Update README.md
Browse files
README.md
CHANGED
|
@@ -1,3 +1,106 @@
|
|
| 1 |
-
---
|
| 2 |
-
license: apache-2.0
|
| 3 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
base_model:
|
| 4 |
+
- meta-llama/Llama-3.1-8B-Instruct
|
| 5 |
+
---
|
| 6 |
+
|
| 7 |
+
|
| 8 |
+
|
| 9 |
+
### Accuracy
|
| 10 |
+
<table>
|
| 11 |
+
<thead>
|
| 12 |
+
<tr>
|
| 13 |
+
<th>Task</th>
|
| 14 |
+
<th>Context Length</th>
|
| 15 |
+
<th>meta-llama/<br>Llama-3.1-8B-Instruct</th>
|
| 16 |
+
<th>Llama-3.1-8B-Instruct-<br>FP8-dynamic-<br>QKV-Cache-FP8-<br>Per-Head</th>
|
| 17 |
+
<th>Llama-3.1-8B-Instruct-<br>FP8-dynamic-<br>QKV-Cache-FP8-<br>Per-Tensor</th>
|
| 18 |
+
<th>Llama-3.1-8B-Instruct-<br>QKV-Cache-FP8-<br>Per-Head</th>
|
| 19 |
+
<th>Llama-3.1-8B-Instruct-<br>QKV-Cache-FP8-<br>Per-Tensor</th>
|
| 20 |
+
</tr>
|
| 21 |
+
</thead>
|
| 22 |
+
<tbody>
|
| 23 |
+
<tr>
|
| 24 |
+
<td rowspan="5"><b>NIAH<br>Single 2</b></td>
|
| 25 |
+
<td>4096</td>
|
| 26 |
+
<td>100.00</td>
|
| 27 |
+
<td>100.00</td>
|
| 28 |
+
<td>100.00</td>
|
| 29 |
+
<td>100.00</td>
|
| 30 |
+
<td>100.00</td>
|
| 31 |
+
</tr>
|
| 32 |
+
<tr>
|
| 33 |
+
<td>16384</td>
|
| 34 |
+
<td>100.00</td>
|
| 35 |
+
<td>100.00</td>
|
| 36 |
+
<td>100.00</td>
|
| 37 |
+
<td>100.00</td>
|
| 38 |
+
<td>100.00</td>
|
| 39 |
+
</tr>
|
| 40 |
+
<tr>
|
| 41 |
+
<td>32768</td>
|
| 42 |
+
<td>100.00</td>
|
| 43 |
+
<td>100.00</td>
|
| 44 |
+
<td>100.00</td>
|
| 45 |
+
<td>100.00</td>
|
| 46 |
+
<td>100.00</td>
|
| 47 |
+
</tr>
|
| 48 |
+
<tr>
|
| 49 |
+
<td>65536</td>
|
| 50 |
+
<td>100.00</td>
|
| 51 |
+
<td>100.00</td>
|
| 52 |
+
<td>100.00</td>
|
| 53 |
+
<td>100.00</td>
|
| 54 |
+
<td>100.00</td>
|
| 55 |
+
</tr>
|
| 56 |
+
<tr>
|
| 57 |
+
<td>131072</td>
|
| 58 |
+
<td>100.00</td>
|
| 59 |
+
<td>100.00</td>
|
| 60 |
+
<td>100.00</td>
|
| 61 |
+
<td>100.00</td>
|
| 62 |
+
<td>100.00</td>
|
| 63 |
+
</tr>
|
| 64 |
+
<tr>
|
| 65 |
+
<td rowspan="5"><b>NIAH<br>Single 3</b></td>
|
| 66 |
+
<td>4096</td>
|
| 67 |
+
<td>100.00</td>
|
| 68 |
+
<td>100.00</td>
|
| 69 |
+
<td>100.00</td>
|
| 70 |
+
<td>100.00</td>
|
| 71 |
+
<td>100.00</td>
|
| 72 |
+
</tr>
|
| 73 |
+
<tr>
|
| 74 |
+
<td>16384</td>
|
| 75 |
+
<td>100.00</td>
|
| 76 |
+
<td>100.00</td>
|
| 77 |
+
<td>100.00</td>
|
| 78 |
+
<td>100.00</td>
|
| 79 |
+
<td>100.00</td>
|
| 80 |
+
</tr>
|
| 81 |
+
<tr>
|
| 82 |
+
<td>32768</td>
|
| 83 |
+
<td>100.00</td>
|
| 84 |
+
<td>100.00</td>
|
| 85 |
+
<td>100.00</td>
|
| 86 |
+
<td>100.00</td>
|
| 87 |
+
<td>100.00</td>
|
| 88 |
+
</tr>
|
| 89 |
+
<tr>
|
| 90 |
+
<td>65536</td>
|
| 91 |
+
<td>100.00</td>
|
| 92 |
+
<td>100.00</td>
|
| 93 |
+
<td>100.00</td>
|
| 94 |
+
<td>100.00</td>
|
| 95 |
+
<td>100.00</td>
|
| 96 |
+
</tr>
|
| 97 |
+
<tr>
|
| 98 |
+
<td>131072</td>
|
| 99 |
+
<td>100.00</td>
|
| 100 |
+
<td>100.00</td>
|
| 101 |
+
<td>100.00</td>
|
| 102 |
+
<td>100.00</td>
|
| 103 |
+
<td>100.00</td>
|
| 104 |
+
</tr>
|
| 105 |
+
</tbody>
|
| 106 |
+
</table>
|