diff --git a/test_inference.py b/test_inference.py
index f7aa1259..c618d99e 100644
--- a/test_inference.py
+++ b/test_inference.py
@@ -467,22 +467,36 @@ def test_ppl_token():
 
         print(f" -- Measuring prompt speed...")
 
+        torch.cuda.synchronize()
+
         current_len = 128
+        step = 128
+        prompt_iters = 3
         while True:
 
-            time_begin = time.time()
+            total_time = 0
+            for i in range(prompt_iters):
 
-            cache.current_seq_len = 0
-            model.forward(ids[:, :current_len], cache, preprocess_only = True)
-            torch.cuda.synchronize()
+                torch.cuda.synchronize()
+                time_begin = time.time()
 
-            time_end = time.time()
-            tps = current_len / (time_end - time_begin)
+                cache.current_seq_len = 0
+                model.forward(ids[:, :current_len], cache, preprocess_only = True)
+
+                torch.cuda.synchronize()
+                time_end = time.time()
+                total_time += time_end - time_begin
+
+            tps = current_len / (total_time / prompt_iters)
 
             print(f" ** Length {current_len:>5} tokens: {tps:>11.4f} t/s")
 
+            if current_len >= 1024: step = 1024
+            if current_len >= 4096: step = 4096
+            if current_len >= 16384: step = 8192
+
             current_len_ = current_len
-            current_len = min(current_len + 128, model.config.max_seq_len)
+            current_len = min(current_len + step, model.config.max_seq_len)
             if current_len == current_len_: break