Update README.md
Browse files
README.md
CHANGED
|
@@ -23,8 +23,6 @@ This was ultimately used to obtain character embeddings for words and used for a
|
|
| 23 |
|
| 24 |
For full details of this anomaly detection task, see [these notebooks](https://github.com/datasig-ac-uk/signature_applications/tree/master/anomaly_detection_language_dataset).
|
| 25 |
|
| 26 |
-
- Loss: 0.0958
|
| 27 |
-
|
| 28 |
## Model description
|
| 29 |
|
| 30 |
Based on the [RoBERTa](https://huggingface.co/docs/transformers/en/model_doc/roberta).
|
|
@@ -37,6 +35,8 @@ This was created for research purposes to investigate the use of embeddings obta
|
|
| 37 |
|
| 38 |
See the [dataset card](https://huggingface.co/datasets/rchan26/english_char_split) for the exact train/validation/test splits of English words.
|
| 39 |
|
|
|
|
|
|
|
| 40 |
## Training procedure
|
| 41 |
|
| 42 |
For full details of training, also see [this notebook](https://github.com/datasig-ac-uk/signature_applications/blob/master/anomaly_detection_language_dataset/train_english_char_bert.ipynb).
|
|
|
|
| 23 |
|
| 24 |
For full details of this anomaly detection task, see [these notebooks](https://github.com/datasig-ac-uk/signature_applications/tree/master/anomaly_detection_language_dataset).
|
| 25 |
|
|
|
|
|
|
|
| 26 |
## Model description
|
| 27 |
|
| 28 |
Based on the [RoBERTa](https://huggingface.co/docs/transformers/en/model_doc/roberta).
|
|
|
|
| 35 |
|
| 36 |
See the [dataset card](https://huggingface.co/datasets/rchan26/english_char_split) for the exact train/validation/test splits of English words.
|
| 37 |
|
| 38 |
+
In our training rum, we obtained a final loss of 0.0958.
|
| 39 |
+
|
| 40 |
## Training procedure
|
| 41 |
|
| 42 |
For full details of training, also see [this notebook](https://github.com/datasig-ac-uk/signature_applications/blob/master/anomaly_detection_language_dataset/train_english_char_bert.ipynb).
|