Commit
·
b39d549
1
Parent(s):
4daeb23
update
Browse files
README.md
CHANGED
|
@@ -118,29 +118,6 @@ python examples/mmlu_llmapi.py --data_dir data/mmlu --hf_model_dir nvidia/DeepSe
|
|
| 118 |
|
| 119 |
Please refer to the [TensorRT-LLM benchmarking documentation](https://github.com/NVIDIA/TensorRT-LLM/tree/main/benchmarks) for details.
|
| 120 |
|
| 121 |
-
#### Evaluation
|
| 122 |
-
The accuracy (MMLU, 5-shot) benchmark results are presented in the table below:
|
| 123 |
-
<table>
|
| 124 |
-
<tr>
|
| 125 |
-
<td><strong>Precision</strong>
|
| 126 |
-
</td>
|
| 127 |
-
<td><strong>MMLU</strong>
|
| 128 |
-
</td>
|
| 129 |
-
</tr>
|
| 130 |
-
<tr>
|
| 131 |
-
<td>FP8
|
| 132 |
-
</td>
|
| 133 |
-
<td>90.8
|
| 134 |
-
</td>
|
| 135 |
-
</tr>
|
| 136 |
-
<tr>
|
| 137 |
-
<td>FP4
|
| 138 |
-
</td>
|
| 139 |
-
<td>86.9
|
| 140 |
-
</td>
|
| 141 |
-
</tr>
|
| 142 |
-
<tr>
|
| 143 |
-
</table>
|
| 144 |
|
| 145 |
|
| 146 |
## Ethical Considerations
|
|
|
|
| 118 |
|
| 119 |
Please refer to the [TensorRT-LLM benchmarking documentation](https://github.com/NVIDIA/TensorRT-LLM/tree/main/benchmarks) for details.
|
| 120 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 121 |
|
| 122 |
|
| 123 |
## Ethical Considerations
|