Update README.md
Browse files
README.md
CHANGED
@@ -277,6 +277,37 @@ The following performance benchmarks were conducted with [vLLM](https://docs.vll
|
|
277 |
</tr>
|
278 |
</thead>
|
279 |
<tbody style="text-align: center">
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
280 |
<tr>
|
281 |
<th rowspan="3" valign="top">A100x1</th>
|
282 |
<th>Qwen/Qwen2.5-VL-7B-Instruct</th>
|
@@ -371,9 +402,40 @@ The following performance benchmarks were conducted with [vLLM](https://docs.vll
|
|
371 |
</tr>
|
372 |
</thead>
|
373 |
<tbody style="text-align: center">
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
374 |
<tr>
|
375 |
<th rowspan="3" valign="top">A100x1</th>
|
376 |
-
<th>Qwen/Qwen2.5-VL-7B-Instruct
|
377 |
<td></td>
|
378 |
<td>0.7</td>
|
379 |
<td>1347</td>
|
|
|
277 |
</tr>
|
278 |
</thead>
|
279 |
<tbody style="text-align: center">
|
280 |
+
<tr>
|
281 |
+
<th rowspan="3" valign="top">A6000x1</th>
|
282 |
+
<th>Qwen/Qwen2.5-VL-7B-Instruct</th>
|
283 |
+
<td></td>
|
284 |
+
<td>4.9</td>
|
285 |
+
<td>912</td>
|
286 |
+
<td>3.2</td>
|
287 |
+
<td>1386</td>
|
288 |
+
<td>3.1</td>
|
289 |
+
<td>1431</td>
|
290 |
+
</tr>
|
291 |
+
<tr>
|
292 |
+
<th>neuralmagic/Qwen2.5-VL-7B-Instruct-quantized.w8a8</th>
|
293 |
+
<td>1.50</td>
|
294 |
+
<td>3.6</td>
|
295 |
+
<td>1248</td>
|
296 |
+
<td>2.1</td>
|
297 |
+
<td>2163</td>
|
298 |
+
<td>2.0</td>
|
299 |
+
<td>2237</td>
|
300 |
+
</tr>
|
301 |
+
<tr>
|
302 |
+
<th>neuralmagic/Qwen2.5-VL-7B-Instruct-quantized.w4a16</th>
|
303 |
+
<td>2.05</td>
|
304 |
+
<td>3.3</td>
|
305 |
+
<td>1351</td>
|
306 |
+
<td>1.4</td>
|
307 |
+
<td>3252</td>
|
308 |
+
<td>1.4</td>
|
309 |
+
<td>3321</td>
|
310 |
+
</tr>
|
311 |
<tr>
|
312 |
<th rowspan="3" valign="top">A100x1</th>
|
313 |
<th>Qwen/Qwen2.5-VL-7B-Instruct</th>
|
|
|
402 |
</tr>
|
403 |
</thead>
|
404 |
<tbody style="text-align: center">
|
405 |
+
<tr>
|
406 |
+
<th rowspan="3" valign="top">A6000x1</th>
|
407 |
+
<th>Qwen/Qwen2.5-VL-7B-Instruct</th>
|
408 |
+
<td></td>
|
409 |
+
<td>0.4</td>
|
410 |
+
<td>1837</td>
|
411 |
+
<td>1.5</td>
|
412 |
+
<td>6846</td>
|
413 |
+
<td>1.7</td>
|
414 |
+
<td>7638</td>
|
415 |
+
</tr>
|
416 |
+
<tr>
|
417 |
+
<th>neuralmagic/Qwen2.5-VL-7B-Instruct-quantized.w8a8</th>
|
418 |
+
<td>1.41</td>
|
419 |
+
<td>0.5</td>
|
420 |
+
<td>2297</td>
|
421 |
+
<td>2.3</td>
|
422 |
+
<td>10137</td>
|
423 |
+
<td>2.5</td>
|
424 |
+
<td>11472</td>
|
425 |
+
</tr>
|
426 |
+
<tr>
|
427 |
+
<th>neuralmagic/Qwen2.5-VL-7B-Instruct-quantized.w4a16</th>
|
428 |
+
<td>1.60</td>
|
429 |
+
<td>0.4</td>
|
430 |
+
<td>1828</td>
|
431 |
+
<td>2.7</td>
|
432 |
+
<td>12254</td>
|
433 |
+
<td>3.4</td>
|
434 |
+
<td>15477</td>
|
435 |
+
</tr>
|
436 |
<tr>
|
437 |
<th rowspan="3" valign="top">A100x1</th>
|
438 |
+
<th>Qwen/Qwen2.5-VL-7B-Instruct</th>
|
439 |
<td></td>
|
440 |
<td>0.7</td>
|
441 |
<td>1347</td>
|