Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
|
@@ -27,46 +27,68 @@ markdown_text = """
|
|
| 27 |
- Input one budget line per time.
|
| 28 |
- Accuracy of the model is ~88%.
|
| 29 |
"""
|
| 30 |
-
# HTML formatted table
|
| 31 |
html_table = """
|
| 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 |
iface = gr.Interface(
|
|
|
|
| 27 |
- Input one budget line per time.
|
| 28 |
- Accuracy of the model is ~88%.
|
| 29 |
"""
|
|
|
|
| 30 |
html_table = """
|
| 31 |
+
<h2 style="text-align: center;">COFOG Budget Classification</h2>
|
| 32 |
+
<p style='text-align: justify'>
|
| 33 |
+
This classifier was developed utilizing the pre-trained BERT
|
| 34 |
+
(Bidirectional Encoder Representations from Transformers) model
|
| 35 |
+
with an uncased configuration, with over 1500 manually
|
| 36 |
+
labeled dataset comprising budget line items extracted from
|
| 37 |
+
various budgetary documents. To balance the data, additional data
|
| 38 |
+
was generated using GPT-4 where categories were not available
|
| 39 |
+
in budget documents. The model training was executed
|
| 40 |
+
on a Google Colab environment, specifically utilizing a Tesla T4 GPU.
|
| 41 |
+
Detailed metrics of the training process are as follows:
|
| 42 |
+
<code>TrainOutput(global_step=395, training_loss=1.1497593360611156,
|
| 43 |
+
metrics={'train_runtime': 650.0119, 'train_samples_per_second':
|
| 44 |
+
9.638, 'train_steps_per_second': 0.608, 'total_flos': 1648509163714560.0,
|
| 45 |
+
'train_loss': 1.1497593360611156, 'epoch': 5.0})</code>. The model
|
| 46 |
+
is designed to predict the primary classification level
|
| 47 |
+
of the Classification of the Functions of Government (COFOG),
|
| 48 |
+
with the predictions from the first level serving as contextual
|
| 49 |
+
input for subsequent second-level classification. The project
|
| 50 |
+
is conducted with an exclusive focus on academic and research
|
| 51 |
+
objectives.
|
| 52 |
+
</p>
|
| 53 |
+
<table style="margin-left: auto; margin-right: auto;">
|
| 54 |
+
<tr>
|
| 55 |
+
<th>Epoch</th>
|
| 56 |
+
<th>Training Loss</th>
|
| 57 |
+
<th>Validation Loss</th>
|
| 58 |
+
<th>Accuracy</th>
|
| 59 |
+
</tr>
|
| 60 |
+
<tr>
|
| 61 |
+
<td>1</td>
|
| 62 |
+
<td>No log</td>
|
| 63 |
+
<td>2.095209</td>
|
| 64 |
+
<td>0.340764</td>
|
| 65 |
+
</tr>
|
| 66 |
+
<tr>
|
| 67 |
+
<td>2</td>
|
| 68 |
+
<td>No log</td>
|
| 69 |
+
<td>1.419945</td>
|
| 70 |
+
<td>0.662420</td>
|
| 71 |
+
</tr>
|
| 72 |
+
<tr>
|
| 73 |
+
<td>3</td>
|
| 74 |
+
<td>No log</td>
|
| 75 |
+
<td>0.683810</td>
|
| 76 |
+
<td>0.850318</td>
|
| 77 |
+
</tr>
|
| 78 |
+
<tr>
|
| 79 |
+
<td>4</td>
|
| 80 |
+
<td>No log</td>
|
| 81 |
+
<td>0.460408</td>
|
| 82 |
+
<td>0.872611</td>
|
| 83 |
+
</tr>
|
| 84 |
+
<tr>
|
| 85 |
+
<td>5</td>
|
| 86 |
+
<td>No log</td>
|
| 87 |
+
<td>0.422096</td>
|
| 88 |
+
<td>0.888535</td>
|
| 89 |
+
</tr>
|
| 90 |
+
</table>
|
| 91 |
+
</div>
|
| 92 |
"""
|
| 93 |
|
| 94 |
iface = gr.Interface(
|