diff --git a/.dev_scripts/benchmark_regression/1-benchmark_valid.py b/.dev_scripts/benchmark_regression/1-benchmark_valid.py
index eb1541a15..3d9fc8d62 100644
--- a/.dev_scripts/benchmark_regression/1-benchmark_valid.py
+++ b/.dev_scripts/benchmark_regression/1-benchmark_valid.py
@@ -48,6 +48,11 @@ def parse_args():
         '--inference-time',
         action='store_true',
         help='Test inference time by run 10 times for each model.')
+    parser.add_argument(
+        '--batch-size',
+        type=int,
+        default=1,
+        help='The batch size during the inference.')
     parser.add_argument(
         '--flops', action='store_true', help='Get Flops and Params of models')
     parser.add_argument(
@@ -86,7 +91,7 @@ def inference(config_file, checkpoint, work_dir, args, exp_name):
         test_dataset.pipeline.insert(1, dict(type='Resize', scale=32))
 
     data = Compose(test_dataset.pipeline)({'img_path': args.img})
-    data = default_collate([data])
+    data = default_collate([data] * args.batch_size)
     resolution = tuple(data['inputs'].shape[-2:])
 
     runner: Runner = Runner.from_cfg(cfg)
@@ -103,7 +108,7 @@ def inference(config_file, checkpoint, work_dir, args, exp_name):
                 start = time()
                 model.val_step(data)
                 torch.cuda.synchronize()
-                time_record.append((time() - start) * 1000)
+                time_record.append((time() - start) / args.batch_size * 1000)
             result['time_mean'] = np.mean(time_record[1:-1])
             result['time_std'] = np.std(time_record[1:-1])
         else: