Updated evals after inference was fixed.
Browse files
README.md
CHANGED
@@ -128,11 +128,29 @@ Current evals out of the Metharme-13b model: <br>
|
|
128 |
<td>27.53407859802246</td>
|
129 |
<td>7.038073539733887</td>
|
130 |
</tr>
|
131 |
-
|
132 |
-
<td>Metharme 13b - 8bit - [
|
133 |
-
<td>
|
134 |
-
<td>
|
135 |
-
<td>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
136 |
</tr>
|
137 |
</tbody>
|
138 |
</table>
|
|
|
128 |
<td>27.53407859802246</td>
|
129 |
<td>7.038073539733887</td>
|
130 |
</tr>
|
131 |
+
<tr>
|
132 |
+
<td>Metharme 13b - 8bit - [true-sequential & 128g]</td>
|
133 |
+
<td>5.2532830238342285</td>
|
134 |
+
<td>27.54250144958496</td>
|
135 |
+
<td>7.038838863372803</td>
|
136 |
+
</tr>
|
137 |
+
<tr>
|
138 |
+
<td>Metharme 13b - 4bit - [true-sequential & 128g]</td>
|
139 |
+
<td>5.420501708984375</td>
|
140 |
+
<td>28.37093734741211</td>
|
141 |
+
<td>7.1930413246154785</td>
|
142 |
+
</tr>
|
143 |
+
<tr>
|
144 |
+
<td>Metharme 7b - 16bit</td>
|
145 |
+
<td>5.7208476066589355</td>
|
146 |
+
<td>41.61103439331055</td>
|
147 |
+
<td>7.512405872344971</td>
|
148 |
+
</tr>
|
149 |
+
<tr>
|
150 |
+
<td>Metharme 7b - 4bit - [act-order]</td>
|
151 |
+
<td>6.2369050979614</td>
|
152 |
+
<td>47.5177230834960</td>
|
153 |
+
<td>7.9044938087463</td>
|
154 |
</tr>
|
155 |
</tbody>
|
156 |
</table>
|