Update README.md

README.md

@@ -174,10 +174,252 @@ The following performance benchmarks were conducted with [vLLM](https://docs.vll
 
 <details>
 <summary>Benchmarking Command</summary>
-guidellm --model neuralmagic/pixtral-12b-quantized.w4a16 --target "http://localhost:8000/v1" --data-type emulated --data prompt_tokens=128,generated_tokens=128,images=1,width=640,height=480 --max seconds 120 --backend aiohttp_server
+```
+  guidellm --model neuralmagic/pixtral-12b-quantized.w4a16 --target "http://localhost:8000/v1" --data-type emulated --data prompt_tokens=<prompt_tokens>,generated_tokens=<generated_tokens>,images=<num_images>,width=<image_width>,height=<image_height> --max seconds 120 --backend aiohttp_server
+```
 
 </details>
 
 ### Single-stream performance (measured with vLLM version 0.7.2)
 
+<table border="1" class="dataframe">
+  <thead>
+    <tr>
+      <th></th>
+      <th></th>
+      <th></th>
+      <th style="text-align: center;" colspan="2" >Document Visual Question Answering<br>1680W x 2240H<br>64/128</th>
+      <th style="text-align: center;" colspan="2" >Visual Reasoning <br>640W x 480H<br>128/128</th>
+      <th style="text-align: center;" colspan="2" >Image Captioning<br>480W x 360H<br>0/128</th>
+    </tr>
+    <tr>
+      <th>Hardware</th>
+      <th>Model</th>
+      <th>Average Cost Reduction</th>
+      <th>Latency (s)</th>
+      <th>QPD</th>
+      <th>Latency (s)th>
+      <th>QPD</th>
+      <th>Latency (s)</th>
+      <th>QPD</th>
+    </tr>
+  </thead>
+  <tbody style="text-align: center">
+    <tr>
+      <th rowspan="3" valign="top">A6000x1</th>
+      <th>mgoin/pixtral-12b</th>
+      <td></td>
+      <td>5.7</td>
+      <td>796</td>
+      <td>4.8</td>
+      <td>929</td>
+      <td>4.7</td>
+      <td>964</td>
+    </tr>
+    <tr>
+      <th>neuralmagic/pixtral-12b-quantized.w8a8</th>
+      <td>1.55</td>
+      <td>3.7</td>
+      <td>1220</td>
+      <td>3.1</td>
+      <td>1437</td>
+      <td>3.0</td>
+      <td>1511</td>
+    </tr>
+    <tr>
+      <th>neuralmagic/pixtral-12b-quantized.w4a16</th>
+      <td>2.16</td>
+      <td>3.2</td>
+      <td>1417</td>
+      <td>2.1</td>
+      <td>2093</td>
+      <td>1.9</td>
+      <td>2371</td>
+    </tr>
+    <tr>
+      <th rowspan="3" valign="top">A100x1</th>
+      <th>mgoin/pixtral-12b</th>
+      <td></td>
+      <td>3.0</td>
+      <td>676</td>
+      <td>2.4</td>
+      <td>825</td>
+      <td>2.3</td>
+      <td>859</td>
+    </tr>
+    <tr>
+      <th>neuralmagic/pixtral-12b-quantized.w8a8</th>
+      <td>1.38</td>
+      <td>2.2</td>
+      <td>904</td>
+      <td>1.7</td>
+      <td>1159</td>
+      <td>1.7</td>
+      <td>1201</td>
+    </tr>
+    <tr>
+      <th>neuralmagic/pixtral-12b-quantized.w4a16</th>
+      <td>1.83</td>
+      <td>1.8</td>
+      <td>1096</td>
+      <td>1.3</td>
+      <td>1557</td>
+      <td>1.2</td>
+      <td>1702</td>
+    </tr>
+    <tr>
+      <th rowspan="3" valign="top">H100x1</th>
+      <th>mgoin/pixtral-12b</th>
+      <td></td>
+      <td>1.8</td>
+      <td>595</td>
+      <td>1.5</td>
+      <td>732</td>
+      <td>1.4</td>
+      <td>764</td>
+    </tr>
+    <tr>
+      <th>neuralmagic/pixtral-12b-FP8-Dynamic</th>
+      <td>1.35</td>
+      <td>1.4</td>
+      <td>767</td>
+      <td>1.1</td>
+      <td>1008</td>
+      <td>1.0</td>
+      <td>1056</td>
+    </tr>
+    <tr>
+      <th>neuralmagic/pixtral-12b-quantized.w4a16</th>
+      <td>1.37</td>
+      <td>1.4</td>
+      <td>787</td>
+      <td>1.1</td>
+      <td>1018</td>
+      <td>1.0</td>
+      <td>1065</td>
+    </tr>
+  </tbody>
+</table>
+ 
+
+
 ### Multi-stream asynchronous performance (measured with vLLM version 0.7.2)
+
+<table border="1" class="dataframe">
+  <thead>
+    <tr>
+      <th></th>
+      <th></th>
+      <th></th>
+      <th style="text-align: center;" colspan="2" >Document Visual Question Answering<br>1680W x 2240H<br>64/128</th>
+      <th style="text-align: center;" colspan="2" >Visual Reasoning <br>640W x 480H<br>128/128</th>
+      <th style="text-align: center;" colspan="2" >Image Captioning<br>480W x 360H<br>0/128</th>
+    </tr>
+    <tr>
+      <th>Hardware</th>
+      <th>Model</th>
+      <th>Average Cost Reduction</th>
+      <th>Maximum throughput (QPS)</th>
+      <th>QPD</th>
+      <th>Maximum throughput (QPS)</th>
+      <th>QPD</th>
+      <th>Maximum throughput (QPS)</th>
+      <th>QPD</th>
+    </tr>
+  </thead>
+  <tbody style="text-align: center">
+    <tr>
+      <th rowspan="3" valign="top">A6000x1</th>
+      <th>mgoin/pixtral-12b</th>
+      <td></td>
+      <td>0.6</td>
+      <td>2632</td>
+      <td>0.9</td>
+      <td>4108</td>
+      <td>1.1</td>
+      <td>4774</td>
+    </tr>
+    <tr>
+      <th>neuralmagic/pixtral-12b-quantized.w8a8</th>
+      <td>1.50</td>
+      <td>0.9</td>
+      <td>3901</td>
+      <td>1.4</td>
+      <td>6160</td>
+      <td>1.6</td>
+      <td>7292</td>
+    </tr>
+    <tr>
+      <th>neuralmagic/pixtral-12b-quantized.w4a16</th>
+      <td>1.41</td>
+      <td>0.6</td>
+      <td>2890</td>
+      <td>1.3</td>
+      <td>5758</td>
+      <td>1.8</td>
+      <td>8312</td>
+    </tr>
+    <tr>
+      <th rowspan="3" valign="top">A100x1</th>
+      <th>mgoin/pixtral-12b</th>
+      <td></td>
+      <td>1.1</td>
+      <td>2291</td>
+      <td>1.8</td>
+      <td>3670</td>
+      <td>2.1</td>
+      <td>4284</td>
+    </tr>
+    <tr>
+      <th>neuralmagic/pixtral-12b-quantized.w8a8</th>
+      <td>1.38</td>
+      <td>1.5</td>
+      <td>3096</td>
+      <td>2.5</td>
+      <td>5076</td>
+      <td>3.0</td>
+      <td>5965</td>
+    </tr>
+    <tr>
+      <th>neuralmagic/pixtral-12b-quantized.w4a16</th>
+      <td>1.40</td>
+      <td>1.4</td>
+      <td>2728</td>
+      <td>2.6</td>
+      <td>5133</td>
+      <td>3.5</td>
+      <td>6943</td>
+    </tr>
+    <tr>
+      <th rowspan="3" valign="top">H100x1</th>
+      <th>BF16</th>
+      <td></td>
+      <td>2.6</td>
+      <td>2877</td>
+      <td>4.0</td>
+      <td>4372</td>
+      <td>4.7</td>
+      <td>5095</td>
+    </tr>
+    <tr>
+      <th>neuralmagic/pixtral-12b-FP8-Dynamic</th>
+      <td>1.33</td>
+      <td>3.4</td>
+      <td>3753</td>
+      <td>5.4</td>
+      <td>5862</td>
+      <td>6.3</td>
+      <td>6917</td>
+    </tr>
+    <tr>
+      <th>neuralmagic/pixtral-12b-quantized.w4a16</th>
+      <td>1.22</td>
+      <td>2.8</td>
+      <td>3115</td>
+      <td>5.0</td>
+      <td>5511</td>
+      <td>6.2</td>
+      <td>6777</td>
+    </tr>
+  </tbody>
+</table>