File size: 2,917 Bytes
a8e765d a8367f4 a8e765d a8367f4 a8e765d a8367f4 a8e765d a137c08 3dba654 a137c08 3dba654 a137c08 3dba654 a137c08 3dba654 a137c08 3dba654 a137c08 3dba654 a137c08 a8e765d 63bf8c3 a00d870 3dba654 346fe28 a8e765d | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 | ---
license: apache-2.0
pipeline_tag: sentence-similarity
language:
- fr
tags:
- embeddings
- french
- feature-extraction
- bfloat16
- sentence-similarity
- text-embeddings
---
<table>
<thead>
<tr>
<th>Evaluation task</th>
<th align="center">Embeddings-Francais-BF16-BASE-50M</th>
<th align="center">Test-Train-Avant-Main-Train</th>
</tr>
<tr>
<td>SICKFr</td>
<td align="center">0.519713</td>
<td align="center"><strong>0.699325</strong></td>
</tr>
<tr>
<td>SyntecReranking</td>
<td align="center">0.313680</td>
<td align="center"><strong>0.328360</strong></td>
</tr>
<tr>
<td>SummEvalFr</td>
<td align="center"><strong>0.306903</strong></td>
<td align="center">0.305028</td>
</tr>
<tr>
<td>AlloProfClusteringS2S</td>
<td align="center"><strong>0.213383</strong></td>
<td align="center">0.209503</td>
</tr>
<tr>
<td>SyntecRetrieval</td>
<td align="center">0.051370</td>
<td align="center"><strong>0.123900</strong></td>
</tr>
<tr>
<td>HALClusteringS2S</td>
<td align="center">Failed</td>
<td align="center"><strong>0.042094</strong></td>
</tr>
<tr>
<th>Hyperparameter</th>
<th align="center">Embeddings-Francais-BF16-BASE-50M</th>
<th align="center">Test-Train-Avant-Main-Train</th>
</tr>
</thead>
<tbody>
<tr>
<td>Training tokens seen</td>
<td align="center">2.46B</td>
<td align="center">61.44M + SFT </td>
</tr>
<tr>
<td>Parameters</td>
<td align="center">169,896,960</td>
<td align="center">21,240,576</td>
</tr>
<tr>
<td>Context length</td>
<td align="center">4096</td>
<td align="center">4096</td>
</tr>
<tr>
<td>Embedding dimension</td>
<td align="center">1536</td>
<td align="center">384</td>
</tr>
<tr>
<td>Vocabulary size</td>
<td align="center">32768</td>
<td align="center">32768</td>
</tr>
<tr>
<td>Layers</td>
<td align="center">4</td>
<td align="center">4</td>
</tr>
<tr>
<td>Heads</td>
<td align="center">12</td>
<td align="center">4</td>
</tr>
<tr>
<td>Head dimension</td>
<td align="center">128</td>
<td align="center">96</td>
</tr>
<tr>
<td>Precision</td>
<td align="center">bfloat16</td>
<td align="center">bfloat16</td>
</tr>
<tr>
<td>Attention backend</td>
<td align="center">SageAttention</td>
<td align="center">SageAttention</td>
</tr>
<tr>
<td>Pooling</td>
<td align="center">Mean pooling</td>
<td align="center">Mean pooling</td>
</tr>
<tr>
<td>Normalization</td>
<td align="center">L2 normalize</td>
<td align="center">L2 normalize</td>
</tr>
</tbody>
</table> |