Add Output TPUT per GPU and Input TPUT per GPU columns to summary table (#178)

* Initial plan

* Add Output TPUT per GPU and Input TPUT per GPU to summary table

Co-authored-by: functionstackx <47992694+functionstackx@users.noreply.github.com>

* Add trailing comma to dictionary for better maintainability

Co-authored-by: functionstackx <47992694+functionstackx@users.noreply.github.com>

---------

Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com>
Co-authored-by: functionstackx <47992694+functionstackx@users.noreply.github.com>

Author: Copilot

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} |"
     )