Upload README.md with huggingface_hub
Browse files
README.md
CHANGED
|
@@ -23,24 +23,44 @@ model-index:
|
|
| 23 |
- type: f1
|
| 24 |
value: 0.9458
|
| 25 |
name: F1 Weighted
|
|
|
|
|
|
|
|
|
|
| 26 |
---
|
| 27 |
|
| 28 |
# my-roberta-RQ3
|
| 29 |
|
| 30 |
This model is a fine-tuned version of [facebook/roberta-base](https://huggingface.co/facebook/roberta-base).
|
| 31 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 32 |
## Training results
|
| 33 |
| Training Loss | Epoch | Step | Validation Loss | Accuracy | F1 Macro | F1 Weighted |
|
| 34 |
|:---:|:---:|:---:|:---:|:---:|:---:|:---:|
|
| 35 |
-
| 0.3710 | 1.0 | 539 | 0.3667 | 0.9399 | 0.2785 | 0.9265 |
|
| 36 |
-
| 0.3190 | 2.0 | 1078 | 0.3273 | 0.9450 | 0.3737 | 0.9394 |
|
| 37 |
-
| 0.3055 | 3.0 | 1617 | 0.3055 | 0.9513 | 0.5054 | 0.9483 |
|
| 38 |
-
| 0.3008 | 4.0 | 2156 | 0.3008 | 0.9529 | 0.6363 | 0.9515 |
|
| 39 |
| ... | ... | ... | ... | ... | ... | ... |
|
| 40 |
-
| 0.4149 | 10.0 | 5390 | 0.
|
| 41 |
|
| 42 |
## Training hyperparameters
|
|
|
|
| 43 |
- learning_rate: 1e-05
|
| 44 |
- train_batch_size: 32
|
| 45 |
- eval_batch_size: 32
|
|
|
|
|
|
|
|
|
|
| 46 |
- num_epochs: 10
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 23 |
- type: f1
|
| 24 |
value: 0.9458
|
| 25 |
name: F1 Weighted
|
| 26 |
+
- type: loss
|
| 27 |
+
value: 0.4223
|
| 28 |
+
name: Loss
|
| 29 |
---
|
| 30 |
|
| 31 |
# my-roberta-RQ3
|
| 32 |
|
| 33 |
This model is a fine-tuned version of [facebook/roberta-base](https://huggingface.co/facebook/roberta-base).
|
| 34 |
|
| 35 |
+
## Test Results
|
| 36 |
+
Dựa trên kết quả đánh giá cuối cùng tại Epoch 10:
|
| 37 |
+
|
| 38 |
+
- **Eval Loss:** 0.4223
|
| 39 |
+
- **Eval Accuracy:** 0.9462
|
| 40 |
+
- **Eval F1 Macro:** 0.5894
|
| 41 |
+
- **Eval F1 Weighted:** 0.9458
|
| 42 |
+
- **Eval Runtime:** 33.1771 seconds
|
| 43 |
+
- **Samples Per Second:** 129.8790
|
| 44 |
+
- **Steps Per Second:** 2.0500
|
| 45 |
+
|
| 46 |
## Training results
|
| 47 |
| Training Loss | Epoch | Step | Validation Loss | Accuracy | F1 Macro | F1 Weighted |
|
| 48 |
|:---:|:---:|:---:|:---:|:---:|:---:|:---:|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 49 |
| ... | ... | ... | ... | ... | ... | ... |
|
| 50 |
+
| 0.4149 | 10.0 | 5390 | 0.4223 | 0.9462 | 0.5894 | 0.9458 |
|
| 51 |
|
| 52 |
## Training hyperparameters
|
| 53 |
+
The following hyperparameters were used during training:
|
| 54 |
- learning_rate: 1e-05
|
| 55 |
- train_batch_size: 32
|
| 56 |
- eval_batch_size: 32
|
| 57 |
+
- seed: 42
|
| 58 |
+
- optimizer: AdamW
|
| 59 |
+
- lr_scheduler_type: linear
|
| 60 |
- num_epochs: 10
|
| 61 |
+
- mixed_precision_training: Native AMP (fp16)
|
| 62 |
+
|
| 63 |
+
## Framework versions
|
| 64 |
+
- Transformers: 4.x.x
|
| 65 |
+
- Pytorch: 2.x.x
|
| 66 |
+
- Tokenizers: 0.x.x
|