fix: use correct rounding (#5)
Browse files- fix: use correct rounding (b74b487796ba80c7f5e9c5ddfd6786eeab271a20)
Co-authored-by: William de Vazelhes <wdevazelhes@users.noreply.huggingface.co>
README.md
CHANGED
|
@@ -159,40 +159,40 @@ Falcon3-7B is trained on 256 H100 nodes (world size 2048).
|
|
| 159 |
<td>MMLU (5-shot)</td>
|
| 160 |
<td>31.1</td>
|
| 161 |
<td>61.0</td>
|
| 162 |
-
<td>50.
|
| 163 |
-
<td>53.
|
| 164 |
<td>42.5</td>
|
| 165 |
</tr>
|
| 166 |
<tr>
|
| 167 |
<td>MMLU-PRO (5-shot)</td>
|
| 168 |
<td>11.7</td>
|
| 169 |
-
<td>28.
|
| 170 |
-
<td>21.
|
| 171 |
<td>22.1</td>
|
| 172 |
-
<td>16.
|
| 173 |
</tr>
|
| 174 |
<tr>
|
| 175 |
<td>IFEval</td>
|
| 176 |
-
<td>14.
|
| 177 |
-
<td>26.
|
| 178 |
<td>24.2</td>
|
| 179 |
-
<td>20.
|
| 180 |
-
<td>25.
|
| 181 |
</tr>
|
| 182 |
<tr>
|
| 183 |
<td rowspan="2">Math</td>
|
| 184 |
<td>GSM8K (5-shot)</td>
|
| 185 |
<td>6.6</td>
|
| 186 |
-
<td>62.
|
| 187 |
-
<td>31.
|
| 188 |
-
<td>25.
|
| 189 |
<td>34.3</td>
|
| 190 |
</tr>
|
| 191 |
<tr>
|
| 192 |
<td>MATH Lvl-5 (4-shot)</td>
|
| 193 |
-
<td>0.
|
| 194 |
-
<td>6.
|
| 195 |
-
<td>1.
|
| 196 |
<td>2.6</td>
|
| 197 |
<td>2.2</td>
|
| 198 |
</tr>
|
|
@@ -203,12 +203,12 @@ Falcon3-7B is trained on 256 H100 nodes (world size 2048).
|
|
| 203 |
<td>54.8</td>
|
| 204 |
<td>54.1</td>
|
| 205 |
<td>53.7</td>
|
| 206 |
-
<td>48.
|
| 207 |
</tr>
|
| 208 |
<tr>
|
| 209 |
<td>GPQA (0-shot)</td>
|
| 210 |
-
<td>24.
|
| 211 |
-
<td>28.
|
| 212 |
<td>28.9</td>
|
| 213 |
<td>25.5</td>
|
| 214 |
<td>28.1</td>
|
|
@@ -217,22 +217,22 @@ Falcon3-7B is trained on 256 H100 nodes (world size 2048).
|
|
| 217 |
<td>MUSR (0-shot)</td>
|
| 218 |
<td>34.5</td>
|
| 219 |
<td>35.5</td>
|
| 220 |
-
<td>34.
|
| 221 |
-
<td>42.
|
| 222 |
<td>41.9</td>
|
| 223 |
</tr>
|
| 224 |
<tr>
|
| 225 |
<td>BBH (3-shot)</td>
|
| 226 |
<td>31.2</td>
|
| 227 |
<td>41.1</td>
|
| 228 |
-
<td>34.
|
| 229 |
<td>36.8</td>
|
| 230 |
-
<td>36.
|
| 231 |
</tr>
|
| 232 |
<tr>
|
| 233 |
<td rowspan="4">CommonSense Understanding</td>
|
| 234 |
<td>PIQA (0-shot)</td>
|
| 235 |
-
<td>74.
|
| 236 |
<td>76.0</td>
|
| 237 |
<td>77.5</td>
|
| 238 |
<td>79.2</td>
|
|
|
|
| 159 |
<td>MMLU (5-shot)</td>
|
| 160 |
<td>31.1</td>
|
| 161 |
<td>61.0</td>
|
| 162 |
+
<td>50.1</td>
|
| 163 |
+
<td>53.0</td>
|
| 164 |
<td>42.5</td>
|
| 165 |
</tr>
|
| 166 |
<tr>
|
| 167 |
<td>MMLU-PRO (5-shot)</td>
|
| 168 |
<td>11.7</td>
|
| 169 |
+
<td>28.4</td>
|
| 170 |
+
<td>21.3</td>
|
| 171 |
<td>22.1</td>
|
| 172 |
+
<td>16.1</td>
|
| 173 |
</tr>
|
| 174 |
<tr>
|
| 175 |
<td>IFEval</td>
|
| 176 |
+
<td>14.8</td>
|
| 177 |
+
<td>26.0</td>
|
| 178 |
<td>24.2</td>
|
| 179 |
+
<td>20.3</td>
|
| 180 |
+
<td>25.2</td>
|
| 181 |
</tr>
|
| 182 |
<tr>
|
| 183 |
<td rowspan="2">Math</td>
|
| 184 |
<td>GSM8K (5-shot)</td>
|
| 185 |
<td>6.6</td>
|
| 186 |
+
<td>62.2</td>
|
| 187 |
+
<td>31.0</td>
|
| 188 |
+
<td>25.5</td>
|
| 189 |
<td>34.3</td>
|
| 190 |
</tr>
|
| 191 |
<tr>
|
| 192 |
<td>MATH Lvl-5 (4-shot)</td>
|
| 193 |
+
<td>0.2</td>
|
| 194 |
+
<td>6.7</td>
|
| 195 |
+
<td>1.4</td>
|
| 196 |
<td>2.6</td>
|
| 197 |
<td>2.2</td>
|
| 198 |
</tr>
|
|
|
|
| 203 |
<td>54.8</td>
|
| 204 |
<td>54.1</td>
|
| 205 |
<td>53.7</td>
|
| 206 |
+
<td>48.1</td>
|
| 207 |
</tr>
|
| 208 |
<tr>
|
| 209 |
<td>GPQA (0-shot)</td>
|
| 210 |
+
<td>24.2</td>
|
| 211 |
+
<td>28.1</td>
|
| 212 |
<td>28.9</td>
|
| 213 |
<td>25.5</td>
|
| 214 |
<td>28.1</td>
|
|
|
|
| 217 |
<td>MUSR (0-shot)</td>
|
| 218 |
<td>34.5</td>
|
| 219 |
<td>35.5</td>
|
| 220 |
+
<td>34.7</td>
|
| 221 |
+
<td>42.7</td>
|
| 222 |
<td>41.9</td>
|
| 223 |
</tr>
|
| 224 |
<tr>
|
| 225 |
<td>BBH (3-shot)</td>
|
| 226 |
<td>31.2</td>
|
| 227 |
<td>41.1</td>
|
| 228 |
+
<td>34.2</td>
|
| 229 |
<td>36.8</td>
|
| 230 |
+
<td>36.0</td>
|
| 231 |
</tr>
|
| 232 |
<tr>
|
| 233 |
<td rowspan="4">CommonSense Understanding</td>
|
| 234 |
<td>PIQA (0-shot)</td>
|
| 235 |
+
<td>74.5</td>
|
| 236 |
<td>76.0</td>
|
| 237 |
<td>77.5</td>
|
| 238 |
<td>79.2</td>
|