Update README.md
Browse files
README.md
CHANGED
|
@@ -19,6 +19,8 @@ It achieves the following results on the evaluation set:
|
|
| 19 |
- Exact: 63.9013
|
| 20 |
- F1: 72.1497
|
| 21 |
|
|
|
|
|
|
|
| 22 |
### Before fine-tuning:
|
| 23 |
|
| 24 |
```
|
|
@@ -48,19 +50,20 @@ DatasetDict({
|
|
| 48 |
})
|
| 49 |
```
|
| 50 |
|
| 51 |
-
## Model description
|
| 52 |
-
|
| 53 |
-
More information needed
|
| 54 |
-
|
| 55 |
## Intended uses & limitations
|
| 56 |
|
| 57 |
-
|
| 58 |
|
| 59 |
## Training and evaluation data
|
| 60 |
|
| 61 |
-
|
|
|
|
|
|
|
| 62 |
|
| 63 |
## Training procedure
|
|
|
|
|
|
|
|
|
|
| 64 |
|
| 65 |
### Training hyperparameters
|
| 66 |
|
|
|
|
| 19 |
- Exact: 63.9013
|
| 20 |
- F1: 72.1497
|
| 21 |
|
| 22 |
+
## More accurate metrics:
|
| 23 |
+
|
| 24 |
### Before fine-tuning:
|
| 25 |
|
| 26 |
```
|
|
|
|
| 50 |
})
|
| 51 |
```
|
| 52 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 53 |
## Intended uses & limitations
|
| 54 |
|
| 55 |
+
This model is fine-tuned to answer questions from the DSPFirst textbook. I'm not really sure what I am doing so you should review it before using it.
|
| 56 |
|
| 57 |
## Training and evaluation data
|
| 58 |
|
| 59 |
+
- `batch_size` of 6 results in 14.82 GB VRAM
|
| 60 |
+
- 4.52 GB RAM
|
| 61 |
+
- 30% of the total questions is dedicated for evaluating.
|
| 62 |
|
| 63 |
## Training procedure
|
| 64 |
+
- The model was trained from Google Colab
|
| 65 |
+
- Utilizes Tesla P100 16GB, took 3.8 hours to train
|
| 66 |
+
- `load_best_model_at_end` is enabled in TrainingArguments
|
| 67 |
|
| 68 |
### Training hyperparameters
|
| 69 |
|