Commit
路
d181bd7
1
Parent(s):
fe1a154
Update README.md
Browse files
README.md
CHANGED
|
@@ -15,7 +15,6 @@ This model is fine-tuned on PQuAD Train set and is easily ready to use. Too long
|
|
| 15 |
|
| 16 |
## Hyperparameters
|
| 17 |
The reason I set batch_size to 4 is limited GPU memory in Google Colab.
|
| 18 |
-
Pre-processing, training and evaluating model took me about 4 hours.
|
| 19 |
```
|
| 20 |
batch_size = 4
|
| 21 |
n_epochs = 1
|
|
@@ -35,14 +34,16 @@ I trained for more than 1 epoch as well, but I get worse results.
|
|
| 35 |
|
| 36 |
### Question Answering On Test Set of PQuAD Dataset
|
| 37 |
|
| 38 |
-
| Metric | Our XLM-Roberta | Our Pars Bert |
|
| 39 |
-
|
| 40 |
-
| Exact Match | 66.56*
|
| 41 |
-
| F1 | 87.31*
|
| 42 |
|
| 43 |
|
| 44 |
|
| 45 |
## How to use
|
|
|
|
|
|
|
| 46 |
```python
|
| 47 |
from transformers import AutoTokenizer, AutoModelForQuestionAnswering
|
| 48 |
path = 'pedramyazdipoor/persian_xlm_roberta_large'
|
|
|
|
| 15 |
|
| 16 |
## Hyperparameters
|
| 17 |
The reason I set batch_size to 4 is limited GPU memory in Google Colab.
|
|
|
|
| 18 |
```
|
| 19 |
batch_size = 4
|
| 20 |
n_epochs = 1
|
|
|
|
| 34 |
|
| 35 |
### Question Answering On Test Set of PQuAD Dataset
|
| 36 |
|
| 37 |
+
| Metric | Our XLM-Roberta Large| Our Pars Bert |
|
| 38 |
+
|:----------------:|:--------------------:|:-------------:|
|
| 39 |
+
| Exact Match | 66.56* | 47.44 |
|
| 40 |
+
| F1 | 87.31* | 81.96 |
|
| 41 |
|
| 42 |
|
| 43 |
|
| 44 |
## How to use
|
| 45 |
+
|
| 46 |
+
## Pytorch
|
| 47 |
```python
|
| 48 |
from transformers import AutoTokenizer, AutoModelForQuestionAnswering
|
| 49 |
path = 'pedramyazdipoor/persian_xlm_roberta_large'
|