Update README.md
Browse files
README.md
CHANGED
@@ -119,7 +119,7 @@ It can be used for [Extractive Q&A](https://huggingface.co/tasks/question-answer
|
|
119 |
|
120 |
[Evaluation notebook](https://github.com/anakin87/electra-italian-xxl-cased-squad-it/blob/main/evaluation.ipynb)
|
121 |
|
122 |
-
## Usage in Transformers
|
123 |
Model checkpoints are available for usage in Pytorch. They can be used directly with pipelines as:
|
124 |
|
125 |
```python
|
@@ -131,7 +131,33 @@ qa(question="Qual è il soprannome di Vasco Rossi?", context="Vasco Rossi, noto
|
|
131 |
```
|
132 |
## Usage in Haystack 🚀🚀🚀
|
133 |
|
134 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
135 |
|
136 |
## Comparison ⚖️
|
137 |
|
@@ -159,4 +185,7 @@ TO DO: link notebook
|
|
159 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
160 |
- lr_scheduler_type: linear
|
161 |
- num_epochs: 2
|
162 |
-
- mixed_precision_training: Native AMP
|
|
|
|
|
|
|
|
119 |
|
120 |
[Evaluation notebook](https://github.com/anakin87/electra-italian-xxl-cased-squad-it/blob/main/evaluation.ipynb)
|
121 |
|
122 |
+
## Usage in Transformers 🤗
|
123 |
Model checkpoints are available for usage in Pytorch. They can be used directly with pipelines as:
|
124 |
|
125 |
```python
|
|
|
131 |
```
|
132 |
## Usage in Haystack 🚀🚀🚀
|
133 |
|
134 |
+
With the [Haystack NLP framework](https://github.com/deepset-ai/haystack), you can use this model and create a scalable Question Answering system that works across millions of documents.
|
135 |
+
|
136 |
+
For a complete walkthrough, see [this notebook](https://github.com/anakin87/electra-italian-xxl-cased-squad-it/blob/main/usage_in_haystack.ipynb).
|
137 |
+
|
138 |
+
```python
|
139 |
+
...
|
140 |
+
print_answers(prediction, details="medium")
|
141 |
+
|
142 |
+
>>> Query: Con chi ha parlato di vaccini il premier Mario Draghi?
|
143 |
+
Answers:
|
144 |
+
[ { 'answer': 'Von der Leyen',
|
145 |
+
'context': " vaccino dell'azienda britannica. Durante la telefonata "
|
146 |
+
'tra Draghi e Von der Leyen, la presidente della '
|
147 |
+
'Commissione Ue ha annunciato al presidente del',
|
148 |
+
'score': 0.9663902521133423},
|
149 |
+
{ 'answer': 'Ursula Von der Leyen',
|
150 |
+
'context': 'colloquio telefonico con la presidente della Commissione '
|
151 |
+
'europea Ursula Von der Leyen. Secondo fonti di Palazzo '
|
152 |
+
'Chigi, dalla conversazione è emerso ch',
|
153 |
+
'score': 0.9063920974731445},
|
154 |
+
{ 'answer': 'Mario Draghi, ha tenuto un lungo discorso alla 76esima '
|
155 |
+
'Assemblea Generale delle Nazioni Unite',
|
156 |
+
'context': 'Il presidente del Consiglio, Mario Draghi, ha tenuto un '
|
157 |
+
'lungo discorso alla 76esima Assemblea Generale delle '
|
158 |
+
'Nazioni Unite, nella notte italiana. Tant',
|
159 |
+
'score': 0.5243796706199646}]
|
160 |
+
```
|
161 |
|
162 |
## Comparison ⚖️
|
163 |
|
|
|
185 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
186 |
- lr_scheduler_type: linear
|
187 |
- num_epochs: 2
|
188 |
+
- mixed_precision_training: Native AMP
|
189 |
+
|
190 |
+
> Created by [Stefano Fiorucci/anakin87](https://github.com/anakin87)
|
191 |
+
> Made with <span style="color: #e25555;">♥</span> in Italy
|