torchinductor.config.batch_fusion = True
+        if args.torchinductor_enable_split_cat_fx_pass:
+            torchinductor.config.split_cat_fx_passes = True
 
         # used for correctness checks, to avoid triton rand() behaving differently from torch rand().
         torchinductor.config.fallback_random = bool(args.torchinductor_fallback_random)