From d227611809a3db194df0d59d04a3c4836f2d49f5 Mon Sep 17 00:00:00 2001
From: "copilot-swe-agent[bot]" <198982749+Copilot@users.noreply.github.com>
Date: Thu, 6 Nov 2025 03:49:31 +0000
Subject: [PATCH 1/3] Initial plan


From f28be6289e2b1c0b16fb6755aedd56844518d744 Mon Sep 17 00:00:00 2001
From: "copilot-swe-agent[bot]" <198982749+Copilot@users.noreply.github.com>
Date: Thu, 6 Nov 2025 03:56:18 +0000
Subject: [PATCH 2/3] Add Output TPUT per GPU and Input TPUT per GPU to summary
 table

Co-authored-by: functionstackx <47992694+functionstackx@users.noreply.github.com>
---
 utils/process_result.py | 3 ++-
 utils/summarize.py      | 8 +++++---
 2 files changed, 7 insertions(+), 4 deletions(-)

diff --git a/utils/process_result.py b/utils/process_result.py
index 94ca30f241..7992bed380 100644
--- a/utils/process_result.py
+++ b/utils/process_result.py
@@ -26,7 +26,8 @@
     'framework': framework,
     'precision': precision,
     'tput_per_gpu': float(bmk_result['total_token_throughput']) / tp_size,
-    'output_tput_per_gpu': float(bmk_result['output_throughput']) / tp_size
+    'output_tput_per_gpu': float(bmk_result['output_throughput']) / tp_size,
+    'input_tput_per_gpu': float(bmk_result['input_throughput']) / tp_size
 }
 
 if mtp_mode:  # MTP
diff --git a/utils/summarize.py b/utils/summarize.py
index 6d926255ed..7668729fef 100644
--- a/utils/summarize.py
+++ b/utils/summarize.py
@@ -12,8 +12,8 @@
 results.sort(key=lambda r: (r['hw'], r.get('framework', 'vllm'), r.get('precision', 'fp8'), r['tp'], r['ep'], r['conc']))
 
 summary_header = f'''\
-| Hardware | Framework | Precision | TP | EP | DP Attention | Conc | TTFT (ms) | TPOT (ms) | E2EL (s) | TPUT per GPU |
-| :-: | :-: | :-: | :-: | :-: | :-: | :-: | :-: | :-: | :-: | :-: |\
+| Hardware | Framework | Precision | TP | EP | DP Attention | Conc | TTFT (ms) | TPOT (ms) | E2EL (s) | TPUT per GPU | Output TPUT per GPU | Input TPUT per GPU |
+| :-: | :-: | :-: | :-: | :-: | :-: | :-: | :-: | :-: | :-: | :-: | :-: | :-: |\
 '''
 print(summary_header)
 
@@ -31,5 +31,7 @@
         f"| {(result['median_ttft'] * 1000):.4f} "
         f"| {(result['median_tpot'] * 1000):.4f} "
         f"| {result['median_e2el']:.4f} "
-        f"| {result['tput_per_gpu']:.4f} |"
+        f"| {result['tput_per_gpu']:.4f} "
+        f"| {result['output_tput_per_gpu']:.4f} "
+        f"| {result['input_tput_per_gpu']:.4f} |"
     )

From ed4639185e5e532caee7e9d8438ba693d1daff1a Mon Sep 17 00:00:00 2001
From: "copilot-swe-agent[bot]" <198982749+Copilot@users.noreply.github.com>
Date: Thu, 6 Nov 2025 03:57:12 +0000
Subject: [PATCH 3/3] Add trailing comma to dictionary for better
 maintainability

Co-authored-by: functionstackx <47992694+functionstackx@users.noreply.github.com>
---
 utils/process_result.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/utils/process_result.py b/utils/process_result.py
index 7992bed380..bb0fff7fac 100644
--- a/utils/process_result.py
+++ b/utils/process_result.py
@@ -27,7 +27,7 @@
     'precision': precision,
     'tput_per_gpu': float(bmk_result['total_token_throughput']) / tp_size,
     'output_tput_per_gpu': float(bmk_result['output_throughput']) / tp_size,
-    'input_tput_per_gpu': float(bmk_result['input_throughput']) / tp_size
+    'input_tput_per_gpu': float(bmk_result['input_throughput']) / tp_size,
 }
 
 if mtp_mode:  # MTP