diff --git a/torchbenchmark/util/backends/torchdynamo.py b/torchbenchmark/util/backends/torchdynamo.py
index 9fac86784e..2eedc89dc7 100644
--- a/torchbenchmark/util/backends/torchdynamo.py
+++ b/torchbenchmark/util/backends/torchdynamo.py
@@ -53,6 +53,11 @@ def parse_torchdynamo_args(model: 'torchbenchmark.util.model.BenchmarkModel', dy
         action='store_true',
         help="enable group fusion in Inductor"
     )
+    parser.add_argument(
+        "--torchinductor_enable_batch_fusion",
+        action='store_true',
+        help="enable batch fusion in Inductor"
+    )
     parser.add_argument(
         "--dynamo_disable_optimizer_step",
         type=distutils.util.strtobool,
@@ -86,6 +91,8 @@ def apply_torchdynamo_args(model: 'torchbenchmark.util.model.BenchmarkModel', ar
             # torchinductor.config.triton.use_bmm = True
         if args.torchinductor_enable_group_fusion:
             torchinductor.config.group_fusion = True
+        if args.torchinductor_enable_batch_fusion:
+            torchinductor.config.batch_fusion = True
 
         # used for correctness checks, to avoid triton rand() behaving differently from torch rand().
         torchinductor.config.fallback_random = bool(args.torchinductor_fallback_random)