Upload README.md with huggingface_hub
Browse files
README.md
ADDED
|
@@ -0,0 +1,51 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: mit
|
| 3 |
+
library_name: transformers
|
| 4 |
+
---
|
| 5 |
+
# sweep-lr5e5-wd001
|
| 6 |
+
|
| 7 |
+
This model was selected as the best checkpoint from a hyperparameter sweep.
|
| 8 |
+
|
| 9 |
+
## Model Architecture
|
| 10 |
+
|
| 11 |
+
<div align="center">
|
| 12 |
+
<img src="figures/arch_diagram.png" width="70%" alt="Architecture Diagram" />
|
| 13 |
+
</div>
|
| 14 |
+
|
| 15 |
+
## Training Configuration
|
| 16 |
+
|
| 17 |
+
| Parameter | Value |
|
| 18 |
+
|---|---|
|
| 19 |
+
| learning_rate | 5e-05 |
|
| 20 |
+
| weight_decay | 0.001 |
|
| 21 |
+
| epochs | 30 |
|
| 22 |
+
| batch_size | 16 |
|
| 23 |
+
| model_arch | deberta-v3-base |
|
| 24 |
+
|
| 25 |
+
## Evaluation Metrics
|
| 26 |
+
|
| 27 |
+
| Metric | Value |
|
| 28 |
+
|---|---|
|
| 29 |
+
| val_loss | 0.198 |
|
| 30 |
+
| val_accuracy | 0.934 |
|
| 31 |
+
| f1_score | 0.921 |
|
| 32 |
+
| inference_latency_ms | 14.2 |
|
| 33 |
+
|
| 34 |
+
## Training Loss Curve
|
| 35 |
+
|
| 36 |
+
<div align="center">
|
| 37 |
+
<img src="figures/loss_curve.png" width="80%" alt="Training Loss Curve" />
|
| 38 |
+
</div>
|
| 39 |
+
|
| 40 |
+
## Usage
|
| 41 |
+
|
| 42 |
+
```python
|
| 43 |
+
from transformers import AutoModelForSequenceClassification, AutoTokenizer
|
| 44 |
+
|
| 45 |
+
model = AutoModelForSequenceClassification.from_pretrained("BestSweepModel-TestRepo")
|
| 46 |
+
tokenizer = AutoTokenizer.from_pretrained("BestSweepModel-TestRepo")
|
| 47 |
+
```
|
| 48 |
+
|
| 49 |
+
## License
|
| 50 |
+
|
| 51 |
+
This model is released under the [MIT License](LICENSE).
|