diff --git a/lama_cleaner/benchmark.py b/lama_cleaner/benchmark.py
new file mode 100644
index 0000000..0e134db
--- /dev/null
+++ b/lama_cleaner/benchmark.py
@@ -0,0 +1,103 @@
+#!/usr/bin/env python3
+
+import argparse
+import multiprocessing
+import os
+import time
+
+import numpy as np
+import nvidia_smi
+import psutil
+import torch
+from tqdm import tqdm
+
+from lama_cleaner.lama import LaMa
+
+try:
+    torch._C._jit_override_can_fuse_on_cpu(False)
+    torch._C._jit_override_can_fuse_on_gpu(False)
+    torch._C._jit_set_texpr_fuser_enabled(False)
+    torch._C._jit_set_nvfuser_enabled(False)
+except:
+    pass
+
+from lama_cleaner.helper import norm_img
+
+NUM_THREADS = str(4)
+
+os.environ["OMP_NUM_THREADS"] = NUM_THREADS
+os.environ["OPENBLAS_NUM_THREADS"] = NUM_THREADS
+os.environ["MKL_NUM_THREADS"] = NUM_THREADS
+os.environ["VECLIB_MAXIMUM_THREADS"] = NUM_THREADS
+os.environ["NUMEXPR_NUM_THREADS"] = NUM_THREADS
+if os.environ.get("CACHE_DIR"):
+    os.environ["TORCH_HOME"] = os.environ["CACHE_DIR"]
+
+
+def run_model(model, size):
+    # RGB
+    image = np.random.randint(0, 256, (size[0], size[1], 3)).astype(np.uint8)
+    image = norm_img(image)
+
+    mask = np.random.randint(0, 255, size).astype(np.uint8)
+    mask = norm_img(mask)
+    model(image, mask)
+
+
+def benchmark(model, times: int, empty_cache: bool):
+    sizes = [
+        (512, 512),
+        (640, 640),
+        (1080, 800),
+        (2000, 2000)
+    ]
+
+    nvidia_smi.nvmlInit()
+    device_id = 0
+    handle = nvidia_smi.nvmlDeviceGetHandleByIndex(device_id)
+
+    def format(metrics):
+        return f"{np.mean(metrics):.2f} ± {np.std(metrics):.2f}"
+
+    process = psutil.Process(os.getpid())
+    # 每个 size 给出显存和内存占用的指标
+    for size in sizes:
+        torch.cuda.empty_cache()
+        time_metrics = []
+        cpu_metrics = []
+        memory_metrics = []
+        gpu_memory_metrics = []
+        for _ in range(times):
+            start = time.time()
+            run_model(model, size)
+            torch.cuda.synchronize()
+            if empty_cache:
+                torch.cuda.empty_cache()
+
+            # cpu_metrics.append(process.cpu_percent())
+            time_metrics.append((time.time() - start) * 1000)
+            memory_metrics.append(process.memory_info().rss / 1024 / 1024)
+            gpu_memory_metrics.append(nvidia_smi.nvmlDeviceGetMemoryInfo(handle).used / 1024 / 1024)
+
+        print(f"size: {size}".center(80, "-"))
+        # print(f"cpu: {format(cpu_metrics)}")
+        print(f"latency: {format(time_metrics)}ms")
+        print(f"memory: {format(memory_metrics)} MB")
+        print(f"gpu memory: {format(gpu_memory_metrics)} MB")
+
+    nvidia_smi.nvmlShutdown()
+
+
+def get_args_parser():
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--device", default="cuda", type=str)
+    parser.add_argument("--times", default=20, type=int)
+    parser.add_argument("--empty-cache", action="store_true")
+    return parser.parse_args()
+
+
+if __name__ == "__main__":
+    args = get_args_parser()
+    device = torch.device(args.device)
+    model = LaMa(device)
+    benchmark(model, args.times, args.empty_cache)