Commit
•
6d7d418
1
Parent(s):
d46cb97
Update README.md (#1)
Browse files- Update README.md (4b91549217e91b742af7eb4b576ad0ff6bf136ec)
Co-authored-by: Qiyang Zhao <qiyang-zhao@users.noreply.huggingface.co>
README.md
CHANGED
@@ -72,8 +72,61 @@ python run_inference.py -m models/Falcon3-10B-1.58bit/ggml-model-i2_s.gguf -p "Y
|
|
72 |
```
|
73 |
|
74 |
# Evaluation
|
75 |
-
|
76 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
77 |
|
78 |
## Useful links
|
79 |
- View our [release blogpost](https://huggingface.co/blog/falcon3).
|
|
|
72 |
```
|
73 |
|
74 |
# Evaluation
|
75 |
+
We report in the following table our internal pipeline benchmarks:
|
76 |
+
|
77 |
+
**Note evaluation results are normalized score from v2 leaderboard tasks - reported results of original models in the blogpost are raw scores**
|
78 |
+
|
79 |
+
<table border="1" style="width: 100%; text-align: center; border-collapse: collapse;">
|
80 |
+
<colgroup>
|
81 |
+
<col style="width: 10%;">
|
82 |
+
<col style="width: 10%;">
|
83 |
+
<col style="background-color: rgba(80, 15, 213, 0.5); width: 7%;">
|
84 |
+
</colgroup>
|
85 |
+
<thead>
|
86 |
+
<tr>
|
87 |
+
<th>Benchmark</th>
|
88 |
+
<th>Llama3-8B-1.58-100B-tokens</th>
|
89 |
+
<th>Falcon3-10B-Instruct-1.58bit</th>
|
90 |
+
</tr>
|
91 |
+
</thead>
|
92 |
+
<tbody>
|
93 |
+
<tr>
|
94 |
+
<td>IFEval</td>
|
95 |
+
<td>17.91</td>
|
96 |
+
<td><b>54.37</b></td>
|
97 |
+
</tr>
|
98 |
+
<tr>
|
99 |
+
<td>MUSR</td>
|
100 |
+
<td><b>4.87</b></td>
|
101 |
+
<td>2.57</td>
|
102 |
+
</tr>
|
103 |
+
<tr>
|
104 |
+
<td>GPQA</td>
|
105 |
+
<td>1.83</td>
|
106 |
+
<td><b>4.27</b></td>
|
107 |
+
</tr>
|
108 |
+
<tr>
|
109 |
+
<td>BBH</td>
|
110 |
+
<td>5.36</td>
|
111 |
+
<td><b>6.59</b></td>
|
112 |
+
</tr>
|
113 |
+
<tr>
|
114 |
+
<td>MMLU-PRO</td>
|
115 |
+
<td>2.78</td>
|
116 |
+
<td><b>6.62</b></td>
|
117 |
+
</tr>
|
118 |
+
<tr>
|
119 |
+
<td>MATH</td>
|
120 |
+
<td>0.26</td>
|
121 |
+
<td><b>2.44</b></td>
|
122 |
+
</tr>
|
123 |
+
<tr>
|
124 |
+
<td>Average</td>
|
125 |
+
<td>5.5</td>
|
126 |
+
<td><b>12.81</b></td>
|
127 |
+
</tr>
|
128 |
+
</tbody>
|
129 |
+
</table>
|
130 |
|
131 |
## Useful links
|
132 |
- View our [release blogpost](https://huggingface.co/blog/falcon3).
|