diff --git a/run.py b/run.py
index 4cc9a31..95ec09d 100644
--- a/run.py
+++ b/run.py
@@ -112,7 +112,7 @@ def get_op_name(executable_path):
     return "unknown"
 
 
-def run_pytorch_benchmark(op_name, N, num_repeats=1, warmup=0):
+def run_pytorch_benchmark(op_name, N, num_repeats=10, warmup=10):
     """
     Runs the equivalent operation in PyTorch and measures execution time.
     Compatible with both CPU and GPU.
@@ -281,6 +281,9 @@ def main(gpu=False):
             print("-" * len(header))
 
             for N in BENCHMARK_SIZES:
+                if HAS_TORCH and torch.cuda.is_available():
+                    torch.cuda.empty_cache()
+
                 # 1. Run C++ Benchmark
                 cpp_res = run_cpp_benchmark(EXECUTABLE_PATH, [N])
                 if cpp_res: